Datasets:

Hendrik-a

/

arxiv

like 0

the results of _ chandra _ snapshot observations of 11 liners ( low - ionization nuclear emission - line regions ) , three low - luminosity seyfert galaxies , and one -liner transition object are presented . our sample consists of all the objects with a flat or inverted spectrum compact radio core in the vla survey of 48 low - luminosity agns ( llagns ) by nagar et al . ( 2000 ) . an x - ray nucleus is detected in all galaxies except one and their x - ray luminosities are in the range to ergs s . the x - ray spectra are generally steeper than expected from thermal bremsstrahlung emission from an advection - dominated accretion flow ( adaf ) . the x - ray to h luminosity ratios for 11 out of 14 objects are in good agreement with the value characteristic of llagns and more luminous agns , and indicate that their optical emission lines are predominantly powered by a llagn . for three objects , this ratio is less than expected . comparing with properties in other wavelengths , we find that these three galaxies are most likely to be heavily obscured agn . we use the ratio(5 ghz)/ , where is the luminosity in the 210 kev band , as a measure of radio loudness . in contrast to the usual definition of radio loudness ((5 ghz)/(b ) ) , can be used for heavily obscured ( .5ex , mag ) nuclei . further , with the high spatial resolution of _ chandra _ , the nuclear x - ray emission of llagns is often easier to measure than the nuclear optical emission . we investigate the values of for llagns , luminous seyfert galaxies , quasars and radio galaxies and confirm the suggestion that a large fraction of llagns are radio loud .

astro-ph0209607

low - ionization nuclear emission - line regions ( liners ; heckman 1980 ) are found in many nearby bright galaxies ( e.g. , ho , filippenko , & sargent 1997a ) . extensive studies at various wavelengths have shown that type 1 liners ( liner 1s , i.e. , those galaxies having broad h@xmath3 and possibly other broad balmer lines in their nuclear optical spectra ) are powered by a low - luminosity agn ( llagn ) with a bolometric luminosity less than @xmath14 ergs s@xmath2 ( ho et al . 2001 ; terashima , ho , & ptak 2000a ; ho et al . 1997b ) . on the other hand , the energy source of liner 2s is likely to be heterogeneous . some liner 2s show clear signatures of the presence of an agn , while others are most probably powered by stellar processes , and the luminosity ratio @xmath6/@xmath15 can be used to discriminate between these different power sources ( e.g. , prez - olea & colina 1996 ; maoz et al . 1998 ; terashima et al . it is interesting to note that currently there are only a few liner 2s known to host an obscured agn ( e.g. , turner et al . this paucity of obscured agn in liners may indicate that liner 2s are not simply a low - luminosity extension of luminous seyfert 2s , which generally show heavy obscuration with a column density averaging @xmath10 @xmath16 @xmath11 @xmath12 ( e.g. , turner et al . 1997 ) . alternatively , biases against finding heavily obscured llagns may be important . for example , objects selected through optical emission lines or x - ray fluxes are probably biased in favor of less absorbed ones , even if one uses the x - ray band above 2 kev . in contrast , radio observations , particularly at high frequency , are much less affected by absorption . although an optical spectroscopic survey must first be done to find the emission lines characteristic of a llagn , follow up radio observations can clarify the nature of the activity . for example , vlbi observations of some llagns have revealed a compact nuclear radio source with @xmath17 k , which is an unambiguous indicator of the presence of an active nucleus and can not be produced by starburst activity ( e.g. , falcke et al . 2000 ; ulvestad & ho 2001 ) . a number of surveys of seyfert galaxies at sub arcsecond resolution have been made with the vla ( ulvestad & wilson 1989 and references therein ; kukula et al . 1995 ; nagar et al . 1999 ; thean et al . 2000 ; schmitt et al . 2001 ; ho & ulvestad 2001 ) and other interferometers ( roy et al . 1994 ; morganti et al . 1999 ) , but much less work has been done on the nuclear radio emission of liners . nagar et al . ( 2002 ) have reported a vla 2 cm radio survey of all 96 llagns within a distance of 19 mpc . these llagns come from the palomar spectroscopic survey of bright galaxies ( ho et al . 1997a ) . as a pilot study of the x - ray properties of llagns , we report here a _ chandra _ survey of a subset , comprising 15 galaxies , of nagar et al s ( 2002 ) sample . fourteen of these galaxies have a compact nuclear radio core with a flat or inverted radio spectrum ( nagar et al . 2000 ) . we have detected 13 of the galactic nuclei with _ chandra_. we also examine the `` radio loudness '' of our sample and compare it with other classes of agn . a new measure of `` radio loudness '' is developed , in which the 5 ghz radio luminosity is compared with the 210 kev x - ray luminosity ( @xmath9=@xmath18 ( 5 ghz)/@xmath19(210 kev ) ) rather than with the b - band optical luminosity ( @xmath20(5 ghz)/@xmath8(b ) ) , as is usually done . @xmath9 has the advantage that it can be measured for highly absorbed nuclei ( @xmath10 up to several times @xmath11 @xmath12 ) which would be totally obscured ( @xmath21 up to a few hundred mag for the galactic gas to dust ratio ) at optical wavelengths , and that the compact , hard x - ray source in a llagn is less likely to be confused with emission from stellar - powered processes than is an optical nucleus . this paper is organized as follows . the sample , observations , and data reduction are described in section 2 . imaging results and x - ray source detections are given in section 3 . section 4 presents spectral results . the power source , obscuration in llagns , and radio loudness of llagns are discussed in section 5 . section 6 summarizes the findings . we use a hubble constant of @xmath22 km s@xmath2 mpc@xmath2 and a deceleration parameter of @xmath23 throughout this paper . our sample is based on the vla observations by nagar et al . their sample of 48 objects consists of 22 liners , 18 transition objects , which show optical spectra intermediate between liners and nuclei , and eight low - luminosity seyferts selected from the optical spectroscopic survey of ho et al . ( 1997a ) . the sample is the first half of a distance - limited sample of llagns ( nagar et al . 2002 ) , as described in section 1 . we selected 14 objects showing a flat to inverted spectrum radio core ( @xmath24 , @xmath25 ) according to nagar et al.s ( 2000 ) comparison with longer wavelength radio data published in the literature . one object ( the liner 2 ngc 4550 ) has a flat spectrum radio source at a position significantly offset from the optical nucleus . this object was added as an example of a liner without a detected radio core . the target list and some basic data for the final sample are summarized in table 1 . the distances are taken from tully ( 1988 ) in which @xmath22 km s@xmath2 mpc@xmath2 is assumed . the sample consists of seven liner 1s , four liner 2s , two seyfert 1s , one seyfert 2 , and one transition 2 object . 12 out of these 15 objects have been observed with the vlba and high brightness temperature ( @xmath26 k ) radio cores were detected in all of them ( falcke et al . 2000 ; ulvestad & ho 2001 ; nagar et al . therefore , these objects are strong candidates for agns . results of archival / scheduled _ chandra _ observations of more liners with a compact flat / inverted spectrum radio core found by nagar et al . ( 2002 ) will be presented in a future paper . a log of _ chandra _ observations is shown in table 1 . the exposure time was typically two ksec each . all the objects were observed at or near the aim point of the acis - s3 back - illuminated ccd chip . eight objects were observed in our guaranteed time observation program and the rest of the objects were taken from the _ chandra _ archives . the eight objects were observed in 1/8 sub - frame mode ( frame time 0.4 s ) to minimize effects of pileup ( e.g. , davis 2001 ) . 1/2 sub - frame modes were used for three objects . ciao 2.2.1 and caldb 2.7 were used to reduce the data . in the following analysis , only events with _ asca _ grades 0 , 2 , 3 , 4 , 6 ( `` good grades '' ) were used . for spectral fitting , xspec version 11.2.0 was employed . an x - ray nucleus is seen in all the galaxies except for ngc 4550 and ngc 5866 . some off - nuclear sources are also seen in some fields . the source detection algorithm `` wavdetect '' in the ciao package was applied to detect these nuclear and off - nuclear sources , where a detection threshold of @xmath27 and wavelet scales of 1 , @xmath28 , 2 , @xmath29 , 4 , @xmath30 , 8 , @xmath31 , and 16 pixels were used . source detections were performed in the three energy bands 0.58 kev ( full band ) , 0.52 kev ( soft band ) , and 28 kev ( hard band ) . the resulting source lists and raw images were examined by eye to exclude spurious detections . the source and background counts were also determined by manual photometry and compared with the results of wavdetect . in the few cases that the two methods gave discrepant results , we decided to use the results of the manual photometry after inspection of the raw images . some sources were detected in only one or two energy bands . in such cases , we calculated the upper limits on the source counts in the undetected band(s ) at the 95% confidence level by interpolating the values in table 2 of kraft , burrows , & nousek ( 1991 ) . table 2 shows the positions , detected counts , band ratios ( hard / soft counts ) , fluxes and luminosities in the 210 kev band of the nuclear sources . the same parameters for off - nuclear sources with signal - to - noise ratios greater than three are summarized in table 3 . for bright objects ( @xmath32 40 counts ) , the fluxes were measured by spectral fits presented in the next section . for faint objects , fluxes were determined by assuming the galactic absorption column density and power law spectra , with photon indices determined from the band ratios . when only lower or upper limits on the band ratio were available , a photon index of 2 was assumed if the limit is consistent with @xmath33 . when the band ratio is inconsistent with @xmath33 , the upper or lower limit is used to determine the photon index . luminosities were calculated only if the source is spatially inside the optical host galaxy as indicated by comparing the position with the optical image of the digitized sky survey . possible identifications for off - nuclear sources are also given in the last column of table 3 . the positions of the x - ray nuclei coincide with the radio core positions to within the positional accuracy of _ chandra_. the nominal separations between the x - ray and radio nuclei are in the range 0.05 0.95@xmath34 . inspection of the images shows that the nucleus in most objects appears to be unresolved , while some objects show faint extended emission . the soft and hard band images of the nuclear regions of ngc 3169 and ngc 4278 are shown in fig 1 as examples of extended emission . in the soft band image of ngc 3169 , emission in the nuclear region extending @xmath35 arcsec in diameter is clearly visible , while the nucleus itself is not detected ( table 2 ) . about 25 counts were detected within a 10 pixel ( 4.9 arcsec ) radius in the @xmath36 kev band . this extended emission is not seen in the hard band . the soft band image of ngc 4278 consists of a bright nucleus and a faint elongated feature with a length of @xmath37 and a position angle of @xmath38 . the hard band image is unresolved . the nuclear regions ( 10 arcsec scale ) of the other galaxies look compact to within the current photon statistics . more extended diffuse emission at larger scales ( @xmath39 ) is seen in a few objects . the nuclei of ngc 2787 and ngc 4203 are embedded in soft diffuse emission with diameters of @xmath40 and @xmath41 , respectively . ngc 4579 shows soft diffuse emission with a similar morphology to the circumnuclear star forming ring , in addition to a very bright nucleus ( see also eracleous et al . ngc 4565 shows extended emission along the galactic plane . ngc 5866 has soft extended emission @xmath42 ( 2 kpc ) in diameter and no x - ray nucleus is detected ( table 2 ) . any diffuse emission associated with the other galaxies is much fainter . spectral fits were performed for the relatively bright objects those with @xmath32 40 detected counts in the 0.58 kev band . the spectrum of one fainter object ( ngc 4548 ) showing a large ( = hard ) hardness ratio was also fitted . some objects are so bright that pileup effects are significant . column 7 of table 1 gives the count rates per ccd read - out frame time and can be used to estimate the significance of pileup . in the two objects ngc 3147 and ngc 4278 , the pileup is mild and we corrected for the effect by applying the pileup model implemented in xspec , where the grade morphing parameter @xmath3 was fixed at 0.5 ( after initially treating it as a free parameter [ davis 2001 ] since @xmath3 is not well constrained ) . the pileup effect for the three objects with the largest count rate per frame ( ngc 4203 , ngc 4579 , and ngc 5033 ) is serious and we did not attempt detailed spectral fits . instead , we use the spectra and fluxes measured with _ asca _ for these three objects ( terashima et al . 2002b and references therein ) in the following discussions . we confirmed that the nuclear x - ray source dominates the hard x - ray emission within the beam size of _ asca _ ( see appendix ) . the other objects in the sample are faint enough to ignore the effects of pileup . x - ray spectra were extracted from a circular region with a radius between 4 pixels ( 2.0@xmath34 ; for faint sources ) and 10 pixels ( 4.9@xmath34 ; for bright sources ) depending on source brightness . background was estimated using an annular region centered on the target . a maximum - likelihood method using the c - statistic ( cash 1979 ) was employed in the spectral fits . in the fit with the c - statistic , background can not be subtracted , so we added a background model ( measured from the background region ) with fixed parameters to the spectral models , after normalizing by the ratio of the geometrical areas of the source and background regions . the errors quoted represent the 90% confidence level for one parameter of interest ( @xmath43=2.7 ) . a power - law model modified by absorption was applied and acceptable fits were obtained in all cases ( fig . the best - fit parameters for the nuclear sources are given in table 4 . the observed fluxes and luminosities ( the latter corrected for absorption ) in the 210 kev band are shown in table 2 . the results for a few bright off - nuclear sources are presented in the appendix . the photon indices of the nuclear sources are generally consistent with the typical values observed in llagns ( photon index @xmath44 , e.g. , terashima et al . 2002a , 2002b ) , although errors are quite large due to the limited photon statistics . the spectral slope of ngc 6500 ( @xmath45 ) is somewhat steeper than is typical of llagns . this may indicate that there is soft x - ray emission from a source other than the agn and/or the intrinsic slope of the agn is steep . one galaxy ( ngc 3169 , a liner 2 ) has a large absorption column ( @xmath10=@xmath46 @xmath12 ) , while two galaxies ( ngc 4548 , a liner 2 , and ngc 3226 , a liner 1.9 ) show substantial absorption ( @xmath47 @xmath12 , and @xmath48 @xmath12 , respectively ) . others have small column densities which are consistent with ` type 1 ' agns . no meaningful limit on the equivalent width of an fe k@xmath3 line was obtained for any of the objects because of limited photon statistics in the hard x - ray band . one object ngc 2787 has only 8 detected photons in the 0.58 kev band and is too faint to obtain spectral information . a photon index of 2.0 and the galactic absorption of @xmath49 @xmath12 were assumed to calculate the flux and luminosity which are shown in table 2 . an x - ray nucleus is detected in all the objects except for ngc 4550 and ngc 5866 . we test whether the detected x - ray sources are the high energy extension of the continuum source which powers the optical emission lines by examining the luminosity ratio @xmath6/@xmath15 . the h@xmath3 luminosities ( @xmath15 ) were taken from ho et al . ( 1997a ) and the reddening was estimated from the balmer decrement for narrow lines and corrected using the reddening curve of cardelli , clayton , & mathis ( 1989 ) , assuming the intrinsic h@xmath3/h@xmath50 flux ratio = 3.1 . the x - ray luminosities ( corrected for absorption ) in the @xmath51 kev band are used . the h@xmath3 luminosities and logarithm of the luminosity ratios @xmath6/@xmath15 are shown in table 5 . the @xmath6/@xmath15 ratios of most objects are in the range of agns ( @xmath52 @xmath6/@xmath15 @xmath16 12 ) and in good agreement with the strong correlation between @xmath6 and @xmath15 for llagns , luminous seyferts , and quasars presented in terashima et al . ( 2000a ) and ho et al . this indicates that their optical emission lines are predominantly powered by a llagn . note that this correlation is not an artifact of distance effects , as shown in terashima et al . ( 2000a ) . the four objects ngc 2787 , ngc 4550 , ngc 5866 , and ngc 6500 , however , have much lower @xmath6/@xmath15 ratios ( @xmath52 @xmath6/@xmath15 .5ex0 ) than expected from the correlation ( @xmath52 @xmath6/@xmath15 @xmath53 ) , and their x - ray luminosities are insufficient to power the h@xmath3 emission ( terashima et al . this x - ray faintness could indicate one or more of several possibilities such as ( 1 ) an agn is the power source , but is heavily absorbed at energies above 2 kev , ( 2 ) an agn is the power source , but is currently switched - off or in a faint state , and ( 3 ) the optical narrow emission lines are powered by some source(s ) other than an agn . we briefly discuss these three possibilities in turn . if an agn is present in these x - ray faint objects and absorbed in the hard energy band above 2 kev , only scattered and/or highly absorbed x - rays would be observed , and then the intrinsic luminosity would be much higher than that observed . this can account for the low @xmath6/@xmath15 ratios and high radio to x - ray luminosity ratios ( @xmath54(5 ghz)/@xmath6 ; table 5 and section 5.3 ) . if the intrinsic x - ray luminosities are about one or two orders of magnitude higher than those observed , as is often inferred for seyfert 2 galaxies ( turner et al . 1997 , awaki et al . 2000 ) , @xmath6/@xmath15 and @xmath54(5 ghz)/@xmath6 become typical of llagns . alternatively , the agn might be turned off or in a faint state , with a higher activity in the past being inferred from the optical emission lines , whose emitting region is far from the nucleus ( e.g. , eracleous et al . also , the radio observations were made a few years before the _ chandra _ ones . this scenario might thus explain their relatively low @xmath6/@xmath15 ratios and their relatively high @xmath55/@xmath6 ratios . if this is the case , the size of the radio core can be used to constrain the era of the active phase in the recent past . the upper limits on the size of the core estimated from the beam size ( @xmath56 2.5 mas ) are 0.16 , 0.19 , and 0.48 pc for ngc 2787 , ngc 5866 , and ngc 6500 , respectively ( falcke et al . therefore , the agn must have been active until @xmath570.52 , @xmath570.60 , and @xmath571.6 years , respectively , before the vlba observations ( made in 1997 june ) and inactive at the epochs ( 2000 jan 2002 jan , see table 1 ) of the x - ray observations . this is an ad hoc proposal and such abrupt declines of activity are quite unusual , but it can not be completely excluded . it may also be possible that the ionized gas inferred from the optical emission lines is ionized by some sources other than an agn , such as hot stars . if the observed x - rays reflect the intrinsic luminosities of the agn , a problem with the agn scenario for the three objects ngc 2787 , ngc 5866 , and ngc 6500 is that these galaxies have very large @xmath18(5 ghz)/@xmath6 ratios , and would thus be among the radio loudest llagns . the presence of hot stars in the nuclear region of ngc 6500 is suggested by uv spectroscopy ( maoz et al . maoz et al . ( 1998 ) studied the energy budget for ngc 6500 by using the h@xmath3 and uv luminosity at 1300 a and showed that the observed uv luminosity is insufficient to power the h@xmath3 luminosity even if a stellar population with the salpeter initial mass function and a high mass cutoff of 120@xmath58 are assumed . this result indicates that a power source in addition to hot stars must contribute significantly , and supports the obscured agn interpretation discussed above . the first possibility , i.e. , an obscured low - luminosity agn as the source of the x - ray emission , seems preferable for ngc 2787 , ngc 5866 and ngc 6500 , although some other source(s ) may contribute to the optical emission lines . additional lines of evidence which support the presence of an agn include the fact that all three of these galaxies ( ngc 2787 , ngc 5866 , and ngc 6500 ) have vlbi - detected , sub - pc scale , nuclear radio core sources ( falcke et al . 2000 ) , a broad h@xmath3 component ( in ngc 2787 , and an ambiguous detection in ngc 5866 ; ho et al . 1997b ) , a variable radio core in ngc 2787 , and a jet - like linear structure in a high - resolution radio map of ngc 6500 with the vlba ( falcke et al . only an upper limit to the x - ray flux is obtained for ngc 5866 . if an x - ray nucleus is present in this galaxy and its luminosity is only slightly below the upper limit , this source could be an agn obscured by a column density @xmath10@xmath59 @xmath12 or larger . if the apparent x - ray luminosity of the nucleus of ngc 5866 is _ much _ lower than the observed upper limit , and the intrinsic x - ray luminosity conforms to the typical @xmath6/@xmath15 ratio for llagn ( @xmath52 @xmath6/@xmath15 @xmath60 ) , then the x - ray source must be almost completely obscured . the optical classification ( transition object ) suggests the presence of an ionizing source other than an agn , so the low observed @xmath6/@xmath15 ratio could alternatively be a result of enhanced h@xmath3 emission powered by this other ionizing source . the x - ray results presented above show that the presence of a flat ( or inverted ) spectrum compact radio core is a very good indicator of the presence of an agn even if its luminosity is very low . on the other hand , ngc 4550 , which does not possess a radio core , shows no evidence for the presence of an agn and all the three possibilities discussed above are viable . if the _ rosat _ detection is real ( halderson et al . 2001 ) , the time variability between the _ rosat _ and _ chandra _ fluxes may indicate the presence of an agn ( see appendix ) . it is notable that type 2 liners without a flat spectrum compact radio core may be heterogeneous in nature . for instance , some liner 2s without a compact radio core ( e.g. , ngc 404 and transition 2 object ngc 4569 ) are most probably driven by stellar processes ( maoz et al . 1998 ; terashima et al . 2000b ; eracleous et al . 2002 ) . in our sample , we found at least three highly absorbed llagns ( ngc 3169 , ngc 3226 , and ngc 4548 ) . in addition , if the x - ray faint objects discussed in section 5.1 are indeed agns , they are most probably highly absorbed with @xmath10@xmath61 @xmath12 . among these absorbed objects , ngc 2787 is classified as a liner 1.9 , ngc 3169 , ngc 4548 , and ngc 6500 as liner 2s , and ngc 5866 as a transition 2 object . thus , heavily absorbed liner 2s , of which few are known , are found in the present observations demonstrating that radio selection is a valuable technique for finding obscured agns . along with heavily obscured llagns known in low - luminosity seyfert 2s ( e.g. , ngc 2273 , ngc 2655 , ngc 3079 , ngc 4941 , and ngc 5194 ; terashima et al . 2002a ) , our observations show that at least some type 2 llagns are simply low - luminosity counterparts of luminous seyferts in which heavy absorption is often observed ( e.g. , risaliti , maiolino , & salvati 1999 ) . however , some liner 2s ( e.g. , ngc 4594 , terashima et al . 2002a ; ngc 4374 , finoguenov & jones 2001 ; ngc 4486 , wilson & yang 2002 ) and low - luminosity seyfert 2s ( ngc 3147 ; section 4 and appendix ) show no strong absorption . therefore , the orientation - dependent unified scheme ( e.g. , antonucci 1993 ) does not always apply to agns in the low - luminosity regime , as suggested by terashima et al . . combination of x - ray and radio observations is valuable for investigating a number of areas of agn physics , including the `` radio loudness '' , the origin of jets , and the structure of accretion disks . low - luminosity agns ( liners and low - luminosity seyfert galaxies ) are thought to be radiating at very low eddington ratios ( @xmath62/@xmath63 ) and may possess an advection - dominated accretion flow ( adaf ; see e.g. , quataert 2002 for a recent review ) . a study of radio loudness in llagns can constrain the jet production efficiency by an adaf - type disk . earlier studies have suggested that llagns tend to be radio loud compared to more luminous seyferts based on the spectral energy distributions of seven llagns ( ho 1999 ) and , for a larger sample , on the conventional definition of radio loudness @xmath64(5 ghz)/@xmath8(b ) ( the subscript `` o '' , which stands for optical , is usually omitted but we use it here to distinguish from @xmath9 see below ) , with @xmath65 being radio loud ( kellermann et al . 1989 , 1994 ; visnovsky et al . 1992 ; stocke et al . 1992 ; ho & peng 2001 ) . ho & peng ( 2001 ) measured the luminosities of the nuclei by spatial analysis of optical images obtained with _ hst _ to reduce the contribution from stellar light . a caveat in the use of optical measurements for the definition of radio loudness is extinction , which will lead to an overestimate of @xmath66 if not properly allowed for . although ho & peng ( 2001 ) used only type 11.9 objects , some objects of these types show high absorption columns in their x - ray spectra . in this subsection , we study radio loudness by comparing radio and hard x - ray luminosities . since the unabsorbed luminosity for objects with @xmath10 .5ex@xmath11 @xmath12 can be reliably measured in the 210 kev band , which is accessible to _ asca _ , _ xmm - newton _ , and _ chandra _ , and such columns correspond to @xmath21 .5ex50 mag , it is clear that replacement of optical by hard x - ray luminosity potentially yields considerable advantages . in addition , the high spatial resolutions of _ xmm - newton _ and especially _ chandra _ usually allow the nuclear x - ray emission to be identified unambiguously , while the optical emission of llagn can be confused by surrounding starlight . in the following analysis , radio data at 5 ghz taken from the literature are used since fluxes at this frequency are widely available for various classes of objects . we used primarily radio luminosities obtained with the vla at .5ex@xmath67 resolution for the present sample . high resolution vla data at 5 ghz are not available for several objects . for four such cases , vlba observations at 5 ghz with 150 mas resolution are published in the literature ( falcke et al . 2000 ) and are used here . for two objects , we estimated 5 ghz fluxes from 15 ghz data by assuming a spectral slope of @xmath68 ( cf . nagar et al . the radio luminosities used in the following analysis are summarized in table 5 . since our sample is selected based on the presence of a compact radio core , the sample could be biased to more radio loud objects . therefore , we constructed a larger sample by adding objects taken from the literature for which 5 ghz radio , 210 kev x - ray , and @xmath69 measurements are available . first , we introduce the ratio @xmath70(5 ghz)/@xmath6 as a measure of radio loudness and compare the ratio with the conventional @xmath69 parameter . the x - ray luminosity @xmath6 in the 210 kev band ( source rest frame ) , corrected for absorption , is used . , which utilizes monochromatic b - band luminosities . this alternative provides completely identical results if the x - ray spectral shape is known and the range of spectral slopes is not large . for example , the conversion factor @xmath8(2 kev)/@xmath6 is 0.31 , 0.26 , and 0.22 kev@xmath2 for photon indices of 2 , 1.8 , and 1.6 , respectively , and no absorption . ] we examine the behavior of @xmath9 using samples of agn over a wide range of luminosity , including llagn , the seyfert sample of ho & peng ( 2001 ) and pg quasars which are also used in their analysis . @xmath69 parameters and radio luminosities were taken from ho & peng ( 2001 ) for the seyferts and kellermann et al . ( 1989 ) for the pg sample . the values of @xmath69 in kellermann et al . ( 1989 ) have been recalculated by using only the core component of the radio luminosities . the optical and radio luminosities of the pg quasars were calculated assuming @xmath71 and @xmath72 ( @xmath73 ) . the x - ray luminosities ( mostly measured with _ asca _ ) were compiled from terashima et al . ( 2002b ) , weaver , gelbord , & yaqoob ( 2001 ) , george et al . ( 2000 ) , reeves & turner ( 2000 ) , iwasawa et al . ( 1997 , 2000 ) , sambruna , eracleous , & mushotzky ( 1999 ) , nandra et al . ( 1997 ) , smith & done ( 1996 ) , and cappi et al . note that only a few objects ( ngc 4565 , ngc 4579 , and ngc 5033 ) in our radio selected sample have reliable measurements of nuclear @xmath8(b ) . 3 . compares the parameters @xmath69 and @xmath9 for the seyferts and pg sample . these two parameters correlate well for most seyferts . some seyferts have higher @xmath69 values than indicated by most seyferts . this could be a result of extinction in the optical band . seyferts showing x - ray spectra absorbed by a column greater than @xmath74 @xmath12 ( ngc 2639 , 4151 , 4258 , 4388 , 4395 , 5252 , and 5674 ) are shown as open circles in fig . at least four of them have a value of @xmath69 larger than indicated by the correlation . the correlation between @xmath75 and @xmath76 for the less absorbed seyferts can be described as @xmath75 = 0.88 @xmath77 + 5.0 . according to this relation , the boundary between radio loud and radio quiet object ( @xmath75 = 1 ) corresponds to @xmath78 . the values of @xmath69 and @xmath9 for a few obscured seyferts are consistent with the correlation , indicating that optical extinction is not perfectly correlated with the absorption column density inferred from x - ray spectra . the pg quasars show systematically lower @xmath69 values than those of seyferts at a given @xmath76 . for the former objects , @xmath79 corresponds to @xmath80 . this apparently reflects a luminosity dependence of the shape of the sed : luminous objects have steeper optical - x - ray slopes @xmath81 ( @xmath82 ; e.g. , elvis et al . 1994 , brandt , laor , & wills 2000 ) , where @xmath83 is often measured as the spectral index between 2200 a and 2 kev , while less luminous agns have @xmath84 ( ho 1999 ) . this is related to the fact that luminous objects show a more prominent `` big blue bump '' in their spectra . fig . 8 of ho ( 1999 ) demonstrates that low - luminosity objects are typically 11.5 orders of magnitude fainter in the optical band than luminous quasars for an given x - ray luminosity . note that none of the pg quasars used here shows a high absorption column in its x - ray spectrum below 10 kev . the definition of radio loudness using the hard x - ray flux ( @xmath85 ) appears to be more robust than that using the optical flux because x - rays are less affected by both extinction at optical wavelengths and the detailed shape of the blue bump , as noted above . further , measurements of nuclear x - ray fluxes of seyferts and llagns with _ chandra _ are easier than measurements of nuclear optical fluxes , since in the latter case the nuclear light must be separated from the surrounding starlight , a difficult process for llagns . 4 shows the x - ray luminosity dependence of @xmath9 . in this plot , the llagn sample discussed in the present paper is shown in addition to the seyfert and pg samples used above . this is an `` x - ray version '' of the @xmath75-@xmath86 plot ( fig . 4 in ho & peng 2001 ) . radio galaxies taken from sambruna et al . ( 1999 ) are also added and we use radio luminosities from the core only . our plot shows that a large fraction ( @xmath87% ) of llagns ( @xmath6@xmath88 ergs s@xmath2 ) are `` radio loud '' . this is a confirmation of ho & peng s ( 2001 ) finding . note , however , that our sample is not complete in any sense , and this radio - loud fraction should be measured using a more complete sample . since radio emission in llagns is likely to be dominated by emission from jets ( nagar et al . 2001 ; ulvestad & ho 2001 ) , these results suggest that , in llagn , the fraction of the accretion energy that powers a jet , as opposed to electromagnetic radiation , is larger than in more luminous seyfert galaxies and quasars . since llagns are thought to have an adaf - type accretion flow , such might indicate that an adaf can produce jets more efficiently than the geometrically thin disk believed present in more luminous seyferts . the three llagns with the largest @xmath9 in fig . 4 are the three x - ray faint objects discussed in section 5.1 ( ngc 2787 , ngc 5866 , and ngc 6500 ) and which are most probably obscured agns . if their intrinsic x - ray luminosities are 12 orders of magnitude higher than those observed , their values of @xmath9 become smaller by this factor and are then in the range of other llagns . even if we exclude these three llagns , the radio loudness of llagns is distributed over a wide range : the radio - loudest llagns have @xmath9 values similar to radio galaxies and radio - loud quasars , while some llagns are as radio quiet as radio - quiet quasars . a comparison with blazars is of interest to compare our sample with objects for which the nuclear emission is known to be dominated by a relativistic jet and thus strongly beamed . the average @xmath89 for high - energy peaked bl lac objects ( hbls ) , low - energy peaked bl lac objects ( lbls ) , and flat spectrum radio quasars ( fsrqs ) are 3.10 , 1.27 , and 0.95 , respectively , where we used the average radio and x - ray luminosities for a large sample of blazars given in table 3 of donato et al . the average @xmath76 for hbls is similar to that for llagns in our sample , while the latter two classes are about two orders of magnitude more radio loud than llagns . although llagns and hbls have similar values of @xmath89 , the spectral slope in the x - ray band is different : llagns have a photon index in the range 1.72.0 ( see also terashima et al . 2002 ) , while hbls usually show steeper spectra ( photon index @xmath32 2 , e.g. , fig . 1 in donato et al . 2001 ) , and the x - ray emission is believed to be dominated by synchrotron radiation . furthermore , blazars with a lower bolometric luminosity tend to have a synchrotron peak at a higher frequency and a steeper x - ray spectral slope than higher bolometric luminosity blazars ( donato et al . 2001 ) . we also constructed an @xmath9-@xmath6 plot ( fig . 5 ) using the _ total _ radio luminosities of the radio source ( i.e. including the core , jets , lobes , and hot spots , if present ) . the radio data were compiled from vron - cetty & vron ( 2001 ) , kellermann et al . ( 1989 ) , and sambruna et al . the pg sample and other quasars are shown with different symbols . this plot appears similar to fig . 4 for llagns , seyferts , and radio - quiet quasars since these objects do not possess powerful jets or lobes and off - nuclear radio emission associated with the agn is generally of low luminosity ( ulvestad & wilson 1989 , nagar et al . 2001 , ho & ulvestad 2001 , kellermann et al . 1989 ) . on the other hand , radio galaxies have powerful extended radio emission and consequently the @xmath9 values calculated using the total radio luminosities become higher than if only nuclear luminosities are used . we used the same x - ray luminosities as in fig . 4 , because jets , lobes , and hot spots are almost always much weaker than the nucleus in x - rays . in fact , in our observations of llagns , we found no extended emission directly related to the agn . thus , the differences between fig . 4 and fig . 5 result from the extended radio emission . fourteen galaxies with a nuclear radio source having a flat or inverted spectrum have been observed with _ chandra _ with a typical exposure time of 2 ksec . an x - ray nucleus is detected in all but one object ( ngc 5866 ) . 11 galaxies have x - ray and h@xmath3 luminosities in good accord with the correlation known for agns over a wide range of luminosity , which indicates that these objects are agns and that the agn is the dominant power source of their optical emission lines . their x - ray luminosities are between @xmath0 and @xmath1 ergs s@xmath2 . the three objects ngc 2787 , ngc 5866 , and ngc 6500 have significantly lower x - ray luminosities than expected from the @xmath6-@xmath15 correlation . various observations suggest that these objects are most likely to be heavily obscured agns . these observational results show that radio and hard x - ray observations provide an efficient way to find llagn in nearby galaxies , even if the nuclei are heavily obscured . one object ( the liner 2 ngc 4550 ) , which does not show a radio core , was also observed for comparison . no x - ray nucleus is detected . if the x - ray source detected in this galaxy with _ rosat _ is indeed the nucleus , the nucleus must be variable in x - rays , which would indicate the presence of an agn . we have used the ratio @xmath90(5 ghz)/@xmath6 as a measure of radio loudness and found that a large fraction of llagns are radio loud . this confirms earlier results based on nuclear luminosities in the optical band , but our results based on hard x - ray measurements are much less affected by obscuration and the detailed shape of the `` big blue bump '' . we speculate that the increase in @xmath9 as @xmath6 decreases below @xmath91 ergs s@xmath2 may result from the presence of an advection - dominated accretion flow in the inner part of the accretion flow in low - luminosity objects . however , the steep x - ray spectra in our sample of llagns rule out high temperature thermal bremsstrahlung as the x - ray emission mechanism . is supported by the japan society for the promotion of science postdoctoral fellowship for young scientists . this research was supported by nasa through grants nag81027 and nag81755 to the university of maryland . in this appendix , we compare our results with previously published results particularly in the hard x - ray band obtained with _ asca _ and _ chandra_. the optical spectroscopic classification is given in parentheses after the object name . _ ngc 3147 ( s2)_. this object was observed with _ asca _ in 1993 september and the observed flux was @xmath92 ergs s@xmath2@xmath12 in the 210 kev band ( ptak et al . 1996 , 1999 ; terashima et al . our _ chandra _ image is dominated by the nucleus and shows that the off - nuclear source contribution within the _ asca _ beam is negligible . therefore , a comparison between the observed _ chandra _ flux ( @xmath93 ergs s@xmath2@xmath12 ) , which is 2.3 times larger than that of _ asca _ , implies time variability providing additional evidence for the presence of an agn . in the _ asca _ spectrum , a strong fe - k emission line is detected at @xmath94 kev ( source rest frame ) with an equivalent width of @xmath95 ev . one interpretation of this relatively large equivalent width is that the nucleus is obscured by a large column density and the observed x - rays are scattered emission ( ptak et al . 1996 ) . however , the luminosity ratios @xmath6/@xmath15 and @xmath6/@xmath96\lambda 5007}$ ] suggest small obscuration ( terashima et al . 2002b ) . the observed variability supports the interpretation that the x - ray emission is not scattered emission from a heavily obscured nucleus . this galaxy is an example of a seyfert 2 with only little absorption in the x - ray band . _ ngc 3226 ( l1.9)_. this galaxy was observed with the _ chandra _ hetg in 1999 december ( george et al . they obtained an intrinsic luminosity of @xmath97 ( @xmath98 ergs s@xmath2 , 68% confidence limit ) in the 210 kev band after conversion to a distance of 23.4 mpc . this luminosity is consistent with our value of @xmath99 ( @xmath100 ergs s@xmath2 , 90% confidence range ) after correction for absorption . _ ngc 4203 ( l1.9)_. a result on the same data set is presented in ho et al . the nucleus of this object has a large x - ray flux and pileup is severe in this observation . a bright source is seen 2@xmath101 se of the nucleus which was also separated from the nucleus with _ asca _ sis observations ( iyomoto et al . 1998 ; terashima et al . the _ chandra _ observations show that there is no source confusing the _ asca _ observation of the nucleus . therefore , we used an _ asca _ flux in the discussions . we analyzed archival _ data observed on 1998 may 24 . the effective exposure times after standard data screening were 19.6 ksec for each sis and 23.9 ksec for each gis . spectrum is well fitted with a power law with a photon index 1.85 ( 1.77@xmath1021.94 ) . the best - fit absorption column is @xmath10=0 , with an upper limit of @xmath103 @xmath12 . the observed flux in the 2@xmath10210 kev band is @xmath104 ergs s@xmath2@xmath12 . our _ chandra _ image in the hard energy band is dominated by the nucleus and no bright source is seen in the field . therefore , the hard x - ray measurement with _ asca _ seems reliable . the _ chandra _ flux in the 210 kev band ( @xmath105 ergs s@xmath2@xmath12 ) is about one - third of the _ asca _ flux indicating variability . _ ngc 4550 ( l2)_. this source is not detected with the present _ chandra _ observation . a detection with the _ rosat _ pspc is reported by halderson et al . the observed _ rosat _ flux in the 0.12.4 kev band is @xmath106 ergs s@xmath2 @xmath12 . rosat _ source is offset from the optical nucleus by 10@xmath34 . if this source is indeed the nucleus , our non detection by _ ( @xmath107 ergs s@xmath2 @xmath12 ) indicates time variability . _ ngc 4565 ( s1.9)_. the nuclear region is dominated by two sources : the nucleus and an off - nuclear source which is brighter than the nucleus . the observed _ chandra _ fluxes of these two source in the 210 kev band ( @xmath108 and @xmath109 ergs s@xmath2@xmath12 ) are slightly lower than those obtained with _ asca _ ( @xmath110 and @xmath111 ergs s@xmath2@xmath12 ; mizuno et al . 1999 ; terashima et al . 2002b ) , respectively . these differences appear not to be significant given the statistical , calibration , and spectral - modeling uncertainties . ( the uncertainties on the _ chandra _ fluxes are dominated by the statistical errors , which are @xmath112 % for the off - nuclear source and @xmath1650% for the nucleus , while the error in the _ asca _ fluxes is dominated by calibration uncertainties of @xmath35 % . ) the _ chandra _ spectrum of the off - nuclear source can be fitted by an absorbed power law model with a photon index of @xmath113 and @xmath10 = @xmath114 @xmath12 . a multicolor disk blackbody model also provides a good fit with best - fit parameters @xmath115 kev and @xmath10 = @xmath116 @xmath12 . _ ngc 4579 ( l1.9/s1.9)_. a result on the same data set is presented by ho et al . the nucleus is significantly piled up in the _ chandra _ observation . hard band image is dominated by the nucleus and no bright source is seen in the field . therefore , we used _ fluxes observed in 1995 and 1998 . detailed _ asca _ results are published in terashima et al . ( 1998 , 2000c ) . a long ( 33.9 ksec exposure ) _ chandra _ observation performed in 2000 may is presented in eracleous et al . the 210 kev flux reported is @xmath117 ergs s@xmath2@xmath12 which is similar to that of the second _ asca _ observation in 1998 ( @xmath118 ergs s@xmath2@xmath12 ) . _ ngc 5033 ( s1.5)_. a result on the same data set is presented in ho et al . the nucleus is significantly piled up in the _ chandra _ observation . the five off - nuclear sources shown in table 3 are located within the _ asca _ beam . the sum of the counts from these sources is less than 44 counts in the 28 kev band , while 380 counts are detected from the nucleus before correction for pileup . therefore , the _ asca _ flux ( @xmath119 ergs s@xmath2@xmath12 ; terashima et al . 1999 , 2002b ) is probably larger than the true nuclear flux by @xmath120% or less , unless the off - nuclear sources show drastic time variability . we used the _ asca _ flux without any correction for the off - nuclear source contribution . the 10% uncertainty does not affect any of the conclusions . we performed a spectral fit to the brightest off - nuclear source ( cxou j131329.7 + 363523 ) . an absorbed power law model was applied , and @xmath10 = 0.40 ( @xmath571.2 ) @xmath121 @xmath12 and a photon index @xmath122 were obtained . _ ngc 5866 ( t2)_. extended emission of diameter @xmath56 30@xmath34 ( @xmath1232 kpc ) is seen . the spectrum of this emission may be represented by a mekal plasma model with @xmath124 1 kev and abundance of 0.15 solar . this component could be identified with a gaseous halo of this s0 galaxy . falcke , h. , lehr , j. , barvainis , r. , nagar , n. m. , & wilson , a. s. 2001 , probing the physics of active galactic nuclei by multiwavelength monitoring , eds . b. m. peterson , r. w. pogge , & r. s. polidan , asp conf . series 224 , p.265 , ( asp : san francisco ) ptak , a. , yaqoob , t. , serlemitsos , p.j . , kunieda , h. , & terashima , y. 1996 , , 459 , 542 quataert , e. 2002 , `` probing the physics of active galactic nuclei '' , eds . b. m. peterson , r. w. pogge , and r. s. polidan , ( san francisco : astronomical society of the pacific ) , asp conference proceedings , vol . 224 , p.71 terashima , y. , ho , l .c . , & ptak , a. f. 2000a , , 539 , 161 terashima , y. , ho , l .c . , ptak , a. f. , mushotzky , r. f. , serlemitsos , p. j. , yaqoob , t. , & kunieda , h. 2000b , , 533 , 729 terashima , y. , ho , l .c . , ptak , a. f. , yaqoob , t. , kunieda , h. , misaki , k. , & serlemitsos , p. j. 2000c , , 535 , l79 terashima , y. , iyomoto , n. , ho , l. c. , & ptak , a. f. 2002b , , 139 , 1 cccccccc name & @xmath125 & class & date & exposure & & notes + & ( mpc ) & & & ( s ) & ( s@xmath2 ) & ( frame@xmath2 ) & + ( 1 ) & ( 2 ) & ( 3 ) & ( 4 ) & ( 5 ) & ( 6 ) & ( 7 ) & ( 8) + ngc 266 & 62.4 & l1.9 & 2001 jun 1 & 2033 & 0.020 & 0.0080 & a + ngc 2787 & 13.3 & l1.9 & 2000 jan 7 & 1050 & 0.0075 & 0.024 & c + ngc 3147 & 40.9 & s2 & 2001 sep 19 & 2202 & 0.54 & 0.21 & a + ngc 3169 & 19.7 & l2 & 2001 may 2 & 1953 & 0.081 & 0.033 & a + ngc 3226 & 23.4 & l1.9 & 2001 mar 23 & 2228 & 0.094 & 0.038 & a + ngc 4143 & 17.0 & l1.9 & 2001 mar 26 & 2514 & 0.063 & 0.025 & a + ngc 4203 & 9.7 & l1.9 & 1999 nov 4 & 1754 & 0.17 & 0.54 & c + ngc 4278 & 9.7 & l1.9 & 2000 apr 20 & 1396 & 0.18 & 0.33 & b + ngc 4548 & 16.8 & l2 & 2001 mar 24 & 2746 & 0.0097 & 0.0039 & a + ngc 4550 & 16.8 & l2 & 2001 mar 24 & 1885 & ... & ... & a + ngc 4565 & 9.7 & s1.9 & 2000 jun 30 & 2828 & 0.045 & 0.081 & b + ngc 4579 & 16.8 & l1.9/s1.9 & 2000 feb 23 & 2672 & 1.1 & 0.47 & a + ngc 5033 & 18.7 & s1.5 & 2000 apr 28 & 2904 & 0.33 & 0.59 & b + ngc 5866 & 15.3 & t2 & 2002 jan 10 & 2247 & ... & ... & a + ngc 6500 & 39.7 & l2 & 2000 aug 1 & 2104 & 0.020 & 0.064 & c + cccccccccl name & ra & dec . & & counts & & hard / soft & flux & luminosity & notes + & ( j2000 ) & ( j2000 ) & ( 0.58 kev ) & ( 0.5 - 2 kev ) & ( 2 - 8 kev ) & band ratio & ( 210 kev ) & ( 210 kev ) + ( 1 ) & ( 2 ) & ( 3 ) & ( 4 ) & ( 5 ) & ( 6 ) & ( 7 ) & ( 8) & ( 9 ) & ( 10 ) + ngc 266 & 0 49 47.81 & 32 16 40.0 & 40.7@xmath1266.4 & 29.8@xmath1265.5 & 10.9@xmath1263.3 & 0.37@xmath1260.13 & 1.6 & 7.5 & a + ngc 2787 & 9 19 18.70 & 69 12 11.3 & 7.9@xmath1262.8 & @xmath5712.0 & @xmath576.4 & ... & 0.25 & 0.053 & b + ngc 3147 & 10 16 53.75 & 73 24 02.8 & 1180.1@xmath12634.4 & 843.4@xmath12629.1 & 333.5@xmath12618.3 & 0.40@xmath1260.03 & 37 & 76 & a , c + ngc 3169 & 10 14 15.05 & 03 27 57.9 & 159.0@xmath12612.6 & @xmath5711.8 & 151.0@xmath12612.3 & @xmath3212.9 & 24 & 26 & a + & & & & & & & 26 & 22 & a , d + ngc 3226 & 10 23 27.01 & 19 53 55.0 & 209.3@xmath12614.5 & 125.2@xmath12611.2 & 80.5@xmath1269.0 & 0.64@xmath1260.09 & 7.6 & 5.5 & a + ngc 4143 & 12 9 36.07 & 42 32 03.0 & 157.4@xmath12612.6 & 121.3@xmath12611.1 & 32.6@xmath1265.7 & 0.27@xmath1260.05 & 3.1 & 1.1 & a + ngc 4203 & 12 15 05.02 & 33 11 49.9 & 294.3@xmath12617.2 & 198.7@xmath12614.1 & 91.7@xmath1269.6 & 0.46@xmath1260.06 & ... & ... & e + ngc 4278 & 12 20 06.80 & 29 16 51.6 & 255.6@xmath12616.4 & 209.6@xmath12614.9 & 52.0@xmath1267.4 & 0.25@xmath1260.04 & 8.1 & 0.91 & a , c + ngc 4548 & 12 35 26.46 & 14 29 46.7 & 26.6@xmath1265.2 & 8.7@xmath1263.0 & 17.7@xmath1264.2 & 2.02@xmath1260.85 & 1.6 & 0.61 & a + ngc 4550 & ... & ... & @xmath127 & @xmath128 & @xmath128 & ... & @xmath129 & @xmath130 & + ngc 4565 & 12 36 20.78 & 25 59 15.7 & 127.3@xmath12611.3 & 92.5@xmath1269.6 & 34.9@xmath1265.9 & 0.38@xmath1260.08 & 3.2 & 0.36 & a + ngc 4579 & 12 37 43.52 & 11 49 05.4 & 3067.9@xmath12655.6 & 2240.5@xmath12647.7 & 812.3@xmath12628.5 & 0.36@xmath1260.01 & ... & ... & e + ngc 5033 & 13 13 27.47 & 36 35 38.1 & 946.5@xmath12630.9 & 562.1@xmath12623.8 & 380.2@xmath12619.5 & 0.68@xmath1260.05 & ... & ... & e + ngc 5866 & ... & ... & @xmath127 & @xmath131 & @xmath132 & ... & @xmath133 & @xmath134 & b + ngc 6500 & 17 55 59.78 & 18 20 18.0 & 42.4@xmath1266.6 & 41.5@xmath1266.5 & @xmath577.6 & @xmath570.18 & 0.28 & 0.55 & a + & & & & & & & 0.69 & 1.3 & a , d + ccccccccccl ngc 2787 & 9 19 23.05 & 69 14 24.4 & j091923.1 + 691424 & 21.0@xmath1264.6 & 17.9@xmath1264.2 & @xmath578.0 & @xmath570.44 & 0.68 & ... & a , b + ngc 3147 & 10 16 51.50 & 73 24 08.9 & j101651.5 + 732409 & 6.7@xmath1262.6 & @xmath576.8 & @xmath579.2 & ... & 0.10 & 0.20 & b + ngc 3169 & 10 14 14.35 & 03 28 10.8 & j101414.3 + 032811 & 6.9@xmath1262.6 & 6.9@xmath1262.6 & @xmath573.0 & @xmath570.43 & 0.11 & 0.051 & b + & 10 14 17.90 & 03 28 55.2 & j101417.9 + 032855 & 10.9@xmath1263.3 & 8.0@xmath1262.8 & @xmath579.4 & @xmath571.18 & 0.18 & 0.084 & b + ngc 3226 & 10 23 26.69 & 19 54 06.8 & j102326.7 + 195407 & 8.9@xmath1263.0 & 7.9@xmath1262.8 & @xmath574.7 & @xmath570.59 & 0.13 & 0.085 & b + ngc 4203 & 12 15 09.20 & 33 09 54.7 & j121509.2 + 330955 & 240.1@xmath12615.5 & 196.4@xmath12614.0 & 40.8@xmath1266.4 & 0.21@xmath1260.04 & ... & ... & a , c , ton 1480 + & 12 15 14.33 & 33 11 04.7 & j121514.3 + 331105 & 11.9@xmath1263.5 & @xmath576.3 & 9.9@xmath1264.3 & @xmath321.57 & 2.0 & ... & a , d j121514.3 + 331105 , g + & 12 15 15.34 & 33 13 54.0 & j121515.3 + 331354 & 6.0@xmath1262.4 & 6.0@xmath1262.4 & @xmath573.0 & @xmath570.50 & 0.10 & ... & a , b + & 12 15 15.64 & 33 10 12.3 & j121515.6 + 331012 & 16.9@xmath1265.2 & 14.9@xmath1265.0 & @xmath576.3 & @xmath570.42 & 0.30 & ... & a , b , star + & 12 15 19.84 & 33 10 12.2 & j121519.8 + 331012 & 15.9@xmath1264.0 & 10.9@xmath1263.3 & @xmath574.7 & @xmath570.43 & 0.29 & ... & a , b + ngc 4550 & 12 35 21.30 & 12 14 04.5 & j123521.3 + 121405 & 6.0@xmath1262.4 & 5.9@xmath1262.4 & @xmath573.0 & @xmath570.50 & 0.10 & ... & a , b + & 12 35 27.76 & 12 13 38.9 & j123527.8 + 121339 & 35.8@xmath1266.0 & 26.6@xmath1265.2 & 8.0@xmath1262.8 & 0.30@xmath1350.12 & 1.1&11000 & a , d , qso 1232 + 125 , h + ngc 4565 & 12 36 14.65 & 26 00 52.5 & j123614.7 + 260052 & 14.9@xmath1265.0 & 8.9@xmath1264.1 & 6.0@xmath1263.6 & 0.67@xmath1260.51 & 0.76 & 0.086 & d , a30 , i + & 12 36 17.40 & 25 58 55.5 & j123617.4 + 255856 & 269.5@xmath12616.4 & 209.7@xmath12614.5 & 59.9@xmath1267.7 & 0.29@xmath1260.04 & 5.8 & 0.67 & e , f , a32 , i + & 12 36 18.64 & 25 59 34.6 & j123618.6 + 255935 & 8.8@xmath1263.0 & @xmath5713.5 & @xmath576.2 & ... & 0.098 & 0.011 & b + & 12 36 19.02 & 25 59 31.5 & j123619.0 + 255932 & 6.9@xmath1262.6 & @xmath5710.7 & @xmath576.2 & ... & 0.077 & 0.009 & b + & 12 36 19.03 & 26 00 27.0 & j123619.0 + 260027 & 19.9@xmath1264.5 & 18.0@xmath1264.2 & @xmath576.4 & @xmath570.36 & 0.22 & 0.025 & b , a33 , i + & 12 36 20.92 & 25 59 26.7 & j123620.9 + 255927 & 5.9@xmath1262.4 & @xmath5710.6 & @xmath574.7 & ... & 0.065 & 0.007 & b + & 12 36 27.39 & 25 57 32.7 & j123627.4 + 255733 & 15.9@xmath1264.0 & 14.9@xmath1263.9 & @xmath574.8 & @xmath570.32 & 0.18 & 0.020 & b , a37 ? , i + & 12 36 28.12 & 26 00 00.9 & j123628.1 + 260001 & 12.9@xmath1263.6 & 11.0@xmath1263.3 & @xmath576.4 & @xmath570.58 & 0.14 & ... & a , b + & 12 36 31.28 & 25 59 36.9 & j123631.3 + 255937 & 12.0@xmath1264.6 & @xmath5718.9 & @xmath574.8 & ... & 0.13 & ... & a , b , a43 ? , i + ngc 5033 & 13 13 24.78 & 36 35 03.7 & j131324.8 + 363504 & 13.9@xmath1263.7 & 10.0@xmath1263.2 & @xmath579.4 & @xmath570.94 & 0.15 & 0.063 & b + & 13 13 28.88 & 36 35 41.0 & j131328.9 + 363541 & 6.7@xmath1262.6 & @xmath5710.2 & @xmath576.3 & ... & 0.072 & 0.030 & b + & 13 13 29.46 & 36 35 17.3 & j131329.5 + 363517 & 34.6@xmath1265.9 & 31.7@xmath1265.7 & @xmath577.9 & @xmath570.25 & 0.37 & 0.16 & b + & 13 13 29.66 & 36 35 23.1 & j131329.7 + 363523 & 47.7@xmath1266.9 & 31.8@xmath1265.7 & 15.9@xmath1265.1 & 0.50@xmath1260.18 & 2.0 & 0.89 & f + & 13 13 35.56 & 36 34 04.4 & j131335.6 + 363404 & 7.0@xmath1262.6 & 6.0@xmath1262.4 & @xmath574.8 & @xmath570.80 & 0.075 & 0.031 & b + ngc 6500 & 17 56 01.59 & 18 20 22.6 & j175601.6 + 182023 & 7.0@xmath1262.6 & @xmath5710.4 & @xmath576.2 & ... & 0.12 & 0.23 & b + ngc 266 & @xmath138 & @xmath139 & 10.6 ( 6 ) + ngc 3147 & @xmath140 & @xmath141 & 45.4 ( 41 ) & a + ngc 3169 & @xmath142 & @xmath143 & 21.3 ( 26 ) + ngc 3226 & @xmath144 & @xmath145 & 8.1 ( 12 ) + ngc 4143 & @xmath146 & @xmath147 & 12.2 ( 11 ) + ngc 4278 & @xmath148 & @xmath149 & 16.9 ( 18 ) & a + ngc 4548 & @xmath150 & @xmath151 & 3.7 ( 5 ) + ngc 4565 & @xmath152 & @xmath153 & 1.6 ( 4 ) + ngc 6500 & @xmath154 & @xmath155 & 8.5 ( 8) + cccccc name & @xmath52@xmath15 & @xmath52@xmath6/@xmath15 & @xmath156(5 ghz ) & @xmath156(5 ghz)/@xmath6 & notes + & ( erg s@xmath2 ) & & ( erg s@xmath2 ) & & + & ( 1 ) & ( 2 ) & ( 3 ) & ( 4 ) & ( 5 ) + ngc 266 & 39.36 & 1.52 & 37.87 & @xmath1023.00 & a + ngc 2787 & 38.56 & 0.16 & 37.22 & @xmath1021.50 & b + ngc 3147 & 40.02 & 1.86 & 38.01 & @xmath1023.87 & c + ngc 3169 & 39.52 & 1.82 & 37.19 & @xmath1024.16 & a + ngc 3226 & 38.93 & 1.81 & 37.20 & @xmath1023.54 & a + ngc 4143 & 38.69 & 1.34 & 37.16 & @xmath1022.87 & b + ngc 4203 & 38.35 & 2.02 & 36.79 & @xmath1023.59 & b , e + ngc 4278 & 39.20 & 0.76 & 37.91 & @xmath1022.05 & b + ngc 4548 & 38.48 & 1.31 & 36.31 & @xmath1023.48 & d + ngc 4550 & 38.50 & @xmath1570.09 & 36.07 & @xmath1582.34 & d + ngc 4565 & 38.46 & 1.10 & 36.15 & @xmath1023.41 & b + ngc 4579 & 39.48 & 1.82 & 37.65 & @xmath1023.59 & b , e + ngc 5033 & 39.70 & 1.67 & 36.79 & @xmath1024.57 & c , e + ngc 5866 & 38.82 & @xmath1570.56 & 36.89 & @xmath1581.18 & b + ngc 6500 & 40.48 & @xmath1020.37 & 38.90 & @xmath1021.21 & a +

a theory of chiral lipid membranes is proposed on the basis of a concise free energy density which includes the contributions of the bending and the surface tension of membranes , as well as the chirality and orientational variation of tilting molecules . this theory is consistent with the previous experiments [ j.m . schnur _ et al . _ , science * 264 * , 945 ( 1994 ) ; m.s . spector _ et al . _ , langmuir * 14 * , 3493 ( 1998 ) ; y. zhao , _ et al . _ , proc . natl . acad . sci . usa * 102 * , 7438 ( 2005 ) ] on self - assembled chiral lipid membranes of dcpc . a torus with the ratio between its two generated radii larger than is predicted from the euler - lagrange equations . it is found that tubules with helically modulated tilting state are not admitted by the euler - lagrange equations , and that they are less energetically favorable than helical ripples in tubules . the pitch angles of helical ripples are theoretically estimated to be about 0 and 35 , which are close to the most frequent values 5 and 28 observed in the experiment [ n. mahajan _ et al . _ , langmuir * 22 * , 1973 ( 2006 ) ] . additionally , the present theory can explain twisted ribbons of achiral cationic amphiphiles interacting with chiral tartrate counterions . the ratio between the width and pitch of twisted ribbons is predicted to be proportional to the relative concentration difference of left- and right - handed enantiomers in the low relative concentration difference region , which is in good agreement with the experiment [ r. oda _ et al . _ , nature ( london ) * 399 * , 566 ( 1999 ) ] .

0705.2707

since tubules were fabricated successfully from chiral lipid molecules , chiral lipid structures have attracted much experimental attention @xcite . most of chiral structures in experiments are self - assembled from dc@xmath3 which is a typical chiral molecule . et al . _ have observed that the spherical vesicles in solution have very weak circular dichroism signal while tubules have strong one @xcite . et al . _ have investigated the chiral lipid tubules formed from various proportions of left- and right - handed dc@xmath3 molecules and found that they are of similar radii @xcite , which reveals that the radii of chiral tubules do not depend on the strength of the molecular chirality . fang s group has carefully resolved the molecular tilting order in the tubules and concluded that the projected direction of the molecules on the tubular surfaces departs 45@xmath2 from the equator of the tubules at the uniform tilting state @xcite . helical ripples in lipid tubules are also observed by the same group with atomic force microscopy @xcite . their pitch angles are found to be concentrated on about 5@xmath2 and 28@xmath2 @xcite . cholesterol is another kind of chiral molecules used in the experiments where helical stripes with pitch angles @xmath4 and @xmath5 are usually observed @xcite . additionally , oda _ et al . _ have reported twisted ribbons of achiral cationic amphiphiles interacting with chiral tartrate counterions @xcite . it is found that the twisted ribbons can be tuned by the introduction of opposite - handed chiral counterions in various proportions @xcite . from the experimental data @xcite , we see that the ratio between the width and pitch of the ribbons is proportional to the relative concentration difference of left- and right - handed enantiomers in the low relative concentration difference region . can we interpret all or at least most of above experimental results within a unified theory ? there are several theoretical discussions on chiral lipid membranes ( clms ) in the previous literature , where the chiral molecules are assumed to be in a smectic c@xmath6 phase at which the direction of the molecules is tilted from the normal of the membranes at a constant angle . the possible free energy of clms is discussed by helfrich and prost from symmetry arguments @xcite . their theory has been further developed and applied in many studies @xcite . nelson and powers have investigated the thermal fluctuations of clms @xcite . et al . _ have discussed tubules with helically modulated tilting state and helical ripples in tubules @xcite . komura and ou - yang have given an explanation to the high - pitch helical stripes of cholesterol molecules @xcite . due to the complicated form of the free energy used in these theories @xcite , it is almost impossible to obtain the general euler - lagrange equations corresponding to the free energy . thus one can not determine whether a configuration , such as a twisted ribbon , a tubule with helically modulated tilting state or a helical stripe , is a true equilibrium structure or not . this difficulty was always ignored in previous discussion of both tubules with helically modulated tilting state and helical stripes @xcite . can we confront this difficulty and construct a more concise theory of clms consistent with the experiments , in which we can unambiguously say which configuration is a genuine equilibrium structure ? we will address these questions in this paper , which is organized as follows : in sec . [ endensity ] , we introduce a concise free energy density of clms which includes the contributions of the bending and the surface tension of the membranes , as well as the chirality and orientational variation of the tilting molecules . in sec . [ closedclms ] , we present the euler - lagrange equations for clms without free edges and use them to explain experimental data @xcite . we predict a torus with the ratio between its two generated radii larger than @xmath1 , which has not yet been observed in the experiments on self - assembled clms . in sec . [ openclms ] , we present the euler - lagrange equations and boundary conditions for clms with free edges and use them to discuss experimental data @xcite . [ summary ] is a brief summary and discussion . in the appendixes , we briefly derive the euler - lagrange equations for clms without free edges and the euler - lagrange equations as well as boundary conditions for clms with free edges through the variational method @xcite developed by one of the present authors . several mathematical details are also put in the appendixes . following the above theories @xcite , we adopt a concise form of free energy density for a clm which consists of the following contributions . \(i ) the bending and surface energy per area is taken as helfrich s form @xcite @xmath7 where @xmath8 and @xmath9 are bending rigidities , and @xmath10 the surface tension . @xmath11 is the spontaneous curvature reflecting the asymmetrical factors between two sides of the membrane . @xmath12 and @xmath13 are the mean curvature and gaussian curvature of the membrane , respectively , which can be expressed as @xmath14 , @xmath15 by the two principal curvature radii @xmath16 and @xmath17 . the curvature energy in eq.([frhelfr ] ) is invariant under the coordinate rotation around the normal of the membrane surface , but will change under the inversion of the normal if @xmath18 . \(ii ) the energy per area originating from the chirality of tilting molecules has the form @xcite @xmath19 where @xmath20 reflects the strength of the molecular chirality which usually determines the handedness of clms the clms with the opposite handedness will be observed in experiments if @xmath20 changes its sign @xcite . without losing the generality , we only discuss the case of @xmath21 in this paper . @xmath22 is the geodesic torsion along the unit vector @xmath23 at some point . here @xmath23 represents the projected direction of the lipid molecules in the experiments @xcite and chiral tartrate counterions in the experiment @xcite on the membrane surface , respectively . if we take a right - handed orthonormal frame @xmath24 with @xmath25 being the normal vector of the membrane as shown in fig . [ figframe ] , @xmath23 can be expressed as @xmath26 , where @xmath27 is the angle between @xmath23 and @xmath28 . at this frame , the curvature tensor can be expressed as a matrix with element @xmath29 , @xmath30 , and @xmath31 shown in appendix [ appdmfm ] . with the curvature tensor , the geodesic torsion along @xmath23 can be expressed as @xcite @xmath32 which is the same as the chiral term in the previous literature , such as the last term of eq.(2 ) in ref . @xcite , eq.(3 ) in ref . @xcite , and the third term of eq.(2.1 ) in ref . in particular , for the principal frame , the above equation is simplified as @xmath33 we will confine @xmath27 to the region ( @xmath34 because of the relation @xmath35 . moreover , it is easy to see that the geodesic torsion along the mirror image of @xmath23 with respect to @xmath28 changes its sign because @xmath36 under the reflection with respect to @xmath28 . thus this term breaks the inversion symmetry in the tangent plane at each point of the membrane , which allows us to distinguish the handedness . additionally , from eq.([torsioninpr ] ) we can see that the minimum of eq.([chiralen ] ) , @xmath37 is reached when @xmath23 departs from the principal direction at a angle @xmath38 or @xmath39 for a given shape of the membrane . here the sign in front of @xmath40 depends on the sign of @xmath41 . the larger difference between @xmath16 and @xmath17 is , the larger absolute value of @xmath42 is reached . in other words , the chiral term favors saddle surfaces ( for example , the twisted ribbons in sec . [ sectwr ] ) whose two principal curvature radii @xmath16 and @xmath17 have opposite signs . right - handed orthonormal frame @xmath24 at any point in a surface where @xmath25 is the normal vector of the surface . ( a ) surface without boundary curve . ( b ) surface with boundary curve where @xmath43 is the tangent vector of the boundary curve , and @xmath44 , in the tangent plane of the surface , is perpendicular to @xmath43.,width=264 ] \(iii ) the energy per area due to the orientational variation of @xmath23 is taken as @xcite @xmath45,\ ] ] where @xmath46 is a constant in the dimension of energy . this is the simplest term of energy cost due to tilting order invariant under the coordinate rotation around the normal of the membrane surface . @xmath47 is the 2-dimensional ( 2d ) gradient operator on the membrane surface , and the 2d cross product `` @xmath48 '' gives a scalar . by defining a spin connection field @xmath49 satisfying @xmath50 @xcite , one can derive @xmath51 through simple calculations . the total free energy density adopted in the present paper , @xmath52 , has the following concise form : @xmath53 where @xmath54 . this special form might arguably be the most natural and concise construction including the bending , chirality and tilting order , for the given vector field @xmath23 and normal vector field @xmath25 . of course , using these two fields , one can also construct more general free energy densities which contain much more terms @xcite . compared with the general form , our special form is simplified in two aspects : ( i ) the contributions of the bending of membranes and the orientational elasticity of @xmath23 are taken as the isotropic forms ; ( ii ) there is no additional coupling between @xmath23 and the curvature of membranes except the chiral term ( [ chiralen ] ) , which makes it a concise theory . although our free energy density is not new relative to the general form appearing in the previous literature , we derive for the first time the corresponding euler - lagrange equations without any assumption on the shapes of membranes before doing the variation , and then check which configuration solves the euler - lagrange equations . more importantly , in the following , we will see that most of the experimental results can be explained in terms of such a concise free energy density . clms without free edges usually correspond to closed vesicles . here a long enough tubule , where the end effect is neglected , is also regarded as a clm without free edges . in this section , we show the euler - lagrange equations of clms without free edges ( derived briefly in appendix [ appendelclose ] ) , and then discuss clms in spherical , tubular and torus shapes . as we will see , our theoretical results are in good agreement with the experiments @xcite . the free energy for a closed clm can be expressed as @xmath55 where @xmath56 is the area element of the membrane and @xmath57 the volume element enclosed by the vesicle . @xmath58 is either the pressure difference between the outer and inner sides of the vesicle or used to implement a volume constraint . for tubular configuration , we usually take @xmath59 . using the variational method @xcite , we obtain the euler - lagrange equations corresponding to the free energy ( [ closedfe ] ) as @xmath60 and @xmath61\nonumber\\ & & \hspace{-0.5cm}+k_{f } [ ( \kappa_{\mathbf{v}}-h ) \mathbf{v}^{2}- \nabla\mathbf{v}\colon\nabla\mathbf{e}_3 ] = 0,\label{el2}\end{aligned}\ ] ] where @xmath62 and @xmath63 are the normal curvature along the directions of @xmath23 and @xmath64 , respectively . they can be expressed as eqs.([normalcur ] ) and ( [ norcurinv ] ) with the curvature tensor . physically , eqs.([el1 ] ) and ( [ el2 ] ) express the moment and force balances along the normal @xmath25 at each point of the membrane . if @xmath20 and @xmath46 vanish , the above two equations degenerate into the shape equation of achiral lipid vesicles , which has been fully discussed in refs . @xcite . two remarks are necessary concerning eq.([el1 ] ) : ( i ) we have selected the proper gauge such that @xmath65 , or else @xmath66 should be replaced with @xmath67 . ( ii ) for closed vesicles different from toroidal topology , the tangent vector field @xmath23 will have singular points . in this case , the right - hand term should be replaced with the sum of @xmath68-function , @xmath69 , where @xmath70 and @xmath71 represent any point and singular point in the clm , while @xmath72 represents the strength of the source or vortex at the singular point @xmath71 . for spherical vesicles of chiral lipid molecules , @xmath22 is always vanishing because @xmath73 . thus the free energy ( [ closedfe ] ) is independent of the molecular chirality and permits the same probability of left- and right - handed spherical vesicles existing in solution . naturally , no evident circular dichroism signal would be observed , which is consistent with the experiment @xcite . of course , we can not exclude the other possible explanation of this experiment that the lipid molecules are non - tilting in the spherical vesicles @xcite . a tubule is regarded as an cylinder with radius @xmath74 and infinite length . then @xmath75 , @xmath76 , and @xmath77 , where @xmath27 is the angle between @xmath23 and the equator of the cylinder . @xmath27 is a constant for the tubule with the uniform tilting state as shown in fig . [ tubules](a ) . thus eq.([el1 ] ) requires @xmath78 we should keep @xmath79 because it corresponds to the local minimum of @xmath80 , which is in good agreement with the experiment by fang s group @xcite where the projected direction of the molecules on the tubular surfaces indeed departing 45@xmath2 from the equator of the tubules with the uniform tilting state . for the cylindrical shape , we set @xmath59 . it is easy to find @xmath81 and @xmath82 both to be vanishing for @xmath79 . thus eq.([el2 ] ) is transformed into @xmath83 this equation indicates that the radius @xmath74 of chiral lipid tubules is independent of the strength of molecular chirality @xmath20 , which is consistent with the experiment @xcite where the tubular radius is insensitive to the proportion of left- and right - handed dc@xmath3 in the tubule . however , we can not exclude the other possible explanation that dc@xmath3 molecules with different handedness are not mixed in the tubule such that each tubule in solution contains merely the one kind of molecules with the same handedness ether left- or right - handed @xcite . tubules : ( a ) uniform tilting state ; ( b ) helically modulated tilting state . arrows represent the projected directions @xmath84 of the tilting molecules on the tubules.,width=264 ] the concept of tubules with helically modulated tilting state was proposed by selinger _ et al . _ the orientational variation of the tilting molecules is assumed to be linear and confined to a small range . here we will directly investigate the orientational variation by using the euler - lagrange equation ( [ el1 ] ) without this assumption . additionally , as mentioned in the introduction , whether the tubules with helically modulated titling state are equilibrium configurations or not is not addressed in ref . however , our theory will unambiguously reveal that they are not equilibrium configurations . let @xmath85 and @xmath86 denote the arc length parameters along the circumferential and axial directions , respectively . if @xmath27 , the angle between @xmath23 and the equator of the cylinder , is not a constant , eqs.([el1 ] ) and ( [ el2 ] ) are then transformed into @xmath87 and @xmath88\nonumber\\&&\hspace{-0.4cm}+\lambda/\rho+k_{c } ( c_{0}^{2}-1/\rho^{2 } ) /2\rho+k_{f } [ ( \phi_{z}^{2}-\phi_{s}^{2})/2\rho+\phi_{sz}/\rho]\nonumber\\&&\hspace{-0.4cm}=0.\label{stripe02}\end{aligned}\ ] ] where the subscripts @xmath85 and @xmath86 represent the partial derivatives respect to @xmath85 and @xmath86 , respectively . the derivations of the above two equations are shown in appendix [ deriveqshelic ] . at the helically modulated tilting state , @xmath27 is invariant along the direction of a fictitious helix enwinding around the tubule as shown in fig . [ tubules](b ) . let @xmath89 be the pitch angle of that helix and apply a coordinate transformation @xmath90 via @xmath91 , where @xmath92 is the coordinate along the helix and @xmath93 is the coordinate orthogonal to @xmath92 . in the new coordinates , @xmath27 depends only on @xmath93 . changing variable @xmath94 and introducing the dimensionless parameters @xmath95 and @xmath96 , we transform eqs.([el1 ] ) and ( [ el2 ] ) , respectively , into @xmath97 and @xmath98 where @xmath99 . the first integral of eq.([stripe1 ] ) is @xmath100 with an unknown constant @xmath101 . substituting eqs.([stripe1 ] ) and ( [ integral ] ) into eq.([stripe2 ] ) , we obtain the right - hand side of eq.([stripe2 ] ) : @xmath102.\label{rhtr}\end{aligned}\ ] ] the necessary condition for validity of eq.([stripe2 ] ) is that the right - hand side @xmath103 is constant , which , in terms of eq.([rhtr ] ) , holds if and only if @xmath104 for varying @xmath105 . therefore , tubules with helically modulated tilting state are not admitted by the euler - lagrange equation ( [ el2 ] ) if @xmath106 . in other words , the orientational variation of the tilting lipid molecules breaks the force balance along the normal direction of the tubule at this state , which might rather induce helical ripples in tubules . assume now that a tubule with radius @xmath74 undergoes small out - of - plane deformations and reaches a new configuration expressed as a vector @xmath107 , where @xmath108 is a function of @xmath85 and @xmath86 . for simplicity , we take @xmath109 , @xmath110 , @xmath111 , and @xmath112 in this subsection . as shown in appendix [ dereqsripp ] , eqs.([el1 ] ) and ( [ el2 ] ) can be transformed into @xmath113 and @xmath114\nonumber\\ & & \hspace{-0.4cm}+(k_c / k_f\rho^2)[(1+\rho^{2}\partial_{ss}+\rho^{2 } \partial_{zz})y/2 - 2\rho^{2}y_{zz}]\nonumber\\ & & \hspace{-0.4cm}+\bar{h}[2(\phi_z^2-\phi_s^2+\phi_{sz})\sin2\phi + ( \phi_{ss}-\phi_{zz}+4\phi_{z}\phi_{s})\cos2\phi]\nonumber\\ & & \hspace{-0.4cm}+(\phi_s^2-\phi_z^2)(y+\rho^{2}y_{ss}-\rho^{2}y_{zz}-1)/2\nonumber\\ & & \hspace{-0.4cm}+\phi_{sz}+y\phi_{s}^{2}-2y\phi_{sz}-y_{zz}+2\rho^2 y_{sz}\phi_{s}\phi_{z}=0,\label{eqripp2}\end{aligned}\ ] ] where @xmath27 is the angle between @xmath23 and the equator of the tubule . small amplitude ripples in a tubule with radius @xmath74 . arrows represent the projected directions @xmath84 of the tilting molecules on the ripples surface.,height=226 ] now we consider helical ripples in the tubule where @xmath27 and @xmath115 are invariant along the direction of a fictitious helix enwinding around the tubule as shown in fig . [ ripplesfig ] . we adopt the same coordinate transformation as the above subsection , and let @xmath116 , where @xmath89 is the pitch angle of the fictitious helix and @xmath105 is governed by eq.([stripe1 ] ) . then the above equations ( [ eqripp1 ] ) and ( [ eqripp2 ] ) are reduced to a matrix equation @xmath117 with @xmath118 and @xmath119 , where the superscript ` @xmath120 ' represents the transpose . the differential operator @xmath121 has four matrix elements : @xmath122 { d^{2}}/ { d\chi^{2}},\end{aligned}\ ] ] where @xmath101 satisfies @xmath123 it is not hard to find a special solution of eq.([rippl2 ] ) as @xmath124 with @xmath125 . considering eq.([rippl1 ] ) , we can prove @xmath126 for @xmath127 and hence @xmath126 for @xmath128 . note that eq.([rippl2 ] ) contains only the first order terms of @xmath129 and @xmath115 , which can also be obtained from the variation of the free energy @xmath130 expanded up to the second order terms of @xmath129 and @xmath115 . the dimensionless mean energy difference between the tubule with helical ripples and that with helically modulated tilting state is expressed as @xmath131 where @xmath132 is the period of the ripples along @xmath93 direction as show in fig . [ ripplesfig ] . substituting the solution ( [ ripplsolution ] ) into the above equation , we have @xmath133 which reveals that a tubule with ripples is energetically more favorable than that in a helically modulated titling state . additionally , we can easily prove from eq.([rippl1 ] ) that the pitch angle obeys @xmath134 . all ripples in tubules resolved in the recent experiment @xcite with atomic force microscope have indeed the pitch angles smaller than @xmath135 . the tubules with ` helically modulated tilting state ' observed in the experiment @xcite might also be the tubules with helical ripples whose amplitudes are below the experimental resolution . in this experiment , all pitch angles are also smaller than @xmath135 . considering @xmath112 , we have @xmath136 from eq.([integral ] ) and then @xmath137 due to @xmath132 being the period of the ripples along @xmath93 direction . if the pitch angle @xmath138 , the fictitious helix is a circle , and eqs.([rippl1 ] ) and ( [ matchcond ] ) requires @xmath139 . if @xmath140 , the fictitious helix is indeed a helix satisfying @xmath141 . it is then easy to derive @xmath142 from eqs.([rippl1 ] ) and ( [ matchcond ] ) . therefore , two kinds of ripples are permitted in our theory : one has pitch angle about @xmath143 ; another satisfies eq.([pitchangle ] ) , which gives @xmath144 for @xmath128 . our theoretical results are thus close to the most frequent pitch angles ( about 5@xmath2 and 28@xmath2 ) of the ripples in tubules observed by fang s group @xcite . ( a ) torus generated by a cycle rotating around an axis in the same plane of the cycle . ( b ) ratio of the generated radii ( @xmath145 ) , eq.([torusrad ] ) , as a function of @xmath146.,width=207 ] a torus is a revolution surface generated by a cycle with radius @xmath74 rotating around an axis in the same plane of the cycle as shown in fig . [ torusfig](a ) . the revolution radius @xmath147 should be larger than @xmath74 . a point in the torus can be expressed as a vector @xmath148 . as shown in appendix [ dereqstorus ] , eq.([el1 ] ) is transformed into @xmath149_\varphi-\frac{\nu h\rho}{k_{f}}\cos2\phi=0,\label{phitorus}\ ] ] where @xmath27 is the angle between @xmath23 and the latitude of the torus , while @xmath150 is the ratio between two generated radii of the torus . it is found that the uniform tilting state @xmath151 satisfies the above equation ( [ phitorus ] ) and that makes @xmath152 to take the minimum by considering eq.([tautours ] ) . as shown in appendix [ dereqstorus ] , with @xmath151 , eq.([el2 ] ) is transformed into @xmath153 if @xmath154 and @xmath59 , the above equation degenerates into eq.([tubradius ] ) obeyed by a tubule with radius @xmath74 at uniform tilting state . if @xmath145 is finite , then eq.([toruseq ] ) holds if and only if the coefficients of @xmath155 vanish . it follows that @xmath156 , @xmath157 and @xmath158 the relation between @xmath145 , the ratio of the generated radii @xmath147 and @xmath74 , and @xmath159 is sketched in fig . [ torusfig](b ) . the ratio @xmath145 increases with @xmath159 . especially , @xmath160 for @xmath161 , which leads to the willmore torus of non - tilting lipid molecules @xcite . since this kind of torus was observed in the experiments @xcite , tori with @xmath162 for @xmath163 might also be observed in some experiments on clms . in this section , we show the euler - lagrange equations and boundary conditions of clms with free edges , and then discuss helical stripes and twisted ribbons . as we will see , our theoretical result on twisted ribbons is consistent with the experiment @xcite . consider a clm with an free edge as shown in fig . [ figframe](b ) . its free energy can be expressed as @xmath164 where @xmath56 is the area element of the membrane and @xmath165 the arc length element of the edge . @xmath166 represents the line tension of the edge . using the variational method @xcite , as shown in appendix [ appendelopen ] we obtain two euler - lagrange equations corresponding to the free energy ( [ openfe ] ) as @xmath167 and @xmath168\nonumber\\ & & + k_{f } [ ( \kappa_{\mathbf{v}}-h ) \mathbf{v}^{2}- \nabla\mathbf{v}\colon\nabla\mathbf{e}_3 ] = 0.\label{el4}\end{aligned}\ ] ] additionally , the boundary conditions obeyed by the free edge are derived as : @xmath169 where @xmath170 , @xmath171 and @xmath172 are the normal curvature , geodesic torsion , and geodesic curvature of the boundary curve ( i.e. , the edge ) , respectively . @xmath44 is in the tangent plane of the membrane surface and perpendicular to @xmath43 , the tangent vector of the boundary curve , as shown in fig . [ figframe](b ) . @xmath173 and @xmath174 are components of @xmath64 in the direction of @xmath44 and @xmath43 , respectively . @xmath175 is the direction derivative of @xmath12 with respect to @xmath44 . a dot represents the derivative with respect to @xmath85 . @xmath176 is the angle between @xmath23 and @xmath43 at the boundary curve . it should be noted that eqs.([el3 ] ) and ( [ el4 ] ) are equivalent to eqs.([el1 ] ) and ( [ el2 ] ) with @xmath59 . eqs.([bc1])([bc4 ] ) describe the force and moment balance relations in the edge . thus they can also be used for a clm with several edges . if @xmath20 and @xmath46 vanish , the above equations ( [ el3])([bc4 ] ) degenerate into the euler - lagrange equations and boundary conditions of open achiral lipid bilayers , which were fully discussed in refs . @xcite . a helical stripe with pitch @xmath177 and radius @xmath74 is shown in fig . [ helixfig ] . in terms of the discussion in sec . [ sechmts ] , we can immediately deduce that a helical stripe with modulated tilting state is not permitted by the euler - lagrange equations ( [ el3 ] ) and ( [ el4 ] ) because it can be thought of as a ribbon wound around a fictitious cylinder . thus here we only have to discuss helical stripes with a uniform tilting state , for which we easily obtain @xmath178 from the euler - lagrange equations . here @xmath27 is the angle between @xmath23 and the equator of the fictitious cylinder . helical stripe . arrows represent the projected directions @xmath84 of the tilting molecules on the stripe s surface.,height=226 ] in the free edge of the helical stripe , we have @xmath179 , @xmath75 , @xmath180 , @xmath181 , @xmath182 , @xmath183 , @xmath184 , where @xmath185 is the pitch angle of the helical stripe . thus the boundary condition ( [ bc1 ] ) is trivial , and eqs.([bc2])([bc4 ] ) can be transformed into @xmath186 if @xmath187 , there exists only a trivial solution @xmath188 to the above equations ( [ helixradius])([helixb3 ] ) . thus our theory does not permit genuine helical stripes with free edges with a uniform tilting state . a twisted ribbon with pitch @xmath177 and width @xmath132 is shown in fig . [ twistrbfig ] , which can be expressed as a a vector @xmath189 with @xmath190 , @xmath191 and @xmath192 . as shown in appendix [ dereqstwist ] , eq.([el3 ] ) is transformed into @xmath193 where @xmath27 is the angle between @xmath23 and the horizontal . twisted ribbons : ( a ) @xmath23 is perpendicular to the edges ; ( b ) @xmath23 parallels the edges . arrows represent the projected directions @xmath84 of the tilting molecules on the ribbons surface.,height=226 ] if we only consider the uniform tilting state , the above equation requires @xmath194 or @xmath195 . it is easy to see that @xmath194 minimizes @xmath196 for @xmath197 while @xmath198 minimizes @xmath196 for @xmath199 from eq.([tautwistr]).thus we should take @xmath194 for @xmath197 and @xmath198 for @xmath199 . the former case corresponds to fig . [ twistrbfig](a ) where @xmath23 is perpendicular to the edges ; the latter corresponds to fig . [ twistrbfig](b ) where @xmath23 is parallel to the edges . as shown in appendix [ dereqstwist ] , both for @xmath194 and @xmath195 , eq.([el4 ] ) is transformed into @xmath200 which requires @xmath109 for non - vanishing @xmath201 . among the boundary conditions ( [ bc1])([bc4 ] ) , only eq.([bc2 ] ) is nontrivial , which gives @xmath202 with @xmath203 . guided by the experimental data @xcite , we may assume @xmath20 to be proportional to @xmath204 , the relative concentration difference of the left- and right - handed enantiomers in the experiment , i.e. , @xmath205 with a constant @xmath206 . in terms of the experimental data , @xmath207 for @xmath208 . thus eq.([bc22 ] ) requires @xmath110 and then @xmath209 to determine the relation between @xmath210 and @xmath204 , we minimize the average energy per area with respect to @xmath211 for given @xmath132 . the average energy per area is calculated as @xmath212 minimizing it with respect to @xmath211 and using eq.([alfeqn ] ) , we obtain @xmath213 with @xmath214 . this relation reveals that the ratio between the width and pitch of the ribbons is proportional to the relative concentration difference of left- and right - handed enantiomers in the low relative concentration difference region . eq.([ratioh ] ) fits well with the experimental data with parameter @xmath215 as shown in fig . [ fignature ] . relation between @xmath216 and @xmath204 . the solid line is the fitting curve @xmath217 and the dots are the experimental data in ref . in this paper , we have focused on a concise free energy from which the euler - lagrange equations of clms without free edges and euler - lagrange equations as well as boundary conditions of clms with free edges can be derived analytically . their implications for extant and future experiments can be summarized as follows . \(i ) our theory predicts the same probability for left- and right - handed spherical vesicles existing in solution . thus no evident circular dichroism signal should be observed , which is consistent with the experiment @xcite . \(ii ) the radius of a tubule at uniform tilting state satisfies eq.([tubradius ] ) . it reveals that the radius is independent of molecular chirality , which is consistent with the experiment @xcite where the tubular radius is insensitive to the proportion of left- and right - handed dc@xmath3 in the tubule . we find that the projected direction of the molecules on the tubular surface departs 45@xmath2 from the equator of the tubule with the uniform tilting state , which is in good agreement with the experiment @xcite . \(iii ) tubules with helically modulated tilting state are not admitted by the euler - lagrange equation ( [ el2 ] ) for non - vanishing @xmath20 . the orientational variation of the tilting lipid molecules breaks the force balance along the normal direction of the tubule . thus a tubule with helically modulated tilting state is not an equilibrium structure within our theoretical framework . \(iv ) helical ripples in tubules are equilibrium structures . the pitch angle of helical ripples is estimated as about 0@xmath2 and 35@xmath2 for @xmath111 , which are close to the experimental values 5@xmath2 and 28@xmath2 observed by fang s group @xcite . \(v ) tori with a ratio of generated radii larger than @xmath1 are predicted by our theory , eq.([torusrad ] ) , which have not yet been observed in the experiments . \(vi ) helical stripes with free edges at either uniform tilting state or helical modulated tilting state are not possible equilibrium configurations . \(vii ) twisted ribbons satisfy euler - lagrange equations ( [ el3 ] ) , ( [ el4 ] ) and boundary conditions ( [ bc1])([bc4 ] ) . the ratio between the width and pitch of the ribbons is proportional to the relative concentration difference of left- and right - handed enantiomers in the low relative concentration difference region , which is in good agreement with the experiment @xcite . finally , we have to list a few open problems which should be addressed in the future work . \(i ) within the present theory , we can not give a simple explanation for the experiments on clms of pure cholesterol @xcite where helical stripes with pitch angles @xmath4 and @xmath5 are usually observed . it has recently been found that the cholesterol helical stripes have good crystal structure @xcite , which is out of the range of our theory . there might be two possible ramifications based on the present theory : one would be to include anisotropic bending effects @xcite in the free energy density ( [ energy2 ] ) ; another one would be to consider a line tension @xmath166 depending on the angle between the directions of the tilting and the free edges @xcite in the free energy ( [ openfe ] ) . \(ii ) we can not yet interpret the twisted ribbon - to - tubule transition with increasing the relative concentration difference of the left- and right - handed enantiomers reported in the recent experiment by oda s group @xcite . a possible reason is that the parameters except @xmath20 in our theory are independent of the relative concentration difference @xcite . additionally , the ribbons and tubules observed in this experiment are usually multi - bilayer structures @xcite . then a decoupling effect @xcite between neighbor bilayers might occur . an extended theory including these two factors might be required to investigate the mechanism responsible for the transition . \(iii ) we have ignored the effect of singular points in clms . although the features of singular points in 2d planar films or achiral nematic spherical , torus vesicles and other manifolds have been fully investigated in previous literature @xcite , it remains a challenge to study the properties of singular points in clms @xcite . zct is grateful to z. c. ou - yang , j. fang and r. oda for instructive discussions and to the alexander von humboldt foundation for financial support . if we take a frame @xmath24 at any point @xmath70 on a surface as shown in fig . [ figframe ] , then the infinitesimal tangential vector at @xmath70 is expressed as @xmath218 and the difference of frame between at points @xmath219 and @xmath70 is denoted as @xmath220 where @xmath221 , @xmath222 , and @xmath223 are 1-forms @xcite . the repeated subscripts in eq.([diffei ] ) and the following contents represent the einstein summation convention . with these 1-forms , the structure equations of a surface can be expressed as @xcite @xmath224 and @xmath225 where the symbol ` @xmath226 ' expresses the wedge product between differential forms and ` @xmath227 ' is the exterior differential operator @xcite . the matrix @xmath228 is called the curvature matrix which is related to the mean and gaussian curvature by @xmath229 for a unit vector @xmath26 , the normal curvature along the direction of @xmath23 can be expressed as @xcite @xmath230 the normal curvature along the direction of an arbitrary vector @xmath64 can be expressed as @xmath231 where @xmath232 and @xmath233 are the components of @xmath64 in the directions of @xmath28 and @xmath234 . in our following derivations , several relations between vector forms and differential forms are used frequently . for convenience , we list them below . @xmath235 where @xmath236 is the area element . @xmath23 is a unit vector and @xmath27 is the angle between @xmath23 and @xmath28 . @xmath49 is the spin connection and @xmath54 . @xmath237 is the hodge star operator @xcite which satisfies @xmath238 and @xmath239 . using eqs.([hkgeneral ] ) , ( [ norcurinv ] ) , ( [ curlm ] ) , ( [ divgm ] ) and ( [ vdefin1 ] ) , we can prove @xmath240 and @xmath241 da\label{ntndmp } \\ & & \hspace{-0.4cm}=d[(v_{2}\cos2\phi- v_{1}\sin2\phi ) \omega _ { 1}+ ( v_{1}\cos2\phi+v_{2}\sin2\phi ) \omega_{2}],\nonumber\end{aligned}\ ] ] where @xmath232 and @xmath233 are the components of @xmath64 in the directions of @xmath28 and @xmath234 . we suggest the reader refer first to ref . @xcite and some textbook on the calculus with the moving frame method before going on if he is interested in the mathematical details . in writing the following contents we assume that the reader has been familiar with the skill in ref . @xcite and good at calculus with the moving frame method . here the clms without free edges will be derived through the variational method developed in ref . @xcite with the aid of the moving frame method and exterior differential forms , which can highly simplify the calculus of variation . assume @xmath242 . through simple calculations , we arrive at @xmath243 and @xmath244 combining with the above two equations and using the integral by parts and stokes theorem , we derive @xmath245 \xi da+k_{f } \oint \xi\ast\mathbf{v}\cdot d\mathbf{r},\label{varyphi}\ ] ] where the second term is the integral along the boundary curve , which vanishes for clms without free edges . since @xmath246 is an arbitrary function , from @xmath247 we derive euler - lagrange equation ( [ el1 ] ) . we have used the assumption @xmath248 when we write eq.([varyphi ] ) . this assumption is indeed satisfied in our discussions on all configurations except ripples in tubules . thus @xmath66 should be replaced with @xmath249 in eq.([varyphi ] ) as well as eq.([el1 ] ) when this condition is not met . in this subsection , let us denote @xmath250 as the functional ( [ closedfe ] ) with @xmath20 and @xmath46 vanishing , and define the additional functional @xmath251 any small deformation of a clm without free edges can always be achieved from small normal displacement @xmath252 at each point @xmath70 in the surface . that is , @xmath253 . the frame is also changed because of the deformation of the surface , which is denoted as @xmath254 where @xmath255 corresponds to the rotation of the frame due to the deformation of the surface . @xmath256 was first dealt with in ref . @xcite which gives @xmath257 following ref . @xcite , and considering eqs.([daudfine ] ) , ( [ vdefin1 ] ) and ( [ vdefin2 ] ) , through somewhat involved calculations , we can derive @xmath258 combining with the above two equations and using the integral by parts and stokes theorem , we derive @xmath259\omega_{3}da\nonumber\\ & & \hspace{0.85cm}+\int k_{f } [ ( \kappa_{\mathbf{v}}-h ) \mathbf{v}^{2}- \nabla\mathbf{v}\colon\nabla\mathbf{e}_3 ] \omega_{3}da,\label{varyomega3}\end{aligned}\ ] ] then the euler - lagrange equation ( [ el2 ] ) follows from @xmath260 because @xmath252 is an arbitrary function . a clm with an free edge can be described as a surface with an boundary curve shown in fig . [ figframe]b . let @xmath242 , we still have eq.([varyphi ] ) whose second term is transformed into @xmath261 with @xmath262 . thus from @xmath247 we can derive the euler - lagrange equation ( [ el3 ] ) and boundary condition ( [ bc1 ] ) . because one can select an arbitrary frame @xmath263 , here we take it such that @xmath28 and @xmath234 align with @xmath43 and @xmath44 in the boundary curve . then any small deformation of a clm with an free edge can always be expressed as the linear superposition of small normal displacement @xmath252 and tangent displacement @xmath264 along @xmath234 at each point @xmath70 in the surface . first , we consider the in - plane deformation mode @xmath265 . the change of the frame is still denoted as eq.([varyei ] ) . in terms of ref . @xcite , we have @xmath266 where @xmath267 is the geodesic curvature of the boundary curve . additionally , we can derive @xmath268 although the derivation of the above equation is somewhat involved , its physical meaning is quite clear . under the displacement @xmath264 along @xmath234 , @xmath269 is similar to the surface tension and the area element near the boundary decreases by @xmath270 . thus the variation of @xmath271 gives eq.([varyomega2 g ] ) and then we obtain the boundary condition ( [ bc2 ] ) from @xmath247 . second , we consider the out - of - plane deformation mode @xmath253 . let us denote @xmath250 as the functional ( [ openfe ] ) with @xmath20 and @xmath46 vanishing , and define the additional functional as eq.([fadd ] ) . @xmath256 is fully discussed in ref . @xcite as @xmath272\omega_{3}da\nonumber\\ & & \hspace{-0.4cm}+\int 2k_{c}(\nabla^{2}h)\omega_{3 } da -\oint[k_{c}(2h+c_{0})-\bar{k}\kappa_{n}]\omega_{23}ds\nonumber\\ & & \hspace{-0.4cm}-\oint[-2k_{c}\partial h/\partial\mathbf{b}+\gamma\kappa_{n}-\bar { k}\dot{\tau}_{g}]\omega _ { 3}ds.\label{bcf0}\end{aligned}\ ] ] here the only difference is that the sign of @xmath9 adopted in the present paper is opposite to that in ref . similar to the derivation of eq.([varyomega3 ] ) from eqs.([vartaum ] ) and ( [ varkappam ] ) by using the integral by parts and stokes theorem , we have @xmath273\omega_{3}da\nonumber\\&+&\int k_{f } [ ( \kappa_{\mathbf{v}}-h ) \mathbf{v}^{2}- \nabla\mathbf{v}\colon\nabla\mathbf{e}_3 ] \omega_{3}da\nonumber\\ & + & \oint[h(v_{t}+\dot{\bar{\phi}})\sin2\bar{\phi}-k_{f}\kappa_{n}v_{t}]\omega_{3}ds \nonumber\\&+&\oint(h/2)\sin2\bar{\phi}\omega_{23}ds,\label{bcfadd}\end{aligned}\ ] ] where @xmath274 . in the above equations ( [ bcf0 ] ) and ( [ bcfadd ] ) , @xmath252 represents the arbitrary small displacement of point on the surface along @xmath25 and @xmath275 is the arbitrary small rotation of @xmath25 around @xmath43 at the edge . thus @xmath247 will give the euler - lagrange equation ( [ el4 ] ) and boundary conditions ( [ bc3 ] ) and ( [ bc4 ] ) . for a cylindrical surface , take a frame such that @xmath28 , @xmath234 and @xmath25 are along the circumferential , axial and radial directions , respectively . then we have @xmath276 thus @xmath277 because @xmath278 and @xmath279 , using the structure equation ( [ structur ] ) , we have @xmath280 . considering eq.([sandomega12 ] ) , we obtain the spin connection @xmath281 and then @xmath282 from eqs.([normalcur ] ) and ( [ laplphi ] ) , we have @xmath283 substituting the above two equations and eq.([hkcylind ] ) into the euler - lagrange equation ( [ el1 ] ) , we can derive eq.([stripe01 ] ) in sec . [ sechmts ] . additionally , using eqs.([diffei ] ) , ( [ omega13 ] ) and ( [ vfieldcylind ] ) , we derive @xmath284 from eq.([nabvnabe3 ] ) we can derive @xmath285 in this derivation , one should note that @xmath286 . from eqs.([kvmhgen ] ) , ( [ ntndmp ] ) , ( [ abccylind ] ) and ( [ vfieldcylind ] ) , we can obtain @xmath287 and @xmath288 substituting eqs.([hkcylind ] ) and ( [ ve3cylind])([ntndmp2 ] ) into the euler - lagrange equation ( [ el2 ] ) , we can obtain eq.([stripe02 ] ) in sec . [ sechmts ] . for the surface expressed as vector form @xmath289 with @xmath108 , the first and second fundamental forms of the surface can be calculated as @xmath290 up to the order of @xmath291 , respectively . the same order is kept in the following expressions in this section . in terms of the correspondence relations @xmath292 and @xmath293 , we have @xmath294 @xmath295 is a function of @xmath85 and @xmath86 , using @xmath296 , we can derive @xmath297 because @xmath298 and @xmath299 , using the structure equation ( [ structur ] ) , we have @xmath300 . considering eqs.([sandomega12 ] ) and ( [ nabds ] ) , we have @xmath301 using eqs.([nabdphi ] ) and ( [ laplphi ] ) , we obtain @xmath302 and then @xmath303\mathbf{e}_{1}+\phi_{z}\mathbf{e}_{2}.\label{vfieldripp}\ ] ] from eqs.([diffei ] ) , ( [ omega13 ] ) , ( [ nabvnabe3 ] ) , ( [ kvmhgen ] ) , and ( [ vfieldripp ] ) , we can derive @xmath304 since we only consider the case of @xmath112 , up to the order @xmath305 , we obtain @xmath306\label{ntndmrip}\\ & & \hspace{-0.4cm}=\bar{h}[2(\phi_{z}^{2}-\phi_{s}^{2}+\phi_{sz})\sin2\phi + ( \phi_{ss}-\phi_{zz}+4\phi_{z}\phi_{s})\cos2\phi]\nonumber\end{aligned}\ ] ] by using eqs.([normalcur ] ) and ( [ ntndmp2 ] ) . substituting the above equations ( [ hkripples])([ntndmrip ] ) into the euler - lagrange equations ( [ el1 ] ) and ( [ el2 ] ) , we can derive the basic equations ( [ eqripp1 ] ) and ( [ eqripp2 ] ) under the assumption @xmath109 , @xmath110 , @xmath111 in sec . [ secripple ] . a torus can be expressed as a vector form @xmath307 with @xmath308 . a frame @xmath24 is taken as @xmath309 by using eqs.([drframe ] ) , ( [ vectorus ] ) and ( [ framtorus ] ) , we can derive @xmath310 from eq.([diffei ] ) and ( [ framtorus ] ) , we obtain @xmath311\omega_1,\label{omega12torus}\\ & & \omega_{13}=-[{\cos \varphi } /{(r+\rho\cos \varphi ) } ] \omega_1,\label{omega13torus}\\ & & \omega_{23}=-(1/\rho)\omega_2.\label{omega23torus}\end{aligned}\ ] ] considering eqs.([sandomega12 ] ) , ( [ nabds ] ) and ( [ omega12torus ] ) , we have @xmath312\mathbf{e}_1,\\ & & \nabla\cdot\mathbf{s}=0.\end{aligned}\ ] ] additionally , ( [ omega13 ] ) , ( [ omega13torus ] ) and ( [ omega23torus ] ) give @xmath313 thus by considering eqs.([torsionexp ] ) , ( [ hkgeneral ] ) , ( [ normalcur ] ) , and ( [ laplphi ] ) , we obtain @xmath314 substituting the above four equations into eq.([el1 ] ) , we can derive eq.([phitorus ] ) in sec . [ sectorus ] . if @xmath151 , then @xmath315 , and @xmath316\mathbf{e}_1,\label{vfieldtorus}\\ & & \mathbf{v}\cdot d\mathbf{e}_3=v_1\omega_{31}=[\sin2\varphi/2(r+\rho\cos\varphi)]d\theta,\end{aligned}\ ] ] where @xmath232 is the component of @xmath64 in the direction of @xmath28 . thus from eq.([nabvnabe3 ] ) we can derive @xmath317 using eqs.([kvmhgen ] ) , ( [ ntndmp ] ) , ( [ abctorus ] ) and ( [ vfieldtorus ] ) , we can obtain @xmath318 and @xmath319 additionally , from @xmath320 , we have @xmath321 substituting the above equations ( [ hktorus ] ) and ( [ nabvnabe3trs])([nab2htorus ] ) into eq.([el2 ] ) , we can derive eq.([toruseq ] ) in sec . [ sectorus ] . a twisted ribbon can be expressed as a vector form @xmath322 by using eqs.([drframe ] ) , ( [ vectortwist ] ) and ( [ framtwistr ] ) , we can derive @xmath324 from eq.([diffei ] ) and ( [ framtwistr ] ) , we obtain @xmath325\omega _ { 2},\label{omega12twst } \\ \omega _ { 13 } & = & -[\alpha /(u^{2}+\alpha ^{2})]\omega _ { 2},\label{omega13twst } \\ \omega _ { 23 } & = & -[\alpha /(u^{2}+\alpha ^{2})]\omega _ { 1}.\label{omega23twst}\end{aligned}\ ] ] considering eqs.([sandomega12 ] ) , ( [ nabds ] ) and ( [ omega12twst ] ) , we have @xmath326\mathbf{e}_{2 } , \\ & & \nabla \cdot \mathbf{s } = 0.\end{aligned}\ ] ] additionally , ( [ omega13 ] ) , ( [ omega13twst ] ) and ( [ omega23twst ] ) give @xmath327 thus by considering eqs.([torsionexp ] ) , ( [ hkgeneral ] ) , ( [ normalcur ] ) , and ( [ laplphi ] ) , we obtain @xmath328 substituting the above four equations into eq.([el3 ] ) , we can derive eq.([eqtwistr1 ] ) in sec . [ sectwr ] . if @xmath194 or @xmath195 , then @xmath315 , and @xmath329\mathbf{e}_{2},\label{vecvtwst}\\ & & \mathbf{v}\cdot d\mathbf{e}_3=v_2\omega_{32}=[u\alpha /(u^{2}+\alpha ^{2})^2 ] du,\end{aligned}\ ] ] where @xmath233 is the component of @xmath64 in the direction of @xmath234 . thus from eq.([nabvnabe3 ] ) we can derive @xmath330 now let us turn to the boundary conditions . if @xmath194 for @xmath197 , we have @xmath334 . at the boundary , @xmath335 and @xmath336 , so @xmath337 from eqs.([hktwist])([tautwistr ] ) , we can obtain @xmath338 additionally , we have @xmath339 in terms of the definition of normal curvature , geodesic curvature and geodesic torsion . from the above equations ( [ vtvbtwst])([taubdtwist ] ) , we find that among the boundary conditions ( [ bc1])([bc4 ] ) , only eq.([bc2 ] ) is nontrivial , which gives eq.([bc22 ] ) . similarly , we obtain the same equation as ( [ bc22 ] ) when @xmath198 for @xmath199 . the average energy per area ( [ fretwistr ] ) can be obtained from @xmath340 , where @xmath341 is the area of the twisted ribbon and @xmath130 is the free energy ( [ openfe ] ) . the calculation is straightforward with @xmath109 , eqs.([hktwist ] ) , ( [ tautwistr ] ) , ( [ vecvtwst ] ) , and @xmath342 j. m. schnur , science * 282 * , 1669 ( 1993 ) . j. m. schnur , b. r. ratna , j. v. selinger , a. singh , g. jyothi , and k. r. k. easwaran , science * 264 * , 945 ( 1994 ) . m. s. spector , j. v. selinger , a. singh , j. m. rodriguez , r. r. price , and j. m. schnur , langmuir * 14 * , 3493 ( 1998 ) . y. zhao , n. mahajan , r. lu , and j. fang , proc . usa * 102 * , 7438 ( 2005 ) . n. mahajan , y. zhao , t. du , and j. fang , langmuir * 22 * , 1973 ( 2006 ) . spector , a. singh , p. b. messersmith , and j. m. schnur , nano lett . * 1 * , 375 ( 2001 ) . d. s. chung , g. b. benedek , f. m. konikoff , and j. m. donovan , proc . usa * 90 * , 11341 ( 1993 ) y. v. zastavker , n. asherie , a. lomakin , j. pande , j. m. donovan , j. m. schnur , and g. b. benedek , proc . usa * 96 * , 7883 ( 1999 ) . r. oda , i. huc , m. schmutz , s. j. candau , f. c. mackintosh , nature ( london ) * 399 * , 566 ( 1999 ) . d. berthier , t. buffeteau , j. lger , r. oda , and i. huc , j. am . soc . * 124 * , 13486 ( 2002 ) . w. helfrich and j. prost , phys . a * 38 * , 3065 ( 1988 ) . o. y. zhongcan and l. jixing , phys . lett . * 65 * , 1679 ( 1990 ) ; phys . a * 43 * , 6826 ( 1991 ) . p. nelson and t. powers , phys . lett . * 69 * , 3409 ( 1992 ) . j. v. selinger and j. m. schnur , phys . lett . * 71 * , 4091 ( 1993 ) . j. v. selinger , f. c. mackintosh , and j. m. schnur , phys . e * 53 * , 3804 ( 1996 ) . j. v. selinger , m. s. spector , and j. m. schnur , j. phys . b * 105 * , 7157 ( 2001 ) . s. komura and o. zhong - can , phys lett . * 81 * , 473 ( 1998 ) . z. hu and v. c. lo , int . b * 19 * , 2999 ( 2005 ) . z. c. tu and z. c. ou - yang , j. phys . a * 37 * , 11407 ( 2004 ) . w. helfrich , z. naturforsch . * 28c * , 693 ( 1973 ) . f. c. mackintosh and t. c. lubensky , phys . lett . * 67 * , 1169 ( 1991 ) ; t. c. lubensky and j. prost , j. phys . ( france ) * 2 * , 371 ( 1992 ) . d. r. nelson and l. peliti , j. phys . france * 48 * , 1085 ( 1987 ) . o. y. zhongcan and w. helfrich phys . lett . * 59 * , 2486 , ( 1987 ) ; phys . a * 39 * , 5280 ( 1989 ) . r. lipowsky , nature ( london ) * 349 * , 475 ( 1991 ) . u. seifert , adv . phys . * 46 * , 13 ( 1997 ) . z. c. ou - yang , j. x. liu and y. z. xie , _ geometric methods in the elastic theory of membranes in liquid crystal phases _ ( world scientific , singapore , 1999 ) . q. du , c. liu and x. wang , _ j. comput . _ 212 , 757 ( 2006 ) . y. m. lvov , r. r. price , j. v. selinger , a. singh , m. s. spector , and j. m. schnur , langmuir * 16 * , 5932 ( 2000 ) . z. c. ou - yang , phys . a * 41 * , 4517 ( 1990 ) . u. seifert , phys . lett . * 66 * , 2404 ( 1991 ) . m. mutz and d. bensimon , phys . a * 43 * , 4525 ( 1991 ) . a. s. rudolph , b. r. ratna , and b. kahn , nature ( london ) * 352 * , 52 ( 1991 ) . z. lin , r. m. hill , h. t. davis , l. e. scriven , and y. talmon , langmuir * 10 * , 1008 ( 1994 ) . r. capovilla , j. guven , and j. a. santiago , phys . e * 66 * , 021607 ( 2002 ) . z. c. tu and z. c. ou - yang , phys . e * 68 * , 061915 ( 2003 ) . t. umeda , y. suezaki , k. takiguchi , and h. hotani , phys . rev . e * 71 * , 011913 ( 2005 ) . x. wang and q. du , e - print arxiv : physics/0605095 . b. khaykovich , c. hossain , j. j. mcmanus , a. lomakin , d. e. moncton , and g. b. benedek , proc . usa 104 , 9656 ( 2007 ) . chen , phys . e * 59 * , 6192 ( 1999 ) . r. c. sarasij and m. rao , phys . * 88 * , 088101 ( 2002 ) ; e - print arxiv : cond - mat/0703182 . a. brizard , c. aim , t. labrot , i. huc , d. berthier , f. artzner , b. desbat , and r. oda , j. am . * 129 * , 3754 ( 2007 ) . r. oda , private communication . u. seifert , j. shillcock , and p. nelson , phys . . lett . * 77 * , 5237 ( 1996 ) . j. b. fournier , eur . j. b * 11 * , 261 ( 1999 ) . s. a. langer , r. e. goldstein , and d. p. jackson , phys . a * 46 * , 4894 ( 1992 ) . d. pettey and t. c. lubensky , phys . e * 59 * , 1834 ( 1999 ) . v vitelli and d. r. nelson , phys . e * 74 * , 021711 ( 2006 ) . h. jiang , g. huber , r. a. pelcovits , and t. r. powers , e - print arxiv : 0707.4004 . the property of singular points on spherical clms is still same as that on achiral spherical vesicles discussed in @xcite . the difference will be evident in non - spherical clms . s. s. chern and w. h. chern , _ lecture on differential geometry _ ( beijing university press , beijing , 1983 ) . c. v. westenholz , _ differential forms in mathematical physics _ ( north - holland , amsterdam , 1981 ) .

a model was developed for the excitation of the uibs by h atom impacts in the interstellar medium . it builds upon the fact that , in the presence of far uv radiation and hydrocarbon grains , the hydrogen gas will be partially dissociated and the grain surface will be partially hydrogenated and partially covered with free carbon bonds . under such a statistical equilibrium , h atoms from the gas will recombine with c atoms at the grain surface at some rate . at each recombination , the h atom deposits an energy of about 5 ev in the grain . half of this is directly converted into vibrational excitation , always distributed in the same way among the most tightly coupled vibration modes of the grain . absent frequent grain - grain collisions , the only outlet for this energy is ir reemission , part of it in the uibs , provided the chemical structure of the grains is adequate , and the other part in the continuum . the partition only depends upon the grain size , all grains being assumed to have the same constitution . only a fraction , about 0.25 , of the grains ( among the smallest ones ) will contribute significantly to the uibs . it is shown quantitatively that h impacts are generally more efficient excitation agents than uv absorption because of the overwhelming abundance of hydrogen relative to uv photons . only very close to young bright stars is this no longer true because photon flux then largely exceeds h atom flux . thus h impacts and fuv absorption are both necessary to understand the variety of observed uib spectra . the model translates into a small number of equations enabling a quantitative comparison of its predictions with available astronomical observations , which have become exquisitely rich and accurate in the last two decades . [ firstpage ]

1110.1452

infrared radiation from molecules and grains in space is usually or mostly attributed to heating by visible and uv light absorption . while this may be justified by the ubiquity of light radiation , it must be remembered that the ism ( interstellar medium ) is also permeated with atomic hydrogen . h atoms are also found in abundance in the limbs of molecular clouds illuminated by bright young stars ( pdrs or photodissociated regions ) . this paper accordingly explores the relative importance of deposition of energy by atomic hydrogen impacts on the same targets . as is well known , the formation of molecular hydrogen by two isolated h atoms is forbidden by energy and momentum conservation ; it is therefore associated instead with recombination in presence of a third body * ( see spitzer 1977 ) . the most readily available such body is a hydrocarbon molecule or grain . the latter is made of a carbon skeleton to which peripheral h atoms are attached . an incident h radical , if moderately energetic , will attract one of the h atoms at the grain periphery forcefully enough to capture it and form an h@xmath0 molecule which readily escapes into space ; this is called h abstraction " in surface physics language . it will be shown below that this process does not leave much energy in the grain . however , it does leave an unoccupied dangling " c - bond . when the next incident h atom meets that free bond , it is most likely to form a strong chemical bond with the host c atom . this recombination " deposits in the grain nearly 4 ev , * half of which is available in the form of kinetic energy to excite the vibrations of the hydrocarbon particle ( the other half going into potential energy ) . although this is no more energy than is carried by a visible photon , it is totally expendable in vibrational excitation , by contrast with photonic energy , which is first delivered to electrons , to relax thereafter into continuum and vibrational radiation . the emission of vibrational bands following h - impact excitation is a type of chemiluminescence . * the basics of this process were first put forward by guillois et al . it may be considered as a special case of the radical reactions " previously envisioned by allamandola and norman @xcite ; but these authors did not treat any process in particular , dwelling instead on relaxation times and emission probabilities . * * * the present work is also motivated by several astronomical observations which hint at the need to complement the photon flux with another excitation agent for the emission of uibs ( unidentified infrared bands ) . one is the fact that the relative intensities of the bands in the spectrum of uibs do not change notably with the average photon energy of the ambient radiation ( uchida et al . 1998 ) ; besides , neither the paucity of uv radiation ( as in galaxy m31 ; pagani et al . 1999 ) nor the absence of starbursts ( see haas et al . 2002 ) precludes uib emission . another hint was the comparison , by onaka et al . @xcite , using the irts satellite , of uib emission ( 5 to 12 @xmath1 m ) at various galactic latitudes throughout the galaxy : the intensity of emission decreases quickly with increasing latitude , but its spectral profile remains unchanged . while there are no hot stars at these latitudes ( and therefore much less far uv photons ) , the uib intensity follows the trend of h atom density , which extends farther from the galactic plane than does molecular hydrogen density ( see imamura and sofue 1997 ) . earlier on , boulanger and prault @xcite , studying the solar environment by means of the iras satellite , found that , away from heating sources and molecular clouds , the ir emission from the cirruses of the ism is well correlated with the column density of hi gas . pagani et al . @xcite reached the same conclusion from the study of a large number of sight lines through the near - by andromeda nebula ; by contrast , they found no correlation of ir emission with uv flux . still another remarkable fact is the occurrence of the uib intensity peak , in pdrs ( photo - dissociation regions ) seen edge - on , in between the peak of recombination radiation ( signaled by h@xmath2 lines , e.g. br @xmath3 ) , on the star - illuminated side ( hii region ) , and the peak of h@xmath0 de - excitation radiation at 2.42 @xmath1 m . this is precisely where maximum atomic h density is to be found . the generic example is the orion bar ( see roche et al . 1989 ) , sellgren et al . 1990 , graham et al . 1993 ) . * an idealized illustration of the succession of regions and of the distribution of densities and radiative fluxes was given by guillois et al . @xcite . * * finally , the need for some excitation process other than photon absorption to understand available observations seems to have been felt by other workers ; e.g. duley and williams @xcite reconsidered briefly another type of chemical excitation of ir vibration : the sudden and violent release , upon mild heating , of potential energy stored in molecules in the form of radicals . * in sec . 2 , i describe the elementary processes which create conditions under which h atom excitation becomes operational . in sec . 3 , rate equations coupling the relevant physical quantities are laid down and the statistical equilibrium values of the variables are determined . the latter are used in sec . 4 to determine the energy build up in a grain , under h atom bombardment . section 5 displays expressions for the uib reemission power . at this stage , it becomes possible to compare uv and h impact excitation of grains , which is done in sec . finally , sec . 7 compares predictions of the chemiluminescence model with observations in various environments and measurements of different physical quantities . when an h atom ( radical ) impinges upon a hydrocarbon target , a number of processes may occur ( see lohmar et al . 2009 , cazaux and tielens 2002 , papoular 2005 ) : a)recoil ; this is usually nearly elastic , i.e. it involves little energy exchange . b)recoil with expulsion of an h atom off the target ; this breaking of a c - h bond requires an unusually energetic h projectile ( @xmath430 000 k ) . c)h abstraction , in which the incoming h atom is attracted by a target h atom to form a h@xmath0 molecule which readily leaves the target with a high velocity ; this leaves a dangling ( unoccupied ) c bond at the target surface , together with a fraction of the c - h bond energy . d)surface recombination of the h projectile with a target dangling bond ( chemisorption ) ; this may occur over a large range of projectile energies , and leaves the whole c - h bond energy in the target . \e ) sticking ( physisorption ) of the incoming h atom at the target surface , over which it can wander until it goes through ( a ) , ( c ) or ( d ) . the steady state coverage of the target depends on all 5 processes but the energy balance is not the same for all , as shown below , * using version 7.5 of hyperchem , the chemical simulation code commercialized by hypercube , inc . earlier versions of this code were used and described in detail by , for instance , papoular @xcite and @xcite . here , i used the semi - empirical pm3 method of molecular dynamics calculation provided by the package . this method uses quantum - mechanical calculations , supported where necessary by empirical constants from the chemistry laboratory , which are kept up to date . in pm3 , the empirical part was tailored by comparison with carbon - bearing molecules , so the energies given by this code agree with experiment to a few percent , better than necessary for present purposes . * first , the c - h bond energy was determined in several different structures . this was done by first optimizing the structure of interest ( at 0 k ) , i.e. minimizing its potential energy as a function of geometry , and measuring this energy . the h atom was then extracted and the new , higher , potential energy measured . the difference is the required bond energy , @xmath5 . for h@xmath0 , this is 109.4 kcal.mol@xmath6 . for an aliphatic chain ( -ch@xmath0-ch@xmath7 ) , @xmath5=98 ; in methyl ch@xmath8 , 101 to 106 ; in benzene , 116 to 118 , all in kcal.mol@xmath6 ( 1 ev=23 kcal.mol@xmath6 ) . * in the following , we take @xmath5=4 ev , uniformly , as higher accuracy is not warranted in describing the general properties of the present model . consider first case ( d ) . the target is assumed to be initially at rest at 0 k. in a simplified semi - classical picture , the representative point of this state is the bottom of a parabolic potential well * as in fig . 1.when the incoming h atom is captured by a dangling bond , the potential well is suddenly depressed by @xmath5 , so the representative point is now found above the new bottom by the same quantity , with zero ke ( kinetic energy ) . it is compelled to slide down the well , rise on the other side , fall down again and repeat this sequence indefinitely . in this new , dynamic , state the average ke and pe ( potential energy ) are both equal to @xmath9 . this energy is initially localized in the new c - h bond in the form of large amplitude stretching vibrations , but it quickly spreads to other , closely coupled , vibrations of the structure * * the average life time of this state before energy is lost through ir radiation is usually longer than 1 ms * ( see radzig and smirnov 1980 ) . so the posibility must be considered that , before this happens , another h radical is captured by the same target . the representative potential well will again be depressed . assume , for simplicity , that the transition to the new , third , state occurs at the moment when the representative point is at its peak height in the well with pe=@xmath5 and ke=0 . its instantaneous pe above the bottom of the new well will therefore be @xmath10 , and the average pe and ke will rise to @xmath5 . thus for each successive h recombination , the target particle gains @xmath11 ev in vibrational energy . if this energy were shared equally between , say n=100 atoms of the target , this would correspond to a * vibrational temperature of @xmath12170 k. * * consider now case ( c ) . here , the gain in pe of the target upon loosing an h atom is partially compensated by the loss of pe associated with the formation of a h@xmath0 molecule , because the h bond energies involved are almost equal . moreover , the h@xmath0 molecule carries away vibrational , translational and rotational kinetic energies . typical values for an aliphatic chain , for instance , are , respectively , 19 , 15 and 0.3 kcal / mol for the h@xmath0 molecule , leaving only a few kcal / mol in the target . for case ( b ) , one can develop the same argument as for ( d ) , except that , here , the target potential well is suddenly raised , instead of depressed , upon losing an h radical . * assume , in fig . 1 , that the system is given more energy than what is necessary to break a ch bond . after a while , half this extra gain in ( potential ) energy is converted into ke , as in case ( d ) . however , this process clearly requires an unusually energetic h projectile . * in estimating the overall vibrational energy gain , below , we will therefore neglect the contributions of recoil and h absraction by comparison with that of h recombination . by hi region is meant , here , a region of space permeated with photons energetic enough to dissociate molecular hydrogen without ionizing the atoms . this excludes dense molecular clouds but includes , for instance , cirruses in the diffuse ism . let @xmath13 be the number densities ( @xmath14 ) of atomic , molecular and nuclear hydrogen , respectively , with @xmath15 . assume the average velocity of the h atoms is @xmath16 cm.s@xmath6 , and the flux of dissociating photons is @xmath17 @xmath18s@xmath6 . let @xmath19 and @xmath20 be the number of c and h atoms in each hydrocarbon grain , and @xmath21 the number density of such grains ( @xmath14 ) . assume , for simplicity , that , in the ism , the mass ratio between grain and gas is 1/300 . ignoring heteroatoms , which are quite rare in fact , and h atoms which are much lighter , this implies @xmath22 , or @xmath23 . since we are interested in h recombination with naked c atoms ( those exhibiting a dangling bond ) at the grain surface , we have to distinguish between the naked and the h - covered " c atoms . this is illustrated in fig . 2 , for an aromatic and an aliphatic skeleton . roughly speaking , a covered c atom may anchor one or two h atoms according to whether the skeleton is aromatic or aliphatic ; so there is 0.5 to 1 covered c atom for every bound h atom . taking 0.75 for simplicity , and assuming the carbon skeleton is not too compact , we may therefore write @xmath24 on average . now , the relative numbers of covered and naked c atoms depend on the density of gaseous h atoms above the grain , and on the various c - s ( cross - sections ) of the gas - grain reactions . in statistical equilibrium , the number of gaseous h atoms captured by a grain is equal to the number of bound h atoms liberated from the grain by h impact . this can be written @xmath25 where @xmath26 are c - s s for recombination , abstraction and expulsion , respectively ; @xmath27 is the c - s for recombination after sticking ( indirect recombination " ) ; a factor @xmath28 was omitted from each side as it is the same for all reactions , so this equilibrium is independent of the ambient gas density . then , defining @xmath29 , we get @xmath30 and , hence , @xmath31 . the c - s s * involved in gas - grain interactions have been estimated , computed , modelled or measured ( see cazaux and tielens 2002 , * mennella et al . 2002 , papoular 2005 , lohmar et al . the results concur only approximately . here , we read from papoular @xcite , with @xmath32 cm.s@xmath6 : @xmath33 @xmath34 , @xmath35 @xmath34 and @xmath36 . equation 2 then gives @xmath37 , indicating that the hydrogen coverage of the target is @xmath38 . * by way of comparison , in a study of gaseous atomic hydrogen interactions with carbon grains , papoular @xcite found that the h coverage of hydrocarbon grains varies between @xmath39 and @xmath40 , depending mainly on the kinetic temperature of the h gas . in a recent study of the hydrogenation of hac ( hydrogenated amorphous carbon ) duley and williams @xcite estimated the coverage at 0.3 - 0.5 . * * * turning to the gas densities , let @xmath41 be the h@xmath0 molecule c - s for dissociation by photons . the rate equations for gaseous atomic and molecular hydrogen densities can be written @xmath42 the terms in the r.h.s . of the first of these equations are for photodissociation , abstraction , recombination and expulsion respectively ; in the second line , the terms are for photodissociation and abstraction respectively . * in statistical equilibrium , the derivatives in the l.h.s . must be set to 0 . defining * @xmath43 , * and using the second rate equation , * @xmath44 * finally , using @xmath15 , we obtain * @xmath45 note that , while @xmath46 depends only on c - s s , @xmath47 also depends on the radiative flux and on the velocity of h radicals . as for the dissociating photon flux which enters the rate equations , its value * varies considerably with the wavelength in the spectral region of interest ( lyman continuum , 900 - 1100 ) , and so does the dissociation rate @xmath48 ( see draine 1978 , fig . 3 , and references therein ) . it is therefore usual to replace the latter by an estimated average value , @xmath49 , in this wavelength range . we adopt @xmath50 s@xmath6 for the galactic ism ( see stecher and williams 1979 , jura 1974 ) . nearer to hot stars , @xmath51 , where @xmath52 is the galactic fuv ( far ultra - violet ) flux , and @xmath53 may reach up to @xmath54 . we shall take @xmath55 , assuming that the * uv spectral profile does not change as the radiation varies in intensity . using these parameter values , eq . 3 then becomes * * @xmath56 thus , _ the ratio of molecules to atoms , z , is proportional to the total number density of h nuclei , and inversely proportional to the uv flux_. this simplistic treatment neglects the extinction of dissociating light by molecules and grains along its path , as well as self - shielding of the h@xmath0 molecules against uv due to their excitation out of their ground level ( see draine and bertoldi 1996 , fig . 8) . when these effects are included , it is found that , for given density and radiation flux , the light is almost unaffected up to a critical column density , beyond which it is quickly extinguished while the molecular fraction increases steeply . the elaborate theory helps understanding the earlier observations of molecular cloud limbs ( see savage et al . 1977 , fig . 6 ) , which set the critical column density at a few times @xmath57 @xmath18 . the fortunate conclusion for our present purposes is that our treatment is acceptable below the critical column density , i.e. in most environments from which the uibs are usually observed to originate : cirruses , molecular limbs and pdrs . assuming the frequency of collisions between grains to be much lower than the frequency of h impacts on a particle , the energy deposited by the latter accumulates in the particle until it is reemitted in the form of ir radiation , which is its sole outlet . a fundamental quantity is , therefore , the total kinetic energy deposited per atom during the radiative lifetime . from eq . 2 , the total number of atoms per particle is @xmath58 from sec . 2 , the vibrational energy deposited by an impinging h atom that recombines on the particle is @xmath9 . the number of such events that occur before ir emission is proportional to the radiative lifetime , @xmath59 , to the ambient h atom influx , to the dangling bond coverage over the particle surface and to the c - s for direct and indirect recombination . on average , the final ( or maximum ) vibrational energy _ per target atom _ will be @xmath60\,,\ ] ] or , * using eq . 2 and 3,@xmath61 * the characteristic behaviour of this function is determined by the factor ( @xmath62 ) . as @xmath47 tends to 0 ( high @xmath63 ) , @xmath64 which is proportional to @xmath65 but no longer depends on @xmath53 . as @xmath47 increases beyond 1 ( low @xmath63 ) , @xmath66 which is proportional to @xmath53 but no longer depends on @xmath65 . for the parameters of our numerical example , and @xmath67 s , @xmath68 and @xmath69 , if @xmath65 is in @xmath14 and @xmath70 in ev for a direct comparison with observations , the total uib emission per unit volume and time is needed . assume that all the chemical energy carried by the impinging h atoms is converted into ir emission . but , for the grains to distinctly emit the uibs , they must be * neither too small ( in which case , they could not contain all the functional groups that carry the uibs ) , nor too large ( in which case most of the energy would go into the continuum ) : this is demonstrated by emittance mesurements on carbonaceous particles of different compositions and sizes ( see , for instance , solomon et al . thus , only a fraction , @xmath71 , in mass , of the is carbonaceous grains ( as defined by these two constraints ) must be included in our calculation . using eq . 8 , the uib emission can then be written * @xmath72 since @xmath70 is proportional to @xmath59 ( see eq . 6 ) , @xmath73 no longer depends on the latter . figure 3 displays @xmath74 , the uib power emitted per ambient h nucleus . again , we distinguish regions of low and high illuminations and write @xmath75 @xmath76 where @xmath65 is in units of @xmath14 and @xmath73 is in w.m@xmath77 ( choices imposed by observational practice ) in a given is environment , it may be helpful to determine which of the two cases applies . equation 8 clearly shows that this depends on whether @xmath47 is much larger or much smaller than 1/2 . the critical condition , @xmath78 , can be expressed in terms of @xmath53 and @xmath65 , using eq . 5 : @xmath79 which is displayed in fig . it is apparent , from eq . 5 and 11 , and fig . 4 , that , in the local ism ( @xmath80 ) , dissociation is rather high and that eq . 9 and 12 are the better approximations . it is instructive to determine the relative contributions of h atoms and dissociative uv photons to grain heating . this is done here by considering the ratio of the rates of h atom recombinations ( on grains ) and uv photon absorptions per unit grain surface , * multiplied by the corresponding energies deposited in the grain , @xmath81 where @xmath82 is the fuv absorption cross - section per c atom . for c - h bonds , this is @xmath83 ( turro 2009 ) , while for c = c bonds it rises to @xmath84 @xmath34 ( see roche et al . @xcite ) . for graphite , it is @xmath85@xmath34 . let us adopt @xmath86 . also take , as above , @xmath9=2 ev , @xmath32 cm.s@xmath6 , @xmath35 @xmath34 and @xmath87=0.5 . the average energy of dissociating ( lyman continuum ) photons will be taken to be 12 ev , and @xmath88 photons . @xmath18.s@xmath6 for the diffuse ism ( @xmath80 ; see draine 1978 and references therein ; mezger et al . 1982 ) . this gives @xmath89 , or , using eq . 4 for @xmath90 , in the two extreme cases ( @xmath47 large or small ) : * @xmath91 @xmath92 the determining factor in eq . 15 is @xmath93 . thus , _ in small fuv fluxes , indirect heating by h atom impacts is much more efficient than radiative heating , independent of @xmath53 ; this is essentially because there are so many fast hydrogen atoms that their flux is larger than that of fuv photons . but , in high fuv fluxes , hydrogen is almost totally dissociated , so @xmath94 decreases because , as @xmath53 increases , @xmath90 no longer increases while radiative heating still increases and ultimately becomes dominant . _ this conclusion holds even if some less energetic photons ( 1100 to 2400 @xmath95 ) are included ; it is even strengthened when it is recalled that photon absorption is intrinsically less efficient than h impact in producing uib emission , as some of its energy is diverted into ionization and another part into continuum emission . the ratio ( 15 ) can be slightly modified to become the ratio of total uib emission to total ir emission . to that effect , assume , as in sec . 5 , that total conversion of the chemical energy into uib emission occurs only in a fraction @xmath71 , in mass , of the carbonaceous grains . then , the numerator of eq . 15 must be multiplied by @xmath71 in order to represent total uib emission . on the other hand , note that not only dissociating photons ( 900 to 1100 @xmath96 ) , but also weaker uv photons ( 1100 to 2400 @xmath96 , according to the definition of the habing uv unit of flux ) , contribute to the radiative heating of grains , and must also be included in the photon flux , @xmath17 ; for the latter , we now take @xmath97 * ( based on fig . 3 of draine 1978 ) , so the denominator now nearly represents the total ir emission . expressions 15 to 17 are thus multiplied by @xmath98 . this will be compared with observations in sec we seek now to compare the predictions of the model with relevant observations . consider first the diffuse infrared emission " from the galaxy , as measured by boulanger and prault @xcite in the solar neighbourhood . they compared the ir measurements of the iras satellite , in its 4 bands , 12 , 25 , 60 and 100 @xmath1 m , with the hi radio emission from the same regions of the solar neighbourhood , away from heating sources and outside dense molecular clouds , where @xmath80 by definition . noting that the radio emission is proportional to the h atoms number density , they deduced the band emission per h atom , in each band , and found 1.1 , 0.7 , 1.1 and 3.2 , respectively , in units of @xmath99 w per h atom , for a total of 6.1 . boulanger and prault @xcite also derived an estimate of the emission per h atom of the diffuse ism , in the uib range ( 2 to 15 @xmath1 m ) : @xmath100 w.h atom@xmath6 . from these data , @xmath71 can be deduced . for the contribution of uv photons to direct grain excitation is negligible in the diffuse ism ( see sec . 6 ) , so both uib and total ir emissions are excited by h impact ( in the present model ) . as a consequence , their ratio is precisely equal to @xmath71 , according to our definition of the latter ( sec . 5 ) , so @xmath101 this considerable fraction is a necessary consequence of the assumption that uib and other ir emissions are excited by the same agent . this constraint can be met with h atom excitation because even relatively large grains can be excited ; the upper limit to their size is several thousands of c atoms per grain , before the continuum confiscates most of the chemical energy . in fact , the grain efficiency in exciting uibs must be a decreasing function of grain size , so the transition between uib - emitting and non - uib - emitting must be gradual rather than a sharp cut - off . pagani et al . @xcite made a similar study of the diffuse ism of the andromeda nebula , m 31 . they considered the signal from the lw2 ( 5 - 8 @xmath1 m ) filter of isocam on board the iso satellite as a tracer of uibs and found that _ it is extremely well correlated , across the galaxy , with the distribution of neutral gas as mapped by the hi ( 21 cm radio ) and co(1 - 0 ) signals ( @xmath65 in our model ) , which they also plotted . by contrast , the correlation is poor with the ionized gas as seen through its h@xmath3 emission , and non - existent with the uv emission_. they conclude that the mid - ir emission can be excited by the ambient visible and near - ir radiation . here , i argue that , if that were the case , there would still be a much tighter correlation with the young stars of the nebular ring than shown by the maps . instead , i set out to interpret quantitatively the findings of pagani et al . @xcite by means of the present model . pagani et al . analyzed their images , pixel by pixel , and plotted their measurements as a graph of i(lw2 ) as a function of the column density , n(h ) , in their fig . 6 . from this , they deduced the regression line @xmath102 * where @xmath103 , and i(lw2 ) is in mjy / pixel and @xmath104 , in @xmath18 . given the size of the pixel ( 6``x6 '' ) , and neglecting the ordinate at the origin , -0.23 , we deduce @xmath105 * since the isocam images cover mainly the general ( diffuse ) ism of m 31 , this number can be compared with the value , 1.5@xmath106 , measured by boulanger and prault @xcite in our galaxy , in the uib range ( 2 to 15 @xmath1 m ) . they are of the same order of magnitude , but the signal from m 31 is weaker by a factor @xmath1074 . this must be due , in large part , to the fact that the lw2 filter covered the range 5 - 8 @xmath1 m , * thus excluding the strong 11.2 and 12.7 @xmath1 m bands , while the measurement of the galaxy covered the larger range 2 - 15 @xmath1 m . the remaining difference is discussed next . * figure 6 of pagani et al . exhibits a significant dispersion about the regression line . this is quantified in their fig . 9 , which plots @xmath108 ( * our @xmath109 ) as a function of the fuv flux in a band of 15 nm width around 200 nm . when translated in the units used here , their measurements extend over * the ranges 0.25 to 0.8@xmath110 erg.@xmath18.@xmath111.arcsec@xmath112 for the fuv flux , and 0.32 to 4.75@xmath113 w/ h atom , for @xmath109 . * boulanger and prault @xcite , had also found considerable patchiness in the ir emission from the solar neighbourhood in all directions ( their fig . this was associated with the notion of cirrus " imagined by low et al . @xcite by analogy with this type of tenuous clouds in our atmosphere . * * * in the present excitation model , this dispersion can be interpreted in terms of h density differences between regions having the same uv illumination . in order to * justify this proposition , we return to eq . 11 , which gives the exact expression of the power reemitted in the uibs per unit volume of space , insert 0.25 for @xmath71 and divide by 2 to take into account the smaller band width of the lw2 filter . the result can be cast as * @xmath114 where @xmath115 , * as in eq . 5 , and @xmath65 is in @xmath14 . * clearly , @xmath116 varies with @xmath65 and @xmath53 . as an example , take the minimum measured uv flux in fig . 9 * of pagani et al . @xcite ( for which the dispersion of plotted points is greatest ) : 2.4 @xmath117 erg.s@xmath6.@xmath118.arcsec@xmath112 , which is close to the value in the solar neighbourhood : 2.1 10@xmath119 ( see draine 1978 ) . for lack of better information , assume that the ratio of dissociating uv flux to the flux at 200 nm is the same in m 31 as in the solar neighbourhood . then , for the selected points , @xmath120 , according to our definition of @xmath53 in sec . 3 . for the corresponding minimum and maximum values of @xmath116 , eq . 19 gives , respectively , about 5 and 77 h at.@xmath14 . for the average @xmath116 ( eq . 20 ) , @xmath121 @xmath14 . these calculations implicitely assume that every measured ir power emanates from a limited region of space ( a cirrus ) , with a relatively homogeneous density distinctly different from that of the environment ( see also sec . * boulanger and prault @xcite also derived an estimate of the emission per h atom of the diffuse ism , within 1 kpc of the sun , in the uib range ( 2 to 15 @xmath1 m ) : @xmath100 w.h atom@xmath6 . in this case , eq . 21 gives @xmath122 @xmath14 . for densities in this range and @xmath80 , @xmath116 is much more sensitive to variations of @xmath65 rather than in @xmath53 . so , although the fuv content in m 31 may be lower than in the solar environment , as suggested by pagani et al . @xcite , the weakness of the measured @xmath116 may , in the present model , be due to the relative paucity of hydrogen rather than to that of the fuv flux . when a young , bright star ( e.g. class o or b ) , happens to be in the vicinity or inside a denser ( and hence cooler ) cloud , its fuv radiation digs into the cloud to form an hii region ( signalled by ionic radiation or h recombination radiation ; e.g. br @xmath3 , p @xmath3 ) , followed farther by an hi region ( pdr proper ; essentially h radicals , signalled by centimeter - band radio emission ) , then by a thin layer of radicals recombining into excited molecules ( signalled by several near ir lines ; e.g. 2.42 @xmath1 m ) and , farther still , by the new edge of the molecular cloud ( signalled by co molecular lines ) . this sequence has often been described ; for a recent analysis , see habart et al . for our present argument , it is essential to note that , in this type of environment , the uib are observed to come from the intermediate , hi region and the adjacent h@xmath123 region * ( where vibrationally excited h is observed ) : see , for instance , roche et al . @xcite for the 11.3- and 12.7-@xmath1 m bands , sellgren et al . @xcite for the 3.3-@xmath1 m band , all observed in the orion bar ; an and sellgren @xcite , studying ngc 7023 , compared the maps of the 3.3-@xmath1 m band with its adjacent , underlying , ir continuum as well as the 1 - 0 s(1 ) line of excited h@xmath0 ; giard et al . @xcite studied m17 at 3.3 and 10 @xmath1 m , at 6 cm and in the c@xmath124o 2 - 1 band . this is a very strong indication of a link between h atoms and uib excitation , especially when it is remembered that , in the present model , those molecules that are formed after an h atom impact on a grain , leave the grain surface with a large amount of vibrational excitation ( see sec . 2 ) and that the h@xmath123 near ir emission is therefore generated at the interface between the hi region and the edge of the molecular cloud . * boulanger et al . @xcite later analyzed the data collected by the iso satellite from several such locations . they were thus able to quantitavely determine the brightnesses of the main uibs , at 6.2 , 7.7 , 8.6 , 11.3 and 12.7 @xmath1 m , and this for several sources for which they also determined the illumination @xmath53 . they found a common trend for all five bands : an initial rise in intensity as a function of illumination , followed by a progressive levelling - off for @xmath125 ( their fig . the brightness of the 7.7 @xmath1 m band , for instance , starts at @xmath126 w. m@xmath112.sr@xmath6 for @xmath80 ( the chamaeleon cirrus cloud in the dffuse ism ) , and terminates at 6.4 w. m@xmath112.sr@xmath6 for @xmath127 ( the omega nebula , m 17 ) . the single uv photon excitation ( also referred to as transient or stochastic heating * in the pah hypothesis ) expects a continuous rise of band brightness with @xmath53 , and the levelling - off is tentatively explained by grain destruction . the present model predicts an initial linear rise in incompletely dissociated regions followed by a levelling - off as a result of high dissociation , and limited amount , of available h gas . * for a more quantitative comparison of our model with observations , consider the emission @xmath73 given here by eq . this ir power is emitted in 4@xmath128 sr . the ir surface brightness of a source of depth @xmath129 along the line of sight is , then , @xmath130 . the results of astronomical measurements are usually expressed in the form @xmath131 , where @xmath132 . now , the spectrum of a finite material structure ( or small grain ) , as opposed to that of a homogeneous medium of infinite extent , is expressed in terms of its discrete line intensities , @xmath133 . there are @xmath134 of these for a total of @xmath135 atoms ( one for each vibrational mode ) ; they can be measured ( absorptivity ) or calculated ( electronic polarizability ) . if the line distribution is sufficiently dense , an adjacent averaging over @xmath136 makes sense , giving a continuous distribution @xmath137 assuming the excitation energy to be uniformly distributed among all vibrational modes , one can write @xmath138 , and the observed astronomical quantity becomes @xmath139 where @xmath129 is in meters , @xmath73 in w.m@xmath77 and @xmath140 in w.m@xmath112.sr@xmath6 . finally , a typical value of the pure number @xmath141 can be derived from a bright source , like ngc1482 ( smith et al . 2007 ) , for each of the main bands . this was done here by interpolating the flux spectrum with 600 points between 5 and 20 @xmath1 m , so that @xmath142 m . for the 7.65- and 11.3-@xmath1 m bands , for instance , one thus obtains 1.4 and 2.1 respectively . inserting @xmath143 in eq . ( 22 ) , together with expression ( 12 ) for @xmath73 , and @xmath144 ( as in sec . 7.2 ) , we obtain @xmath145 thus , for @xmath146 m ) to comply with the observed value of 10@xmath147 with @xmath80 , @xmath148 should be 3.3@xmath149 @xmath150 with , for instance , @xmath129=1 pc and @xmath151 @xmath14 , which seems reasonable . for this choice , @xmath152 and , in fig . 4 , the representative point falls above the critical line , which would justify the use of approximations 9 and 12 . it is tempting to use the relation between hydrogen column density and visible extinction along the same line of sight ( if available ) : @xmath153 ( see bohlin et al . 1978 ) . here , @xmath154=3@xmath155 , so @xmath156 ; but , in the case of chamaeleon , laureijs et al . @xcite measured @xmath157=1 . the discrepancy may be due to a number of causes , the most probable of which is that , for this value of a@xmath158 , the molecular hydrogen fraction in the column density may be quite high due to light extinction by molecules ( see savage et al . 1977 , fig . 6 ; draine and bertoldi @xcite , fig . 8) ; in this case , our expression 3 would underestimate @xmath47 , thus leading to a lower estimate of @xmath65 . at the high @xmath53 end of boulanger et al.s observations , for m 17 with @xmath127 , @xmath140(7.7@xmath1m)=6@xmath159 w.m@xmath112.sr@xmath6 . equation ( 26 ) then gives @xmath160 cm.@xmath161 . here , the geometry of the emitting region must be very different than that of a thin cloud , because the very strong illumination by a young , bright , star near by drives a shock in a thick cloud , thus considerably increasing the density in a region of limited extent : a bar " whose thickness along the direction to the star is a small fraction of a parsec ( see , for instance , the extensive study of the orion bar by bregman et al . @xcite ) . a more likely order of magnitude for @xmath129 , therefore , is now 0.1 pc , for which one gets @xmath162@xmath14 , again within reasonable range for this type of object . more recently , analyzing the data of the _ spitzer _ satellite , smith et al . @xcite produced a large number of good quality uib spectra from star - forming galaxies . the peak brightnesses of the main bands cover the range between 10@xmath163 and @xmath164 w.m@xmath112.sr@xmath6 , similar to the galactic data of boulanger et al . , confirming the remarkable uniformity of the uib * spectral profiles , independant of their intensity . * a crucial result of boulanger et al.s @xcite observations , among others , is that the relative intensities of the main uibs do not vary much , nor systematically , over a @xmath53 range of @xmath54 . this insensitiveness was noticed by several authors , observing various environments ( out of galactic plane , other galaxies , etc . ) : see , for instance , murakami et al . @xcite , chan et al . @xcite , mattila et al . @xcite , kahanp et al . @xcite , smith et al . it is in agreement with h - impact excitation , which , in nanoparticles , is followed by non - thermal , quasi uniform , redistribution of the same chemical energy ( 5 ev ) among those vibrational modes of the target that are strongly coupled . by contrast , the insensitiveness to photon flux is at variance with what one would expect from single uv photon excitation as noted by sellgren et al . @xcite . in this model , the grain temperature is supposed to decrease with the photon energy so the relative intensities of shorter - wavelength bands would also decrease ; the uibs should even fade out precipitously for illuminating star temperatures below @xmath165 k , as the fuv content of the illumination is then too small . but this was not borne out by their observations of several reflection nebulae illuminated by stars with temperatures ranging from 5000 to 21000 k ( see also uchida et al . 1998 ) . mattila et al . @xcite also remarked that their and others observations implied that the uib carriers are very resistant to different environments ( rne , pne and hii regions ) . none of the above is incompatible with the small spectral variations in position , intensity or width , which led to the classification of uib spactra in different classes ( see peeters et al . 2002 , tokunaga 1997 ) . in some cases , the variations are more radical , such as for the 11 - 13 @xmath1 m massif across ngc7023 ( boulanger et al . still more striking is the detection or non - detection of the 3.3-@xmath1 m band , independent of illuminating star temperature or ambiant uv content over 4 orders of magnitude of @xmath53 ( smith et al . this poses a common problem to all models , but appears to be linked with particular chemical processes or environments , rather than with the excitation process ; as such , it falls outside the scope of this work . in most uib spectra , an underlying continuum is clearly apparent , rising towards long wavelengths and often very weak at short wavelengths ( see , for instance , cesarsky et al . 2000 , peeters et al . 2004 , zhang and kwok 2010 ) . from their analysis of measurements with the iso satellite in the 6 - 12 @xmath1 m band and at 100 @xmath1 m , kahanp et al . @xcite concluded that both emissions were strongly correlated at small and large scales , which implies a strong connection ( origin , chemical composition , environment ) between the small grains responsible for the former , and the large ones , responsible for the latter . mattila et al . @xcite arrived at the same conclusion from their study of the disk of ngc 891 , which they observed at 8 wavelengths characteristic of different ism components . this conclusion is in line with the present treatment , as well as with the roughly constant ratio of uib to total ir emission ( sec . . however , it seems to be at odds with the large variations of the spectral band - to - continuum ratio , or contrast , @xmath166 , from object to object ( see also boersma et al . if the grain composition does not change with size , this may be explained as follows . the continuum brightness contribution to the band at @xmath167 can be written @xmath168 where @xmath169 is the grain emissivity , @xmath170 is planck s black - body law ( w.m@xmath112.@xmath1m@xmath6 ) , @xmath129 the object depth along the line of sight ( m ) , @xmath21 the grain number density ( m@xmath77 ) , and @xmath171 the grain optical c - s ( m@xmath172 ) . the grain temperature is estimated by assuming thermal equilibrium and equating the power absorbed by a grain from the illuminating flux @xmath173 , to the power reemitted by the same grain in the ir . for simplicity , assume the grain absorptivity ( emissivity ) varies as @xmath174 and that all the reemitted power is at the same wavelength , viz . the peak of the bb ( black body ) emission , @xmath175 , and is given by the stefan - boltzmann law , @xmath176 where @xmath177w.@xmath18.k@xmath178 . also assume the flux that heats the large grains is @xmath179 , where @xmath53 is the same as above , but @xmath180 is much larger than the dissociating flux , in order to include lower energy photons . following boulanger and prault @xcite , we take @xmath181w.@xmath18 . also , defining the photon wavelength as 0.1 @xmath1 m , one gets @xmath182 . then , equating absorbed and reemitted powers , the grain diameter and its emissivity ( at the reference wavelength ) factor away , and it is easily deduced that @xmath183 equation ( 26 ) still requires an estimate of the emissivity , @xmath169 ; it can be obtained from the consideration of the total far ir emission measured by boulanger and prault @xcite in the iras 60- and 100-@xmath1 m bands : @xmath184 w per h atom . this is to be equated with the stefan - boltzmann law . now , spitzer @xcite estimated the grain c - s at @xmath185@xmath34 per h atom . assuming the fraction associated with carbonaceous dust to be @xmath186 , then @xmath187@xmath188 . finally , taking @xmath189k , we obtain @xmath190 at the peak of the black body emission , 150 @xmath1 m and , for an opacity in @xmath174 , @xmath191 using the spitzer result again to express @xmath192 in eq . 26 , the latter becomes @xmath193 finally , the contrast becomes @xmath194 where @xmath73 is taken from eq . 11 and @xmath195 , from eq . 23 . as an illustration , fig . 5 plots @xmath196 as a function of @xmath53 , as well as @xmath166 for @xmath197 m and 11.3 @xmath1 m , both for @xmath198 @xmath14 . our model predicts that , for most environments in the ism , @xmath53 and @xmath65 are such as to deliver high contrasts , as usually observed . it also predicts that the contrast decreases as the uib wavelength increases , as observed , except for the diffuse ism ( see onaka et al . 1996 ) : this is because , there , hydrogen is dissociated although the illumination is so low that the larger grains are very cold and the continuum very weak . finally , the model predicts that the transition from high to low contrast is very swift as @xmath53 increases , due to the steep short wavelength edge of planck s law . as a result , the uibs paradoxically disappear " at high illuminations ! in the present model , this is not due to their destruction , but to the fact that the underlying continuum quickly increases , even as the uib emission levels off because all available hydrogen is already dissociated ( cf . observational examples of this behaviour of the band - to - continuum ratio can be found in giard et al . @xcite ( 3.3-@xmath1 m uib , 10-@xmath1 m continuum ) and cesarsky et al . @xcite ( 6.2-@xmath1 m uib , 16-@xmath1 m continuum ) , both observing m 17 . the predictions of our model concur with the conclusions of an and sellgren @xcite , based on their observations of the reflection nebula ngc 7023 with high spatial resolution , in the 3.29-@xmath1 m uib and its neighboring continuum at 2.18 @xmath1 m . they find the two emissions to coexist in a vast hi region between the hii region enclosing the illuminating star , and the thin h@xmath123 filament at the edge of the molecular cloud . however , the intensity distributions are distinctly different , with the 3.29-@xmath1 m peaking within the filament , while the 2.18-@xmath1 m peaks at about half this distance from the star . moreover , they measured a feature - to - continuum ratio peaking at @xmath199 , in the filament , and decreasing with the distance , @xmath94 , from the illuminating star as @xmath200 , i.e. roughly as @xmath201 ( here , @xmath53 is proportional to the star flux heating the grains ) . remarkably , this trend is very smooth , in stark contrast with the very dissimilar spatial distributions of the two emissions . as noted by the authors , both these facts are hardly understandable if both emissions are prompted by absorption of the star radiation . on the other hand , the present model predicts just this behaviour of the band contrast . however , fig . 5 suggests a much faster variation with @xmath53 . note that the @xmath200 trend is only a regression line over a large set of highly dispersed points , covering only a small range of @xmath53 : a factor 2 , which suggests high density heterogeneity . so the discrepancy may be due to the lines in fig . 5 being drawn for constant @xmath65 . but it could also be due to a number of assumptions made for the sake of simplicity : -assuming too tight relationships between the densities of large grains ( emitting the continuum ) , uib - emitting grains and h radicals ( sec . these relationships hold best in the diffuse ism , and may be invalidated in the presence of very intense uv radiation able to drive a shock front into the h cloud . -assuming that the illuminating spectrum does not change with @xmath53 , while , in fact it varies considerably from the diffuse ism to the neighbourhood of a bright star . ever since pipher @xcite discovered the diffuse ir emission ( dire ) originating in the galactic plane , the problem of its excitation has been open for discussion . based on observations by low et al . @xcite and rouan et al . @xcite , respectively , mezger @xcite estimated the total dire at @xmath202w , while ryter and puget @xcite set it at 7.5@xmath203w . given reasonable star formation rates , it has been hard to argue that this emission can be accounted for by the luminous o stars alone . mezger estimated that o stars fuelling radio hii regions contribute only 20 % of the total . he also concluded that , for uv absorption and ir reemission by the galactic grains in the diffuse medium to account for @xmath204 , these grains should be able to absorb 8 times more uv photons than they can do in fact ( this is the ir excess " ) . section 6 above suggests that h atom impacts might do the job . in order to explore this possibility in the present model , note that the total number of h atoms in the galaxy was estimated by ryter and puget @xcite and mezger @xcite , again excluding the galactic center , at @xmath205m@xmath206 , which amounts to @xmath207 h atoms . using the values of @xmath204 mentioned above , we deduce the total ir power reemitted per h atom , @xmath116=4.5 and 14 @xmath99 w / h atom . these are comparable to the value deduced from the iras data by boulanger and prault @xcite : 6.1 @xmath99 . now , @xmath204 must be compared with the value predicted by our model , i.e. eq . 11 , which gives the total ir power reemitted per unit volume of space , if @xmath71 is replaced by 1 . here , as in sec . 7.1 , we deal with a tenuous environment , so approximation ( 12 ) may be used instead . integrating this quantity over the volume of the galaxy ( excluding star forming regions ) , we obtain @xmath208 where @xmath65 is in @xmath14 . the total volume , v , of the galaxy is the ratio of the total number of h nuclei divided by the average density of the latter . this is about 1 @xmath14 ( see spitzer 1977 ) , so v@xmath209 @xmath210 . the root mean square density can then be deduced : @xmath211 where @xmath212 was set at @xmath213 w. the square root of this is 14 @xmath14 , which is typical of clouds containing optically thin h@xmath0 in the dism ( see jura 1975 ) ; the model thus proves to be sound . the difference between @xmath214 and @xmath215 is a measure of the patchiness of the diffuse ism . a model was developed for the excitation of the uibs by h atom impacts in the interstellar medium . it builds upon the fact that , in the presence of far uv radiation and hydrocarbon grains , the hydrogen gas will be partially dissociated and the grain surface will be partially hydrogenated and partially covered with free carbon bonds . under such a statistical equilibrium , h atoms from the gas will recombine with c atoms at the grain surface at some rate . at each recombination , the h atom deposits an energy of about 5 ev in the grain . half of this is directly converted into vibrational excitation , always distributed in the same way among the most tightly coupled vibration modes of the grain . absent frequent grain - grain collisions , the only outlet for this energy is ir reemission , part of it in the uibs , provided the chemical structure of the grains is adequate , and the other part in the continuum . the partition only depends upon the grain size , all grains being assumed to have the same constitution . only a fraction , @xmath216 , of the grains ( the smallest ones ) will contribute significantly to the uibs . it is shown quantitatively that h impacts are generally more efficient excitation agents than uv absorption because of the overwhelming abundance of hydrogen relative to uv photons . only very near young bright stars is this no longer true because photon flux then largely exceeds h atom flux . thus h impacts and fuv absorption are both necessary to understand the variety of observed uib * spectral intensities . * the model translates into a small number of equations enabling a quantitative comparison of its predictions with available astronomical observations , which have become exquisitely rich and accurate in the last two decades . section 7 collects some of the most crucial tests of the model : - it predicts uib emission even far from star bursts and in poor - fuv environments . - since the excitation agent and the deposited energy are always the same , uib reemission is independent of the temperature of the neighbouring illuminating stars . - in pdrs , it becomes possible to understand the observed very near spatial coincidence of hi and h@xmath123 emission with the peak of the uibs . - the model predicts the measured average power in the uibs ( 2 - 15 @xmath1 m ) emitted per ambient h atom ( in statistical equilibrium in the solar neighbourhood ) , 1.5 @xmath99 w , provided that 1/4 of the cosmic carbon be locked in uib - emitting grains , and the average h nuclei density be about 100 @xmath14 in the average emitting cloud . - agreement between computed and measured uib intensities is obtained for a given fuv illumination , provided the unknown h column densities are given adequate values , which are found to be quite plausible . this is true even for m 31 , which is presumed to be poor in fuv flux . - the assumption of a common carrier for the uibs and the underlying continuum makes it possible to quantitatively account for the observed large variations of the band - to - continuum contrast ratio : they are simply due to the steep variation of wien s law , at a given wavelength , for small variations of illumination . - the ir excess " paradox , posed long ago by the comparison of the overall galactic ir emission with the uv flux available for grain excitation , is resolved quantitatively by h atom excitation , provided the r.m.s . density of h nuclei throughout the galaxy be of order 200 @xmath217 . * in order to allow quantitative comparisons to be made with observations , a number of assumptions were necessary . the most restrictive one is the neglect of uv extinction and self - shielding of hydrogen ; the calculations are progressively invalidated as column densities along the sight lines exceed a few times 10@xmath218 @xmath18 . fortunately , this is not the case for most of the mesured uib spectra . * for lack of relevant experimental results or explicit values from theoretical computations , i relied heavily on numerical chemical simulation to obtain the values of cross - sections of grain - gas interactions . as the heating is proportional to the c - h recombination c - s ( eq . 8) , the latter needs measurement or at least confirmation by other means . it was also assumed that the importance of the dissociating uv flux relative to total uv does not depend on the total uv flux ; this only impacts the degree of dissociation of hydrogen , and most of our calculations are weakly sensitive to it . none of these assumptions notably affects the general nature and performance of this model . the notion of a fraction @xmath71 should not be taken to correspond to definite lower and upper size limits ; for the transition between emitting and non - emitting grains is progressive and likely to depend on grain structure and composition . some of the physical processes involved here are evidenced by molecular dynamic simulations . even for a 100-atom molecule , an important broadening mechanism is already at work and directly observable in chemical simulations : this is the interaction between modes and constant redistribution , between them , of the initially deposited energy . this widens lines into bands and confers wings to each line , so plateaus " appear below bands " in crowded spectral regions , and even an underlying continuum is visible ( see papoular 2000 ) . this effect is observed to increase quickly with the size , much more so than the number of modes . as the grain size increases , the number of skeletal vibration modes ( involving the whole structure ) increases more rapidly than the number of modes of functional groups ( involving only a small number of atoms of adequate chemical nature , and with adequate bonds to other atoms ) . the uibs are mainly due to the latter , while the continuum ( mainly in the farther ir ) is due to the former . thus , the fraction of energy deposited in the grain that goes into uib emission progressively decreases as the size increases . this is observed up to the largest structures attainable by the usual simulations ( of order of 10@xmath219 atoms ) . tailoring the composition and structure of the grains so that they carry the uibs is outside the scope of this paper ( it was treated at length by papoular 2011 ) . the validity of our conclusions do not depend on these properties , provided the grains bear enough carbon dangling bonds at their periphery , in steady state , so they can capture ambient h atoms . i thank the reviewer , dr c. boersma , for several useful comments . an j. and sellgren k. 2003 , apj 599 , 312 allamandola l. and norman c. 1978 , a&a 66 , 129 boersma c. , bauschlicher jr . c. , allamandola l. , ricca a. , peeters e. and tielens a. 2010 , a&a 511 , a32 bohlin r. , savage b. and drake j. 1978 , apj 224 , 132 boulanger f. and prault m. 1988 , apj 330 , 964 boulanger f. , abergel a. , bernard j. , cesarsky d. , puget j .- , reach w. , ryter c. et al . 1998 , asp conf . ser . 132 , p15 boulanger f. 1999 , les houches lxx , no 11 p253 bregman . j. , larson k. , rank d. and temi p. 1994 , apj 423 , 326 cazaux s. and tielens a. 2002 , apj lett . 575 , l29 cesarsky d. , lequeux j. , abergel a. et al . 1996 , a&a 315 , l309 cesarsky d. , lequeux j. , ryter c. and grin m. 2000 , a&a 354 , l87 chan k .- w . , rorllig t. , onaka t. et al . 2001 , apj 346 , 273 duley w. and williams d. 2011 , apj lett . 737 , l44 draine b. 1978 , apj suppl . 36 , 595 draine b. and bertoldi f. 1996 , apj 468 , 269 giard m. , bernard j.p . and dennefeld m. 1992 , a&a 264 , 610 guillois o. et al . 1998 , faraday discuss . 109 , 335 graham et al . 1993 , apj 105 , 250 haas m. , klass u. and bianchi s. 2002 , a&a 385 , l23 habart e. , boulanger f. and verstraete l. 2011 , a&a 527 , a122 imamura k. and sofue y. 1997 , a&a 319 , 1 jura m. 1974 , apj 191 , 375 jura m. 1975 , apj 197 , 575 and 581 kahanp j. , mattila k. , lehtinen k. , leinert c. and lemke d. 2003 , a&a 405 , 999 laureijs r. , chlewicki g. , wesselius p. and clark f. 1989 , a&a 220 , 226 lohmar i. , krug j. and biham o. 2009 , a&a 504 , l5 low f. et al . 1977 , apj lett . 214 , l115 low f. et al . 1984 , apj lett . 278 , l19 mattila k. , lehtinen k. and lemke d. 1999 , a&a 342 , 643 mennella v. , brucato j. , colangeli l. and palumbo p. 2002 , apj 569 , 531 mezger p. 1978 , a&a 70 , 565 mezger p. 1982 , a&a 105 , 372 murakami h. , freund m. , ganga k. et al . 1996 , pasj 48 , l41 onaka t. , yamamura i. , tanabe t. , roellig t. and yuen l. 1996 , 48 , l59 pagani l. , lequeux j. , cesarsky d. , donas j. , millard b. , loinard l. and sauvage m. 1999 , a&a 351 , 447 papoular r. 2000 , spectrocimica acta 57a/4 , 947 papoular r. 2005 , mnras 359 , 683 papoular r. 2011 , mnras 415 , 494 ; see also 2008 , arxiv 1008.5136 peeters e. , hony s. , van kerckhoven c. , tielens a. , allamandola l. , hudgins d. and bauschlicher c. 2002 , a&a 390 , 1089 peeters e. , spoon h. and tielens a. 2004 , apj 613 , 986 pipher j. 1973 , iau sympos . 52 radzig a. and smirnov b. 1980 , reference data on atoms , molecules and ions , springer - verelag berlin heidelberg roche p. , aitken d. and smith c. 1989 , mnras 236 , 485 rouan d. , lena p. , puget j .- de boer k. and wijnbergen , j. 1977 , apj lett . 213 , l39 ryter c. and puget j.l . 1977 , apj 215 , 161 savage b. , bohlin r. , drake j. and budich w. 1977 , apj 216 , 291 sellgren k. , tokunaga a. and nakada y. 1990 , apj 349 , 120 sellgren k. 2001 , spectrochimica acta 57a , 627 smith t. , clayton j. and valencic l. 2004 , astron . j. 128 , 357 smith j. , draine b. , dale d. et al . 2007 , apj 656 , 770 solomon p. , carangelo r. , best p. , markham j. and hamblen d. 1986 , 21st sympos . on combustion , 437 spitzer l. 1977 , physical processe in the interstellar medium , j. wiley & sons , n. y. stecher t. and williams d. 1979 , a&a 67 , 115 tokunaga a. 1997 , asp conf . vol 124 , 149 turro n. et al . 2009 , principles of molecular photochemistry uchida k. , sellgren k. and werner m. 1998 , apj lett . 493 , l109 zhang y. , kwok s. and hrivnak b. 2010 , apj 725 , 990

previous computations of the potential landscape with the shapes parameterized in terms of cassini ovaloids are extended to collective dynamics at finite excitations . taking fission as the most demanding example of large scale collective motion , transport coefficients are evaluated along a fission path . we concentrate on those for average motion , namely stiffness , friction and inertia . their expressions are formulated within a locally harmonic approximation and the help of linear response theory . different approximations are examined and comparisons are made both with previous studies , which involved different descriptions of single particle dynamics , as well as with macroscopic models . special attention is paid to an appropriate definition of the deformation of the nuclear density and its relation to that of the single particle potential . for temperatures above the inertia agrees with that of irrotational flow to less than a factor of two , but shows larger deviations below , in particular in its dependence on the shape . also friction exhibits large fluctuations along the fission path for small excitations . they get smoothed out above where attains values in the range of the wall formula . for the inverse relaxation time turns out to be rather insensitive to the shape and increases with . addtoresetequationsection pacs numbers : 21.60.ev , 21.60.cs , 24.10pa , 24.75+i * to appear in prc *

nucl-th9701032

one of the oldest but still most challenging problems of nuclear physics is an adequate description of collective motion at finite excitations . as the prime example one may quote nuclear fission which has attracted the attention of both experimentalists as well as theoreticians since its discovery . to date it is still an open question which type of configurations the system undergoes on its way from the potential minimum over the saddle region down to scission . whereas in the early days those of the compound model were clearly favoured in theoretical pictures , after the discovery of the shell model that of independent particle motion came into fashion more and more . this development was enhanced after computers got fast enough such that hartree - fock type computations could be done in every lab . however , there can be little doubt that this picture fails to describe collective motion at finite excitations where one is compelled almost by experimental evidence that the dynamics shows irreversible behaviour , not only by the very nature of the decay process itself but by the appearance of frictional forces . it is more than questionable that this feature can adequately be met by introducing simple minded collision terms . decent descriptions of fission in terms of the one body density operator most likely require to consider correlations beyond the independent particle picture , together with non - markovian effects . this is a difficult problem in itself , not to mention the computational task of solving this equation of motion for the one body density . for these reasons it may still be interesting and worth while to start from a more phenomenological point of view introducing the shape parameters as collective variables . it is true that in this way again the picture of independent particles will serve as a starting point , in the form of the deformed shell model . however , the latter is simple and flexible enough to allow one considering residual interactions , in one way or other . as we shall see , it may be possible to gain insight into their importance by studying dynamical aspects . likewise , we may be able to get information on the complexity of the configurations which are to be considered . such a task becomes more feasible in case collective motion is sufficiently slow . then one may exploit the quasi - static picture which reduces the complexity of the full problem drastically . under such circumstances one may actually linearize the problem and treat collective motion locally within a harmonic approximation . in this way one may take advantage of the benefits of linear response theory . one of the major problems in theories of this type is to find a decent guess for the relevant macroscopic variables , a problem which is familiar almost from all transport theories . for nuclear fission there exists some kind of guiding principle through the liquid drop model . the latter is known to represent the static energy for temperatures above @xmath8 . since at these temperatures one expects motion to be strongly damped , it will most likely follow somehow the line of steepest decent . possible shapes which a fissioning nucleus may assume on its way to scission have been looked for in @xcite by minimizing the liquid drop energy . this minimization has been done for some realistic energy density functional under the constraint of fixing a parameter which measures the distance between the evolving fragments . incidentally , it is the same parameter which we are going to exploit later on in our approach . it so turns out that the shapes found in this way can be approximated fairly well by the cassini ovaloids introduced to nuclear physics in @xcite . later in @xcite a single particle model has been constructed for such a parameterization of shapes , which was based on the woods - saxon potential . in this paper we are going to use this model for computations of transport coefficients , after some suitable modifications which are necessary to incorporate the effects mentioned above . one of our goals will be to study average motion along the fission path for different temperatures , as it is reflected in the associated transport coefficients of inertia , friction and local stiffness . in this sense the aim of our present work is similar to the one of @xcite , where a two center shell model was used . the latter feature renders the previous model simpler on the computational level . on the other hand , the parameterization of the shape by means of cassini ovaloids opens the possibility of treating more realistic shapes , which are perhaps better suited to describe the later stages of a fission process . furthermore , it is fair to say that the woods - saxon potential may be supposed to resemble more the `` true '' mean field . for cassini ovaloids commonly a few parameters suffice to treat in simple terms a whole variety of realistic shapes including very compact ones as well as strongly deformed ones with a well developed neck , or even those corresponding to separated fragments . in this sense this parameterization may be considered superior to expansions in terms of spherical harmonics ( see @xcite ) . in the ideal case one would then be able to compute transport tensors for all the parameters , the collective degrees of freedom , one claims to be relevant . this is a tremendous task and so far has been carried through only for a two dimensional model @xcite , without utilizing though the full microscopic potential of linear response theory . in this paper we want to restrict ourselves to the one dimensional case . the main reason for that is found , of course , in the simplification one gains by this restriction . however , it may be said that at present most of the applications of macroscopic equations of motion to fission at finite excitation adhere to a similar confinement , see e.g. @xcite - @xcite . evidently one then needs to rely on the `` right '' guess of the fission path . as said before and for arguments given there we presume it to be represented well enough by the line along the valley of the static energy . possible improvements have to be left for future studies . we follow the suggestion put forward in @xcite , but would like to repeat the most important elements for convenience . the cassini ovaloids are obtained by rotating the curve @xmath9^{1/2}\ ] ] around the @xmath10-axis , with @xmath10 and @xmath11 being cylindrical coordinates . the constant @xmath12 is defined by volume conservation , implying that the family of shapes ( [ ovals ] ) depends only on one deformation parameter @xmath13 . as is easily recognized from ( [ ovals ] ) the value of @xmath14 corresponds to a sphere . for @xmath15 the form resembles very much that of a spheroid with the ratio of the axes given by @xmath16 at @xmath17 a neck appears and at @xmath18 the nucleus separates into two fragments . a few examples of the family ( [ ovals ] ) are shown in fig.1 . it is possible to describe a more general class of axially symmetric shapes by exploiting an expansion about the surface given by ( [ ovals ] ) . we may introduce two new coordinates @xmath19 and @xmath20 such that @xmath21 corresponds to the cassini ovals ( [ ovals ] ) . the coordinate @xmath20 specifies the position of a point on the line given by ( [ ovals ] ) , see @xcite for details . with these two variables at our disposal we may parameterize a new shape . the latter is meant to express the deviation from the ovaloid given by ( [ ovals ] ) by means of an expansion into a series of legendre polynomials @xmath22 the full set of collective variables or parameters then includes the coefficients @xmath23 in addition to the @xmath24 from before . sometimes it is convenient to introduce a measure for the overall elongation of the nucleus instead of @xmath24 . one may chose , for instance , the distance @xmath25 between the left and right center of masses . to have a dimensionless quantity one may divide @xmath25 by the diameter @xmath26 of the sphere ( of identical volume ) to get @xmath27 with @xmath28 being the @xmath10-coordinate of the center of mass of the the whole complex . the integration is carried out over the volume within the sharp surface specified by ( [ alphas ] ) . asymptotically , the @xmath29 turns into half of the distance between centers of mass of the fission fragments . incidentally , this variable is defined uniquely for any parameterization of the shape and has been used in the past by many authors . this feature facilitates comparisons to theories which are based on shell models with different shape variables . the relation between @xmath29 and @xmath13 is demonstrated in fig.2 . the single - particle hamiltonian @xmath30 will be constructed like in @xcite . it has terms for the kinetic energy @xmath31 , the radial potential @xmath32 , the spin - orbit coupling @xmath33 and the coulomb potential @xmath34 : @xmath35 the radial part @xmath32 is represented by a finite depth woods - saxon potential @xmath36^{-1}\ ] ] where @xmath37 is the shortest distance from the point @xmath38 to the sharp surface and @xmath12 is the diffuseness parameter which is assumed to be constant along the surface . the spin - orbit potential may be written in a way which makes apparent that it is proportional to the gradient of the potential given in ( [ wspot ] ) , @xmath39\nabla v\ ] ] here @xmath40 and @xmath41 stand for the nucleon s momentum and the spin . the coulomb potential is calculated for a charge distributed uniformly inside the sharp surface ( [ alphas ] ) or ( [ ovals ] ) . the single - particle energies and wave functions are determined by diagonalizing the matrix of the hamiltonian ( [ hipm ] ) calculated with the wave functions of a deformed axially symmetric oscillator potential , see @xcite . an example of the deformation dependence of the single - particle energies is shown in fig.3 . as the result of diagonalization one obtains not only energies and wave functions of bound states but also those of discrete states of positive energy , which for the woods - saxon potential lie in the continuum . the density of these states depends on the number of oscillator shells included into the basis . in the computations within the shell correction method the number of oscillator shells is optimized by the requirement that the states with positive energies provide a smooth extrapolation of the density of bound states into the continuum . accounting for such states with positive energy improves considerably the `` plateau '' of the shell correction as a function of the averaging interval . in the present paper we do so not only when calculating deformation energies but also in the computation of transport coefficients . the cut - off in the single - particle energy was set equal to @xmath42 . we have checked that a variation of the cut - off energy within the interval @xmath43 does not change much the values of the transport coefficients . this may be understood from the fact that these transport coefficients reflect the truly low frequency behavior of the system . the deformation energy @xmath44 at zero temperature is calculated according to the shell correction method @xcite,@xcite as the sum of the liquid drop energy @xmath45 and the shell correction @xmath46 ( including the one for the pairing energy ) @xmath47 the liquid drop energy is computed as the contributions from the coulomb @xmath48 and surface @xmath49 energies according to @xcite,@xcite @xmath50 where @xmath51 and @xmath52 are the corresponding energies of the spherical shape . as an example , fig.4 exhibits the results of the calculation of @xmath44 and @xmath45 at zero temperature for @xmath53 as function of parameters @xmath13 and @xmath54 . the temperature dependence of coulomb and surface energy is accounted for by using the forms @xmath55 with @xmath56 and @xmath57 @xmath58 @xmath59 @xcite . to compute the shell correction at finite temperature we use the phenomenological ansatz proposed in @xcite @xmath60 { \tau \over \sinh \tau}\ ] ] with @xmath61 and @xmath62 , @xmath63 for temperatures larger than @xmath64 ( as considered in the present paper ) the shell effects are strongly suppressed . already at @xmath65 the minimum of the total deformation energy almost coincides with the bottom of the liquid drop valley . as said before the latter can be approximated rather well just by the cassini ovaloids . thus we restricted our set of deformation parameters to the one parameter @xmath13 only , with all the @xmath23 put equal to zero . however instead of @xmath13 we prefer to use @xmath29 defined by [ rr12 ] . the @xmath29-dependence of the total deformation energy and that of the liquid drop is shown in fig.5 . being the mass number of the fissioning nucleus . in the following we will assume to be given a many body hamiltonian @xmath66 which depends parametrically on the collective variable @xmath67 which specifies the shape of nuclear surface . although for the computations to be presented below , it will mostly be identical to @xmath29 , in this section we still prefer to use the general notation @xmath67 instead , last but not least to indicate the general validity of the discussion to come . this hamiltonian is assumed to represent the system s total energy . on the level of the shell model this means to add some c - number terms to the sum over those single particle hamiltonians introduced in ( [ hipm ] ) ( see@xcite ) . as will be discussed in the next section , later on we want to account for collisional damping , which from a principal point of view requires to add a two body interaction @xmath68 . for the moment it is not very important to know details about the way it will be handled besides the fact that we claim this interaction to be independent of @xmath67 . as a consequence of the latter feature , the generator for collective motion , namely @xmath69 @xmath70 , is of pure one body nature . this operator defines the main source of the coupling between the collective degree of freedom @xmath71 and the nucleonic ones . indeed , within the locally harmonic approximation ( lha ) the effective hamiltonian can be written as @xmath72 in the second order term the `` nucleonic '' part appears only as an average of the corresponding operator . consistently with the harmonic approximation , this average is to be built with that density operator @xmath73 which in the quasi - static picture is to be calculated with the hamiltonian at @xmath74 , namely @xmath75 . it is here where thermal concepts come into play . in this `` unperturbed '' density operator ( for the nucleons ) one needs to specify the amount of heat the nucleonic ( or intrinsic ) degrees of freedom have at the given configuration parameterized by @xmath74 . the simplest possibility is offered by the canonical distribution @xmath76 , to which our computations will be restricted . clearly this picture has to rely on the assumption of a quick relaxation of the relevant internal degrees of freedom ; we will come back to this question later on . details about this lha can be found in many references , see e.g. @xcite,@xcite , @xcite : there it is also described how this local dynamics can be handled within a suitable application of linear response theory . for this reason , we will only recall the most important theoretical issues . the local motion in the @xmath67-variable can be described in terms of the so called collective response function @xmath77 . it can be derived by introducing a ( hypothetical ) external force @xmath78 and by evaluating how the deviation of @xmath79 from some properly chosen static value reacts to this external field in linear order : @xmath80 as shown in @xcite and @xcite the @xmath81 can be brought to the form @xmath82 here a response function @xmath83 for `` intrinsic '' motion appears : the @xmath83 measures how , at some given shape @xmath74 and for some temperature @xmath84 , the nucleonic degrees of freedom react to the coupling @xmath85 . its time dependent version reads @xmath86 ) \equiv 2i\theta(t - s)\tilde{\chi}^{\prime\prime}(t - s)\ ] ] in this expression the time development of the field operators is defined by the same hamiltonian @xmath87 which appears in the density @xmath88 . the function @xmath89 on the very right stands for the so called dissipative part . in fourier space the full response separates into real and imaginary parts like @xmath90 , the @xmath91 sometimes being called reactive part . as one may guess from the very construction , the derivation of ( [ rfcoll ] ) relies on quasi - static properties of the nucleonic degrees of freedom . for instance , such properties appear in the coupling constant @xmath92 which is to be determined from @xmath93 moreover , the nucleonic degrees of freedom are assumed to behave ergodic in the sense of having the adiabatic susceptibility @xmath94 be identical to the isolated one , the static response @xmath95 : @xmath96 as has been demonstrated in @xcite this condition is not fulfilled in the deformed shell model , which implies that special measures are to be taken to which we will come to below . it should be noted that the derivation of ( [ rfcoll ] ) involves a self - consistency relation between the deformation @xmath67 of the mean field and the one of the density . the latter may be measured by the expectation value @xmath97 . for linearized dynamics this self - consistency condition reduces to the equation @xmath98 well known from the case of undamped vibrations @xcite , @xcite . realize , please , that the quantity @xmath99 is to be calculated with the actual dynamical nuclear states accounting for their appropriate occupations . before concluding this subsection we like to write down a more convenient form for the stiffness @xmath100 appearing in ( [ appa ] ) . it is defined as the second derivative of the internal energy @xmath101 with respect to deformation at fixed entropy @xmath102 . since it is not easy to calculate such a derivative it is better to reexpress it by the one of the free energy @xmath103 at fixed temperature ( see eqs.(a.18)-(a.19 ) of @xcite ) @xmath104 in @xcite it was found that for temperatures larger than @xmath105 the change of @xmath7 with the collective coordinate @xmath67 is small such that the second term on the right hand side of eq.([stiff0 ] ) can be neglected . below we would like to use the shell correction method when calculating static energies . in this method usually the intrinsic energy is involved , rather than the free energy . thus one needs to relate the derivatives of the free and the intrinsic energy both taken at at fixed temperature . this can be done by differentiating the relation @xmath106 with respect to deformation and obeying that the entropy can be expressed as @xmath107 . as the result one gets @xmath108 like before , the expression on the very right is justified for larger temperatures , where the change of @xmath109 with @xmath7 is small . this approximation is used in the computations presented below . in fig.6 the stiffness ( [ stiffe ] ) of the energy ( [ enedef])-([tempshco ] ) is shown as function of the deformation parameter @xmath29 . it is seen that its liquid drop part becomes negative for @xmath110 and the total stiffness exhibits rather strong fluctuations due to shell structure . in general the frequency dependence of @xmath77 exhibits a complex structure . its dissipative part @xmath111 represents the strength distribution over all possible modes of the whole nucleus which can be excited by an external force like the one introduced above , namely @xmath78 . rewriting ( [ eom ] ) as @xmath112 it follows that the inverse of the collective response function can be interpreted as an integral kernel for the effective equation of motion for the time dependent quantity @xmath113 . evidently , this time dependent form of the equation of motion must be expected to contain non - markovian effects . however , there may be situations for which it becomes possible to reduce this complicated structure to differential form . the clue to this simplification can be found by recalling the case of the damped oscillator for which ( [ nonmarkeom ] ) takes on the form @xmath114 here the following effective `` forces '' appear : an inertial one , a friction force and the conservative one which is related to the derivative of the effective collective potential , calculated in linearized form . the associated `` transport coefficients '' have been marked by an upper index `` f '' , to indicate that they are associated to the quantity @xmath115 . now the strategy of how to handle the general case is clear : whenever there is a pronounced peak in the strength distribution @xmath111 we may approximate it by a lorentzian . since the latter is defined by three quantities , like the width , the position of the maximum and its height , one may deduce the three transport coefficients which appear in the oscillator response , for instance as given by ( [ eomosc ] ) . in principle , such a procedure may be applied to any one of the peaks of the original strength distribution , which allows one to deduce these transport coefficients for all the possible collective modes , the low frequency ones as well as the high frequency ones . as a matter of fact , such a scheme is commonly adopted for the description of collective vibrations , as discussed in @xcite , @xcite . the difference to the present application is seen in the fact that we want to apply this procedure to describe global dynamics in locally harmonic fashion . the latter aspect puts an additional constraint . as mentioned earlier , our application of linear response theory goes along with the assumption that collective motion is slow , such that the nucleonic degrees of freedom follow closely a thermal equilibrium . we may recall that in some formulas given above , this equilibrium has been parameterized by the quasi - static density operator @xmath116 . estimates of the time scale on which such a relaxation can be expected to occur will be given below . but already on this level of information , it is clear that the whole concept would probably not work if high frequency collective modes would be important . for this reason we are bound to concentrate on low frequency ones , if the construction of the transport coefficients is to be consistent with the basic assumptions for the applicability of the quasi - static picture . in practice the `` fit '' of the strength distributions of the oscillator model to the `` correct '' one involves the full response functions , not only their dissipative parts . in short such an adjustment may be characterized by @xmath117 . a more correct form can be written as a variational procedure @xmath118 with the variation to be performed with respect to the coefficients @xmath119 and @xmath120 . in this way both real and imaginary parts of @xmath77 are fitted simultaneously . fortunately , for the practical applications to be discussed below it turns out that the coefficients @xmath121 and @xmath120 are rather insensitive to the upper integration limit @xmath122 . the value of the latter was fixed to be @xmath123 . the reader will have noticed the appearance of the factor @xmath124 in ( [ lorfit ] ) , as well as the fact that we have left out the upper index @xmath125 in the transport coefficients . this change is easily understood by referring to the self - consistency condition ( [ selfcond ] ) . the latter can be interpreted as a transformation from the @xmath125-mode to the @xmath67-mode . it implies a corresponding transformation both for the response functions as well as for the transport coefficients ( for details see @xcite ) . in this sense the @xmath126 and the transport coefficients @xmath127 stand for the collective response and the coefficients of the @xmath67-mode , respectively , with @xmath128 . it may turn out that collective motion is so slow , as compared to the dynamics of the nucleons , that the transport coefficients can be deduced by expanding the response functions around @xmath129 . at first such a procedure has been applied to the intrinsic response ( cf.@xcite ) . in this way one may get an approximate solution of the secular equation for the position of the poles of the collective response ( [ rfcoll ] ) . one may write @xmath130 this form invites to define coefficients for friction @xmath131 and inertia @xmath132 by @xmath133 and @xmath134 which in the past have been called the coefficients in `` zero frequency '' limit . the effective stiffness @xmath100 is seen to be identical to the local stiffness of the quasi - static energy ( mind ( [ appa ] ) ) , as one would expect to hold true for slow motion , indeed . incidentally , the expression for the inertia can be shown to be a generalization of the one of the cranking model to the case of damped motion @xcite . unfortunately , for strong damping this @xmath132 becomes very small , and sometimes even negative . for practical applications this does not always lead to problems , as usually damping may be so strong that inertia must drop out of the macroscopic equations of motion . nevertheless , this behaviour of the inertia @xmath132 is very unpleasant , but fortunately , one can do better . rather than to concentrate just on the denominator of ( [ rfcoll ] ) , it is better to take into account the full information contained in this expression . as we may recall this was done implicitly when constructing the transport coefficients @xmath135 and @xmath136 by approximating ( [ nonmarkeom ] ) by ( [ eomosc ] ) . an equation like ( [ eomosc ] ) may be obtained by expanding @xmath137 in ( [ nonmarkeom ] ) to second order in @xmath138 . in this way one gets @xmath139 @xmath140 and @xmath141 as compared to the zero frequency limit defined above , there are two modifications : whereas all three coefficients obtain a factor of proportionality , only the inertia gets an additional contribution . in most practical applications this proportionality factor is not very important , as usually the static response is much larger than the static stiffness : @xmath142 . it is only for small excitations and for deformed oscillator shell models , in particular , that the size of @xmath143 becomes comparable to that of @xmath144 . however , the additional term in ( [ colzermass ] ) ensures that the modified inertia does not drop indefinitely anymore with increasing damping . later on we will demonstrate with the help of numerical results that ( [ colzerfric ] ) and ( [ colzermass ] ) approximate the selfconsistent friction @xmath145 and inertia @xmath120 very well at temperatures @xmath146 . to distinguish from the zero frequency limit ( [ zerofrilim ] ) and ( [ zeromass ] ) we will associate the approximation ( [ colzerfric ] ) and ( [ colzermass ] ) to `` the zero frequency limit for the collective response function '' . finally , we should like to mention that relations similar to the ones given in ( [ colzerstif])-([colzermass ] ) were obtained earlier in @xcite , namely for the model case that the collective response function just consists of one ( approximately lorentzian ) peak . solving the equations for @xmath147 and @xmath132 given in @xcite with respect to @xmath148 and @xmath2 one gets back to the equations ( [ colzerstif])-([colzermass ] ) . in the previous section we found microscopic expressions of transport coefficients for large scale motion . for many reasons it is desirable to compare them with those of `` macroscopic models '' @xcite , like there are the liquid drop model for the inertia and the wall formula for dissipation not to mention stiffness , which we have seen to become identical to the one of the static energy anyway , as soon as collective motion becomes sufficiently slow . in @xcite many points of principle nature have been clarified about how the macroscopic limit can be obtained in microscopic theories , concentrating largely on vibrations around stable configurations . here we like to look at another , more practical , albeit very important issue . by its very nature , these macroscopic models assume the nuclear density to be constant inside some surface at which the density drops from the nuclear matter value down to zero within zero range . commonly this surface is parameterized with the same set of shape variables which in our description define the deformation of the mean field , for which in the present discussion we have chosen the one collective coordinate @xmath67 . contrary to the macroscopic picture , the microscopic calculation leads to a density distribution with a soft surface , which in addition depends essentially on the occupation of states . take the simple case of a spherical potential : if a shell with fixed angular momentum @xmath149 is not filled completely the corresponding density distribution will not be spherically symmetric . on the other hand , it is clear that the transport coefficients @xmath150 will depend sensibly on the distribution of matter . thus one needs to employ some specific transformation to relate them from one case to the other . this is best done looking at the mean value ( or moment ) of the operator @xmath151 , as a representative for the average density . in the case of our microscopic picture , we have seen the self - consistency condition ( [ selfcond ] ) to imply the relation @xmath128 between the transport coefficients of the @xmath125 and @xmath67-mode . all we need to do is to search for a similar condition which translates the @xmath125-motion into that of the sharp density distribution , with the latter being expressed through the corresponding shape parameter . such a relation can be found by applying the hypothesis , which we will substantiate below , that the average @xmath152 , calculated with a density distribution having a sharp surface , can be approximated well by applying an appropriate strutinsky smoothing to the shell model density . the latter is defined as @xmath153 where the @xmath154 are smoothed occupation numbers @xcite , @xcite . how they may be used not only to calculate static expectation values like in ( [ strmom ] ) , but of corresponding response functions as well has been studied in @xcite and @xcite . the desired relation between @xmath155 and @xmath67 may now be obtained simply by applying to ( [ strmom ] ) the derivative with respect to @xmath67 . one finds @xmath156 which may be used to deduce @xmath157 @xmath158 . here , @xmath159 represents the transport coefficients for the sharp density distribution , but calculated for the associated @xmath67-mode . combining this relation with the previous one , we get @xmath160 as the only one further approximation we have assumed that both density distributions lead to the same averaged value of the field operator @xmath151 . in ( [ qtransp ] ) there appear on the very left and on the very right the transport coefficients for the @xmath67-mode , once for the sharp surface of the macroscopic models and once for the collective coordinate specifying the mean field . we may add here that the quantity @xmath161 has a physical meaning similar to that of a static susceptibility , hence the choice of this symbol . the only difference to the isothermal susceptibility is found in using the smoothed occupation numbers @xmath162 instead of @xmath163 . let us turn now to `` prove '' the hypothesis made . this can be done explicitly for the simple case of the deformed oscillator potential . for the more general case we want to appeal to physical intuition , to the extent of accepting the idea that by its very construction strutinsky smoothing commonly does lead to the macroscopic picture . suppose we are given a spheroid whose deformation @xmath67 is fixed by the ratio between the semi - axes in @xmath10-direction and the one perpendicular to it : @xmath164 the sharp surface @xmath102 is then given by @xmath165 the deformation of the sharp density may be classified by the average of the quadrupole operator @xmath166 calculated as @xmath167 here , @xmath168 measures the density of homo- geneous nuclear matter representing @xmath63 nucleons distributed uniformly within this surface . in fig.7 the quadrupole moment ( [ quamom ] ) is plotted by the line with stars as function of the deformation parameter @xmath169 . conversely , the surface ( [ spheroids ] ) may be interpreted as an equipotential surface of the corresponding deformed oscillator potential would be related to our field operator @xmath170 taken at spherical shape by @xmath171 . ] . for such a potential one may compute single - particle wave functions @xmath172 and from them the moment of the microscopic density as @xmath173 obviously this moment @xmath174 depends on the occupation numbers @xmath175 . in fig.7 we show curves corresponding to three choices of @xmath175 : dotted line : the occupation numbers are fixed at the spherical shape , where the lowest energy states are filled , and kept constant , independent of @xmath67 ( `` diabatic '' situation ) ; dashed curve : at each crossing of states the particles and holes are redistributed in such a way that always the lowest states are occupied ( `` adiabatic '' situation ) ; solid curve : here , the smoothed occupation numbers of the shell correction method are used . we see that for the diabatic occupation numbers the quadrupole moment ( [ qudens ] ) differs substantially from that of the sharp density distribution @xmath176 ( line with stars ) , approximately by a factor of two . recall that this sharp density distribution reflects the deformation of the potential . on the contrary , the density computed with the adiabatic occupation numbers follows the deformation of the potential _ on average_. it can be said that this feature is one of the basic elements of the copenhagen picture of collective motion @xcite ( see also @xcite ) . indeed , such a redistribution of particles leads to a consistency between the shapes of the potential and the density . for static situations , this leads to the well known relation between the occupation numbers in the various directions and the corresponding frequencies of the potential . in the treatment of @xcite this relation is fulfilled for specific deformations . in fig.7 , the latter correspond to the points where the broken line crosses that with the stars . for this broken line it is hardly possible to define precisely the derivative of @xmath174 with respect to the deformation parameter . on the other hand , the condition just mentioned can be fulfilled everywhere using the strutinsky smoothed occupation numbers . in this way the derivative is well defined . furthermore , the quadrupole moment of the density computed with the smoothed occupation numbers practically coincides with that of the sharp surface distribution . in this section we are going to specify further details of our treatment of nucleonic dynamics . it has already been mentioned that the mere shell model is not sufficient . at finite excitations the effects of collisions can not be neglected ; one even expects them to become the more important the higher the nucleonic temperature will be . to treat collisions on the basis of an explicit form of a two body interaction @xmath177 is hardly possible . therefore we follow another path and parameterize the effect it would have on the single particle energies . details of this method can be found in the publications mentioned before , in particular in @xcite ( see also @xcite , @xcite , @xcite ) . the fourier transform of the intrinsic response function given in ( [ tdepresp ] ) can be expressed as the sum over single - particle states @xmath178 with @xmath179 here @xmath180 is the fermi function determining the occupation of single - particle levels . the @xmath181 represents the distribution of single - particle strength over more complicated states . it may be parameterized by @xmath182 in terms of the real and imaginary part of the self - energy @xmath183 which are assumed to have the following forms : @xmath184}\ ] ] and @xmath185/c^2}\ ] ] both are connected to each other by a kramers - kronig relation . the @xmath186 in ( [ imselfenomt])-([reselfenomt ] ) is the chemical potential and the cut - off parameter @xmath187 accounts for the fact that the imaginary part of the self - energy does not increase indefinitely when the excitations get away from the fermi energy . in the present calculation we choose @xmath188 and @xmath189 . the @xmath190 appearing in ( [ respfun ] ) is the one - body green function @xmath191 which is related to the spectral density @xmath192 by @xmath193 details about the evaluation of the integral in ( [ respfun ] ) are given in the appendix . for future purpose we want to use this form ( [ respfun ] ) of the response function and write down a more detailed expression for the friction coefficient in zero frequency limit : @xmath194 it is obtained ( see @xcite ) by substituting eqs.([respfunsum ] ) - ( [ respfun ] ) into ( [ zerofrilim ] ) . in section 3.1 it was mentioned that the condition of ergodicity ( [ ergodic ] ) is hard to fulfill in the deformed shell model . this statement refers to the study presented in @xcite , where it was shown that even collisional damping does not help , at least in the version as used to date . one important reason was seen to lie in the fact that our renormalized single particle energies have the same degeneracies as those of the pure shell model . but these degeneracies are by far bigger and happen much more often than one would expect for configurations of the compound nucleus . if one believes the latter to be important which should be the case for a fissioning system which has to overcome a large barrier one needs to take special measures to cure the problem just mentioned . one way to make the deficiencies apparent is to look at the heat pole which shows up either in the correlation function @xmath195 , or in the relaxation function @xmath196 . both are related to the dissipative response as @xmath197 ( see @xcite or @xcite for more details as well as for references to the original literature ) . both functions have a peak at @xmath129 , whose width @xmath198 was seen to be twice the single particle width ( [ imselfenomt ] ) calculated at the chemical potential , i.e. @xmath199 . on very general grounds , the height of this peak can be seen to be proportional to the difference of isothermal and isolated susceptibilities , @xmath200 . numerical calculations in @xcite showed this height to be large . rewriting this difference as @xmath201 and recalling that in the nuclear case the difference between the isothermal and adiabatic susceptibility , @xmath202 , is small @xcite , @xcite , the origin is identified to come from a large violation of ergodicity : @xmath203 . this discussion indicates what we can do to cure this problem : cut the contributions of the heat pole to all functions mentioned previously down to the magnitude it would have in case the system were ergodic . this means to reduce the height of this peak at @xmath129 by the factor @xmath204 . in @xcite a system was studied were @xmath205 vanishes identically , such that the reduction of the heat pole amounted to neglect contributions from all states having the same energy . such a correction can easily be done . in expressions like ( [ respfunsum ] ) one simply has to restrict the summations in a proper way . here we like to adopt a similar procedure , in the sense of neglecting the influence of a finite difference @xmath205 . at the temperatures considered it will be very small , indeed . on the other hand we go one step further and neglect also contributions from neighboring states whose energy is finite but smaller than the collisional width the particles , which as we just saw also reflects the width of the heat pole . the consequences of such a manipulation are shown in fig.8 for the intrinsic response ( left part ) , the corresponding relaxation function @xmath206 ( upper right part ) and the collective strength distribution @xmath207 ( lower right part ) . all of them have been computed for @xmath208 . the fully drawn lines correspond to calculations where all matrix elements are taken into account . for the dashed curves matrix elements @xmath209 between states of energy difference @xmath210 have been discarded . their contribution to the relaxation function is exhibited in the upper right part by the dotted line . this plot demonstrates nicely that the heat pole can be associated to a lorentzian of width @xmath198 around @xmath129 , and thus corresponds to a pole on the imaginary axes @xcite , @xcite . it is interesting to note that for the present calculation @xmath211 of this lorentzian is made out only by two _ pseudo - crossings _ of single - particle states which takes place close to the fermi energy . by pseudo - crossing we mean a situation where as function of @xmath67 two levels come close but never cross ; one such event is encircled in fig.3 . the big contribution to friction in the zero frequency limit of ( [ zerofrilim ] ) ( or ( [ colzerfric ] ) ) which results from this heat pole can be estimated looking at the upper right part of fig.8 : the slope of the fully drawn line is much bigger than the one of the dashed line . on the other hand , contributions to @xmath131 from _ real crossings _ of levels is very small . this is due to the fact that the matrix elements @xmath212 vanish exactly at the crossing points and are very small in the vicinity of such a ( real ) crossing . the influence of the heat pole ( in the intrinsic system ) on the collective strength distribution can be inferred from the lower right part of fig.8 . the fully drawn line shows a small peak at very small frequencies , say @xmath213 . compared to the much bigger strength found in the peak at about @xmath214 one is inclined to just `` forget '' the small peak when one wants to define the transport coefficients . indeed , it somehow looks very natural to associate the bigger peak to the genuine low frequency mode . this may be understood as another argument to leave out the contribution of the heat pole to the transport coefficients , besides the one involving ergodicity for intrinsic motion . all computations to be reported below were done along this line , i.e. contributions to response function from the states with the energy difference @xmath210 were not taken into account . the whole formulation of our theory is based on the assumption that the nucleonic degrees of freedom stay close to a thermal equilibrium . the latter is not fixed , however ; rather it continuously gets disturbed by collective motion itself , forgetting for the moment a possible evaporation of light particles or gammas . it should thus be of interest to have some estimate of an appropriate relaxation time . the best candidate for this is offered by the `` generator '' of collective motion , namely the @xmath215 , which defines the coupling of the collective variable to the nucleons . it is predominately this quantity which `` decides '' which kind of modes of the intrinsic degrees of freedom get excited . we may recall from the discussion in sect.3 the close relation of this @xmath151 to the nucleonic response function appearing in our theory : the @xmath83 parameterizes that average `` excitation '' @xmath216 which comes about through a change of the collective variable @xmath217 . if we are just interested to estimate the time @xmath218 after which the @xmath219 has decayed to its static value , we may study the time dependent function @xmath220 given in ( [ tdepresp ] ) . in literal sense it represents the @xmath221 if excited by a sharp pulse like @xmath222 . notice please , that for @xmath223 the @xmath220 is proportional to the derivative of the fourier transformed relaxation function , namely @xmath224 ( see ( [ tdepresp ] ) and ( [ correlfun ] ) ) . thus the information contained in @xmath220 is equaivalent to that of @xmath225 up to an additive constant . the latter measures the long time limit of @xmath225 ; it is related to the strength of the heat pole . at the moment we are interested only in the behaviour for finite times . in fig.9 . results of two different computations of @xmath220 are shown . one case just refers to the deformed shell model , for the other one collisional damping is taken into account . it is clearly seen that only by way of such collisions we may speak of genuine relaxation . it is also observed that the latter does not depend much on the shape . however , the relaxation time @xmath218 decreases considerably with increasing temperature . the latter effect is expected of course from the very form by which the temperature appears in the single particle widths ( [ imselfenomt ] ) . from the fully drawn lines of fig.9 one may deduce for @xmath218 values like @xmath226 for @xmath227 . ( this may for instance be done by approximating the envelope of @xmath220 by an expontial such that @xmath218 may be defined through the `` width '' at half maximum . ) later on in sect.5.3 we are going to compare them with typical time scales of collective motion , but we may say already here that this microscopic time @xmath218 is at least one order of magnitude smaller than the one which measures motion in the collective variable @xmath67 . finally , we should like to mention that the results found from the present computation are in accord with those reported in @xcite . in this chapter we will discuss the numerical results for transport coefficients @xmath228 computed along the fission path of @xmath53 . as it was already mentioned , for temperatures above @xmath64 the fission path is parameterized in terms of cassini ovaloids by only one deformation parameter @xmath13 . since we prefer to use the parameter @xmath29 instead of @xmath13 we have to relate the transport coefficients accordingly . this can be done by exploiting the following relation , obeying that the @xmath23 s are fixed , @xmath229 which simply follows from general properties of coordinate transformations . the derivative @xmath230 is obtained by differentiating ( [ rr12 ] ) , the result is shown by dashed line in fig.2 . recall please , that the deformation dependence of the transport coefficients is defined essentially by the choice of the collective variables . for example , @xmath231 decreases with @xmath29 but @xmath232 is increases as a function of @xmath13 . both in the figures as well as in the text below we will omit the indices @xmath233 , keeping in mind that the transport coefficients are defined with respect to @xmath29 ( even if sometimes they will be shown as function of @xmath13 ) . the friction coefficient @xmath234 and mass parameter @xmath120 defined according to ( [ lorfit ] ) are shown in fig.10 by solid lines as function of the deformation parameter @xmath235 for temperatures between @xmath236 and @xmath237 . they are compared with calculations for which the approximation ( [ colzerfric])-([colzermass ] ) is used for friction and inertia , respectively . the latter results are marked by dashed curves . as can be seen , this approximation is quite accurate for @xmath238 . this implies that for such temperatures one may avoid the time consuming computation of the frequency dependence of collective response function . one may compute friction and inertia directly from ( [ colzerfric ] ) and ( [ colzermass ] ) . as for the big fluctuations seen at @xmath239 we expect them to become much smaller as soon as pairing correlations will be taken into account , which shall be subject of future studies . for comparison we also show in fig.10 ( by the lines with stars ) wall friction @xmath240 and the inertia of irrotational flow . according to @xcite wall friction @xmath241 is proportional to the squared normal velocity @xmath242 of the surface , integrated over the nuclear surface . following @xcite this may be deduced from the loss of ( collective ) energy which is given by @xmath243 where @xmath11 and @xmath244 are the nucleons density and fermi velocity . for axially symmetric shapes the surface velocity @xmath245 can be expressed in terms of the profile function @xmath246 from eq . ( [ ovals ] ) as @xmath247 the mass parameter of an incompressible irrotational fluid has been computed as suggested in @xcite . it can be written as @xmath248 with the potential @xmath249 for the velocity field expressed by the potential of some `` surface charge '' distribution @xmath250 ) into the neumann equation @xmath251 leads to some integral equation for the density of the `` surface charge '' @xmath252 which was solved iteratively starting with @xmath253 as zeroth approximation to @xmath252 , for details see @xcite . we have checked that for the particular case of the shape family ( [ ovals ] ) the werner - wheeler method @xcite turns out to be a very accurate approximation to the mass parameter ( [ mirr ] ) . both results coincide within the thickness of the lines in the figure . in figs.11 and 12 the friction coefficient and the mass parameter are presented as function of the deformation parameter @xmath13 for five different temperatures @xmath254 . they have been calculated from the oscillator fit ( [ lorfit ] ) , but transformed to sharp densities according to ( [ qtransp ] ) . the parameter @xmath13 is chosen for reasons to be given below . in both figures results of the commonly adopted macroscopic models , namely @xmath255 and @xmath256 are shown by the lines with stars . several observations can be made . + @xmath257 to some extent the @xmath255 and @xmath256 can be said to be reached at the higher temperatures . the very fact that this statement is more true for the inertia but less so for friction can be understood as follows . as shown in @xcite , @xmath255 may be considered the macroscopic limit of our model only if such subtleties as collisional damping are left out . conversely , the high temperature limit of the inertia is related to the value of the energy weighted sum , and the latter is known to be associated to @xmath256 . this is true at least when one treats vibrations within simple models ( see @xcite , @xcite for the situations of @xmath258 and @xcite at @xmath259 ) . + @xmath257 as has been demonstrated in @xcite , in the general case the evaluation of the sum rule value becomes somewhat delicate for collisional damping . the nice feature of reaching the @xmath256 automatically can be considered a proof of the usefulness of the transformation ( [ qtransp ] ) . + @xmath257 like in the calculations of @xcite , on average friction is seen to increase with @xmath7 whereas the inertia decreases . + @xmath257 again similarly to @xcite , both transport coefficients decrease with deformation on average . + @xmath257 the coordinate dependence shows fluctuations around the average trend , which are more pronounced at smaller temperatures . they appear to be even larger than those reported in @xcite , but they are similar to the results of @xcite . as for the friction coefficient , the dependence on deformation is particularly strong around the spherical shape . it reaches a kind of local maximum there , or perhaps at a slightly oblate shape . these features can be exhibited more clearly for parameterization in terms of @xmath13 , for which reason we have made this choice here . many of these features point to the importance of shell effects , in particular the peak for friction around the spherical configuration , which is clearly visible for @xmath260 but which disappears at @xmath261 . we like to elaborate on this statement by trying to split the friction coefficient up into a smooth and a fluctuating part . for this study we take the zero frequency limit . let us suppose for a moment that the matrix elements @xmath262 considered as function of the single - particle energies have some smooth average component @xmath263 and the oscillating component can be neglected . in this case one may rewrite eq.([zerofric ] ) in the form @xmath264 the @xmath265 in ( [ zerofric ] ) are peaked functions with their maximum at @xmath266 so that the sum of @xmath192 over @xmath92 may be interpreted as the density @xmath267 of single - particle states . as usual the latter can be split into the smooth and oscillating components @xmath268 with @xmath269 and @xmath270 where the brackets stand for an averaging over the single - particle spectrum . inserting this decomposition into ( [ zerofrappr ] ) , and noting that an integration performed with the bell - like function @xmath271 can be understood like the average introduced above , we will get @xmath272 { \cal f}^2(\mu , \mu)\ ] ] assuming the oscillating component of the density to be periodic in the energy with some period @xmath273 and amplitude @xmath274 , viz @xmath275 , it is easy to convince oneselves that @xmath276 . the quantity @xmath274 is determined by the magnitude of the shell correction . it is a smooth function of particle number ( see @xcite ) but still depends on deformation . it is maximal at that deformation where the shell structure is more pronounced . for the woods - saxon potential this happens to be so at the spherical shape . so it is due to the shell structure that the friction coefficient gains additional contributions around the sphere . for the case of the woods - saxon potential this specific feature seems to be responsible for the dip one sees in fig.10 both for friction as well as for the inertia around @xmath277 , even at the somewhat larger temperatures of @xmath278 . it may be said , however , that such a behavior is not seen in computations performed with the model of @xcite . there the change of the friction coefficient with @xmath29 , for instance , resembles more the smooth one given by the wall formula . it has become customary to parameterize friction in terms of the ratio @xmath279 , although most of the interest has concentrated on friction alone ; often the inertia was simply taken to be the reduced mass of the fissioning system . as we have seen above , for a microscopic theory both quantities will vary not only with the collective variable but with temperature as well . it may be expected , of course , that for the ratio the dependence on shape is much weaker than that of the individual quantities . this follows simply from the observation that any common , purely geometrical factor will drop out . in fig.13 we present the results for @xmath280 obtained from those for inertia and friction discussed before . indeed , this quantity is essentially constant over the whole deformation region , for all computations but @xmath239 , a case for which the fluctuations seen in our results are too big because of our neglecting pairing . however , there is a marked dependence on excitation : @xmath280 increases strongly with @xmath7 . this is in clear distinction to the result one gets from applying the wall formula for friction and that of irrotational flow for the inertia . interestingly enough , these macroscopic estimates lead to some @xmath67-dependence , which in a sense is even bigger than suggested by the trend of our results . it can be said that the latter are very close to those obtained in the computations with a two - center shell model potential @xcite . physically , the inverse of @xmath280 can be interpreted as the relaxation time ( @xmath281 ) to the maxwell distribution for collective motion . in full glory this feature can only be understood looking at the dynamics in collective phase space ( see e.g.@xcite ) . however , one may grasp its content recalling the local equation for average motion of the damped oscillator : @xmath282 . from this equation it also becomes apparent that yet another relaxation time can be defined , namely @xmath283 . for overdamped motion the latter is the only relevant one . actually , such a situation is given for temperatures above @xmath284 . let us infer this feature comparing the macroscopic relaxation times with the microscopic @xmath218 found in sect.4.3 . there we got values in the range between @xmath285 and @xmath286 for temperatures between @xmath287 and @xmath288 , discarding the case of @xmath239 for the moment . taking for the @xmath280 of fig.13 typical values like @xmath287 and @xmath289 for @xmath290 and @xmath288 , respectively , we recognize the @xmath291 to be comparable or only slightly larger than @xmath218 . this means that in this range of temperatures the friction force is so strong that it leads to an instant damping or disappearance of the kinetic energy . this is not in contradiction to the basic assumption behind the quasi - static picture . to justify the latter all one needs to have is the motion in @xmath67 to be slow compared to that for the intrinsic degrees of freedom . for this question it is the @xmath292 which becomes relevant . its value can be estimated with the help of figs.6 and 10 . let us concentrate on deformations for which the local stiffness does not get zero . since for @xmath293 shell effects are not important anymore we may estimate @xmath0 as the one of the liquid drop energy . the right part of fig.6 . tells us the @xmath294 to be of the order of @xmath295 for @xmath29 larger than about @xmath296 . in this range the value of @xmath1 is about @xmath297 , as seen from fig.10 ( for @xmath298 ; it is even bigger for larger temperatures ) . for @xmath292 this implies values of the order of @xmath299 , which are larger than @xmath218 by more than one order of magnitude . in this paper we have applied the single particle model of @xcite to describe large scale motion at finite excitations . to this end this model had to be modified to include effects of collisional damping and it had to be adapted to the formulation of collective motion in the spirit of the locally harmonic approximation ( for a review see @xcite ) . numerical computations have been performed for the transport coefficients of average motion along a fission path . the latter was identified by the valley in the potential landscape obtained for the liquid drop model . it was argued that for the range of temperatures considered in the present study this path may be expected to represent fairly well the actual situation . first of all , because of the evidence one has from static considerations that cassini ovaloids describe well the shapes of the fissioning nucleus . secondly , for the large damping one expects to be given , the system will be creeping down the collective potential and thus will stay close to the line of steepest descent . for the transport coefficients values were found which are in accord with previous studies , in particular with the ones of @xcite . this is especially so for their dependence with temperature , and to lesser extent for their variation with the nuclear shape . for instance , it turns out that typical effects of single particle motion become more apparent here than they did in @xcite , like there are fluctuations of both friction and inertia with the collective variable . to large amount they disappear when building ratios like for the @xmath300 or the @xmath301 , two quantities which are commonly used to parameterize collective dynamics . a new development has been achieved with respect to comparisons with macroscopic models , like that of irrotational flow for the inertia and that of wall friction . as one knows , both are calculated for sharp density distributions which may differ considerably from those which correspond to the nucleons densities in the shell model . a transformation was suggested which allows one to connect both density distributions and thus enables one to connect the transport coefficients accordingly . this transformation was found by applying a strutinsky smoothing procedure to evaluate the average static value of the relevant one - body operator @xmath215 . it is this operator which by a self consistency argument is related to the deformation of the mean field . finally , we briefly like to turn to comparisons with experimental findings . first of all , we may mention that the values for @xmath280 shown above concur with the range suggested by fission experiments @xcite . two other relevant parameters are the @xmath302 and the @xmath303 mentioned above , which for instance appear in kramers famous formula for the decay rate of a one - dimensional fission model : @xmath304 . here , @xmath305 measures the barrier height , @xmath306 determines the vibrational frequency in the potential minimum and @xmath307 is to be evaluated at the saddle . the latter two quantities are shown in figs.14 and 15 , respectively , for both points . they have been calculated by averaging the @xmath308dependent transport coefficients in the neighborhood of the minimum and the barrier , where both are those of the @xmath7-dependent potential energy . these results agree very well with those already shown in @xcite . it is remarkable that @xmath309 does not change much with @xmath7 , neither at the minimum nor at the barrier . conversely , like the @xmath280 shown before , the @xmath303 too definitely increases with temperature . this behavior is in at least qualitative agreement with the one found in @xcite and the numbers for @xmath303 lie in the range of values deduced only recently in @xcite from experimental evidence . it is true , of course , that for more quantitative analyses one would have to consider both the effects of pairing as well as of angular momentum , not to mention the fact that there is some freedom @xcite in adjusting the two parameters which define collisional damping . however , there is probably little doubt about the temperature dependence of our coefficients , say above @xmath310 . this marked change with @xmath7 is in clear distinction to the macroscopic models mentioned . in the light of this feature agreements between experimental findings and theoretical descriptions appear somewhat questionable if they are only based on these macroscopic models @xcite , @xcite . likewise the somewhat peculiar behavior of @xmath280 with @xmath29 suggested in @xcite is not confirmed by our results ( see also @xcite ) . it is perhaps fair to say that some of the difficulties one still encounters at present when comparing theory with experiment is due to the high complexity of the problem itself as well that of the analysis of the data . to underline this statement , we just like to take up a point raised in @xcite , namely that common statistical codes evaluate the fission decay rate not by the transition state result , to which kramers formula reduces to for @xmath311 , but by that of the `` statistical model '' where @xmath306 is replaced by @xmath7 . the difference between these two variants can easily be inferred from fig.14 . * acknowledgments*. the authors want to thank the deutsche forschungsgemeinschat for financial support . two of us ( f.a.i and v.v.p . ) would like to thank the physics department of the tum for the hospitality extended to them during their stay in garching . the computation of the intrinsic response function is somewhat involved . due to the frequency dependence both of fermi distribution @xmath312 and width @xmath313 the integral in ( [ respfun ] ) can not be calculated analytically . the numerical integration is rather time consuming since @xmath181 are sharply peaked functions which width varies by two order of magnitude depending on the values of @xmath314 . fortunately the integration in ( [ respfun ] ) can be carried out by means of residues theorem closing the integration limit in the lower half plane . for this one needs to find the poles and residues of all the terms in the integrand of ( [ respfun ] ) . let us first look for the poles of @xmath315 . substituting ( [ dspspecdensgam]),([imselfenomt ] ) into ( [ green ] ) and introducing notation @xmath316 ( [ green ] ) can be brought to the form @xmath317\over 2\gamma_0 \delta /c^4 } + { i[(\hbar \omega -\mu^+)(\hbar \omega -\mu^-)-c^2]\over 2\gamma_0 /c^2 } \big\rbrace^{-1 } \label{greenp}\end{aligned}\ ] ] it is not difficult to note that the last terms in square brackets cancel with each other and both numerator and denominator can be divided by @xmath318 . in this way @xmath315 becomes @xmath319\ ] ] where @xmath320 are solutions of the equation @xmath321 namely @xmath322^{1/2 } \big\rbrace\ ] ] the both poles of @xmath315 lie in the lower half - plane . the residues of @xmath315 are simple functions of @xmath320 and @xmath323 as it is seen from ( [ greenpp ] ) . from the definition ( [ green ] ) it is easy to seen that poles and residues of @xmath324 are complex conjugated to that of @xmath315 . the pole representation for @xmath181 is easily obtained from ( [ greenpp ] ) . besides the poles of @xmath325 and @xmath181 one should account also for the poles of fermi function @xmath312 in the plane of complex @xmath138 ( so - called matzubara frequencies ) @xmath326 in principle , the sum extends over infinite many terms , but in practice the summation in ( [ respfun ] ) is cut at such frequencies @xmath327 which contribute less than determined by the desired accuracy . 99 v.m.strutinsky , n.ya.lashchenko and n.a.popov , nucl.phys . * 46 * , 639 ( 1963 ) v.s . stavinsky , n.s . rabotnov and a.a . seregin , yad.fiz . * 7 * , 1051 ( 1968 ) v.v . pashkevich , nucl . a169 * , 275 ( 1971 ) s. yamaji , f.a . ivanyuk and h. hofmann , nucl.phys . * a * , in press a. bohr and b. mottelson . nuclear structure , * vol . * * 2 * , ( benjamin , london , 1975 ) r. samhammer , doctoral thesis , technical university munich ( 1988 ) and r. samhammer , h. hofmann and s. yamaji , nucl . a503 , ( 1989 ) 404 t. wada , n. carjan and a. abe , nucl.phys . * a538 * , 283c ( 1992 ) ; phys.rev 70 * , 3538 ( 1993 ) k. pomorski , j. bartel , j. richert and k. dietrich , nucl . phys . * a605 * , 87 ( 1996 ) i.i . gontchar and p. frbrich , physics of atomic nuclei . * 57 * , 1181 ( 1994 ) m. thoennessen , nucl . phys . * a 599 * , 1 ( 1996 ) j. damgaard , h.c . pauli , v.v . pashkevich and v.m . strutinsky , nucl.phys . * a135 * , 432 ( 1969 ) v.m . strutinsky , nucl.phys . * a95 * , 420 ( 1967 ) ; * a122 * , 1 ( 1968 ) m. brack , j. damgaard , h.c . pauli , a.s . jensen , v.m . strutinsky and c.y . wong , rev.mod.phys . * 44 * , 420 ( 1972 ) c. guet , e. strumberger and m. brack . phys.lett . * 205 b * , 427 ( 1988 ) p.j . siemens and a.s . jensen , `` elements of nuclei : many - body physics with the strong interaction '' , addison and wesley , 1987 h. hofmann , phys . reports , in press d. kiderlen , h. hofmann and f.a . ivanyuk , nucl . phys . * a550 * , 473 ( 1992 ) h.hofmann , f.a.ivanyuk and s.yamaji , nucl . phys . * a598 * , 187 ( 1996 ) h. hofmann , phys.lett . * 61 b * , 423 ( 1976 ) h. hofmann , r. samhammer and s. yamaji , phys . lett . * 229 b * , 309 ( 1989 ) w.j . swiatecki , nucl.phys . * a574 * , 233c ( 1994 ) h. hofmann and f.a . ivanyuk , z.phys . a , hadrons and nuclei . * 344 * , 285 ( 1993 ) r. alkofer , h. hofmann and p.j . siemens , nucl . a476 * , 213 ( 1988 ) s. yamaji , h. hofmann and r. samhammer , nucl . phys . * a475 * , 487 ( 1988 ) a. s. jensen , j. leffers , h. hofmann and p.j . siemens , physica scripta . * t5 * , 186 ( 1983 ) j. blocki , y. boneh , j.r . nix , j. randrup , m. robel , a.j . sierk , w.j . 113 * , 330 ( 1978 ) f.a . ivanyuk , v.m . kolomietz and a.g . magner , phys.rev . * c52 * , 678 ( 1995 ) k.t.r . davis , a.j . sierk and j.r . nix , phys.rev . * c13 * , 2385 ( 1976 ) f.a . ivanyuk and k. pomorski , phys.rev . * c53 * , 1861 ( 1996 ) h. hofmann , s. yamaji and a.s . jensen , phys . * 286 b * , 1 ( 1992 ) f.a.ivanyuk and v.m.strutinsky , z. phys . * a293 * , 337 ( 1979 ) d. hilscher , i.i . gontchar and h. rossner , physics of atomic nuclei . * 57 * , 1187 ( 1994 ) d.j . hofman , b.b . back , i. diszegi , c.p . montoya , s. schadmand , r.varma , and p. paul , phys . rev.let . * 72 * , 470 ( 1994 ) ; see also : p. paul and m. thoennessen , ann.rev.part.nucl.sci . * 44 * , 65 ( 1994 ) j.p . lestone , tours symposium on nuclear physics ii , pg 458 - 469 , world scientific publishing co. ( 1995 ) isbn 981 - 02 - 2156 and preprint 1996 ; see also j.p . lestone , phys . ( 1993 ) 2245

we present optical longslit spectroscopic observations of 21 low - luminosity , extreme late - type spiral galaxies . our sample is comprised of sc - sm local supercluster spirals with moderate - to - low optical surface brightnesses and with luminosities at the low end for spiral disk galaxies ( ) . for each galaxy we have measured high spatial resolution position - velocity ( p - v ) curves using the h emission line , and for 15 of the galaxies we also derive major axis rotation curves . in% of our sample , the p - v curves show significant asymmetries in shape , extent , and/or amplitude on the approaching and receding sides of the disk . a number of the p - v curves are still rising to the last measured point , or reach a clear turnover on only one side . in most instances we find good agreement between the kinematic centers of extreme late - type spirals as defined by the global emission profile and by their optical continuum , although in a few cases we see evidence of possible real offsets . in spite of their shallow central gravitational potentials , at least 6 of the galaxies in our sample possess semi - stellar nuclei that appear to be compact nuclear star clusters ; in 5 of these cases we see kinematic signatures in the p - v curves at the location of the nucleus . finally , we find that like giant spirals , our sample galaxies have higher specific angular momenta than predicted by current cold dark matter models . 0.5 cm

astro-ph0203188

matthews & gallagher ( 1997 ; hereafter mg97 ) used the term `` extreme late - type spirals '' to refer to the lowest - luminosity rotationally - supported disk galaxies that still exhibit regular disk structure and well - defined optical centers . typically these are moderate - to - low surface brightness sc - sm spirals that represent the low - end extremes of properties such as mass , luminosity , and size for a given hubble type . extreme late - type spirals are among the dynamically simplest of disk galaxies : they tend to be pure disk systems , often with weak or absent spiral arm structure . some extreme late - type spirals also appear to have undergone only minimal dynamical heating ( e.g. , matthews 2000 ) . however , in spite of their apparent simplicity , there is evidence of a variety of intriguing structural and dynamical phenomena in these small galaxies , including : multi - component disk structure ( matthews 2000 ) ; kinematically lopsided disks ( e.g. , matthews , van driel , & gallagher 1998 ; swaters et al . 1999 ; matthews & uson 2002 ) ; and compact starcluster nuclei ( e.g. , matthews et al . 1999 ; bker et al . 2001,2002 ) . high - quality kinematic information for extreme late - type spirals is clearly desirable in order to further explore these trends . recently , the acquisition of kinematic information , particularly disk rotation curves , for low - luminosity and low surface brightness ( lsb ) spirals has also become of considerable interest for exploring broader issues relevant to our overall understanding of galaxy formation and the nature of galaxian dark matter . from numerous studies , it has now been found that the dynamics of these types of galaxies often appear to be dominated by dark matter even at very small galactocentric radii ( e.g. , jobin & carignan 1990 ; ct , carignan , & sancisi 1991 ; martimbeau , carignan , & roy 1994 ; meurer et al . 1996 ; mcgaugh & de blok 1998 ; de blok et al . 2001 ) . because the contribution to the overall dynamics from the visible matter in these faint disk systems is consequently quite small , this means that extreme late - type / lsb spirals can provide us with some of the most stringent constraints on dark matter halo properties of galaxies , and may hold a key to testing important classes of galaxy formation paradigms and cosmological models ( e.g. , kravtsov et al . 1998 ; van den bosch et al . 2000 ; iliev & shapiro 2001 ; van den bosch , burkert , & swaters 2001 ) . until recently , most of the rotation curves that have been obtained for extreme late - type spirals , especially lsb systems , were derived from aperture synthesis observations ( e.g. , carignan , sancisi , & van albada 1988 ; bosma , van der hulst , & athanassoula 1988 ; van der hulst et al . 1993 ; de blok , mcgaugh , & van der hulst 1996 ; swaters 1999 ) . unfortunately , many of the rotation curves available for extreme late - type spirals suffer from poor angular resolution , leading in some cases to significant beam smearing , whereby the slope of the inner , rising portion of the rotation curve is underestimated ( see swaters 1999 ; swaters , madore , & trewhella 2000 ; van den bosch et al . this in turn hampers attempts to perform accurate rotation curve mass decompositions and to meaningfully constrain the shape of the dark matter halo ( e.g. , blais - ouellette et al . 1999 ; van den bosch et al . beam smearing can be particularly problematic for the extreme late - type spirals , since these systems are physically small , and at distances of 10 - 20 mpc , have angular sizes of only a few arcminutes or less . for this reason , _ optical _ kinematic data ( i.e. , longslit emission line spectroscopy or fabry - perot interferometry ) , which have higher spatial and spectral resolution , offer a more accurate means to map the inner rotation curve of extreme late - type spirals . these observations can thus serve as an extremely valuable complement to observations in mapping the full disk rotation curves ( e.g. , de blok et al . 2001 ) . the combination of optical and data is also powerful for constraining disk angular momenta ( e.g. , van den bosch et al . 2001 ; see also section [ angmom ] ) and for comparing the kinematics of the ionized versus the neutral gas . finally , high - resolution optical kinematic data provide unique information in their own right , such as kinematically confirming the presence of compact star cluster nuclei in small spirals ( see section [ nukes ] ) . mg97 presented optical @xmath3 and @xmath4 ccd photometry for a sample of 49 extreme late - type spiral galaxies with @xmath5 . their sample consisted of sc - sm field spirals , all within the local supercluster ( @xmath63000 . ) . although these objects are of relatively low luminosity and span a variety of morphological types ( see figure 11 of mg97 ) , all have regular disk structures and well - defined optical centers , distinguishing them structurally from dwarf irregular ( dirr ) galaxies of similar luminosities . most , but not all of the galaxies are objects of low mean optical surface brightness ( @xmath7 mag arcsec@xmath8 ) . generally these galaxies are -rich , with high @xmath9 ratios for their hubble types ( mg97 ; matthews , van driel , & gallagher 1998 ) . finally , all of the galaxies in the mg97 sample exhibit significant rotational broadening in their global profiles , with all but few objects showing double - peaked spectra ( fouqu et al . 1990 ; matthews et al . 1998 ) . in this paper we use optical longslit emission line ( h@xmath1 ) spectroscopy to explore the rotational properties of 21 of the mg97 galaxies . in section [ sample ] we describe our sample selection . in section [ obs ] & [ reduction ] we describe our observations and data reduction , respectively . we present position - velocity ( p - v ) diagrams derived from the new h@xmath1 data in section [ results ] , and compare them to the global profiles of the galaxies . for the 15 galaxies with highest - quality optical data , we also derive major axis rotation curves . we discuss characteristics of our p - v plots and rotation curves further in section [ results ] , including the presence of kinematic asymmetries ( section [ asymm ] ) . in section [ nukes ] , we investigate the dynamical signatures of the semi - stellar nuclei on the rotation curves of several of the galaxies and use these to derive crude mass estimates for the nuclei , while in section [ angmom ] we briefly comment on the specific angular momenta of the disks in our sample . finally , in the appendix we provide comments on our individual longslit spectra . our optical disk rotation profiles are measured using the h@xmath1 emission line , which is in general only easily observable to within @xmath10 in most normal , bright galaxies , and to only a fraction of @xmath11 in faint , extreme late - type systems ( e.g. , table 2 ) . therefore , ultimately the observations we present should be combined with extended rotation curves to gauge the full disk rotation of our sample galaxies . currently rotation curve data are not available for the bulk of the galaxies in our sample , hence it is intended that the results presented here will provide an aid for selecting objects suitable for future mapping . for this reason , we are also making available in electronic format all of our measured p - v data ( table 3 ) . for our spectroscopic measurements we selected from mg97 twenty - one extreme late - type spiral galaxies covering a range of optical morphologies . where possible , preference was given to galaxies with moderate - to - high inclinations . all of our targets have radial velocities @xmath122800 and therefore reasonable spatial resolution for these systems can be achieved from the ground under average seeing conditions . some global properties of our targets are given in tables 1 & 2 . sample broadband images of 4 of the galaxies are shown in figure [ fig : elts ] . our longslit spectroscopic observations were obtained during two nights in 1995 january using the 4-m telescope and ritchey - chrtien ( r - c ) spectrograph at the cerro tololo inter - american observatory ( ctio ) . all spectra were taken under clear conditions . seeing throughout the run was @xmath2@xmath13-@xmath14 . our r - c spectrograph set - up included the folded schmidt camera at the _ f_/7.8 focus . the detector was a tek 1024@xmath151024 ccd with arcon controller . with the arcon controller , the ccd is divided into four quadrants , each with its own amplifier . this results in a slightly different gain and readnoise for each of the four sections of the chip . the mean gain was 1.25 @xmath16/adu and the mean readnoise was 4.3 @xmath16 rms . at the _ f_/7.8 focus , the folded schmidt camera yielded a plate scale of @xmath13 per 24@xmath17 m pixel on the tek ccd . our slit length and width were 328@xmath18 and @xmath19 , respectively . we used filter gg495 with grating kpglg , which has 860 _ l_/mm , a blaze of 11000 at first order . this yielded a dispersion of 0.68 per pixel and spectral coverage @xmath20 of roughly 615 centered on the redshifted h@xmath1 @xmath216562.82 line . this @xmath20 interval also includes the [ ] @xmath226583.41 , and [ ] @xmath216718.26,6732.64 emission lines . using position angles from the literature or estimated from previously obtained ccd images , we generally strove to place the spectrograph slit along the major axis of each target ( or in two cases , along the galaxy bar ; see table 1 ) . unfortunately , in a few cases , less optimal slit position angles were achieved ( see table 1 and the appendix ) . hereafter we denote the offset in angle between the photometric major axis and the observed slit position angle as @xmath23 . all of our target galaxies have small enough optical angular sizes such that they were fully sampled with the 328@xmath18 slit length . our exposure times for each target ranged from 500 - 2000@xmath24 ( see table 1 ) . whenever possible , two independent exposures were obtained to aid in cosmic ray rejection . during each observation , a thar comparison lamp exposure was also acquired at the same telescope position as the target galaxy for use in correction for geometric distortions ( see below ) . our data were reduced using the standard spectral reduction and analysis packages available in _ iraf_. overscan corrections were applied to each frame using the quadproc routine . sequences of bias frames taken each afternoon showed the bias level of the tek ccd to be extremely unstable . count levels in the bias images ranged from 3 to 8 adu , but the underlying structure varied widely for frames obtained only minutes apart . in addition , a ripple pattern with an amplitude @xmath20.5 adu was visible in most bias frames , and this pattern was found to shift in location from one bias frame to another . for this reason , we were not able to reliably subtract the bias level from our images . since the count levels in the bias can be significant relative to some of the faint emission line signals from our targets , we were unable to obtain accurate absolute line fluxes or line ratios from the present data . this also introduces an additional source of error in determining the centroid of the galaxy in cases where the galaxy continuum is fairly weak ( see below ) . however , our absolute wavelength calibration should not be affected . the _ iraf _ background task was used to remove scattered light from our images , and the data were flatfielded using normalized domeflats . the task illum was then used to apply an illumination correction derived from twilight sky exposures . this procedure yielded images that were flat to a fraction of a per cent . a sample of one of our flatfielded spectra is shown in figure [ fig : image ] . cosmic rays were identified by blinking the multiple object exposures when available , otherwise by visual inspection . in cases where two object exposures were available , cosmic ray events in the vicinity of galaxian emission features were removed by replacing the affected pixels with the corresponding pixel values from the other image . when only one exposure was available , cosmic rays were removed by interpolating from the surrounding pixels ( in cases where they landed on background regions ) , or by simply masking the affected pixel(s ) ( in instances where they were superimposed on an emission line ) . to derive a correction for geometric distortions , we utilized our thar comparison spectra and followed a procedure similar to that outlined by massey , valdes , & barnes ( 1992 ) . we identified the lines in the thar lamp spectra , and then applied to our galaxy spectra the geometric transformation of @xmath22 as a function of the position on the ccd by fitting a sixth order , two - dimensional chebyshev function using the task fitcoords . based on the dispersion solutions found by fitcoords , we transformed all rows along our object spectra to identical linear wavelength scales using the task transform , which performed the necessary interpolations using a third order spline function . the rms residuals for the wavelength solutions were typically 0.05 pixel , corresponding to 0.034 ( @xmath21.6 at h@xmath1 ) . as a check , we also attempted the corrections for geometric distortions using lines identified from night sky spectra . we found the results of the two approaches to be indistinguishable . aside from the expected flexure shifts ( which affect absolute , but not relative wavelength calibrations ; see below ) , we found our dispersion solutions to be very stable over the course of the run . during our reductions we found the multiple standard star exposures obtained for the purpose of correcting the tilt of the spatial axis relative to the edge of the detector to be inadequate to map the trace along the full extent of the slit , hence we were unable to properly correct for this effect . the tilt is approximately 2 pixels across the full 1024-pixel width of the ccd . this does not affect our absolute wavelength calibration , but means that when a spectrum is extracted parallel to the edge of the ccd , the h@xmath1 , [ ] , and [ ] emission lines will all sample slightly different parts of the galaxy rotation curve . this effect matters little for our purposes , since the [ ] and [ ] emission lines were often quite weak in our galaxies and were generally not useful for mapping the galaxy rotation curve over significant angular extents . in addition , these lines were frequently too noisy to provide a reliable check on the radial velocitues derived from the h@xmath1 features . to derive p - v curves for our galaxies , we used the _ iraf _ apall task to extract a series of spectra along the slit for each object . because most of the galaxies have at least some degree of asymmetry , and because many of the rotation profiles are rising to the last measured point , we were unable to use a point of symmetry to determine the systemic velocity . however , all of the galaxies had sufficient continuum emission such that the peak of the continuum could be used for the determination of the galaxy center . galaxy centroid determination was therefore accomplished by plotting the average of 10 columns on either side of the galaxy h@xmath1 emission , and adopting as the center the position of the peak value of the continuum . we estimate this method yields a galaxy center good to roughly one - third of a pixel ( i.e. @xmath210 ) , accounting for the bias level uncertainties described above . we measured the spatial profile width of a standard star to be roughly 10 pixels , and adopted this as our extraction aperture width . for each extraction , the h@xmath1 line s centroid was determined by averaging anywhere between 8 and 19 dispersion lines , depending on the emission line strength and velocity dispersion in a given object . a series of extraction apertures was then placed by hand along the full extent of the slit , with roughly 3 pixels overlap between consecutive apertures . background regions for each aperture were set interactively , and subtracted by fitting a second order chebyshev function over the region . a by - product of the apall task for each extracted object spectrum is the spectrum of the subtracted sky background , which contains numerous oh night sky lines . because of our moderately long exposures , these lines were of high signal - to - noise . osterbrock et al . ( 1996 ) compiled an accurate list of reference wavelengths for these lines , thus we were able to utilize them to obtain a final dispersion solution for our spectra . use of this wavelength calibration technique eliminates the need to separately correct the zero - point solution for the effects of spectrograph flexure . as a final step , we measured the wavelengths of the emission lines in the extracted spectra using splot to perform gaussian fits to the line profiles . the resulting wavelengths at each point were converted to radial velocities and corrected for heliocentric motion . the uncertainties in each data point ( @xmath25 ) were computed as @xmath26 , where @xmath27 is the velocity resolution per pixel and @xmath28 is the mean signal - to - noise ratio of the three brightest pixels in each line profile . the sizes of the error bars plotted on figure [ fig : pv ] ( discussed below ) thus give an indication of the relative strength of the h@xmath1 line in our various target galaxies . figure [ fig : comp ] shows a plot comparing the heliocentric radial velocities derived from our new optical rotation curves with the values derived from global spectra by fouqu et al . ( 1990 ) and matthews et al . overall the agreement is excellent . this agreement suggests that most extreme late - type spirals seem to have well - defined centers in spite of their often diffuse stellar disks and shallow central potentials . consequently , this can offer an additional means of distinguishing extreme late - type spirals from true irregular galaxies . in our sample we have only 6 instances where we see a discrepancy of more than @xmath210 between the radial velocities derived from optical versus radio measurements . because of the small formal measurement errors on both the optical and radial velocities ( @xmath2910 and @xmath295 , respectively ) , it is possible that in at least some of these cases , these discrepencies may a consequence of true offsets between the centroid of the global profile and the center of the galaxy defined by the optical continuum . similar offsets have been reported in other moderate - to - low - mass , late - type spirals ( colin & athanassoula 1981 ; minniti , olszewski , & reike 1993 ; rownd , dickey , & helou 1994 ) . in 4 of the 6 galaxies where @xmath30 and @xmath31 differ by more than 10 , asymmetries are observed in the global profiles ( see figure [ fig : pv ] ; section [ asymm ] ) . together these two trends may hint that these could be galaxies where the stellar disk is in libration about the minimum of the or dark matter potential ( kornreich et al . 1998 ; see also levine & sparke 1998 ; noordermeer , sparke , & levine 2001 ) . levine & sparke ( 1998 ) and noordermeer et al . ( 2001 ) have shown through numerical simulations and modelling that such configurations can reproduce the lopsidedness observed in many extreme late - type spirals ( see also below ) . nonetheless , as we discuss below , approximately one - third of our sample appears to exhibit kinematic lopsideness , hence in the majority of lopsided galaxies , the peak of the stellar density correlates surprisingly well with the centroid of the global profile . the centroid of the global profile thus generally seems to be a good indicator of the location of the minimum of the disk potential , even in most asymmetric galaxies . this would seem to argue that the overall potential of the galaxies ( established mainly by the dark matter halo in the small systems considered here ) are symmetric on large scales , even in cases where asymmetries in the visible matter or velocity field are observed ( cf . rix & zaritsky 1995 ; jog 1999 ) . the position - velocity ( p - v ) curves derived from our final spectra are presented in figure [ fig : pv ] . when available , we plot next to the h@xmath1 p - v curves the corresponding high resolution global spectra from matthews et al . parameters measured from the optical data , including systemic velocities and the peak measured rotational velocities , are presented in table 2 along with photometric and global parameters from the literature . from figure [ fig : pv ] and table 2 it can be seen that among the galaxies where @xmath32 ( i.e. , cases where the galaxy s major axis and the spectrograph slit were well - aligned ) , the disk rotation velocities measured in the plane of the sky from our new optical data ( @xmath33 ) agree to within measurement errors with those derived from global data ( @xmath34 ) about half of the time . in the remaining cases , regions were not observable to sufficiently large radii to reach the peak of the rotation curves in our longslit measurements of these optically faint galaxies . for the 15 galaxies in our sample with @xmath35 and @xmath32 , we derived deprojected rotation curves along the disk major axis ( figure [ fig : rc ] ) . the disks were assumed to be in circular rotation , and the measured heliocentric radial velocities at each point were projected onto the plane of the galaxy using the relations given by rubin , ford , & thonnard ( 1980 ) to correct for disk inclination and small position angle misalignments . in cases where the systemic velocities derived from global spectra differed from our new optically derived values , we used whichever value minimized the asymmetry of the folded rotation curve . distances for the galaxies were adopted from mg97 . we see from figures [ fig : pv ] & [ fig : rc ] that most of our sample galaxies show a leisurely rise in rotational velocity with increasing radius , and many have rotation profiles that continue to rise to the last measured point . similar rotation curve shapes are frequently seen among the extreme late - type spiral rotation curves measured by other workers using the h@xmath1 emission line ( e.g. , goad & roberts 1981 ; karachentsev 1991 ; makarov , burenkov , & tyurina 1999 ; swaters 1999 ; dalcanton & bernstein 2000 ; mcgaugh , rubin , & de blok 2001 ) and have become well - established as the hallmark of small , dark matter - dominated disk galaxies . these are in stark contrast to giant spirals , where rotation curves typically rise rapidly to @xmath36 within the inner portion of the stellar disk ( cf . casertano & van gorkom 1991 ) . in 3 of the most extreme cases ( eso 418 - 008 ; eso 504 - 017 ; eso 440 - 049 ) , we can partially attribute the shallow velocity gradients in our observed p - v plots to position angle misalignments between the spectrograph slit and the galaxy major axis ( @xmath37 ; see table 1 ) . however , in all other instances , the slow rise of the rotation velocity as a function of @xmath38 appears to be real , and thus be indicative of low matter densities within the inner disks of these galaxies . for example , eso 358 - 060 has a particularly shallow rotation curve in spite of being nearly edge - on ( @xmath39=85@xmath40 ) . this galaxy has a very low deprojected central surface brightness [ @xmath4125.5 mag arcsec@xmath8 ] and would likely be nearly invisible if seen close to face - on in an optical broadband survey image . as demonstrated by matthews & wood ( 2001 ) , such a shallow rotation curve can not be explained solely by projection effects or internal extinction , even in edge - on systems . in only a few of our sample galaxies do we see a clear turnover or flattening of the optical p - v curve on both the approaching and receding sides : eso 482 - 005 , eso 504 - 025 , eso 440 - 049 , eso 443 - 079 . these galaxies are some of the `` earlier '' hubble type objects in our sample ( sc - sd ) , and include some of the sample galaxies with the most clearly - defined spiral structure . in some other cases , we see an apparent turnover on only one side of the p - v curve ( eso358 - 020 ; eso 425 - 008 ; eso 438 - 005 ; eso 380 - 025 ; eso 508 - 034 ; eso 444 - 033 ; see also section [ asymm ] ) . we find that even in instances where a turnover or flattening is seen , our observed p - v curves rarely extend significantly beyond this turnover . moreover , while the sc - scd galaxies in our sample seem more prone to exhibiting flattening of their p - v curves , we otherwise see no absolute correlation between hubble type , galaxy morphology , or luminosity and the p - v or rotation curve shape in our sample . for example , eso 504 - 025 and eso 358 - 020 both have @xmath43 - 17.5 and @xmath44 , but have very different inner rotation curve shapes and amplitudes . at the same time , we also see cases where optically rather dissimilar galaxies ( e.g. , eso 380 - 025 and eso 508 - 034 ) have similar rotation curve shapes and amplitudes . these rotation curves signatures may hold important clues as to the evolutionary histories of these galaxies for example , the possibility that morphological tranformations may occur in late - type disk galaxies as a result of viscous evolution , interactions , or tidally - induced starbursts ( e.g. , gallagher & matthews 2001 ; noguchi 2001 ; matthews & uson 2002 ) . rotation curves that do not flatten within the optical galaxy imply that the bulk of the stellar disk is in solid body rotation . it has long been known that such rotation curves are common in gas - rich dirr galaxies ( e.g. , carignan & freeman 1988 ; carignan & puche 1990 ; ct et al . 1991 ; casertano & van gorkom 1991 ; ct , carignan , & freeman 2000 ) . however , while most such galaxies are still predominantly rotationally supported , many of the dirrs with apparent solid - body rotation curves also have single - peaked global spectra , implying that solid body rotation persists to the outermost observable regions of the gas disk ( e.g. , skillman 1996 ) . in contrast , most of the galaxies in our sample have _ double - peaked _ global profiles ( figure [ fig : pv ] ; see also fouqu et al . 1990 ; matthews et al . 1998 ) , which in turn implies that the outer rotation curve contains a flat or relatively flat region ( cf . giovanelli & haynes 1988 ) , although it may lie outside the stellar disk . data will clearly be needed to trace this portion of the disk rotation profiles in our present sample . it is evident from figures [ fig : pv ] & [ fig : rc ] that asymmetries are the rule rather than the exception for our extreme late - type spiral rotation profiles . many of the p - v plots and rotation curves show significant differences in their radial extents on the approaching and receding sides . often , the amplitudes of the two sides of the rotation curves also differ ( sometimes by tens of kilometers per second ) , and in some cases , the rotation curves show differing velocity gradients on the two sides of the galaxy . although there is frequently some uncertainty in defining the center of any given rotation curve , our sample nonetheless appears to contain some galaxies where the velocity differences between the approaching and receding sides are greater than 20% , in contrast to swaters et al . ( 1999 ) , who suggested that perhaps such significant ( kinematic ) asymmetries may not exist . based on statistical studies of high - quality global data , richter & sancisi ( 1994 ) and haynes ( 1998 ) found that profile asymmetries occur in approximately 50% of normal spirals . matthews et al . ( 1998 ) explored the frequency of profile asymmetries in extreme late - type spirals and found that the frequency of lopsidedness may be even higher among small late - type spirals than among giant galaxies . unfortunately , from asymmetries in the global profiles alone , it is not possible to assess whether these asymmetries arise from asymmetric gas distributions ( e.g. as a result of accretion events ; rix & zaritsky 1995 ; zaritsky & rix 1997 ) or from true kinematic lopsidedness , which in turn may indicate lopsidedness in the overall galaxy potential ( e.g. , schoenmakers , franx , & de zeeuw 1997 ; levine & sparke 1998 ; jog 1999 ) . our new optical spectra allow us to obtain additional insight into this question . six of the 17 galaxies in the present sample for which matthews et al . ( 1998 ) obtained high precision global profiles were catagorized by matthews et al . as having strong asymmetries ( eso 358 - 020 , eso 422 - 005 ; eso 502 - 016 ; eso 504 - 017 ; eso 380 - 025 ; eso 444 - 033 ) . in 5 out of 6 of these cases , we find the optical rotation curve to have a greater amplitude on the same side as the `` excess '' in the global profile . several of these cases also exhibit different velocity gradients on the approaching and receding sides of the disk . this latter effect is particularly pronounced in eso 422 - 005 and eso 444 - 033 . this suggests that these asymmetries result from true lopsidedness in the disk kinematics rather than non - symmetric matter densities ( see swaters et al . 1999 ; noordermeer et al . 2001 ) . among the 11 remaining galaxies in our sample whose profiles were analyzed by matthews et al . and not classified as strongly asymmetric , we see a significant rotation curve _ amplitude _ asymmetry in only one other case ( eso 358 - 015 ) , and a significant difference in the rates at which the two sides of the rotation curve rise in only eso 358 - 015 and eso 359 - 029 . strong asymmetries in the global profiles therefore appear to be good , although not perfect , predictors of lopsidedness in the disk kinematics . as noted above , we see additional galaxies in our sample whose rotation curves show asymmetries in terms of their _ extents _ on the two sides of the disk ( e.g. , eso 305 - 009 , eso 425 - 008 , eso 438 - 005 ) . however , in these latter cases , it is possible that these asymmetries arise from asymmetric matter distributions as opposed to true lopsided kinematics . we conclude that the results from our small sample are consistent with at least one - third of extreme late - type spirals exhibiting true kinematic asymmetries ( see also swaters et al . 1999 ) , and hence possibly asymmetries in their overall potentials ( but see section [ voptcomp ] ) and with approximately 50% of extreme late - type spirals exhibiting either kinematic _ or _ structural asymmetries . since all of the objects in our sample are field spirals , generally lying on the peripheries of loose groups or associations ( gallagher , littleton , & matthews 1995 ) , it would seem unlikely that the lopsided natures of all of our sample galaxies are exclusively the result of very recent interactions with other galaxies , although there may be some exceptions ( e.g. , matthews & uson 2002 ) . although asymmetries are common among disk galaxies covering a wide range of masses and luminosities ( e.g. , baldwin , lynden - bell , & sancisi 1980 ; richter & sancisi 1994 ; haynes et al . 1998 ) , our new data , and the findings of matthews et al . ( 1998 ) , hint that they are equally , or perhaps more common in low - mass spirals . indeed , based on the models of noordermeer et al . ( 2001 ) , pronounced asymmetries are more persistent in galaxy systems where the dark matter halo strongly dominates the potential , as is expected in typical extreme late - type spirals ( see also levine & sparke 1998 ) . using models with a disk lying off - center in a dark halo , noordermeer et al . produced model rotation curves whose shapes were very similar to some of those seen in the present sample , i.e. , showing differing slopes on the approaching and receding sides , and flattening of the rotation profile on one side only . swaters et al . ( 1999 ) noted similar signatures in the rotation curves of the two asymmetric extreme late - type spirals that they analyzed . five of the extreme late - type spirals in the present sample were noted by mg97 to possess a compact , semi - stellar nucleus ( eso 359 - 029 ; eso 358 - 015 ; eso 305 - 009 ; eso 504 - 025 ; eso 505 - 013 ; see also figure [ fig : elts ] ) . the nucleus of eso 359 - 029 was subsequently observed with the planetary camera 2 on the _ hubble space telescope _ and demonstrated by matthews et al . ( 1999 ) to be a true compact star cluster nucleus analogous to those at the centers of many brighter galaxies . subsequently , one additional galaxy in our sample ( eso 418 - 008 ) has also been found to possess a compact star cluster nucleus based on _ hst _ imaging ( bker et al . 2002 ; windhorst et al . although the formation mechanism for these compact nuclei at the centers of such diffuse galaxies with shallow central potentials remains a mystery , these types of `` naked '' nuclei are now known to be commonplace at the centers of late - type , pure disk spirals ( see also van den bergh 1995 ; phillips et al . 1996 ; bker et al . 2001,2002 ) . in our present sample , in five of the cases where the presence of a compact nuclear feature is seen in the ccd images of mg97 , we also see a signature of this nucleus in our spectroscopic data i.e . the p - v curves exhibit reversals , breaks , or disturbances near the location of the nucleus ( see figure [ fig : pv ] ) . this strongly suggests that in all of these cases , these nuclei are compact , massive star clusters rather than simply small nuclear regions . if we assume that the ionized gas surrounding these nuclei is in keplerian rotation , we can make very rough estimates of their masses from our present data @xmath45 , where @xmath46 is the gravitational constant , @xmath38 is the galactocentric distance along the major axis , and @xmath47 is the orbital semi - amplitude of the material orbiting the nucleus in the reference frame of the disk , after correction for inclination . because the velocity and spatial resolution of our data are rather coarse relative to the amplitudes and spatial extents of these central rotation curve features , and because it is unclear whether the nuclear disk material will share the same inclination as the main disk of the galaxy , these calculations are necessarily very crude . nonetheless , we estimate nuclear masses @xmath48 , comparable to the masses of low - luminosity star cluster nuclei derived for other late - type spirals [ e.g. , m33 : @xmath49 ( kormendy & mcclure 1993 ) ; ngc 4449 : @xmath50 ( bker et al . 2001 ) ; ic 342 : @xmath51 ( bker , van der marel , & vacca 1999 ) ] . as a final note , we draw attention to the fact that some of the nuclei in our present sample appear to be offset from the dynamical centers of the galaxies as defined by either the optical continuum and/or the centroid of the global profile ( see also the appendix ) . the frequent displacement of compact nuclei from the dynamical centers of galaxies has been discussed previously by miller & smith ( 1992 ; see also levine & sparke 1998 ) and may offer an important clue to the formation mechanism of the nuclei . higher resolution spectroscopy of more extensive samples of these types of nuclei is clearly desirable in order to better understand their natures , to improve these mass estimates , and to help us understand their dynamical relationships to the surrounding disk material . in cold dark matter ( cdm ) galaxy formation models , protodisk baryons tend to efficiently lose angular momentum to their dark matter halos , resulting in a situation where the predicted sizes of galaxy disks are too compact by substantial factors ( e.g. , navarro , frenk , & white 1995 ; navarro & steinmetz 2000 ) . since our sample contains a variety of small disk galaxies , it provides an opportunity to constrain disk angular momenta at moderate galactic masses . these systems are of particular interest , since their moderate mass , yet gas - rich and relatively unevolved disks pose a challenge for proposed solutions to the angular momentum problem via internal evolution , such as intense star - forming events ( cf . efstathiou 2000 ; silk 2001 ) . we calculated specific angular momenta @xmath52 for the stellar disks in our present sample using the approximate disk scale lengths in mg97 and assuming @xmath53 , where the inverse exponential scale length is @xmath1 and @xmath54 is the peak of the linearly rising rotation curve . this is a lower bound , as the gas will have higher specific angular momentum , as will the stellar disks of galaxies where @xmath54 is approached rapidly within 1 - 2 scale lengths . the result is that we find typical angular momenta of 100 - 200 km s@xmath55 kpc for systems with @xmath56 100 km s@xmath55 . thus this excercise reveals that our sample galaxies appear to fall on the normal angular momentum - rotational velocity relationship found for giant spirals ( e.g. , navarro & steinmetz 2000 ) and that our data agree with results from rotation curve studies in showing that disks of gas - rich , moderate mass galaxies have normal specific angular momenta ( see van den bosch et al . our findings therefore are consistent with the specific angular momenta of small , pure disk systems having been set during their formation ( see fall & efstathiou 1980 ) , in contradiction to the predictions of standard cdm models ( cf . navarro & steinmetz 2000 ) . we have presented new optical longslit kinematic measurements for 21 low - luminosity , extreme late - type spiral galaxies using the h@xmath1 emission line . we have derived position - velocity ( p - v ) curves for all 21 galaxies , as well as deprojected major axis rotation curves for 15 of the objects . to facilitate future synthesis with rotation curve measurements , we have also made all of our p - v data available in electronic format . most of the extreme late - type spirals in our sample exhibit slowly rising , nearly linear p - v curves throughout a significant fraction of their stellar disks . many of the p - v curves continue to rise to the last measured point in the optical galaxy , implying low central matter densities . the disks of these galaxies appear to follow the trend between specific angular momentum and rotational velocity seen in giant spirals , suggesting that the angular momenta of disks are not closely tied to the details of their structures or star formation histories , and are intrinsically larger than predicted by current cold dark matter models . a significant fraction of our measured optical p - v curves are not symmetric , and our observations are consistent with @xmath250% of extreme late - type spirals possessing kinematic and/or structural asymmetries . in some cases , this may also be manifested as an offset between the @xmath57 values measured from global spectra and that measured from the optical data , although in most cases those values show excellent agreement . we find that lopsided global profiles appear to be a good , although not a perfect predictor of kinematic asymmetries in extreme late - type spiral disks . five of the extreme late - type spirals in our sample possess compact , semi - stellar nuclei that produce kinematic signatures in the galaxy rotation curves . rough mass estimates help to confirm that these are massive , highly compact star nuclei star clusters analogous to those seen in brighter galaxies . such nuclei appear to be a common feature of extreme late - type spiral galaxies in spite of their apparently weak central gravitational potentials . we are grateful to the late r. schommer as well as the ctio staff for their assistance with the rc spectrograph observations . ldm completed a portion of this work while supported by a clay fellowship from the harvard - smithsonian center for astrophysics . jsg thanks the university of wisconsin graduate school and for their partial support of this research . baldwin , j. e. , lynden - bell , d. , & sancisi , r. 1980 , mnras , 193 , 313 blais - ouellette , s. , carignan , c. , amran , p. , & ct , s. 1999 , aj , 118 , 2123 bker , t. , laine , s. , van der marel , r. p. , sarzi , m. , rix , h .- w . , ho , l. c. , & shields , j. c. 2002 , aj , in press ( astro - ph/0112086 ) bker , t. , van der marel , r. p. , mazzuca , l. , rix , h .- w . , rudnick , g. , ho , l. c. , & shields , j. c. 2001 , aj , 121 , 1473 bker , t. , van der marel , r. p. , & vacca , w. d. 1999 , aj , 118 , 831 bosma , a. , van der hulst , j. m. , & athanassoula , e. 1988 , a&a , 198 , 100 carignan , c. & freeman , k. c. 1988 , apj , 347 , 760 carignan , c. & puche , d. 1990 , aj , 100 , 641 carignan , c. , sancisi , r. , & van albada t. s. 1988 , aj , 95 , 37 casertano s. & van gorkom j. h. 1991 , aj , 101 , 1231 colin , j. & athanassoula , e. 1981 , a&a , 97 , 63 ct , s. , carignan , c. , & sancisi , r. 1991 , aj , 102 , 904 ct , s. , carignan , c. , & freeman , k. c. 2000 , aj , 120 , 3027 dalcanton , j. j. & bernstein , r. a. 2000 , in dynamics of galaxies : from the early universe to the present , ed . by f. combes , g. a. mamon , and v. charmandaris , ( san franciso : asp ) , 161 de blok , w. j. g. , mcgaugh , s. s. , bosma , a. , & rubin , v. c. 2001 , apj , 552 , l23 de blok , w. j. g. , mcgaugh , s. s. , & van der hulst , j. m. 1996 , mnras , 283 , 18 efstathiou , g. 2000 , mnras , 317 , 697 fall , s. m. & efstathiou , g. 1980 , mnras , 193 , 189 fouqu , p. , bottinelli , l. , durand , n. , gouguenheim , l. , & paturel , g. 1990 , a&as , 86 , 473 gallagher , j. s. , littleton , j. e. , & matthews , l. d. 1995 , aj , 109 , 2003 gallagher , j. s. & matthews , l. d. 2001 , in modes of star formation and the origin of field star populations , ed . by e. grebel & w. brandner , ( san francisco : asp ) , in press giovanelli , r. , & haynes , m. p. 1988 , in galactic and extragalactic radio astronomy , edited by g. l. verschuur and k. i. kellermann ( new york : springer - verlag ) , 526 goad , j. w. & roberts , m. s. 1981 , apj , 250 , 79 haynes , m. p. , van zee , l. , hogg , d. e. , roberts , m. s. , & maddalena , r. j. 1998 , aj , 115 , 62 iliev , i. t. , shapiro , p. r. 2001 , apj , 546 , l5 jobin , m. & carignan , c. 1990 , aj , 100 , 648 jog , c. j. 1999 , apj , 522 , 661 karachentsev , i. d. 1991 , sov . , 17 , 206 kormendy , j. & mcclure , r. d. 1993 , aj , 105 , 1793 kornreich , d. a. , haynes , m. p. , & lovelace , r. v. e. 1998 , aj , 116 , 2263 kravtsov , a. v. , klypin , a. a , bullock , j. s. , & primack , j. r. 1998 , apj , 502 , 48 levine , s. e. & sparke , l. s. 1998 , apj , 496 , l13 martimbeau , n. , carignan , c. , & roy , j .- 1994 , aj , 107 , 543 makarov , d. i. , burenkov , a. n. , & tyurina , n. v. 1999 , astron . , 25 , 706 massey , p. , valdes , f. , & barnes , j. 1992 , a user s guide to reducing slit spectra with iraf ( http://iraf.noao.edu/docs/spectra.html ) matthews , l. d. 2000 , aj , 120 , 1764 matthews , l. d. et al . 1999 , aj , 118 , 208 matthews , l. d. & gallagher , j. s. 1997 , aj , 114 , 1899 ( mg97 ) matthews , l. d. & uson , j. m. 2002 , in seeing through the dust : the detection of and the exploration of the ism in galaxies , ed . by r. taylor , t. landecker , & t. willis ( san francisco : asp ) , in press matthews , l. d. , van driel , w. , & gallagher , j. s. 1998 , aj , 116 , 1169 matthews , l. d. & wood , k. 2001 , apj , 548 , 150 2001 mcgaugh , s. s. & de blok , w. j. g. 1998 , apj , 499 , 41 mcgaugh , s. s. , rubin , v. c. , & de blok , w. j. g. 2001 , aj , 122 , 2381 meurer , g. r. , carignan , c. , beaulieu , s. f. , & freeman , k. c. 1996 , aj , 111 , 1551 miller , r. h. & smith , b. f. 1992 , apj , 393 , 508 minniti , d. , olszewski , e. w. , & rieke , m. 1993 , apj , 410 , l79 navarro , j. f. , frenk , c. s. , & white , s. d. m. 1995 , mnras , 275 , 56 navarro , j. f. & steinmetz , m. 2000 , apj , 538 , 477 noguchi , m. 2001 , apj , 555 , 289 noordermeer , e. , sparke , l. s. , & levine , s. e. 2001 , mnras , 328 , 1064 osterbrock , d. e. , fulbright , j. p. , martel , a. r. , keane , m. j. , & trager , s. c. 1996 , pasp , 108 , 277 phillips , a. c. , illingworth , g. d. , mackenty , j. w. , & franx , m. 1996 , aj , 111 , 1566 richter , o .- g . & sancisi , r. 1994 , a&a , 290 , 9 rix , h .- w . & zaritsky , d. 1995 , apj , 447 , 118 rownd , b. k. , dickey , j. m. , & helou , g. 1994 , aj , 108 , 1638 rubin , v. c. , ford , w. k. jr . , & thonnard , n. 1980 , apj , 238 , 471 schoenmakers , r. h. m. , franx , m. , & de zeeuw , p. t. 1997 , mnras , 292 , 349 silk , j. 2001 , mnras , 324 , 313 skillman , e. d. 1996 in the minnesota lectures on extragalactic neutral hydrogen , asp conference series , vol . 106 , ed . by e. d. skillman ( san francisco : asp ) , 208 swaters , r. a. 1999 , ph.d . thesis , university of groningen swaters , r. a. , madore , b. f. , & trewhella , m. 2000 , apj , 531 , l107 swaters , r. a. , schoenmakers , r. h. m. , sancisi r. , & van albada t.s . 1999 , mnras , 304 , 330 van den bergh , s. 1995 , aj , 110 , 613 van den bosch , f. c. , burkert , a. , & swaters , r. a. 2001 , mnras , 326 , 1205 van den bosch , f. c. , robertson , b. e. , dalcanton , j. j. , & de blok , w. j. g. 2000 , aj , 119 , 1579 van der hulst , j. m. , skillman , e. d. , smith , t. r. , bothun , g. d. , mcgaugh s. s. , & de blok w. j. g. 1993 , 106 , 548 windhorst , r. a. et al . 2002 , submitted to apjs zaritsky , d. & rix , h .- w . 1997 , apj , 477 , 118 llccccccc eso418 - 008 & sbdm & 03 29 28 & -30 22 54 & 2 & 1500 & 0 & 140@xmath58 & 40 eso482 - 005 & sbd & 03 30 52 & -24 18 06 & 2 & 1800 & 79 & 80 & 1 eso358 - 015 & sdm & 03 31 10 & -34 58 30 & 2 & 2000 & 16 & 180 & 16 eso358 - 020 & sbdm & 03 32 58 & -32 48 18 & 2 & 1500 & 165 & 167 & 2 eso358 - 060 & sm & 03 43 18 & -35 43 30 & 2 & 2000 & 102 & 102 & 0 eso359 - 029 & sdm & 04 10 56 & -33 07 42 & 2 & 2000 & 20 & 20 & 5 eso422 - 005 & sdm & 04 50 07 & -28 40 30 & 1 & 1000 & 35 & 15 & 20 eso305 - 009 & sabc & 05 06 26 & -38 22 30 & 2 & 2000 & 90 & 63@xmath58 & 27 eso425 - 008 & sbm : & 06 04 38 & -27 52 18 & 2 & 1800 & 80 & 80 & 0 am0605 - 341 & sbdm & 06 05 31 & -34 11 49 & 1 & 750 & 90 & 90 & 0 eso502 - 016 & sbd & 11 02 48 & -26 21 18 & 1 & 1200 & 82 & 79 & 3 eso438 - 005 & sd & 11 06 33 & -28 06 00 & 2 & 2000 & 62 & 60 & 2 eso504 - 010 & sbdm & 11 40 32 & -23 09 18 & 1 & 1200 & 14 & 15 & 1 eso504 - 017 & scd / bcd & 11 46 15 & -27 06 00 & 1 & 500 & 60 & 103 & 43@xmath59 eso504 - 025 & sd & 11 51 18 & -27 04 18 & 2 & 1800 & 30 & 20 & 10 eso440 - 049 & sc & 12 02 59 & -31 08 42 & 2 & 1800 & 0 & 60 & 60 eso505 - 013 & sabc & 12 03 33 & -22 34 18 & 1 & 900 & 135 & 45 & 90@xmath59 eso380 - 025 & sbdm & 12 21 56 & -35 07 54 & 1 & 1000 & 15 & 13 & 2 eso443 - 079 & sabd & 13 07 38 & -27 42 18 & 2 & 2000 & 0 & 180 & 0 eso508 - 034 & spec & 13 14 13 & -25 04 24 & 1 & 900 & 132 & 129 & 3 eso444 - 033 & sbdm & 13 23 16 & -31 52 12 & 1 & 1000 & 89 & 88 & 1 lcccrrrcrrc eso418 - 008 & 1.91 & @xmath6017.4 & 55.0 & 153.5 & 125.0 & 1188.9 & 0.64 & 79.6 & 1193.4 & 1 eso482 - 005 & 2.60 & @xmath6017.0 & 81.3 & 170.8 & 153.7 & 1914.2 & 1.54 & 178.1 & 1918.4 & 1 eso358 - 015 & 1.53 & @xmath6016.3 & 46.7 & 100.2 & 67.8 & 1402.9 & 0.59 & 98.8 & 1386.4 & 1 eso358 - 020 & 1.93 & @xmath6017.7 & 37.1 & 117.7 & 98.1 & 1754.2 & 0.35 & 64.8 & 1762.2 & 1 eso358 - 060 & 2.24 & @xmath6014.3 & 85.1 & 82.0 & 54.0 & 801.2 & 1.03 & 61.2 & 794.4 & 1 eso359 - 029 & 1.92 & @xmath6015.8 & 56.4 & 141.1 & 134.2 & 872.6 & 0.68 & 140.8 & 866.3 & 1 eso422 - 005 & 1.70 & @xmath6016.7 & 41.7 & 126.2 & 99.0 & 1460.9 & 0.73 & 65.3 & 1481.3 & 1 eso305 - 009 & 4.63 & @xmath6017.2 & 52.9 & 129.8 & 113.1 & 1018.8 & 1.56 & 112.4 & 999.4 & 1 eso425 - 008 & 1.20 & @xmath6014.1 & 78.8 & 106.5 & 91.3 & 1475.5 & 0.57 & 94.3 & 1473.9 & 1 am0605 - 341 & 1.27 & @xmath6015.4 & 26.0 & 124.2 & 95.3 & 775.2 & 0.44 & 55.8 & 768.9 & 1 eso502 - 016 & 2.46 & @xmath6017.2 & 67.1 & 155.0 & 102.8 & 1511.0 & 0.85 & 106.7 & 1540.3 & 1 eso438 - 005 & 3.40 & @xmath6016.7 & 80.0 & 163.9 & 145.5 & 1499.4 & 1.86 & 145.9 & 1497.2 & 1 eso504 - 010 & 1.44 & @xmath6016.2 & 67.2 & 145 & 127 & 1932 & 0.64 & 118.3 & 1940.8 & 2 eso504 - 017 & 1.10 & @xmath6018.0 & 30.8 & 220.1 & 169.3 & 1888.3 & 0.55 & 123.2 & 1891.5 & 1 eso504 - 025 & 2.42 & @xmath6017.2 & 39.9 & 137 & 118 & 1637 & 1.06 & 105.2 & 1631.6 & 2 eso440 - 049 & 2.22 & @xmath6018.5 & 32.0 & 151 & 131 & 2264 & 0.78 & 54.5 & 2273.0 & 2 eso505 - 013 & 2.96 & @xmath6018.8 & 18.3 & 132.2 & 117.3 & 1726.1 & 1.62 & 82.6 & 1719.4 & 1 eso380 - 025 & 1.63 & @xmath6018.6 & 76.9 & 180.6 & 163.9 & 2788.1 & 0.84 & 186.1 & 2799.2 & 1 eso443 - 079 & 2.18 & @xmath6017.2 & 72.8 & 157.7 & 136.3 & 2124.4 & 1.16 & 132.5 & 2127.8 & 1 eso508 - 034 & 1.53 & @xmath6017.5 & 64.6 & 167 & 150 & 1910 & 0.79 & 162.2 & 1911.5 & 2 eso444 - 033 & 2.03 & @xmath6017.9 & 74.3 & 169.0 & 147.0 & 2428.2 & 0.92 & 182.9 & 2408.7 & 1 lccc eso482 - 005 & 70.5 & 2003.0 & 18.5 & 68.1 & 2001.5 & 19.4 & 65.7 & 1997.5 & 17.9 & 53.3 & 1977.6 & 15.4 & 51.2 & 1987.1 & 10.1 & 49.0 & 1994.7 & 8.2 & 46.9 & 1990.7 & 8.2 & 44.8 & 1984.0 & 7.8 & 42.4 & 1979.8 & 7.5 & 40.3 & 1978.5 & 9.0 & 37.8 & 1984.2 & 10.4 & 36.0 & 1984.8 & 14.0 & 34.2 & 1987.3 & 12.7 & 31.8 & 1979.9 & 5.6 & 29.4 & 1990.1 & 3.5 & 27.6 & 1987.6 & 3.3 & 25.7 & 1987.3 & 3.3 & 23.0 & 1977.8 & 5.1 & 20.3 & 1974.2 & 6.2 & 17.2 & 1972.8 & 7.5 & 14.2 & 1966.4 & 8.2 & 11.2 & 1960.1 & 6.8 & 8.5 & 1955.5 & 7.1 & 5.1 & 1946.4 & 6.7 & 2.4 & 1929.8 & 5.0 & -0.5 & 1918.4 & 4.8 & -2.9 & 1914.9 & 4.9 & -6.3 & 1902.8 & 6.7 & -9.0 & 1885.9 & 7.8 & -12.0 & 1869.0 & 8.6 & -15.0 & 1861.7 & 11.8 & -17.5 & 1859.2 & 13.7 & -23.8 & 1868.3 & 12.1 & -26.2 & 1869.2 & 12.2 & -29.0 & 1867.5 & 12.9 & -31.7 & 1857.8 & 16.1 & -59.8 & 1824.9 & 13.5 & -63.2 & 1832.7 & 11.7 & -66.2 & 1837.1 & 10.9 & -68.9 & 1846.7 & 16.2 & -75.3 & 1831.8 & 19.2 & -77.7 & 1836.5 & 18.8 _ eso 418 - 008 : _ the shallow amplitude of the h@xmath1 p - v curve relative to the global profile likely partially arises from the positioning of the spectrograph slit along the galaxy bar , which is displaced @xmath240@xmath40 from the galaxy major axis . _ eso 358 - 015 : _ the break in this galaxy s p - v curve near @xmath57 appears to correspond to the location of the galaxy s semi - stellar nucleus , which is displaced from the dynamical center of the galaxy as defined by both the global profile and the optical continuum emission . the h@xmath1 p - v curve is strongly asymmetric , as is the stellar disk of the galaxy . these strong asymmetries suggest that the offset between the measured optical and radio radial velocities may be real , and a consequence of a true offset between the stellar and the disks . near the galaxy s nucleus , the h@xmath1 emission lines contain a broad wing on the redward side ; some of the h@xmath1 emission features near the galaxy nucleus also appear double - peaked . _ eso 358 - 060 : _ the h@xmath1 p - v curve extends significantly farther on the approaching side than on the receding side . this galaxy does not appear to have spiral arm structure or strong internal absorption , hence the `` wiggles '' in its p - v curve likely result from a very patchy distribution of regions . _ eso 359 - 029 : _ the reversal seen near the center of this rotation curve appears to correspond to the location of this galaxy s semi - stellar nucleus ( see also matthews et al . weak asymmetries are seen in both the optical p - v curve and the global profile . possible h@xmath1 absorption may be present . _ eso 422 - 005 : _ it is unlikely that a misalignment of 20@xmath40 between the spectrograph slit and the disk major axis can account for the discrepancy between the optical and radio @xmath57 values , or the odd shape of the optical p - v curve , which on the receding side shows a rise , then falls to near @xmath57 before once again rising . this galaxy shows an asymmetry in the global profile in terms of the shape of the rotation peaks , as well as a weak asymmetry in the stellar disk ( see also mg97 ) . _ eso 305 - 009 : _ strong nuclear continuum present . a reversal in the optical p - v curve near @xmath57 occurs at the location of this galaxy s semi - stellar nucleus . a misalignment of @xmath227@xmath40 between the slit placement and the galaxy major axis may partially account for the odd shape of the rotation curve , which on the approaching side shows a rise , then falls to near @xmath57 before once again rising . _ am 0605 - 341 : _ this galaxy shows very little rotation in its stellar disk . the global profile is significantly broader than the amplitude of the h@xmath1 p - v curve , indicating this galaxy must have extended and reach its maximum rotational velocity well outside of its stellar disk . the h@xmath1 lines near the center of this galaxy are rather broad . our spectra suggest presence of a possible starburst nucleus . _ eso 502 - 016 : _ the asymmetries in the stellar disk and the global profile of this galaxy suggest that the observed offset between the optical and radio @xmath57 values may be real . the galaxy does not show any obvious features in the optical disk to explain the wavy appearance of the p - v curve . _ eso 504 - 017 : _ the slit was misaligned by @xmath243@xmath40 from the galaxy major axis during the observations of this galaxy . however , since the disk has a relatively low inclination ( @xmath6131@xmath40 ) , this probably did not affect the observed p - v curve significantly , which shows a very low amplitude in the inner regions of the galaxy . this galaxy was shown by matthews et al . ( 1998 ) to have rather extended . _ eso 504 - 025 : _ the slight `` glitch '' seen in the optical p - v curve near @xmath57 appears to be due to this galaxy s semi - stellar nucleus . the p - v curve is observed to flatten off , but distinct dips are seen on the flat part of the p - v curve near radii of @xmath62 from the galaxy center . _ eso 505 - 013 : _ strong nuclear continuum present . this galaxy contains spiral arms clearly delineated by regions , as well as a semi - stellar nucleus . the ripples in the optical p - v curve may be due to the spiral arm pattern , while the rotation curve reversal seen near @xmath57 is likely due to the semi - stellar nucleus . the rotation curve is spatially more extended on the receding side , while the global profile contains slightly more flux on the approaching side . the slit was placed near the galaxy s minor axis during these observations , but because of the galaxy s small inclination ( @xmath63 ) , the observed shape of the p - v curve was probably not significantly affected . _ eso 380 - 025 : _ the shape of the h@xmath1 p - v curve shows significant differences on the receding and approaching sides , with the receding side flattening off , and the approaching side continuing to rise to the last measured point . _ eso 444 - 033 : _ both the approaching and receding sides of the h@xmath1 p - v curve are quite linear in shape , but have differing slopes . the @xmath57 value measured from the global spectrum appears to better correspond to the rotation curve center than the optically - derived value . both the optical galaxy and the global profile are very asymmetric , thus the observed offset between the optical and radio @xmath57 values may be real .

there is good evidence that the centers of massive early - type galaxies have a bottom - heavy stellar initial mass function ( imf ) compared to the imf of the milky way . here we study the radial variation of the imf within such galaxies , using a combination of high quality keck spectroscopy and a new suite of stellar population synthesis models that cover a wide range in metallicity . as in the previous studies in this series , the models are fitted directly to the spectra and treat all elemental abundance ratios as free parameters . using newly obtained spectroscopy for six galaxies , including deep data extending to for the galaxies ngc1407 , ngc1600 , and ngc2695 , we find that the imf varies strongly with galactocentric radius . for all six galaxies the imf is bottom - heavy in the central regions , with average mass - to - light ratio `` mismatch '' parameter at . the imf rapidly becomes more bottom - light with increasing radius , flattening off near the milky way value ( ) at . a consequence is that the luminosity - weighted average imf depends on the measurement aperture : within we find , consistent with recent lensing and dynamical results from slacs and atlas . our results are also consistent with several earlier studies that were based on analyses of radial gradients of line indices . the observed imf gradients support galaxy formation models in which the central regions of massive galaxies had a different formation history than their outer parts . finally , we make use of the high signal - to - noise central spectra of ngc1407 and ngc2695 to demonstrate that we are indeed measuring imf effects , not abundance effects .

1611.09859

there is strong evidence that the stellar initial mass function ( imf ) in the centers of massive early - type galaxies is bottom - heavy with respect to the imf in the milky way disk . the evidence primarily comes from three distinct observations . first , recent stellar population synthesis ( sps ) modeling consistently indicates a relatively large contribution of low mass stars to the integrated light ( e.g. , cenarro et al . 2003 ; van dokkum & conroy 2010 , 2011 ; conroy & van dokkum 2012b ; smith , lucey , & carter 2012 ; spiniello et al . 2012 ; la barbera et al . 2013 ; la barbera , ferreras , & vazdekis 2015 ) . second , the dynamics of massive galaxies are better fitted when a salpeter ( 1955 ) imf is assumed for the stellar component than when a kroupa ( 2001 ) or chabrier ( 2003 ) imf is used ( e.g. , thomas et al . 2011 ; cappellari et al . 2012 ; dutton , mendel , & simard 2012 ) . dutton et al . ( 2012 ) show that this also holds for very compact galaxies , which are thought to have only a minor contribution from dark matter within their effective readius . third , gravitational lensing studies show a behavior similar to the dynamical studies , requiring a salpeter - like imf for the highest mass lenses ( e.g. , treu et al . 2010 ; spiniello et al . 2011 ; sonnenfeld et al . 2015 ) . it is encouraging that all three methods are qualitatively consistent , and that comparisons of different techniques generally show the same trends . conroy et al . ( 2013 ) show that dynamical and stellar population synthesis modeling of compact galaxies are consistent with each other , and posacki et al . ( 2015 ) combine lensing with stellar population modeling . lyubenova et al . ( 2016 ) find good agreement between dynamical and stellar population measurements of galaxies in the califa survey , carefully controlling for systematic differences in the methodology . taking these studies a step further , spiniello et al . ( 2015a ) combine lensing , dynamics , and sps modeling to constrain both the slope and the cutoff of the imf below 1 for a sample of nine galaxies . however , not all studies are in agreement , and the question whether the imf varies between galaxies is not yet settled . as shown by smith ( 2014 ) a direct comparison of published results of the same individual galaxies using different techniques shows very large scatter . in addition , a detailed study of several very nearby strong lenses suggests tension between the lensing masses and the sps - inferred ones ( smith , lucey , & conroy 2015 ) . another example of possible tension is the lack of variation in the number of x - ray binaries with galaxy velocity dispersion ( peacock et al . 2014 ) , although this constrains the imf at high masses , not low masses . it is likely that these disagreements , and the large differences between some independent measurements of the same galaxies ( see smith 2014 ) , are due to a combination of factors . first , it seems that there is considerable galaxy - to - galaxy scatter in the imf ( e.g. , conroy & van dokkum 2012b ; leier et al . furthermore , it is certainly the case that the random errors in all methods underestimate the true uncertainty ( see , e.g. , tang & worthey 2015 ) . a striking illustration of this is fig . 12 of conroy & van dokkum ( 2012b ) , which shows that the sps - derived imf can vary between milky way - like and super - salpeter for different model assumptions and spectral fitting regions . also , different techniques are sensitive to different stellar mass ranges , as explored effectively in spiniello et al . ( 2015a ) . whereas lensing and dynamics measure the total mass , which includes stellar remnants , dark matter , and gas , sps methods are sensitive to the light of stars in specific mass ranges ( see fig . 17 of conroy & van dokkum 2012a , and conroy et al . 2016 ) . another possible explanation for the variation between studies is that the imf may not only vary _ between _ galaxies but also _ within _ galaxies . if this is the case , the use of different effective apertures will introduce scatter even if the same methodology is applied to the same objects . also , for a given projected aperture , lensing is sensitive to the mass in a cylinder , dynamics to the mass in a sphere , and stellar population synthesis to the projected light . imf gradients may be expected , because the velocity dispersion , surface mass density , metallicity , age , and @xmath8-enhancement all change with radius ( e.g. , mehlert et al . 2003 ; kuntschner et al . 2010 ) , and the imf may correlate with these parameters ( conroy & van dokkum 2012b ; hopkins 2013 ) . there are also reasons to expect an imf gradient from the formation history of massive galaxies . there is evidence that the centers of many massive galaxies were assembled in a short period of intense star formation at @xmath9 ( bezanson et al . 2009 ; oser et al . 2010 ; barro et al . 2013 ; nelson et al . 2014 ) that is distinct from their later growth . the physical conditions inside these star forming cores were very different from those in the milky way disk today ( e.g. , zolotov et al . 2015 ; van dokkum et al . 2015 ; barro et al . 2016 ) . in the previous papers in this series ( van dokkum & conroy 2012 ; conroy & van dokkum 2012b ) we used an effective aperture of @xmath10 , where @xmath11 is the projected half - light radius . this small aperture was largely determined by the signal - to - noise ratio ( s / n ) that is required to measure the imf - sensitive band ( wing & ford 1969 ) . for comparison , dynamical studies typically quote results at @xmath12 . in this paper we extend this analysis to larger radii , using newly obtained data from the kecki telescope in combination with an updated suite of stellar population synthesis models . this is not a new topic : studies of gradients in imf - sensitive spectral features go back at least to boroson & thompson ( 1991 ) , who found that the doublet increases toward the centers of early - type galaxies . the difficulty is that stellar abundances also change with radius ; boroson & thompson ( 1991 ) could not clearly distinguish an imf gradient from a gradient in the sodium abundance ( see also worthey , ingermann , & serven 2011 ) . recent work has provided somewhat conflicting results , even with high data quality and sophisticated modeling . martn - navarro et al . ( 2015a ) find evidence for strong imf gradients ( from bottom - heavy in the center to milky way - like at large radii ) in two massive early - type galaxies , and a constant imf in a low mass galaxy . similarly , la barbera et al . ( 2016 ) derive a strong imf gradient in a single massive elliptical galaxy using deep optical and near - ir spectroscopy with the very large telescope . in contrast , mcconnell , lu , & mann ( 2016 ) suggest that the observed line index gradients of two massive early - type galaxies can be fully explained by abundance variations , and zieleniewski et al . ( 2016 ) use spatially - resolved spectroscopy to derive a milky way imf for two of the three brightest galaxies in the coma cluster . adopting a different approach , davis & mcdermid ( 2017 ) find significant variation in the imf gradients among seven galaxies using their molecular gas kinematics . in this paper , we build on this previous work using optical spectroscopy extending to @xmath13 m obtained with the dual - beam low resolution imaging spectrograph ( lris ; oke et al . 1995 ) on keck . we obtained high quality spectra out to @xmath14 , thanks to a custom long slit mask and the fact that we spent half of the observing time on empty sky fields . following the methods we used in previous papers in this series , and in contrast to other studies of imf gradients , we fit state of the art models directly to the spectra rather than to indices . we obtained spatially - resolved spectroscopy of six early - type galaxies on december 1920 2014 , using lris on the kecki telescope . for three of the galaxies we obtained deep data with a special long slit , and interspersed the science images with blank sky exposures of equal integration time . cccccc ngc1407 & 23.3 & 1.26 & 292 & maj & 7800 + ngc1600 & 45.8 & 1.12 & 340 & maj & 9000 + ngc2695 & 35.3 & 0.42 & 229 & maj & 3000 + ngc3414 & 23.5 & 0.49 & 240 & min & 1800 + ngc4552 & 16.0 & 0.52 & 264 & min & 1800 + ngc4564 & 17.0 & 0.24 & 163 & min & 1800 the three primary targets of our study are the giant elliptical galaxies ngc1407 and ngc1600 and the s0 galaxy ngc2695 . ngc1407 is the dominant member of the eponymous ngc1407 group at 24mpc , which also includes the elliptical galaxy ngc1400 ( brough et al . 2006 ; romanowsky et al . ngc1600 , at @xmath15mpc , is generally considered an isolated elliptical , with an extensive system of satellite galaxies ( smith et al . ngc2695 is one of the brightest galaxies in a group that also contains the elliptical galaxy ngc2699 . distances are averages taken from ned . ngc1407 and ngc1600 are very similar : both are slowly rotating , very large and very massive galaxies . their central velocity dispersions are 292 and 340 respecively ( see [ starpops.sec ] ) . they likely have massive central black holes , with the black hole in ngc1600 recently claimed to be among the most massive in the local universe ( thomas et al . their major axis half - light radii are @xmath16 and @xmath17 ( li et al . 2011 ) , corresponding to 8.8kpc and 16kpc . ngc1407 is nearly round , with an ellipticity of @xmath18 . ngc1600 has @xmath19 . ngc2695 is a rotating s0 galaxy . it was chosen largely because of its availability at the end of the night . in addition to these galaxies we observed three others , with the slit oriented along the minor axis : ngc3414 , ngc4552 , and ngc4564 . these galaxies were selected because of their relatively high [ na / fe ] values in data we had obtained previously . exposure times were shorter for these objects ( although still substantially longer than in van dokkum & conroy 2012 ) , and we can not measure their imf gradients to the same distance as for the primary galaxies . the six galaxies are shown in fig . [ galimages.fig ] . the images were scaled to the same distance so their spatial extent can be compared directly . basic information for the galaxies is provided in table [ sample.tab ] . the semi - major axis effective radii of ngc1407 and ngc1600 were taken directly from li et al . the effective radii of the other galaxies were determined from the circularized values and ellipticities listed in cappellari et al . ( 2007 ) . lris is a dual beam optical spectrograph , providing simultaneous high sensitivity observations from the far blue to the far red . the beams were split with the d680 dichroic , which has a 50% reflectance wavelength of 6640 and a 50% transmission wavelength of 6800 . in the blue arm , we used the 300lmm@xmath20 grism blazed at 5000 . this is a departure from the strategy of van dokkum & conroy ( 2012 ) and mcconnell et al . ( 2016 ) , who used the higher resolution 600lmm@xmath20 grism . the 300lmm@xmath20 grism covers the full spectral range from the atmospheric cutoff to the dichroic . the 600lmm@xmath20 grism covers wavelengths @xmath21 , missing the doublet . as shown in fig . 12 of conroy & van dokkum ( 2012a ) ( and in [ imftest.sec ] ) , the nad line is important for distinguishing imf effects from variations in the sodium abundance . in the red arm the 600lmm@xmath20 gold - coated grating blazed at 10000 was used , as in previous work . the lris - red detector has fully depleted , high - resistivity ccds ( see rockosi et al . 2010 ) . these devices are sensitive to wavelengths @xmath22 m , and have no appreciable fringing . this makes it possible to do accurate , sub - percent spectroscopy in the far red . the standard long slit of lris covers only approximately half of the field of view . furthermore , in both the blue and the red beams the middle of the slit falls in the gap between the two mosaiced detectors . as a result the usable contiguous length of the slit on a single chip is only @xmath23 . we designed a custom slit mask , comprised of a @xmath24 wide , @xmath25 long slit that is broken into four pieces to ensure mechanical stability . this slit is approximately twice as long as the standard long slit . it has three gaps ; the central one coincides with the detector gap . for the three primary targets we used the following observing strategy . we used the `` special '' long slit that we designed , aligned with the major axis of the galaxy . the red side data were binned on - chip by a factor of two to reduce the read - out time , providing pixel scales of @xmath26 in the red and @xmath27 in the blue . we obtained a series of 600s exposures , alternating on - target exposures with off - target exposures . these off - target exposures are used in the reduction to enable very accurate sky subtraction over the entire spatial range of interest . they are @xmath28 removed from the galaxies , and carefully chosen so that no stars or other contaminating objects fall in the slit . for each object an equal number of on - target and off - target exposures was obtained . this strategy is similar to that of kelson et al . ( 2002 ) , who used lris to study the kinematics of a brightest cluster galaxy to large radii . mcconnell et al . ( 2016 ) also took off - target exposures but less frequently ( one sky exposure for every 24 science exposures ) . the total on - target exposure times were 7800s , 9000s , and 3000s for ngc1407 , ngc1600 , and ngc2695 respectively . the other three galaxies were observed in a classical way , dispensing with the off - target exposures . the slit was aligned with the minor axis , to facilitate standard sky subtraction techniques . three 600s exposures were obtained for each galaxy , moving the telescope along the slit by @xmath29 in between exposures . the total exposure time was therefore 1800s for each object . the slit positions are indicated on fig . [ galimages.fig ] for all six galaxies . the data reduction was done with a custom pipeline written in the python programming environment . we focus here on the reduction of the three primary targets , as the analysis of the other galaxies largely follows that described in van dokkum & conroy ( 2012 ) . the data were divided into sets of three exposures , consisting of a science exposure and the two adjacent blank sky exposures . for each science exposure , a two - dimensional sky frame was created by averaging the adjacent sky exposures . this produces a good match to the sky in the science exposure if changes in the sky line intensities were linear over the @xmath30min that elapsed during the three exposures . to account for non - linearity , the averaged sky frame was scaled to match the science frame by measuring the fluxes of the brightest sky lines in both images . this scaling was typically @xmath31% , and only important for data taken in the evening or morning when approaching 18 degree twilight . next , the spectra were wavelength calibrated . the initial solution is based on arc line exposures of the custom long slit taken in the afternoon . we used all lamps . the cadmium , zinc , and mercury lamps are the main calibrators in the blue and the neon and argon lamps have many lines in the red . in the standard line list there is a rather large gap in wavelength coverage between the ar lines at 9787 and 10473 , a wavelength regime that contains the band . fortunately there is a faint ar line at 10054.81 , and we added this line to the list . in both the blue and the red , the detector consists of two chips with two amplifiers each . the data from each of the four amplifiers ( two chips , and two amplifiers per chip ) were fitted separately . in the red , a polynomial of 6@xmath32 order in the spatial direction and 5@xmath32 order in the wavelength direction gives an rms scatter of @xmath33 for each of the segments . the full observed wavelength range is approximately 7450 10750 . as a check on the wavelength calibration in the far red we collapsed the 2d arc spectra in the spatial direction and measured the location of the very weak @xmath34 ar line . the measured wavelength is 10335.5 , which demonstrates that the polynomial fit accurately captures the transformation from pixel coordinates to wavelengths in this regime . in the blue , a polynomial of 6@xmath32 order in both the spatial direction and the wavelength direction gives an rms scatter of @xmath35 . lris has considerable flexure , and the exact wavelength solution is a complex function of the pointing of the telescope and other factors . for each individual science exposure we applied a zero - order offset to the high order arc line solutions . the applied offsets are medians of the offsets calculated from multiple sky emission lines . the offsets are typically @xmath36 in the blue and @xmath37 in the red , with the exact effect dependent on the field and time of night , and with no obvious residual wavelength dependence . , or @xmath38 , for ngc1600 . the top panel shows the observed spectrum . the middle planel is the ratio between the galaxy spectrum and the sky . at @xmath39 , the galaxy flux is only 1% 10% of the sky emission in the far red . the bottom panel shows the sky subtracted spectrum . ] for the stellar population fitting it is important to accurately measure the instrumental resolution . in the red @xmath40 , much smaller than the velocity dispersions of the galaxies , and it can be measured accurately from sky emission lines ( see fig . [ resol.fig ] ) . in the blue the situation is more complex , and requires careful treatment . first , due to our choice of a low resolution grism that covers nad , @xmath41 , comparable to the velocity dispersions of the galaxies . second , there are no sky emission lines that can be used blueward of the @xmath42[oi ] line . third , we find that there are significant focus variations in the blue , particularly redward of @xmath43 . the variations seem random and may be caused by temperature changes during the nights . we measured the instrumental resolution in the blue in the following way . all six galaxies were previously observed with the higher resolution 600lmm@xmath20 grism ( in 2012 ; see conroy et al . , in preparation ) . the instrumental resolution of those data varies between @xmath44 at 3700 to @xmath45 at 5500 . using the same extraction aperture , we fit the new spectra with the high resolution 2012 data in six wavelength intervals between 3700 and 5500 , with the velocity dispersion @xmath46 as the only free parameter . the instrumental resolution of the new data is then @xmath47 in each wavelength interval . the results are shown by the red points in fig . [ resol.fig ] . for each galaxy the red points connect smoothly to the @xmath48 blue points , demonstrating that our methodology is consistent with direct measurements from sky emission lines . solid lines show the third - order polynomials that are used in the stellar population modeling . the wavelength - calibrated individual amplifiers of each science exposure were placed in a common 2d image , taking the detector gap into account . next , a zero - order residual sky subtraction was performed to account for small differences in wavelength calibration and sky line intensity between the science exposures and the adjacent sky frames . this is particularly important for the broad o@xmath50(@xmath51 ) band at @xmath52 , which is independent of the oh lines and varies on short time scales . to do this , a small region at the bottom of the frame was used to measure the residual sky spectrum in each science exposure . as the center of the galaxy was placed on the top detector away from the chip gap ( the standard slit pointing origin ) , this region is at a radius of approximately @xmath53 from the center . we verified that the galaxy flux at this radius is sufficiently low that its subtraction has a negligible effect on the analysis in this paper . the individual science frames were combined , scaling by the collapsed galaxy flux and rejecting high and low pixels . the distortion in the spatial direction ( the @xmath49distortion ) was determined by measuring the central position of the galaxy as a function of wavelength and fitting these positions with a 3@xmath54 order polynomial . the spectra were shifted , ensuring that the center of the galaxy falls on the center of a pixel in the corrected frame . the corrections for telluric absorption and the instrument response function follow the same procedures as described in detail in van dokkum & conroy ( 2012 ) . briefly , a theoretical atmospheric transmission spectrum was fitted to the observed central spectrum of the galaxy in the wavelength interval 9300 9700 , where there are many strong h@xmath50o lines . the fit also includes a polynomial to account for the variation in the galaxy spectrum in this spectral range . the fits converge quickly and provide a near - perfect removal of the telluric absorption lines . we note that for the galaxies in this paper there is no ambiguity , as the telluric lines are a factor of @xmath55 narrower than the galaxy absorption lines . the spectra were corrected for the instrument response using observations of the white dwarf feige110 . special care was taken to correct for the broad hydrogen paschen lines in the observed white dwarf spectrum . a graphic illustration of the reduction in the outermost spectral bins is shown in fig . [ demo.fig ] . accurate modeling of the sky is critical , as the galaxy flux is only 1% 10% of the sky emission in the far red . the reduced 2d spectra of the three primary galaxies are shown in fig . [ 2dspec.fig ] . the spectra were divided by a polynomial in the wavelength direction to reduce the dynamic range and highlight the absorption lines at all radii . with a few isolated exceptions ( such as the strong @xmath565578[oi ] line ) the spectra are very clean with no obvious systematic issues . for all six galaxies we extracted one - dimensional ( 1d ) spectra from the two - dimensional ( 2d ) reduced spectra . the apertures are defined in units of binned ( @xmath26 ) pixels . with the exception of the inner apertures they are spaced quadratically , following the relation @xmath57 , with @xmath58 an integer and @xmath59 the aperture radius in pixels . this scaling is a compromise between having a sufficiently fine sampling of the full radial range and maximizing the s / n ratio in each bin . the central aperture is 3 pixels ( @xmath60 ) , corresponding to the approximate seeing . the galaxy spectrum at each radius is defined as the sum of all image rows between @xmath61 and @xmath62 , not including rows that were masked because of missing data ( due to the chip gap ) or because of contaminating objects . for each aperture in each galaxy we calculated the luminosity - weighted mean radius @xmath63 , properly taking masked rows into account . except for the central aperture ( @xmath2 ) each radius occurs twice , as spectra are independently extracted from each side of the galaxy . because the weighting and masking are not the same on each side of the galaxy , the positive and negative distances from the center are not identical . these luminosity - weighted radii are the ones that are used in the remainder of this paper . extracted spectra at @xmath64 and @xmath65 are shown in figs . [ full_spectra1.fig ] [ full_spectra3.fig ] , for the three primary galaxies . the @xmath65 spectrum is the average of the three radial bins that are closest to @xmath66 . as we show later , this is the approximate radius where the imf is no longer bottom - heavy but consistent with that of the milky way . the spectra were de - redshifted and smoothed to a common resolution of 450 , so they can be compared directly . the chosen wavelength ranges do not cover the full extent of the spectra but correspond approximately to the regions that were used in the stellar population modeling ( see conroy et al . there are several obvious differences between the inner spectra and outer spectra . the na lines , and particularly , are stronger in the center than at @xmath67 . for ngc1407 and ngc1600 the band is also more prominent in the central aperture . the spectra were fitted with stellar population synthesis models , following the procedures described in conroy & van dokkum ( 2012b ) . only spectra with a median s / n ratio @xmath68 in the red were used . for the three primary galaxies there are typically 20 radial bins that satisfy this criterion ; for the other three galaxies there are typically 10 . in total , 92 spectra were used . almost every aspect of the models has been updated since the analysis in conroy & van dokkum ( 2012b ) . the changes are briefly summarized here , and described in more detail in conroy et al . the most important difference is that the models cover a large range of metallicity and age , owing to the use of a greatly expanded stellar library ( villaume et al . this is important as we aim to separate abundance gradients from imf gradients . the model ages range from 1gyr to 13.5gyr and [ z / h ] ranges from @xmath69 to @xmath70 . furthermore , the mist stellar isochrones are used ( choi et al . 2016 ) , which cover a wide range of ages , masses , and metallicities . the expanded stellar library uses the m dwarf library of mann et al . ( 2015 ) and newly obtained near - ir spectroscopy of 283 stars ( which already had optical spectra from the miles library snchez - blzquez et al . a spectral interpolator is used to provide spectra on a continuous grid of @xmath71 , @xmath72 , and metallicity . the library and interpolator are presented in villaume et al . finally , the elemental response functions have been revised , using updated atomic and molecular line data . the models have 39 free parameters , including 18 individual elemental abundances and several nuisance parameters . details are given in conroy et al . ( 2016 ) . among several `` data '' parameters is a multiplicative factor that is applied to all the errors and a scaling factor that is applied to a residual telluric absorption spectrum . following our previous papers the imf has two free parameters , @xmath73 and @xmath74 , which are the logarithmic slopes of the imf in the mass ranges @xmath75 and @xmath76 respectively . the imf is assumed to have the salpeter ( 1955 ) slope of 2.35 at @xmath77 . in conroy et al . ( 2016 ) we analyze more complex forms of the imf . the models are fit using a markov chain monte carlo algorithm ( foreman - mackey et al . 2013 ) , after broadening them to the ( wavelength - dependent ) instrumental resolution ( see 2.3.2 ) . the best - fitting models are shown in red in figs.[full_spectra1.fig ] [ full_spectra3.fig ] , after shifting them to the rest - frame and smoothing to a resolution of @xmath78 . note that the fits were performed on the original spectra , not on the smoothed spectra . the fits are generally excellent , but there are systematic differences between the models and the data that exceed the expected photon noise . this is demonstrated in fig . [ res.fig ] , which shows the ratio between the data and the best - fitting model , for all six galaxies and all radial bins . the residuals are not consistent with random noise but are highly correlated , with the residuals from each of the 92 spectra showing the same rest - frame wavelength dependence . the green line shows the median as a function of wavelength . the rms of these systematic residuals at a resolution of @xmath79 is 0.50% from 4000 5000 , 0.23% from 5000 6300 , 0.22% from 8000 8900 , and 0.20% from 9650 10070 . we also show the residual after subtracting the green line . there are virtually no features in these `` residuals of the residuals '' , which means that we model the _ variation _ of the spectra ( from galaxy to galaxy and as a function of radius ) extremely well . qualitatively similar behavior was seen in conroy & van dokkum ( 2012b ) for individual early - type galaxies and in conroy , graves , & van dokkum ( 2014 ) for sdss stacks . the residuals are probably not due to problems in the data , as the six galaxies have different radial velocities : if the systematic residuals were related to the sky subtraction , the telluric absorption correction , or the response function they would line up in the observed frame , not the rest - frame . it is also unlikely that they are caused by errors in the line profiles . although the assumption of gaussian profiles is a simplification ( see , e.g. , van der marel & franx 1993 ) , there is no correlation between the strength of absorption features and the amplitude of the residuals . the most likely cause is deficiencies in the stellar population synthesis models at the @xmath80% level . the strongest residuals are at @xmath81 and @xmath82 . these are probably related to tio ; note that the feature at @xmath81 is redward of the doublet and coincides with a tio band head at @xmath83 . the interplay between this tio feature and nai is demonstrated explicitly in the inset of fig . 9 in van dokkum & conroy ( 2012 ) . it is remarkable , and important , that the residuals do not seem to correlate with stellar population parameters . as shown in [ trends.sec ] the galaxies span a wide range in stellar abundance ratios and imf ( although not in age ) , and yet the residuals are nearly identical for all 92 spectra . the residuals in the present study do not correlate very well with those in conroy & van dokkum ( 2012b ) or conroy et al . for example , the main deficiency in the red in conroy & van dokkum ( 2012b ) was that the models underpredicted the strength of the calcium triplet lines , whereas in the present study the main residuals are at the locations of tio bandheads . this is perhaps not surprising given the many changes to the stellar population synthesis models since our previous papers . in this paper we are primarily concerned with the _ variation _ in the spectra as a function of radius . in fig . [ radvar.fig ] we show the ratio of the central spectrum to that at @xmath84 , that is , the ratio of the two spectra shown for each galaxy in figs . [ full_spectra1.fig ] [ full_spectra3.fig ] . these ratio spectra show which spectral features have strong radial gradients and which are approximately constant . expressed as a ratio , shows the largest variation of all individual spectral features . its equivalent width in the ratio spectrum ranges from 0.8 for ngc1600 to 2.3 for ngc2695 . nad is a well - known interstellar medium ( ism ) line , but its increase toward small radii is almost certainly stellar in origin . in hst images ngc1600 has no visible dust absorption in its central regions ( van dokkum & franx 1995 ) , and neither has ngc1407 . furthermore , the decrease of nad with radius is gradual rather than abrupt , and there is no obvious kinematic difference between this line and the other absorption lines ( see schwartz & martin 2004 ) . the doublet also varies strongly with radius , as do many other spectral features , particularly in the blue . the feature and the calcium triplet lines do _ not _ show much variation . the band changes significantly with radius in ngc1407 but only weakly in ngc2695 . it is striking how complex the ratio spectra are , and how different from the actual spectra . this illustrates the difficulty of interpreting spectral index measurements and the power of full spectrum fitting . for each galaxy , the red line shows the ratio of the best - fitting models . the models capture the variation in the data very well . they not only reproduce the changes ( and lack of changes ) in strong features such as mg and nad , but also the detailed behavior of the spectra on all scales . we now turn to the measured values for the kinematics , stellar abundance ratios , ages , and mass - to - light ( @xmath85 ) ratios of the galaxies , as a function of radius . the model has 39 free parameters ; 11 of these are shown in fig . [ starpops.fig ] . from left to right and top to bottom , the panels show the velocity dispersion ( taking instrumental broadening and the model resolution into account ) ; the degree of rotational support ; the age ; the iron abundance ; the abundances of mg , o , c , ca , and na with respect to iron ; the @xmath86 ratio for a kroupa ( 2001 ) imf ; the @xmath86 ratio for the best - fit imf ; and the ratio of these @xmath85 ratios @xmath8 . errorbars indicate the 16@xmath32 and 84@xmath32 percentiles of the posterior probabilities . the galaxies span a range of a factor of two in central dispersion , from 163 to 340 . five of the six galaxies are slow rotators ; the exception is ngc2695 , which is an s0 galaxy with a prominent disk . the ages are uniformly high : all galaxies are older than 10gyr at all measured radii . a full analysis of the metal line gradients is beyond the scope of the present paper . briefly , the gradients are very similar from one galaxy to the next , and broadly consistent with previous work ( e.g. , trager et al . 2000 ; kuntschner et al . 2010 ; greene et al . the iron abundance decreases with radius , from @xmath87}\approx 0.1 $ ] in the center to @xmath87}\approx -0.3 $ ] at @xmath67 . the [ mg / fe ] and [ o / fe ] ratios increase with radius . the increase with radius in the @xmath8-elements is somewhat more pronounced than previously found for ngc1407 ( spolaor et al . 2008b ) ; we do not have an explanation for this possible tension . the ca abundance tracks fe , as was also seen in previous studies ( e.g. , saglia et al . 2002 ; graves et al . 2007 ; conroy et al . the [ na / fe ] ratio shows a steep negative gradient , again consistent with previous work ( boroson & thompson 1991 , and many other studies ) . as [ fe / h ] also decreases with radius , the gradient in the sodium abundance [ na / h ] is even steeper ; it decreases from @xmath88 in the center to @xmath89 at @xmath90 . this is important as the doublet is a key imf diagnostic . we will return to this in [ imftest.sec ] . the last three panels of fig . [ starpops.fig ] show the key result of this study . panel @xmath91 shows the @xmath85 ratio as a function of radius when assuming a kroupa ( 2001 ) imf . the @xmath85 ratio gradually decreases with radius , from @xmath92m@xmath93l@xmath94 in the center to @xmath95m@xmath93l@xmath94 at @xmath67 , again consistent with many previous studies ( e.g. , tortora et al . panel @xmath96 shows the @xmath85 ratio when allowing the imf to vary . for all six galaxies , the central @xmath85 ratio is much higher for a varying imf than for a kroupa ( 2001 ) imf . that is , all galaxies prefer a bottom - heavy imf that is steeper than that of the milky way at low masses . panel @xmath97 shows the imf `` mismatch '' parameter @xmath8 , defined as @xmath98 . the imf parameter falls off steeply , from @xmath99 at @xmath64 to @xmath100 at @xmath101 . the @xmath8 gradients for the individual galaxies are shown in fig . [ imf_ind.fig ] . only measurements with average uncertainties @xmath102 are shown . the @xmath103axis is logarithmic , to show the individual points more clearly . for the purpose of this figure the central aperture ( @xmath64 ) was placed at @xmath104 . at most radii there are two measurements , one for each side of the galaxy . it is reassuring that the measurements on each side of the galaxy are generally consistent within the uncertainties . the six galaxies show the same trends : the imf is bottom - heavy in the center , and gradually becomes more bottom - light . the last - measured point is consistent with the imf of the milky way for all galaxies . we parameterize the imf variation in the following way . we fit a model of the form @xmath105 at small radii , reaching a plateau of @xmath106 when @xmath107 . gradients in early - type galaxies are usually expressed in @xmath108 rather than @xmath63 , but the advantage of a simple linear function is that it does not diverge at @xmath64 . using all data points for the six galaxies we find @xmath109 this fit is shown in fig . [ alpha_fit.fig ] . it is remarkable that the best - fitting value for @xmath110 is only 10% larger than the milky way imf ; there is no known aspect of our modeling that prefers the kroupa ( 2001 ) or chabrier ( 2003 ) imf over other forms . the rms of the residuals of the fit is 0.41 , a factor of 1.3 higher than the expected scatter from the formal errors . this is probably due to a combination of systematic errors in the models and galaxy - to - galaxy variation in the imf ( see [ imftest.sec ] ) . the transition to the plateau value of @xmath111 occurs at @xmath112 . based on these six galaxies , we conclude that bottom - heavy imfs are a phenomenon that is unique to the centers of massive galaxies , on physical scales of @xmath31kpc . our results are qualitatively consistent with several other recent studies ( e.g. , martn - navarro et al . 2015a ; la barbera et al . 2016 ) , but not with all . mcconnell et al . ( 2016 ) and zieleniewski et al . ( 2016 ) suggest that observed gradients in imf - sensitive features can be fully explained by abundance variations rather than by a change in the imf . we note that there is no direct conflict , as we have not studied the imf gradients of the same galaxies . also , as discussed in conroy & van dokkum ( 2012a ) and la barbera et al . ( 2016 ) , a gradient in the imf is not necessarily accompanied by observed gradients in imf - sensitive features ; even the band can be nearly constant with radius due to compensating changes in [ fe / h ] and other effects . nevertheless , it is a valid question whether the subtle effects of a changing initial mass function can really be reliably detected given the ( sometimes dramatic ) changes in the stellar abundances . the high s / n spectra presented in this paper offer the opportunity for a `` semi - empirical '' demonstration that we can separate imf effects from abundance effects . we compared the derived stellar population parameters for all 92 spectra and looked for pairs of spectra where the abundance pattern and age are a close match but the imf is very different . this analysis is similar in spirit to the comparison between massive elliptical galaxies in virgo and m31 globular clusters that we did in 2011 ( van dokkum & conroy 2011 ) . using the criteria @xmath113gyr , @xmath114}<0.1 $ ] , @xmath115}<0.1 $ ] , @xmath116}<0.1 $ ] , @xmath117}<0.1 $ ] , @xmath118}<0.1 $ ] , and @xmath119 we found a single match : the central 3-pixel aperture of ngc1407 and a slightly larger central aperture ( of 5 pixels ) of ngc2695 . the age is 13.3gyr for the ngc1407 spectrum and 13.7gyr for the ngc2695 spectrum . the abundance ratios of eight elements are shown in fig . [ index_comp.fig ] . they are very similar , although it should be noted that the differences in [ x / fe ] are somewhat smaller than the differences in [ x / h ] , given the 0.05dex offset in [ fe / h ] . the best - fitting imf parameters are ( by selection ) quite different , with @xmath120 for ngc1407 and @xmath121 for ngc2695 . the two spectra are shown in the top panels of fig.[matched_specs.fig ] ( black and grey ) . they are , as expected , very similar . the best - fitting models are overplotted in red and blue . rows 2 , 3 , 4 and 5 of fig . [ matched_specs.fig ] all show the ratio of the two spectra in black , with the errors in grey . the ratio spectrum is flat to @xmath122% over most of the wavelength range . the two most prominent exceptions are the doublet and the doublet , which are both stronger in ngc1407 than in ngc2695 . there are also systematic differences at the 0.5% level at the location of the @xmath123 ca triplet line , near the band , and in the blue near 4600 . } < 0.1 $ ] for all measured elements . the age difference is also very small at @xmath124gyr . however , the derived imfs are quite different : @xmath125 for the ngc1407 spectrum and @xmath126 for the ngc2695 spectrum . ] we created ad hoc models for ngc1407 and ngc2695 to investigate whether we can isolate the imf features in the ratio spectrum . first , we made two models that differ only in their imf . that is , the model for ngc1407 has @xmath127 and the model for ngc2695 has @xmath126 but in all other respects they are identical , with all model parameters except the imf set to the averages of the two galaxies . the red line in the second row of fig . [ matched_specs.fig ] shows the ratio of these two models . these imf - only models do well in the red , fitting the nai doublet , the ca triplet region , and the feh band reasonably well . the residuals in the blue are not well matched by this model ; this is expected as the blue spectrum is not very sensitive to variation in the contribution from low mass stars . more importantly , imf - only models fail to account for the large difference in nad between the two galaxies : only @xmath128 of its depth in the ratio spectrum can be accounted for by the imf . although the spectra were selected to have _ similar _ abundance ratios , they are of course not _ identical_. of particular relevance is that the sodium abundance of ngc1407 is slightly higher than that of ngc2695 . ngc1407 has @xmath129}=0.47 $ ] and ngc2695 has @xmath129}=0.44 $ ] . taking the [ fe / h ] ratios into account , we have @xmath130}=0.57 $ ] for ngc1407 and @xmath130}=0.49 $ ] for ngc2695 , a difference of 0.07dex . in the third row of panels of fig . [ matched_specs.fig ] we show the ratio of two models that only differ in their sodium abundance . the model for ngc1407 has @xmath130}=0.57 $ ] and the model for ngc2695 has @xmath130}=0.49 $ ] , and all other model parameters , _ including the imf _ , are identical and set to the average of the best - fitting models of the two galaxies . the red line fits the nad doublet well , but underpredicts the dwarf - sensitive nai doublet by a factor of @xmath36 . this sodium - only model also fails to fit the feh region of the ratio spectrum . the blue line in these panels reflects an attempt to fit the nai doublet by artificially increasing the na abundance . the response is approximately linear , and we fit a model ratio spectrum of the form @xmath131 , with @xmath132 the ratio spectrum for the actual na abundances of the two galaxies and @xmath133 a free parameter . the blue line has @xmath134 , and it provides a good fit to , by construction . however , the fit to is catastrophic ; unsurprisingly , given that the red line provided a good fit to this feature , it overpredicts nad by a factor of @xmath133 . we conclude that the difference in the observed strength of the nai line between the two galaxies can not be attributed to a difference in sodium abundance . spectral coverage of the nad doublet is critical : both nad and nai are senstive to the imf and the na abundance , but nad is _ mainly _ sensitive to the na abundance and nai is _ mainly _ sensitive to the imf ( as shown in the second and third row of fig . [ matched_specs.fig ] ) . we note that any ism contribution to nad would imply that the imf is even more bottom - heavy than what we infer in this paper : it would mean that the na abundance is lower , which in turn implies that more low mass stars are needed to reproduce the observed strength of . the models in the fourth row of panels show explicitly that the combination of a varying imf and a small difference in the na abundance can reproduce both nai and nad , as well as the feh region . we now created a model for each galaxy that has the best - fit imf and the best - fit na abundance for that galaxy , with all other model parameters set to the averages for the two galaxies . the ratio of these models fits both na lines , as well as the overall shape of the spectrum in the red . finally , in the bottom row of panels we show the ratio of the full models , with the parameters for each galaxy set to the best fitting values for that galaxy . the model ratio spectrum is an excellent fit throughout the spectral range . it is now clear that the residuals in the blue are not noise but due to the small differences between the galaxies in [ fe / h ] , [ ti / fe ] , and the other model parameters . the rms residuals are 0.19% from 4500 6200 , 0.13% from 8000 8900 , and 0.11% from 9650 10070 . throughout this paper we have analyzed the imf as a function of @xmath135 , the absolute distance from the center of the galaxy in units of the half - light radius . this is the most straightforward choice given that we extract spectra as a function of radius and the galaxies all have different sizes . however , as we have not only measured the local imf but also the local velocity dispersion , age , and elemental abundances we can ask whether any of these parameters correlates better with the imf mismatch parameter than @xmath136 does ( see also , e.g. , la barbera et al . 2016 ) . in fig . [ control.fig ] we show the imf mismatch parameter @xmath8 as a function of six different parameters . the first is @xmath136 , for reference . the red curve is eq . [ imf.eq ] ; the residuals from this fit have an rms scatter of @xmath137 . the grey line is a fit of the form @xmath138 . this fit has a higher scatter of 0.48 , but it enables us to compare the predictive power of @xmath136 to the other parameters . the uncertainty in @xmath139 is 0.04 , as derived from the formal error bars combined with the sample size . the top right panel shows the relation between @xmath8 and physical radius . the scatter is similar to the relation between @xmath8 and @xmath136 . as a function of @xmath136 , radius in kpc , velocity dispersion , [ mg / fe ] , [ fe / h ] , and [ na / h ] . the red line is eq . [ imf.eq ] . grey lines are simple powerlaw fits . the scatter in the residuals from these fits is indicated in each panel . the metallicity predicts the local imf better than the velocity dispersion does . note that the local [ mg / fe ] ratio is anti - correlated with the local imf in this sample . ] the second row of fig . [ control.fig ] shows correlations with velocity dispersion and [ fe / h ] . the central velocity dispersion has often been found to correlate with the central imf ; examples are treu et al . ( 2010 ) , conroy & van dokkum ( 2012b ) , cappellari et al . ( 2013 ) , and la barbera et al . ( 2015 ) . here we find , somewhat surprisingly , that the local dispersion is not a very good predictor of the local imf . the correlation with velocity dispersion has much larger scatter than the correlation with radius . the correlation with [ fe / h ] , on the other hand , is remarkably tight , with a scatter of 0.41 . in the bottom panels we show relations with [ mg / fe ] and [ na / fe ] . in conroy & van dokkum ( 2012b ) it was found that the [ mg / fe ] ratio correlates strongly with @xmath8 , and we suggested that the star formation time scale might be a key driver of the form of the imf . here we find an _ anti-_correlation between @xmath8 and [ mg / fe ] . and the central value of @xmath8 . ] this reflects the mild increase of [ mg / fe ] with radius in fig . [ starpops.fig ] . we note that the existence of a strong positive correlation between @xmath8 and [ mg / fe ] had been called into question by smith ( 2014 ) and by la barbera et al . the [ na / fe ] ratio also shows large scatter ; in fact , all relations between @xmath8 and [ x / fe ] show much larger scatter than the relations between @xmath8 and [ x / h ] . this suggests that the overall metallicity , rather than a specific element , is coupled with the form of the imf ( see also martn - navarro et al . 2015a ) . finally , we caution that these results may partially be driven by scatter between the six galaxies in our small sample ; as an example , the variation in velocity dispersion between the galaxies is larger than the typical variation within each individual galaxy . larger samples are needed to explore these issues further . an implication of the existence of imf gradients is that the measured imf depends on the aperture that is used in the analysis . as noted in 1 this complicates comparisons between different techniques , such as gravitational lensing and stellar population synthesis modeling . using the form of the gradient given in eq . [ imf.eq ] we calculate the average luminosity - weighted imf within apertures of different radii . the galaxies span a range in sersic index ( spolaor et al . 2008a ; krajnovi et al . 2013 ) , and we calculate the average imf for @xmath140 , @xmath141 , and @xmath142 . the results are shown in fig.[imf_int.fig ] . within an aperture of radius @xmath63 , calculated from eq . [ imf.eq ] . the solid red curve is for a sersic ( 1968 ) index @xmath141 and the broken curves are for @xmath140 ( dotted ) and @xmath142 ( dashed ) . large open symbols are determined from published relations between @xmath143 and @xmath144 , using the average dispersion of the six galaxies in our sample . the square is derived from strong gravitational lenses ( treu et al . the circle is from dynamical modeling of nearby early - type galaxies ( cappellari et al . the triangle is from ( posacki et al . 2015 ) , who combined atlas@xmath7 and slacs data . small open symbols are for a single massive galaxy , and are taken from la barbera et al . overlapping points are offset slightly in @xmath63 , for clarity . all results are in reasonable agreement , given the differences in methods . ] the average luminosity - weighted imf parameter is @xmath145 ( i.e. , heavier than the salpeter form ) only for apertures @xmath146 . within @xmath147 we find @xmath6 , depending on the form of the surface brightness profile . our results are consistent with those of la barbera et al . ( 2016 ) , who analyzed a single massive galaxy in a similar way ( small circles in fig . [ imf_int.fig ] ) . we conclude that the imf in apertures that contain most of the light is lighter than the salpeter form , and only mildly heavier than a kroupa ( 2001 ) or chabrier ( 2003 ) imf . this aperture - dependence may explain some of the discrepancies in imf studies of massive galaxies in the literature . in particular , dynamical and lensing studies typically find values for @xmath8 that are in between the kroupa ( 2001 ) and the salpeter ( 1955 ) forms ( @xmath148 ) , whereas stellar population studies in much smaller apertures have found @xmath149 and even higher ( e.g. , conroy & van dokkum 2012b ; conroy et al . we can test whether our results are consistent with dynamical and lensing studies . first , we determine the luminosity - weighted velocity dispersion as a function of aperture radius . for each galaxy the observed relation between @xmath150 and @xmath151 is fitted with a powerlaw and extrapolated to @xmath5 . the luminosity - weighted dispersion is then obtained by integrating these fits , weighted by an @xmath152 law . the mean luminosity - weighted dispersion of the six galaxies is @xmath153 within @xmath11 ( and very similar within @xmath154 ) . next , we use published relations between @xmath155 and @xmath156 to compare our results to lensing and dynamical studies . the open square in fig . [ imf_int.fig ] is for the slacs sample of strong lenses in treu et al . ( 2010 ) ; the einstein radius of these objects is @xmath157 . the circle is from a dynamical analysis of the atlas@xmath7 sample of nearby early - type galaxies ( cappellari et al . the triangle is from a recent joint analysis of the slacs and atlas@xmath7 samples ( posacki et al . the errorbars reflect only the formal uncertainty in the average luminosity - weighted velocity dispersion , and neglect all other random and systematic errors . nevertheless , the dynamical and lensing results are in excellent agreement with the gradients that we derive in this paper . four galaxies in our sample are part of the atlas@xmath7 survey , enabling a direct comparison . for the galaxies ngc2695 , ngc3414 , ngc4552 , and ngc4564 cappellari et al . ( 2013 ) find @xmath158 with very little scatter . this is , again , in very good agreement with the red curve in fig . [ imf_int.fig ] . if , instead , we had compared the atlas@xmath7 values to our results for the centers of the galaxies , we would have concluded that the studies disagree with one another ( see also lyubenova et al . using a new suite of stellar population synthesis models and high quality data for six galaxies we find that the stellar imf in massive galaxies is a strong function of radius . the imf is bottom - heavy in the centers of the galaxies and reaches the approximate milky way form at @xmath159 . this result is consistent with several recent studies that used different models and were based on index measurements rather than on full spectrum fitting ( see martn - navarro et al . 2015a ; la barbera et al . it is also consistent with earlier work on the centers of the most massive galaxies ( van dokkum & conroy 2010 ; conroy & van dokkum 2012b ) . given all the changes in our models and the vast improvements in data quality it is remarkable that we reached similar conclusions in 2010 as we do today . the reason is probably that we began by targeting the most extreme galaxies , which are least sensitive to the details of the modeling . the agreement is not universal , however . mcconnell et al . ( 2016 ) use very similar data , but conclude that the observed radial trends in two early - type galaxies can be explained entirely by abundance gradients . it may be that those two galaxies happen to have milky way imfs throughout , but that is not a very satisfying explanation . modeling of the mcconnell et al . ( 2016 ) data using similar techniques as employed here may shed more light on this . abundance gradients are certainly a concern in all these studies , as they can mimic imf effects . it is well known that the doublet is sensitive to the na abundance , and as discussed by conroy & van dokkum ( 2012a ) and la barbera et al . ( 2016 ) the band depends on [ fe / h ] and , in certain regimes , on age and [ @xmath8/fe ] . the upshot is that the two strongest gravity - sensitive features in the optical window are both difficult , or even impossible , to interpret in isolation . allowing non - solar abundance ratios is critical , particularly for na . furthermore , only full spectrum fitting , or a carefully chosen combination of different indices ( see conroy & van dokkum 2012a ) , can isolate imf effects . as the modeling gets more complex it becomes more difficult to identify the key data that produce a particular result ; the high degree of sophistication in this field carries the risk that the modeling begins to resemble a black box . some of these issues are addressed in conroy et al . ( 2016 ) , where we use the exquisite ngc1407 data to analyze what specific stellar mass ranges are constrained by the data . in the somewhat stand - alone 5.1 we address this issue in a different way , analyzing the ratio of two spectra with similar elemental abundances but a different imf . this `` semi - empirical '' approach follows earlier work on m31 globular clusters and massive ellipticals ( van dokkum & conroy 2011 ) . in the case of ngc1407 and ngc2695 we can point to specific features that drive the outcome , and the analysis also provides understanding of what spectral features constrain the imf in the rest of the sample . it is fortuitous that we have two such spectra in our small sample ; future studies could specifically single out such matched pairs for deep follow - up spectroscopy to increase the sample . in [ imf_int.sec ] we show that the luminosity weighted imf within an aperture depends strongly on the aperture size . lensing and dynamical studies typically have an effective aperture of @xmath160 , and we demonstrate in fig . [ imf_int.fig ] that the gradients we measure are entirely consistent with the slacs and atlas@xmath7 constraints for galaxies with similar velocity dispersions . these results are consistent with the recent study of lyubenova et al . ( 2016 ) , who found that there is good agreement between dynamical and stellar population measurements of @xmath8 when measured in a self - consistent way and in the same aperture . we emphasize here that our entire analysis applies to massive galaxies only , with @xmath161 . as even the central imf is milky way - like for less massive galaxies ( conroy & van dokkum 2012b ) , they probably have much weaker gradients than the galaxies studied here ( see martn - navarro et al . ( 2015a ) and spiniello et al . ( 2015b ) for studies of imf gradients in two such galaxies ) . the fact that the galaxy - integrated imf ( to @xmath162 ) has @xmath163 for our massive galaxy sample may mean that the galaxy - wide imf is not very different from that of the milky way for nearly all galaxies in the universe . possible exceptions are the most compact massive galaxies ( martn - navarro et al . 2015b ; conroy et al . 2013 ) , and ultra - faint dwarf galaxies which may have a bottom - light imf ( geha et al . 2013 ) . the strong gradients that we find have implications for various areas of astrophysics , and we touch on a few here . first , the large gradient in the @xmath85 ratio ( see panel @xmath96 of fig . [ starpops.fig ] ) needs to be taken into account when measuring the masses of supermassive black holes . as a case in point , thomas et al . ( 2016 ) use stellar kinematics to infer the existence of a black hole in ngc1600 with a mass of @xmath164 , among the most massive black holes ever found . the sphere of influence of this black hole , defined as the radius within which the black hole mass equals the stellar mass , is @xmath165kpc , or @xmath166 . within this radius we find @xmath167 for ngc1600 , which means that the dynamics are fully consistent with the imf that we derive and no `` ultra - massive '' black hole may be required . more detailed modeling of the central kinematics , using the stellar @xmath85 constraints derived here , is required to understand the full effect on derived black hole masses . another implication is that the stars that are in the centers of massive galaxies today formed in a very different way than the stars that are at large radii . as discussed in 1 there is fairly good evidence that the central @xmath168kpc are indeed unique environments , in that they assembled in a short period of intense star formation at high redshift . the outer regions were likely accreted at later times , and could therefore have abundance ratios and imfs that more closely resemble low mass satellite galaxies than the galaxy centers . turning this around , the presence of imf gradients could be viewed as evidence for the kind of two - phase formation models that were proposed by oser et al . ( 2010 ) and others . in the context of these models , and studies such as fang et al . ( 2013 ) , van dokkum et al . ( 2014 ) , and martn - navarro et al . ( 2015b ) , one might expect that the imf correlates better with the physical radius than with @xmath136 . both parameters correlate equally well in our small sample ( see 5.2 ) . our results do strongly suggest that the imf in compact , massive galaxies at @xmath9 which are thought to be the progenitors of the cores of today s massive galaxies should be heavier than that of the milky way . although galaxies obviously evolve over this timeframe , both mixing due to mergers ( sonnenfeld , nipoti , & treu 2016 ) and projection effects are expected to decrease , not increase , the observed central value of @xmath8 over time . there may be some tension with observations : the dynamics of both star forming and quiescent massive compact galaxies seem to point to relatively low @xmath85 ratios ( belli , newman , & ellis 2013 ; van de sande et al . 2013 ; van dokkum et al . the uncertainties are currently too large to place strong constraints , but this should change with improved measurements of high redshift galaxies in the jwst era . it may turn out that the high redshift galaxies have gradients too , and/or that we are misinterpreting their galaxy - integrated kinematics ( as may be indicated by recent observations ; newman et al . 2015 ; belli et al . it will also be important to have better calibrated stellar @xmath85 ratios at young ages ( see dutton et al . 2012 ) . finally , our results have implications for star formation theory . probably the most important result in this context is the finding that the imf can be even heavier than the salpeter ( 1955 ) form , something that had been suggested in recent studies ( e.g. , chabrier , hennebelle , & charlot 2014 ) . these implications are discussed in paper iv in this series ( conroy et al . 2016 ) . we thank nacho martn - navarro for comments on the manuscript . support from nasa grant nnx15ak14 g , stsci grant go-13681 , and nsf grants ast-1313280 , ast-1518294 , and ast-1616598 is gratefully acknowledged . c. c. acknowledges support from the packard foundation . a. v. is supported by an nsf graduate research fellowship . based on data obtained with the w. m. keck observatory , on maunakea , hawaii . the authors wish to recognize and acknowledge the very significant cultural role and reverence that the summit of maunakea has always had within the indigenous hawaiian community . we are most fortunate to have the opportunity to conduct observations from this mountain . , c. , stover , r. , kibrick , r. , lockwood , c. , peck , m. , cowley , d. , bolte , m. , adkins , s. , et al . 2010 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol . 7735 , society of photo - optical instrumentation engineers ( spie ) conference series

we construct the cubic interaction vertex and dynamically generated supercharges in light - cone superstring field theory in the pp - wave background . we show that these satisfy the pp - wave superalgebra at first order in string coupling . the cubic interaction vertex and dynamical supercharges presented here differ from the expressions previously given in the literature . using this vertex we compute various string theory three - point functions and comment on their relation to gauge theory in the bmn limit . hep - th/0210246 + aei-2002 - 086 + spin-2002/33 + itp-2002/54 0.5 cm 1.0 cm a. pankiewicz and b. stefaski , jr . 0.5 cm _ max - planck - institut fr gravitationsphysik , albert - einstein institut + am mhlenberg 1 , d-14476 golm , germany _ 0.5 cm _ spinoza institute , university of utrecht + postbus 80.195 , 3508 td utrecht , the netherlands _ 1.0 cm october 2002 1.0 cm

hep-th0210246

following earlier work on maximally supersymmetric solutions of eleven - dimensional supergravity @xcite , a new maximally supersymmetric solution of type iib supergravity @xcite was found . this paved the way for quantisation in light - cone gauge of superstring theory in a constant flux @xcite . the spectrum consists of a unique massless groundstate on which a tower of massive states is constructed @xcite using creation operators whose masses are of order @xmath2 where @xmath3 @xmath4 is momentum along the @xmath5 light - cone direction and @xmath6 is the flux . this new string background merits further investigation . since the string spectrum in the pp - wave background is now known , string interactions are the next step in the study of the pp - wave background . in this background ten - dimensional lorentz invariance is broken by the non - zero flux , and hence it is no longer possible to set @xmath4 to zero in general scattering amplitudes . this obstruction significantly hinders the vertex operator approach to string interactions . there is only one other known way of studying string interactions in light - cone gauge pioneered by mandelstam @xcite for the bosonic string . in this approach an interaction vertex for the scattering of three strings can be constructed by requiring continuity of string fields on the worldsheet depicted in figure [ fig1 ] . this continuity is enforced by a delta functional on string fields @xmath7 for computational purposes it is essential to express the delta functional in fourier modes @xcite . the interaction vertex may then be written as an exponential of creation operators which enforce the delta functional conditions mode by mode . the functional approach @xcite can be extended to the superstring @xcite . the cubic interaction vertex is a first order in string coupling @xmath8 correction to the free hamiltonian . in order to satisfy the superalgebra with this new hamiltonian , dynamical supercharges also receive corrections . this essential difference from the bosonic string modifies the form of the vertex @xcite . in particular @xcite , in the oscillator basis , the superstring interaction vertex as well as the dynamical supercharges not only have a part exponential in creation operators , but also have a prefactor which is polynomial in creation operators . the exact form of the prefactors is determined by two requirements . firstly , they should not destroy worldsheet continuity enforced by the exponential part of the vertex . secondly , the superalgebra should be satisfied order by order in string coupling . in @xcite an oscillator expression for the flat spacetime superstring vertex and dynamical supercharges was constructed and shown to satisfy all said consistency conditions . in this paper we apply the formalism of @xcite to the pp - wave background in order to construct an interaction vertex and dynamical supercharges in the oscillator basis . we prove explicitly that our expressions satisfy all the aforementioned consistency conditions . we use this vertex to compute various three - point functions in string theory . the pp - wave interaction vertex has also been recently investigated in @xcite . our results differ from the expressions presented in @xcite in several ways . firstly , our bosonic prefactor contains a piece , not present in @xcite , proportional to @xmath9 ( _ cf_. equation ) . this piece is crucial in proving ( see section [ sec4 ] ) that the dynamically generated supercharges and interaction hamiltonian satisfy the pp - wave superalgebra at first order in string coupling . further , without it , three - point functions of states containing only fermionic creation operators would vanish ( see section [ sec52 ] ) . secondly , we also present the oscillator - basis expression for the fermionic prefactor ; in particular , the @xmath10-dependent normalisation ( _ cf_. equation ) is essential in proving that the dynamical superalgebra is satisfied . the exponential part of our vertex also contains a piece linear in fermionic zero - modes @xcite ( _ cf_. equation ) and our expression for the coefficients of the negatively moded bosonic creation operators in the prefactor ( _ cf_. equation ) differs by a factor of @xmath11 @xcite from the original one in @xcite . discrepancy . ] since we prove that our expressions for the dynamical supercharges and hamiltonian satisfy the superalgebra , we believe that the vertex constructed here is the correct one . the pp - wave background can be obtained as a penrose limit @xcite of @xmath12 @xcite . through the ads / cft correspondence @xcite , this has led to a conjectured duality between string theory in the pp - wave background and a double scaling limit of su(@xmath13 ) , @xmath14 super yang - mills theory @xcite . explicitly one considers a sector of ( bmn ) operators with @xmath15 r - charge @xmath16 and conformal dimension @xmath17 such that the difference @xmath18 remains fixed in the limit @xmath16 , @xmath19 , @xmath20 and @xmath21 fixed @xcite . the bmn operators are field theory duals of perturbative string states in the pp - wave background . the sym theory is believed to be expandable in a double series in the effective genus counting parameter @xmath22 and effective coupling @xmath23 @xcite @xmath24 moreover , the finite quantity @xmath18 is a function of @xmath23 @xcite and @xmath22 @xcite . the expansion parameters @xmath25 and @xmath23 are related to string theory parameters by @xcite @xmath26 one of the most appealing aspects of this duality is that both string theory and the bmn sector of gauge theory are simultaneously perturbatively accessible . while some further investigation seems necessary to clarify the exact form of this correspondence , at the planar ( interacting ) field theory two - point function / free string level , there is increasing evidence @xcite that the bmn operators do correspond to the free string theory . the extension of the duality beyond this limit , by considering string interactions and the non - planar sector of ( interacting ) gauge theory respectively , needs further clarification . indeed , the exact notion of the penrose limit in a cft needs to be better understood @xcite . by focusing on a certain subclass of the bmn operators , several proposals for such an extension of the bmn correspondence have been made and were studied in @xcite . in this paper we use our interaction vertex to compute three - point functions for various string states . in particular , for the class of amplitudes studied so far in the literature the term in the cubic vertex proportional to @xmath27 does not contribute and we recover the results of @xcite . we show that so - called impurity preserving three - point functions of states dual to a class of protected operators vanish in string theory , as would be expected by the extension of the duality recently proposed in @xcite . further , we compute three - point functions for string states having only fermionic creation operators . here only the @xmath27 term contributes and these amplitudes may provide a check of any extension of the string - bit dynamical supercharges to terms bilinear in fermions , which are not known at present @xcite . we hope our results can shed some light on the bmn correspondence and that the three - string vertex can be used to investigate the duality further . finally if the bmn duality is to be believed , d - branes in the pp - wave background @xcite should play an important role in gauge theory . this paper is organized as follows . in section [ sec2 ] we briefly review the free string in the pp - wave background and set our notation . in sections [ sec3 ] and [ sec4 ] we construct the superstring vertex in the pp - wave background . section [ sec3 ] focuses on the exponential part of the vertex . most of the results of this section have appeared in @xcite and are included here for completeness . in section [ sec4 ] we construct the prefactors and show that the superalgebra is satisfied with these modified generators . using the vertex , we compute some three - point functions and compare them to field theory in the bmn limit in section [ sec5 ] . several appendices , containing computational details are also included . the free string in the maximally supersymmetric pp - wave background in light - cone gauge is described by @xmath28 and @xmath29 in position space or by @xmath30 and @xmath31 in momentum space , where @xmath32 is a transverse so(8 ) vector index , @xmath33 is a so(8 ) spinor index . the index @xmath34 denotes the @xmath35th string ( see figure [ fig1 ] ) . the bosonic part of the light - cone action in the pp - wave background is @xcite @xmath36\,,\ ] ] where @xmath37 in a collision process @xmath38 will be positive for an incoming string and negative for an outgoing one . the mode expansions of the fields @xmath39 and @xmath40 at @xmath41 are @xmath42\ , . \end{split}\ ] ] the fourier modes can be re - expressed in terms of creation and annihilation operators as @xmath43 canonical quantization of the bosonic coordinates @xmath44=i\d^{ij}\d_{rs}\d(\s_r-\s_s)\ ] ] yields the usual commutation relations @xmath45=\d^{ij}\d_{nm}\d_{rs}\,.\ ] ] the fermionic part of the light - cone action in the pp - wave background is @xcite @xmath46\,,\ ] ] where @xmath47 is a complex , positive chirality so(8 ) spinor and @xmath48 is symmetric , traceless and squares to one .. ] the mode expansion of @xmath47 and its conjugate momentum @xmath49 at @xmath41 is @xmath50\ , . \end{split}\ ] ] the fourier modes satisfy the reality condition @xmath51 and , due to the canonical anti - commutation relations for the fermionic coordinates @xmath52 they obey the following anti - commutation rules @xmath53 it is convenient to define a new set of fermionic operators @xcite @xmath54\,,\ ] ] which break the @xmath55 symmetry to @xmath56 . here @xmath57 these modes satisfy @xmath58 the free string light - cone hamiltonian is @xmath59 in the above the zero - point energies cancel between bosons and fermions . since the hamiltonian only depends on two dimensionful quantities @xmath10 and @xmath60 , @xmath61 and @xmath38 should not be thought of as separate parameters . the vacuum @xmath62 is defined as @xmath63 the isometries of the pp - wave superalgebra are generated by @xmath64 , @xmath65 , @xmath66 , @xmath67 , @xmath68 and @xmath69 . the latter two are angular momentum generators of the transverse @xmath56 symmetry of the pp - wave background . the 32 supersymmetries split into kinematical supercharges @xmath70 , @xmath71 and dynamical ones @xmath72 , @xmath73 . the latter square to the light - cone hamiltonian and hence get corrected by interactions . a subset of the superalgebra that will be of importance throughout this paper is @xcite @xmath74 & = -i\m^2j^{+i}\,,\qquad [ h , q^+]=-\m\pi q^+\,,\\ \{q^-_{\da},\bar{q}^-_{\db}\ } & = 2\d_{\dot{a}\dot{b}}h - i\m\bigl(\g_{ij}\pi\bigr)_{\dot{a}\dot{b}}j^{ij } + i\m\bigl(\g_{i'j'}\pi\bigr)_{\dot{a}\dot{b}}j^{i'j'}\ , . \end{split}\ ] ] the dynamical supercharges can be expressed in terms of world - sheet fields as @xmath75\ ] ] and @xmath76^{\dg}$ ] . expanding in modes one finds @xmath77\l_{0(r ) } \nonumber\\ & + \sum_{n\neq 0}\sqrt{|n|}\left[a_{n(r)}p_{n(r)}^{-1}b_{n(r)}^{\dg } + e(\a_rn)a_{n(r)}^{\dg}p_{n(r)}b_{-n(r)}\right]\bigr)\,,\end{aligned}\ ] ] where @xmath78 in interacting light - cone string field theory some generators of the supersymmetry algebra , such as @xmath16 , @xmath79 and @xmath70 , are quadratic in fields . these kinematical generators correspond to symmetries which are manifest in the theory and do not receive @xmath80 corrections . on the other hand , dynamical generators such as @xmath64 , @xmath72 and @xmath73 , depend on the interactions and hence involve higher powers of fields . for example @xmath64 , the full hamiltonian of the interacting theory , has an expansion in string coupling @xmath81 where @xmath82 creates or destroys a single string , while the @xmath83 conserves the number of strings . despite such corrections to the generators , the supersymmetry algebra has to be satisfied order by order in @xmath80 . this consistency condition gives rise to two types of constraints : the kinematic constraints arise from demanding that ( anti)commutation relations of kinematical with dynamical generators are realized at order @xmath84 ; the dynamic constraints arise from the ( anti)commutation relations of dynamical generators alone . the former are relatively easy to satisfy , and it is the latter relations that severely constrain the form of the dynamical generators and will be used to determine the interaction vertex below . in this section we review the exponential part of the cubic vertex which enforces the kinematic constraints . in the following section we construct the prefactor of the interaction vertex by imposing the dynamic constraints . the bosonic contribution @xmath85 to the exponential part of the three - string interaction vertex has to satisfy the kinematic constraints @xcite @xmath86 these are the same as in flat space and arise from the commutation relations of @xmath64 with @xmath66 ( _ cf_. equation ) and @xmath67 . they guarantee momentum conservation and continuity of the string worldsheet in the interaction . the coordinates of the three strings are parameterized by @xmath87 and @xmath88 , @xmath89 . the solution to the constraints in equation can be constructed as a functional integral and is given by @xcite @xmath90 where @xmath91 is annihilated by @xmath92 , @xmath93 . the determinant factor coming from the functional integral will be cancelled by the fermionic determinant except for the zero - mode part which is @xmath94 . in the equation above the non - vanishing elements of the bosonic neumann matrices for @xmath95 , @xmath96 are @xcite @xmath97 here @xmath98 and @xmath99 where @xmath100 the matrices @xmath101 arise from the fourier transformation of the constraints in equation and are given in appendix [ appe ] . notice that in the neumann matrices the inverse of the infinite dimensional matrix @xmath102 appears . in contrast to flat space @xcite , in the pp - wave explicit expressions for the neumann matrices valid for all @xmath10 , are not known . it is however possible to obtain their large @xmath10 behavior using the approach of @xcite ( see also appendix [ appe ] ) . we also define @xmath103 an explicit expression for the vector @xmath104 is given in appendix [ appe ] . the quantities @xmath102 , @xmath105 and @xmath106 manifestly reduce to their flat space counterparts , defined in @xcite , as @xmath107 . the only non - vanishing matrix elements with negative indices are @xmath108 . they are related to @xmath109 via @xcite @xmath110 a very useful formula relating @xmath109 to @xmath111 is @xcite @xmath112_m\left[u_{(s)}^{-1}c_{(s)}^{1/2}c\bar{n}^s\right]_n\,.\end{aligned}\ ] ] this factorization theorem can be used to verify @xcite that @xmath85 satisfies equation . it will also prove essential throughout the next section . analogously to the bosonic case , the fermionic exponential part of the interaction vertex has to satisfy @xcite @xmath113 these constraints arise from the commutation relations of @xmath64 with @xmath70 and @xmath71 , _ cf_. equation . the solution is @xcite @xmath114|e_b^0\ra\,,\ ] ] where @xmath115 and @xmath116 is the pure zero - mode part of the fermionic vertex @xmath117 is not the vacuum defined to be annihilated by the @xmath118 . rather , it satisfies @xmath119 and @xmath120 . in the limit @xmath107 it coincides with the flat space state that generates the massless multiplet by acting with @xmath121 on it . the fermionic neumann matrices can be expressed in terms of the bosonic ones as @xcite @xmath122_{mn}\,,\\ \label{qm } q^r_n & = \frac{e(\a_r)}{\sqrt{|\a_r|}}(1 - 4\m\a k)^{-1}(1 - 2\m\a k(1+\pi))\bigl[p_{(r)}c_{(r)}^{1/2}c^{1/2}\bar{n}^r\bigr]_n\,.\end{aligned}\ ] ] in summary , the part of the cubic contribution to the dynamical generators satisfying the kinematic constraints is @xmath123 in the previous section we reviewed the exponential part of the vertex , which solves the kinematic constraints . the remaining dynamic constraints are much more restrictive and are solved by introducing a prefactor @xcite , polynomial in creation operators , in front of @xmath124 . and @xmath125 , defined below , where @xmath126 and @xmath125 are linear in creation operators . apart from an important factor of @xmath11 ( see discussion in the introduction ) , the oscillator expressions for @xmath127 and @xmath128 were first derived in @xcite . the corresponding expression for @xmath125 was first derived in @xcite . ] within the functional formalism , the prefactor can be re - interpreted as an insertions of local operators at the interaction point @xcite ( see also appendix [ appf ] ) . in this section we present expressions for the dynamical generators and prove that they satisfy the superalgebra up to order @xmath84 . define the linear combinations of the free supercharges ( @xmath129 ) @xmath130 which satisfy @xmath131 since @xmath68 and @xmath69 are not corrected by the interaction , it follows that at order @xmath84 the dynamical generators have to satisfy @xmath132 in order to derive equations that determine the full expressions for the dynamical generators one has to compute ( anti)commutators of @xmath133 and @xmath134 with the constituents of the prefactors . moreover , the action of the supercharges on @xmath124 has to be known in terms of these constituents . here the factorization theorem for the bosonic neumann matrices and the relation between the bosonic and fermionic neumann matrices given in equations and prove to be essential . so as not to destroy the constraints in equation , the bosonic part of the prefactors ( that we will collectively denote by @xmath135 ) has to satisfy @xcite @xmath136=0 = \bigl[\,\sum_{r=1}^3e(\a_r)x_r(\s_r),\mc{p}\bigr]\,.\ ] ] making an ansatz linear in creation oscillators and taking the fourier transform of the above equations one finds that negative and non - negative modes decouple from each other . the resulting constraint equations on the infinite component vectors appearing in the ansatz can be solved using flat space identities @xcite ; for details , see @xcite . the solution of these constraints involving non - negatively moded bosonic creation oscillators is @xcite @xmath137 where @xmath138=i\ ] ] and in terms of the zero - mode creation oscillators @xmath139 the explicit expression for @xmath140 is given in appendix [ appb ] , equation . the solution involving negatively moded creation oscillators is @xmath141 where @xmath142 the fermionic parts of the prefactor have to satisfy the conditions @xmath143 that are solved in a similar way to the bosonic case . the solution involving non - negatively moded creation oscillators is @xmath144 where the infinite component vector @xmath145 can be expressed in terms of the bosonic one @xmath140 @xcite ( see appendix [ appb ] , equation ) . as in flat space @xcite , it turns out that the prefactors do not involve negatively moded fermionic creation oscillators . below we present the results necessary to verify equations and , given the ansatz - for the cubic vertex and dynamical supercharges . we need @xmath146\,|v\ra= \sqrt{2}\bar{\eta}\sum_{r=1}^3[\wt{q}_{(r)},{\mc k}]\,|v\ra= \m\g(1+\pi){\mc y}|v\ra\,,\ ] ] where @xmath147 and @xmath148 notice , that the above identities are only valid when both sides of the equation act on @xmath124 . the action of the supercharges on @xmath124 is @xmath149 the proof of equations - is given in appendix [ appb ] . the results of the previous subsection motivate the following ansatz for the explicit form of the dynamical supercharges and the three - string interaction vertex @xmath150 substituting this into equations and and using equations - , we get the following equations for @xmath151 , @xmath152 and @xmath153 denotes symmetrization in @xmath154 , @xmath155 . ] @xmath156 which originate from terms proportional to @xmath157 and are identical to the flat space equations of @xcite . two additional equations , arising from terms proportional to @xmath158 are @xmath159^a\bigr)s^i_{\dot{b})}\,,\\ -\d_{\dot{a}\dot{b}}v^{ii } & = -\frac{i}{\sqrt{2}}\frac{\a'}{\a } \g^i_{a(\dot{a}}\bigl(\bar{d}^a - i\bigl[\pi d\bigr]^a\bigr)\tilde{s}^i_{\dot{b})}\ , . \end{split}\ ] ] as in flat space @xcite we define @xmath160 recall first the solution of the flat space equations @xcite . one introduces the following functions of @xmath161 @xmath162 here @xmath163}\,,\qquad u^i_{abc\dot{a}}\equiv -\g^{ij}_{[ab}\g^j_{c]\dot{a}}\,.\ ] ] notice that @xmath164 is traceless and symmetric in @xmath32 , @xmath16 , hence @xmath165 is a symmetric tensor of @xmath55 , whereas @xmath166 is antisymmetric . the solution of equations is @xcite @xmath167 next consider the additional equations . using the flat space solution , these can be rewritten as @xmath168^as^i_{\dot{b})}\,\qquad 0=\g^i_{a(\dot{a}}\bigl[\pi d\bigr]^a\tilde{s}^i_{\dot{b})}\,.\ ] ] we prove in appendix [ appd ] that these equations are also satisfied by equation . the proof of equation is more involved and provides an important consistency check of the ansatz . we show in appendix [ appc ] that it leads to the equations @xmath169 here @xmath170 and @xmath171}_{cd]}\,.\ ] ] the first three equations are identical to those in flat space we correct some minor typos present in @xcite . ] and were proven in @xcite . the additional equation is proved in appendix [ appd ] . we have not yet fixed the overall normalisation of the dynamical generators which can depend on @xmath10 and the @xmath60 s . this is more difficult to do than in flat space since there is no @xmath172 generator in the pp - wave background . a comparison with a supergravity calculation fixes the normalisation for small @xmath10 to be @xmath173 @xcite . however , this does not completely fix the normalisation ; one may still multiply by a function of @xmath10 and @xmath60 that goes to one as @xmath107 . on the other hand , the definition of @xmath174 in equation and the fact that the terms @xmath175 and @xmath27 in equation are multiplied by different powers of @xmath176 are fixed by demanding that one recovers the flat space equations , - . in order to obtain the supergravity expressions for the dynamical generators from equations , one should set @xmath106 to zero , since it originates from massive string modes . together with @xmath177=i\d^{ij}$ ] , one can check that the supergravity vertex obtained in this way agrees with the supergravity vertex presented in section 4 of @xcite . we would like to stress that the nontrivial relation between @xmath174 and @xmath125 in equation , as well as the part in the cubic interaction vertex proportional to @xmath27 , were not present in @xcite and played an essential role in the above proof of the superalgebra in the interacting string field theory . in @xcite , functional expressions for the constituents of the prefactors were used to argue that the vertex of @xcite satisfied the superalgebra . it is well known @xcite that these functional expressions do not in general agree with the oscillator basis expressions given in equations , and . as explained in appendix [ appf ] , it would appear that this subtlety is the origin for the absence of the @xmath27 term in @xcite . in flat space the functional expression agrees with the one in the oscillator basis only when acting directly on @xmath124 @xcite . in the pp - wave however , they agree only up to non - trivial functions of @xmath10 @xcite ( see also appendix [ appf ] ) . this leads to equation . in summary the oscillator expression , which is used in the following section to compute three - point functions , has been shown explicitly to satisfy the superalgebra at @xmath84 . in this section we use @xmath82 to compute three - string amplitudes . for three general string states @xmath178 in the number - basis representation the amplitude is @xmath179 we restrict ourselves to the case , where either all states have bosonic excitations or fermionic ones . we compare our expressions to gauge theory results in the bmn limit . recall that the full interacting hamiltonian @xmath180 in string theory is identified with the operator @xmath18 in field theory @xcite . a comparison of matrix elements of the two operators requires an identification of the free string basis used in string theory with a dual basis in field theory . at the planar level , multi - trace bmn operators are identified with multi - string states @xcite . however , non - planar corrections lead to a mixing of single and multi trace bmn operators @xcite and this effect has to be taken into account when comparing matrix elements of @xmath82 to field theory . the dual basis was proposed @xcite to be the one , in which the matrix of two - point functions of redefined bmn operators ( _ i.e. _ now mixtures of single and multi - trace ) is diagonal in free field theory . as emphasized in @xcite , the required basis transformation is not unique . to facilitate the comparison with field theory , we change our oscillator basis to the bmn basis . for @xmath96 the transformation is @xmath181 using equation we have @xmath182 where @xmath183 in this basis the neumann matrices are @xmath184 we compute amplitudes involving the string states @xmath185 are related to @xmath186 via @xmath187 eight of the zero - modes in equation , namely @xmath188 , @xmath189 and , say , @xmath190 , @xmath191 are saturated by @xmath116 , so , to give a non - zero contribution the remaining four zero - modes must be contracted with the @xmath192 term in @xmath193 . hence , the fermionic contribution is @xmath194 where @xmath195 ; the factor @xmath196 is due to equation . the bosonic zero - mode determinant can be computed using results in appendix b of @xcite @xmath197 as discussed at the end of section [ sec4 ] the overall normalisation of the cubic interaction vertex is still not completely fixed . we will take it to be @xmath198 by comparing a three - string amplitude ( _ e.g. _ equation ) in the limit @xmath199 to a known field theory result . strictly speaking , from the comparison with supergravity and field theory we only know that equation is correct for @xmath200 and @xmath199 . since the function @xmath201 plays such a prominent role in pp - wave string field theory , we conjecture that it is the correct choice for all @xmath10 . combining the various contributions we conclude that for amplitudes of string states build out of bosonic oscillators acting on the @xmath62 , the interaction vertex is . ] @xmath202 in particular , the term proportional to @xmath27 does not contribute , since @xmath203 is traceless . using the factorization theorem this can be further simplified to @xcite @xmath204 recall that @xmath205 . here @xmath206 is related to @xmath207 , the ratio of r - charges conventionally used in sym , by @xmath208 . for amplitudes involving two zero - mode oscillators and one massive string state we find of the @xmath35th string and the factor of @xmath209 . ] @xmath210_n \left(\bar{n}^3_{|n|}\right)^2\nonumber\\ \label{00n1 } & = -\m g_2\l'\sqrt{-\b(\b+1)}\frac{\sin^2n\pi\b}{2\pi^2 } \left[1-\frac{1}{4}(3 - 16a_r)\l'n^2+{\mc o}(\l^{\prime\,2})\right]\,,\\ g_{\text{s}}\la n_{1,1}|h_3|0_1\ra|0_1\ra & = -2g_2\m\a_1\a_2(-\b(\b+1))^{3/2}\left[cu_{(3)}^{-1}c_{(3)}\right]_n\left(\bar{n}^3_{|n|}\right)^2 \left[1+\frac{1}{4\m\a_3}\left[cu_{(3)}^{-1}\right]_n\right]\nonumber\\ \label{00n2 } & = -\m g_2\l'\sqrt{-\b(\b+1)}\frac{\sin^2n\pi\b}{\pi^2 } \left[1-\frac{1}{4}(7 - 16a_r)\l'n^2+{\mc o}(\l^{\prime\,2})\right]\,.\end{aligned}\ ] ] the leading order result given in equation has also been computed in @xcite and is in agreement with field theory @xcite . the subleading terms are predictions for the 2-loop results in field theory from which one should be able to fix the yet unknown constant @xmath211 ( _ cf . _ appendix [ appe ] ) . an amplitude of two massive string states and the pp - wave vacuum is @xmath212 this amplitude has also been computed in @xcite and agrees with the field theory result @xcite . finally , consider amplitudes involving only zero - mode oscillators . from equation it follows , that all amplitudes with , say , @xmath213 oscillators for the first string , @xmath95 oscillators for the second string , @xmath214 oscillators for the third string and @xmath215 vanish for all @xmath10 . this is in agreement with the proposal of @xcite , since the operators in field theory dual to single and double string states at order @xmath216 are constructed in such a way that they do not mix in free field theory . since the operator dual of the single string is protected this is an exact result for all @xmath217 , thereby matching the string theory result . as in the bosonic case , we change the oscillator basis to the bmn basis ( @xmath96 ) @xmath218 and compute amplitudes of the states @xmath219 only terms of order @xmath220 and @xmath221 contribute . note that these amplitudes are identically zero for the vertex constructed in @xcite , because the trace over @xmath222 comes from the term proportional to @xmath27 . as an example , we consider @xmath223 which are fermionic analogues to the bosonic amplitudes in equations and . for the above amplitudes to be non - zero it is essential that the fermionic exponential part of the vertex contains a term linear in zero - modes . counting fermionic zero - modes shows that in this case only the @xmath220 term in @xmath224 gives a contribution . we find @xmath225_n \left(\bar{n}^3_{|n|}\right)^2\nonumber\\ & = \m g_2\l'\sqrt{-\b(\b+1)}\frac{\sin^2n\pi\b}{2\pi^2}\left[1-\frac{1}{4}(3 - 16a_r ) \l'n^2+{\mc o}(\l^{\prime\,2})\right]\,,\\ \label{fer2 } g_{\text{s}}\la n_{5,6}|h_3|m_{7,8}\ra|v\ra & = -\frac{1}{2}\m g_2\a_1 ^ 2\a_2 ^ 2(\b+1)(1 - 4\m\a k)^{-2 } \bigl[cu_{(1)}^{-1}c_{(1)}\bigr]_m\bigl[cu_{(3)}^{-1}c_{(3)}\bigr]_n \left(\bar{n}^1_{|m|}\bar{n}^3_{|n|}\right)^2\nonumber\\ & = \m g_2\l'(\b+1)\frac{\sin^2n\pi\b}{2\pi^2}+{\mc o}(\l^{\prime\,2})\,.\end{aligned}\ ] ] up to a sign , the first amplitude in equation is identical to its bosonic analogue for all @xmath10 , whereas the second one , equation , is the same at leading order . these results reflect that the vertex constructed in section [ sec4 ] satisfies the superalgebra . one can check that equations and are unchanged if all spinor indices are in @xmath226 and ( up to permutations ) vanish otherwise . we are grateful for discussions with g. arutyunov , n. beisert , b. de wit , m. b. green , s. minwalla , a. petkou , f. quevedo and m. taylor . in particular , we would like to thank j. plefka and s. theisen for many useful comments on the manuscript and for sharing their insights . part of this work was performed during the workshop `` the quantum structure of spacetime and the geometric nature of fundamental interactions '' in leuven . we would like to thank the organizers for the stimulating atmosphere they created . this work was supported by gif , the german - israeli foundation for scientific research , fom , the dutch foundation for fundamental research on matter and the european community s human potential programme under contract hprn - ct-2000 - 00131 in which a. p. is associated to the university of bonn . here we give the details leading to equations , , and . the explicit expression for the infinite component vector @xmath227 appearing in the bosonic constituent of the prefactors is @xmath228 the fermionic quantity @xmath229 can be expressed in terms of @xmath227 as @xcite @xmath230 the following identities prove very useful ( @xmath231 ) @xmath232|v\ra\,,\\ \label{theta } \theta|v\ra & = -\sqrt{2}\sum_{r , n}q^r_nb_{-n(r)}^{\dg}|v\ra\,.\end{aligned}\ ] ] using the mode expansions of @xmath233 , @xmath234 , @xmath235 , @xmath236 and @xmath125 given in equations , , and one gets @xmath237_na_{-n(r)}^{\dg}\,,\\ \sum_{r=1}^3\{\bar{q}^-_{(r)},{\mc y}\ } & = \mathbb{p}\g - i\frac{\m\a}{\a'}\mathbb{r}\g\pi + \g\sum_{r=1}^3\frac{1}{\sqrt{|\a_r|}}\sum_{n=1}^{\infty}\bigl[p_{(r)}^{-1}c^{1/2}g_{(r)}\bigr]_na_{n(r)}^{\dg}\,,\end{aligned}\ ] ] @xmath238 & = \m\g(1+\pi)\sqrt{\frac{2}{\a'}}\l+ \g\sum_{r=1}^3\frac{e(\a_r)}{\sqrt{|\a_r|}}\sum_{n=1}^{\infty}\bigl[p_{(r)}^{-1}c^{1/2}f_{(r)}\bigr]_nb_{n(r)}^{\dg}\,,\\ \sum_{r=1}^3[q^-_{(r)},{\mc k}_- ] & = i\g\sum_{r=1}^3\frac{e(\a_r)}{\sqrt{|\a_r| } } \sum_{n=1}^{\infty}\bigl[p_{(r)}^{-1}c^{1/2}u_{(r)}f_{(r)}\bigr]_nb_{-n(r)}^{\dg}\,,\\ \sum_{r=1}^3[\bar{q}^-_{(r)},{\mc k}_0+{\mc k}_+ ] & = -\frac{\m\a}{\sqrt{2\a'}}\g(1-\pi)\theta+ \g\sum_{r=1}^3\frac{e(\a_r)}{\sqrt{|\a_r|}}\sum_{n=1}^{\infty}\bigl[p_{(r)}c^{1/2}f_{(r)}\bigr]_nb_{-n(r)}^{\dg}\,,\\ \sum_{r=1}^3[\bar{q}^-_{(r)},{\mc k}_- ] & = -i\g\sum_{r=1}^3\frac{e(\a_r)}{\sqrt{|\a_r| } } \sum_{n=1}^{\infty}\bigl[p_{(r)}c^{1/2}u_{(r)}f_{(r)}\bigr]_nb_{n(r)}^{\dg}\,.\end{aligned}\ ] ] using equations , and we find equations equations and . the action of the supercharges on @xmath124 given in equations and can be proven similarly . one needs @xmath239_n\bigl[u_{(s)}^{-1}c_{(s)}^{1/2}c\bar{n}^s\bigr]_m\,,\\ \bar{n}^{rs}_{-n ,- m}+e(\a_r)\left(\frac{m}{n}\left|\frac{\a_r}{\a_s}\right|\right)^{1/2}p_{n(r)}p_{m(s)}q^{rs}_{nm } & = 0\ , , \end{split}\ ] ] @xmath240_n\bigl[c_{(s)}^{1/2}\bar{n}^s\bigr]_m\,,\\ \bar{n}^{rs}_{-n ,- m}-e(\a_s)\left(\frac{m}{n}\left|\frac{\a_r}{\a_s}\right|\right)^{3/2}p_{n(r)}^{-1}p_{m(s)}^{-1}q^{rs}_{nm } & = \frac{\a}{\a_s}(1 - 4\m\a k)^{-1 } \bigl[p^{-2}_{(r)}c_{(r)}^{1/2}\bar{n}^r\bigr]_n\bigl[c_{(s)}^{1/2}c\bar{n}^s\bigr]_m \end{split}\ ] ] which follow from and . in this appendix we prove that @xmath241^as^i_{\dot{b } ) } & = 0\,,\\ \label{2 } \g^i_{a(\dot{a}}\bigl[\pi d\bigr]^a\wt{s}^i_{\dot{b } ) } & = 0\,,\\ \label{3 } \bigl(\g^i_{a\dot{a}}\bar{d}_b\wt{s}^i_{\dot{b}}+ \g^i_{a\dot{b}}d_bs^i_{\dot{a}}\bigr)(1-\pi)^{ab } & = 0\,.\end{aligned}\ ] ] equations and are equivalent to @xmath242 the first equation has terms of order @xmath243 and @xmath244 , whereas the second one has terms of order @xmath220 , @xmath192 and @xmath221 . there are two contributions to the order @xmath243 in equation , both vanish separately . the first one is @xmath245 whereas the second one is @xmath246}\pi^{ab}y^cy^d= \frac{1}{24}\left(\g^{ij}\g^{kl}\right)_{(\da\db ) } \left(\g^{ij}\pi\g^{kl}\right)_{cd}y^cy^d=0\,.\end{aligned}\ ] ] here we have used equations and . from the fourier identities @xcite @xmath247 it follows that the terms of order @xmath244 vanish as well . this proves equation . the @xmath220 term in equation is @xmath248 and the order @xmath221 term vanishes by . the terms of order @xmath192 in equation are @xmath249}\left(y^by^cy^dy^e+\frac{1}{24}{\e^{cdeb}}_{ghij } y^gy^hy^iy^j\right ) \nonumber\\ & = -\frac{1}{16}{\pi^a}_b\left(\g^{ijkl}_{\da\db}-2\d_{\da\db}\d^{ik}\d^{jl}\right ) \g^{ij}_{a[c}\g^{kl}_{de]}\left(y^by^cy^dy^e+\frac{1}{24}{\e^{cdeb}}_{ghij } y^gy^hy^iy^j\right ) \nonumber\\ & = -\frac{1}{16}{\pi^a}_b\g^{ijkl}_{\da\db}t^{ijkl}_{acde } \left(y^by^cy^dy^e+\frac{1}{24}{\e^{cdeb}}_{ghij}y^gy^hy^iy^j\right)=0\,.\end{aligned}\ ] ] in the last step we used that @xmath250 is symmetric and traceless and @xmath251 this proves equation . finally , equation is equivalent to @xmath252}^i+\frac{\a}{\a'}\g^i_{a[\da}\frac{\p}{\p y^b}s_{1\,\db]}^i \bigr)(1-\pi)^{ab}=0\,.\end{aligned}\ ] ] the first equation is symmetric in @xmath154 , @xmath155 and contains terms of order @xmath243 and @xmath244 . these vanish for the same reason as those in equation . the second equation is antisymmetric in @xmath154 , @xmath155 and contains terms of order @xmath220 , @xmath192 and @xmath221 . the @xmath220 contribution to equation is @xmath253}(1-\pi)^{ab}=\frac{1}{4}\g^{ij}_{\da\db}\g^i_{ab}(1-\pi)^{ab}=0\,.\ ] ] from equation it follows that the term of order @xmath221 vanishes as well . finally , there are two contributions to the terms of order @xmath192 , both of them vanish separately . the first one is @xmath254}^i(1-\pi)^{ab}= -\frac{1}{3}\g^i_{a[\da}u^i_{cde\db]}{(1-\pi)^a}_by^by^cy^dy^e=\nonumber\\ & \frac{1}{12}\left(\g^{ij}_{\da\db}\d_{a[c}\g^{ij}_{de ] } + \frac{1}{4}\left(\g^{ij}\g^{kl}\right)_{[\da\db]}\g^{ij}_{a[c}\g^{kl}_{de]}\right ) { ( 1-\pi)^a}_by^by^cy^dy^e=\nonumber\\ & \frac{1}{12}\g^{ij}_{\da\db}\g^{ik}_{a[c}\g^{kj}_{de]}{(1-\pi)^a}_by^by^cy^dy^e= \frac{1}{6}\g^{ij}_{\da\db}\g^{ij}_{bc}(1-\pi)_{de}y^by^cy^dy^e=0\,.\end{aligned}\ ] ] in the last step we have used equation . the second contribution of order @xmath192 then vanishes by equation . this concludes the proof of equation . apart from symmetry and tracelessness of @xmath250 we have used the following identities @xmath255 } -\frac{1}{16}\left(\g^{ij}\g^{kl}\right)_{\da\db}\g^{ij}_{a[b}\g^{kl}_{cd]}\,,\\ \label{ij } \g^{ik}_{a[b}\g^{jk}_{cd ] } & = t^{ij}_{abcd}-2\d_{a[b}\g^{ij}_{cd]}\,,\\ \label{ijij } \g^{ij}_{ab}\g^{ij}_{cd } & = 8\bigl(\d_{ac}\d_{bd}-\d_{ad}\d_{bc}\bigr)\,,\\ \label{klpij } \gamma^{ijkl}_{\da\db}\left(\g^{kl}\pi\g^{ij}\right)_{[ab]}&=0 \,.\end{aligned}\ ] ] here we demonstrate that equation leads to the constraints given in section [ sec4 ] . to this end , we adopt a trick introduced in @xcite . namely , we associate the world - sheet coordinate dependence with the oscillators as @xmath257 then we integrate the constraint equation over the @xmath258 . in dealing with the resulting expressions we can integrate by parts since the integrand is periodic . in addition to the identities in equations , . ] and we have to calculate the commutator of @xmath259 with @xmath127 and and its tilded counterpart . we get @xmath260\,|v\ra=-2i(1 - 4\m\a k)^{-1/2}\g \left[\dot{y}+y'+\frac{i}{2}\m(1-\pi)\left(y-2y_0\right)\right]|v\ra\,,\\ \sqrt{2}\bar{\eta}&\sum_{r=1}^3[\wt{q}_{(r)},\wt{\mc k}]\,|v\ra=-2i ( 1 - 4\m\a k)^{-1/2}\g\left[\dot{y}-y'+\frac{i}{2}\m(1-\pi)\left(y-2y_0\right)\right]|v\ra\ , . \end{split}\ ] ] here @xmath261 is the zero - mode part of @xmath174 , we suppressed the @xmath262 , @xmath258 dependence and @xmath263 the fact that the above equations have a term which only depends on the zero mode @xmath261 is important . combining the various contributions to equation yields @xmath264 \nonumber\\ & -\frac{\m}{\sqrt{2}}(1 - 4\m\a k)^{-1}\bigl [ \left(\g^i_{a\dot{a}}\bar{d}_b\tilde{s}^i_{\dot{b}}+\g^i_{a\dot{b}}d_bs^i_{\dot{a}}\right)(1-\pi)^{ab } + 2i\left(\g^i_{a\dot{a}}\eta\tilde{s}^i_{\dot{b}}-\g^i_{a\dot{b}}\bar{\eta}s^i_{\dot{a}}\right){(1-\pi)^a}_by_0^b \bigr]\nonumber\\ & \label{qtq1}+\frac{i}{\sqrt{2}}\frac{\a'}{\a}\left [ { \mc k}^i{\mc k}^j\g^j_{a\dot{a}}d^a\tilde{s}^i_{\dot{b}}-\wt{\mc k}^i\wt{\mc k}^j\g^j_{a\dot{b}}\bar{d}^as^i_{\dot{b } } \right]\biggr)|v\ra=0\,.\end{aligned}\ ] ] next we remove the @xmath258 derivatives from @xmath174 by partial integration . this requires the identity @xcite ; on the r.h.s . of equations ( d.26 ) and ( d.27 ) it should be @xmath265 instead of @xmath266 ; the r.h.s . of equation ( e.8 ) should be multiplied by @xmath267 . this corrections modify equation ( e.9 ) of @xcite slightly , which is now in agreement with equations of this paper . these typos do not affect the proof presented in appendix e of @xcite . ] @xmath268 and @xmath269 using the above identities equation becomes after integrating by parts @xmath270\bigl(\dot{y}^a-\dot{y}^a_0\bigr)\\ & -\frac{\m}{\sqrt{2}}(1 - 4\m\a k)^{-1 } \bigl(\g^i_{a\dot{a}}\bar{d}_b\tilde{s}^i_{\dot{b}}+\g^i_{a\dot{b}}d_b s^i_{\dot{a}}\bigr)(1-\pi)^{ab } \\ & -i{\mc k}^i{\mc k}^j\bigl[\d^{ij}m_{\dot{a}\dot{b}}-\frac{\a'}{\sqrt{2}\a } \g^j_{a\dot{a}}d^a\tilde{s}^i_{\dot{b}}\bigr ] + i\wt{\mc k}^i\wt{\mc k}^j\bigl[\d^{ij}m_{\dot{a}\dot{b}}-\frac{\a'}{\sqrt{2 } \a}\g^j_{a\dot{b } } \bar{d}^as^i_{\dot{a}}\bigr]\biggr)|v\ra=0\ , . \end{split}\ ] ] the bosonic constituents of the prefactor can be obtained via the operators @xcite @xmath271 acting on the exponential part of the vertex these satisfy @xmath272 here we defined @xmath273 which is equal to one for @xmath200 @xcite . for the fermionic constituent of the prefactor one considers @xcite @xmath274 which satisfies @xcite @xmath275 comparing this identity with equation , we see that @xmath276 we conjecture that this is indeed the case . unfortunately , it seems that there is no way of proving this conjecture , at least for large-@xmath10 , without knowing @xmath105 for all @xmath10 . this is because in equation , the large-@xmath10 expansion does not commute with the infinite sum over @xmath214 . one can write the functional equivalent of equations , and as @xmath277 here @xmath278\d^8\bigl[\sum_sp_s(\s)\bigr]\ ] ] is the functional expression leading to the exponential part of the vertex @xcite . finally , let us mention the following subtlety . using equation it is easy to see that for example @xmath279\,|v\ra= \m\g\pi y|v\ra\,.\ ] ] however , this is not equal to the commutator of @xmath280 with @xmath281 . to see this , rewrite @xmath282|v\ra= \sqrt{2}\bar{\eta}\sum_{r=1}^3[\wt{q}_{(r)},\lim_{\s\to\pi\a_1}k(\s)]\,|v\ra \nonumber\\ \label{qkfunc } & -\frac{\a'}{\a}\g y(1 - 4\m\a k)f^{-1}(\m)[\lim_{\s\to\pi\a_1}k(\s),\wt{k}]|v\ra= \m\g\pi y|v\ra+\m\g y|v\ra=\m\g(1+\pi)y|v\ra\,.\end{aligned}\ ] ] here we used equation and @xmath283|v\ra= -\frac{\m\a}{\a'}(1 - 4\m\a k)^{-1}f(\m)|v\ra\,.\ ] ] equation is equivalent to equation of section [ sec4 ] . it is this appearance of the matrix @xmath284 as opposed to just @xmath250 , that is responsible for the term proportional to @xmath27 in the cubic interaction vertex . in this appendix we collect some results on the large-@xmath10 expansion . we need the explicit expressions for the matrices @xmath101 and the vector @xmath104 . for @xmath285 they are @xmath286 writing @xmath287 it was shown in @xcite that @xmath288 with @xmath211 an undetermined number . the next term in the expansion would be of order @xmath289 . it follows that @xmath290b\,.\ ] ] furthermore the large-@xmath10 expansion of @xmath291 is @xcite @xmath292 the next term in the expansion would be of order @xmath293 . here , @xmath294 is undetermined as well , however it was shown in @xcite that @xmath295 from the definition of @xmath109 it is easy to see that @xmath296 at leading order . here and below @xmath35 , @xmath297 . using the factorization theorem equation , the above formulae are sufficient to determine the leading large-@xmath10 behavior of all the other neumann matrices as well . we find @xmath298 and therefore @xmath299 99 j. kowalski - glikman,_vacuum states in supersymmetric kaluza - klein theory _ , phys . b * 134 * ( 1984 ) 194 . c. m. hull , _ exact pp wave solutions of 11-dimensional supergravity _ , phys . b * 139 * ( 1984 ) 39 . j. figueroa - ofarrill and g. papadopoulos , _ homogeneous fluxes , branes and a maximally supersymmetric solution of m - theory _ , jhep * 08 * ( 2001 ) 036 [ arxiv : hep - th/0105308 ] . m. blau , j. figueroa - ofarrill , c. hull and g. papadopoulos , _ a new maximally supersymmetric background of iib superstring theory _ , jhep * 01 * ( 2002 ) 047 [ arxiv : hep - th/0110242 ] . r. r. metsaev , _ type iib green - schwarz superstring in plane wave ramond - ramond background _ , b * 625 * ( 2002 ) 70 [ arxiv : hep - th/0112044 ] . r. r. metsaev and a. a. tseytlin , _ exactly solvable model of superstring in plane wave ramond - ramond background _ , phys . d * 65 * ( 2002 ) 126004 [ arxiv : hep - th/0202109 ] . s. mandelstam , _ interacting string picture of dual resonance models _ , nucl . b * 64 * ( 1973 ) 205 ; _ ibid . _ , _ dual - resonance models _ , phys . * 13 * ( 1974 ) 259 . e. cremmer and l. gervais , _ combining and splitting relativistic strings _ , nucl . b * 76 * ( 1974 ) 209 ; _ ibid . _ , _ infinite component field theory of interacting relativistic strings and dual theory _ , nucl . phys . b * 90 * ( 1975 ) 410 . m. kaku and k. kikkawa , _ the field theory of relativistic strings , pt . 1 : trees _ , phys . rev . d * 10 * ( 1974 ) 1110 ; _ ibid . _ , _ the field theory of relativistic strings . 2 : loops and pomerons _ , phys . rev . d * 10 * ( 1974 ) 1823 . s. mandelstam , _ interacting string picture of the neveu - schwarz - ramond model _ , nucl . phys . b * 69 * ( 1974 ) 77 . m. green and j. schwarz , _ superstring interactions _ b * 218 * ( 1983 ) 43 . m. green , j. schwarz and l. brink , _ superfield theory of type ( ii ) superstrings _ , nucl . b * 219 * ( 1983 ) 437 . m. spradlin and a. volovich , _ superstring interactions in a pp - wave background _ , hep - th/0204146 . m. spradlin and a. volovich , _ superstring interactions in a pp - wave background ii _ , hep - th/0206073 . a. pankiewicz , _ more comments on superstring interactions in the pp - wave background _ , jhep * 09 * ( 2002 ) 056 [ arxiv : hep - th/0208209 ] . r. penrose , _ any space - time has a plane - wave as a limit _ in ` differential geometry and relativity ' , m. cahen and m. flato eds . , reidel , dordrecht , netherlands 1976 , pp.271 - 275 . j. m. maldacena , _ the large @xmath13 limit of superconformal field theories and supergravity _ , adv . * 2 * ( 1998 ) 231 [ int . j. theor . * 38 * ( 1999 ) 1113 ] [ arxiv : hep - th/9711200 ] . s. s. gubser , i. r. klebanov and a. m. polyakov , _ gauge theory correlators from non - critical string theory _ , phys . b * 428 * ( 1998 ) 105 [ arxiv : hep - th/9802109 ] . e. witten , _ anti - de sitter space and holography _ , adv . * 2 * ( 1998 ) 253 [ arxiv : hep - th/9802150 ] . d. berenstein , j. maldacena and h. nastase , _ strings in flat space and pp waves from @xmath300 super yang mills _ , jhep * 04 * ( 2002 ) 013 [ arxiv : hep - th/0202021 ] . c. kristjansen , j. plefka , g. w. semenoff and m. staudacher , _ a new double - scaling limit of @xmath300 super yang - mills theory and pp - wave strings _ , nucl . b*643 * ( 2002 ) 31 [ arxiv : hep - th/0205033 ] . n. r. constable , d. z. freedman , m. headrick , s. minwalla , l. motl , a. postnikov and w. skiba , _ pp - wave string interactions from perturbative yang - mills theory _ , jhep * 07 * ( 2002 ) 017 [ arxiv : hep - th/0205089 ] . d. j. gross , a. mikhailov and r. roiban , _ operators with large r charge in @xmath300 yang - mills theory _ , annals phys . * 301 * ( 2002 ) 31 [ arxiv : hep - th/0205066 ] . a. santambrogio and d. zanon , _ exact anomalous dimensions of @xmath300 yang - mills operators with large r charge _ , hep - th/0206079 . g. arutyunov and e. sokatchev , _ conformal fields in the pp - wave limit _ , jhep * 08 * ( 2002 ) 014 [ arxiv : hep - th/0205270 ] . h. verlinde , _ bits , matrices and 1/n _ , arxiv : hep - th/0206059 . j. g. zhou , _ pp - wave string interactions from string bit model _ , hep - th/0208232 . n. beisert , c. kristjansen , j. plefka , g. w. semenoff and m. staudacher , _ bmn correlators and operator mixing in @xmath300 super yang - mills theory _ , hep - th/0208178 . n. r. constable , d. z. freedman , m. headrick and s. minwalla , _ operator mixing and the bmn correspondence _ , hep - th/0209002 . h. verlinde and d. vaman , _ bit strings from @xmath14 gauge theory _ , hep - th/0209215 . r. a. janik , _ bmn operators and string field theory _ , hep - th/0209263 . j. pearson , m. spradlin , d. vaman , h. verlinde and a. volovich , _ tracing the string : bmn correspondence at finite @xmath20 _ , hep - th/0210102 . j. gomis , s. moriyama and j. park , _ sym description of sft hamiltonian in a pp - wave background _ , hep - th/0210153 . k. skenderis and m. taylor , _ branes in ads and pp - wave spacetimes _ , jhep * 06 * ( 2002 ) 025 [ arxiv : hep - th/0204054 ] . o. bergman , m. r. gaberdiel and m. b. green , _ d - brane interactions in type iib plane - wave background _ , hep - th/0205183 . j. schwarz , _ comments on superstring interactions in a plane - wave background _ , jhep * 09 * ( 2002 ) 058 [ arxiv : hep - th/0208179 ] . y. kiem , y. kim , s. lee and j. park , _ pp - wave / yang - mills correspondence : an explicit check _ , nucl . b*642 * ( 2002 ) 389 [ arxiv : hep - th/0205279 ] . p. lee , s. moriyama and j. park , _ a note on cubic interactions in pp - wave light cone string field theory _ , phys . d * 67 * , 086001 ( 2003 ) [ arxiv : hep - th/0209011 ] .

we address the electronic structure and magnetic properties of vacancies and voids both in graphene and graphene ribbons . using a mean field hubbard model , we study the appearance of magnetic textures associated to removing a single atom ( vacancy ) and multiple adjacent atoms ( voids ) as well as the magnetic interactions between them . a simple set of rules , based upon lieb theorem , link the atomic structure and the spatial arrangement of the defects to the emerging magnetic order . the total spin of a given defect depends on its sublattice imbalance , but some defects with can still have local magnetic moments . the sublattice imbalance also determines whether the defects interact ferromagnetically or antiferromagnetically with one another and the range of these magnetic interactions is studied in some simple cases . we find that in semiconducting armchair ribbons and two - dimensional graphene without global sublattice imbalance there is maximum defect density above which local magnetization disappears . interestingly , the electronic properties of semiconducting graphene ribbons with uncoupled local moments are very similar to those of diluted magnetic semiconductors , presenting giant zeeman splitting .

0802.2029

magnetic order occurs , in most instances , in materials with partially filled @xmath2 or @xmath3 shells . there is , however , a recent awareness that the possibility of magnetic order can also occur in materials without open @xmath2 or @xmath3 shells@xcite . experimental evidence of this new type of magnetism has been found in thin films of certain oxides ( hfo@xmath4 , zno , tio@xmath4)@xcite , as well as irradiated graphite@xcite , and thiol - capped gold nanoparticles @xcite . although more experimental work is probably necessary to confirm and understand magnetism in these systems , there are at least two scenarios for which theory provides a mechanism for the appearence of magnetism without @xmath2 or @xmath3 open shells . on one side , in some lattices intrinsic lattice defects like vacancies lead to the formation local magnetic moments , a preliminary condition for the existence of magnetic order . this is the case in graphite@xcite , graphene@xcite and ii - vi semiconductors@xcite . on the other side , it has been recently found that clusters with specific shapes , like triangular graphene islands@xcite or icosahedral@xcite gold clusters , have large degeneracies at the fermi energy in their single particle spectra . these degeneracies are related to the symmetry of the nanostructure and , in words of luo _ et al._@xcite , they behave like superatoms , with magnetic ground states that comply with atomic - like hund s rules . importantly , both vacancy induced@xcite and superatomic magnetism@xcite occur in graphene structures and , as we show here , have the same origin . in this work we present extensive numerical work to understand vacancy induced magnetism in graphene and graphene ribbons and we analyze our results in the context of a broader theoretical framework that unifies superatomic@xcite and vacancy induced magnetism@xcite in graphene . in part our motivation stems from the recently shown possibility of fabricating high mobility graphene based field effect transistors@xcite which has created enormous interest in graphene based electronics@xcite . additional possibilities arise from the fabrication of semiconducting graphene ribbons@xcite and graphene nanoislands@xcite with top - down techniques as well as the growth of graphene islands with bottom - up techniques@xcite . magnetic order in patterned or nanostructured graphene would bring up new opportunities of research in spintronics . graphene honeycomb structure is a bipartite lattice , formed by two interpenetrating triangular sublattices , a and b , such that the first neighbors of an atom a belong to the sublattice b and viceversa@xcite . the low - energy electronic structure of graphene can be described by a single - orbital ( @xmath5 ) nearest - neighbor hopping hamiltonian@xcite . this model correctly describes two dimensional graphene as a zero gap semiconductor with linear bands around the fermi energy . the single particle spectrum of a nearest - neighbor tight - binding model in a bipartite lattice has particle - hole symmetry@xcite . the magnetic properties of both graphene - based nanostructures and defective graphene are intimately related to the appearance of midgap states and how they are affected by electron - electron interactions . the existence of zero - energy states in disordered bipartite lattices was proved by inui et al.@xcite . within the first - neighbor tight - binding model , a sufficient condition@xcite for the existence of midgap states is the existence of a finite sublattice imbalance , @xmath6 , where @xmath7 and @xmath8 are the number of atoms belonging to each sublattice or missing from each sublattice in an otherwise perfect system . thus , whereas ideal graphene has @xmath9 and no midgap states , both defective graphene and some graphene islands , such as triangles , can present finite sublattice imbalance and @xmath10 midgap states . the result of inui et al.@xcite has been used in a recent work on disorder in graphene by pereira _ et al._@xcite . incidentally , the existence of zero - energy or midgap states in uncompensated graphene structures was known long ago in the context of chemical studies of hydrocarbons as the longuet - higgis conjecture@xcite . because of the particle - hole symmetry , midgap states are half filled for neutral graphene and the appearence of magnetic moments is expected in analogy with hund s rule in atomic magnetism . the hubbard model extends the single - particle tight - binding model including the effect of coulomb repulsion between two electrons in the same atomic site . importantly , a theorem by lieb , valid for the exact ground state of the hubbard model and neutral bipartite lattices@xcite states that the total spin @xmath0 of the ground state is given by @xmath11 . lieb s theorem provides a rigorous connection between vacancies in the graphene lattice and the emergence of magnetism . as as result , sublattice unbalanced neutral graphene will always present a finite total magnetic moment . although lieb s theorem provides the total spin of the ground state , it does not say much about the actual local magnetic order or spin texture . for instance , @xmath1 does not preclude the existence of local magnetic moments coupled antiferromagnetically or presenting compensated ferrimagnetic order . the most notorious example of compensated ferrimagnetic order can be found in zigzag ribbons@xcite , where each edge presents ferromagnetic order antiparallel to each other for a total vanishing magnetic moment . other examples can be found in hexagonal graphene islands , where , beyond a critical size , contiguous sides alternate the direction of the ferromagnetically ordered magnetic moments@xcite . the rest of this paper is organized as follows . in sec . [ method ] we review the single orbital hubbard model and the different methodologies used to describe the electronic structure of defective graphene and graphene ribbons . the underlying non - interacting spectrum and associated magnetic textures can be anticipated following some basic rules which are presented in sec . [ rules ] . we illustrate the validity of the rules by numerical calculations in the case of semiconducting armchair ribbons ( sec . [ single ] ) with vacancies , voids , or notches , both in the non - interacting ( sec . [ single ] ) and interacting ( [ u ] ) cases . the results for bulk graphene are discussed in sec . ( [ bulk ] ) . summary and conclusions are presented in sec . [ conclusions ] . we consider the low - energy physics that takes place in the subspace expanded only by the single @xmath12 orbital ( the one perpendicular to the graphene plane ) . next - to - near neighbor hopping is neglected and the electron - electron interactions are included locally in the form of an on - site repulsion or hubbard model . when the interactions are turned off this reduces to the widespread one orbital tight - binding model@xcite . the hubbard term is treated in a mean field approximation@xcite . comparison between the results so obtained and density functional theory ( dft ) calculations yield very good agreement for two - dimensional graphene@xcite , carbon nanotubes@xcite , zigzag@xcite and armchair graphene ribbons@xcite , as well as graphene islands@xcite . we model vacancies and voids in perfect graphene or graphene ribbons by removing atoms , actually , by removing the representing @xmath12 orbitals in the tight - binding model . this results in a reduction of the coordination of the atoms adjacent to the missing atoms . we ignore the lattice distorsion and we assume that the on - site energy is the same for edge and bulk atoms . the single - orbital hamiltonian implicitly assumes full hydrogen passivation of the @xmath13 dangling bonds of the atoms without full coordination . this assumption , which might not be completely realistic in the case of actual vacancies@xcite , does not invalidate our model ; for we can consider an alternative physical realization : the chemisorption of a hydrogen atom on top of a bulk graphene atom@xcite effectively removes a @xmath12 orbital from the low - energy hamiltonian . in our one - orbital model there is no difference between these two scenarios . dft calculations on graphene ribbons@xcite , graphene islands@xcite , and bulk graphene with vacancies@xcite have shown that the results follow the predictions of lieb s theorem , even though dft calculations go beyond the first - neighbor hopping , short - range interaction hubbard model on which the theorem is based . in other words , second neighbour hopping and inter - site coulomb repulsion , present in the dft calculations , do not modify the relation between lattice imbalance and total spin of the ground state warranted for the hubbard model for which these couplings are absent . from this point of view these couplings are irrelevant . it is thus justified to consider the following mean - field hamiltonian : @xmath14 where @xmath15 runs over all lattice sites and the non - interacting hamiltonian reads @xmath16 where the sum runs over nearest - neiborgh lattice sites @xmath17 and @xmath18 ev . without loss of generality , we have set the diagonal terms of the hamiltoninan to zero . for neutral graphene we can rewrite the mean - field hamiltonian ( up to a constant ) as the sum of two terms : @xmath19 where @xmath20 and @xmath21 . the second term in @xmath22 represents the non - trivial contribution of interactions . the calculations have been performed considering three different types of boundary conditions . for the evaluation of the non - interacting density of states ( dos ) in ribbons we compute the green s function projected on the region where the defects are located . the perfect regions of the ribbon away from the defects are included in the green s function by means of a self - energy . we refer the reader to ref . for more details on this methodology . when the interactions are turned on ( @xmath23 ) we consider ribbons with periodic boundary conditions in one direction or , for bulk graphene , periodic boundary conditions in both directions . more details will be given in the respective sections . in this section we provide some general rules to understand the appearance of midgap states and magnetic textures due to single - atom vancacies or , more generally , voids in an otherwise sublattice balanced graphene structure , for instance , an infinite defect - free semiconducting armchair ribbon . a void in graphene can be partially characterized by the number and type , a or b , of atoms removed from the otherwise perfect structure . we will label voids as one would do for chemical compounds , a@xmath24b@xmath25 . voids will be unbalanced when the are created by removing @xmath7 and @xmath8 atoms such that @xmath26 the sublattice imbalance @xmath27 , which can be either positive or negative , can be interpreted as an imbalance `` charge '' . this quantity is central to our discussion , although the exact formula of the void is also important since it gives an idea of the size and shape of the void . in the case of ribbons the voids can be close to the edges , becoming thus notches . ( for the most part we will refer to voids in the bulk of the ribbon , but the conclusions apply equally to the case of notches on the edges . ) for a single unbalanced void characterized by @xmath27 , @xmath10 zero - energy states appear in the non - interacting spectrum with weight on only one sublattice@xcite . when the graphene structure presents a gap @xmath28 , as in armchair ribbons , these states are normalizable and localized around the void ( in contrast to two - dimensional graphene where the zero - energy states are not normalizable@xcite ) . figure [ voids ] shows various examples of voids in a ribbon with different sublattice imbalance @xmath27 . [ hbt ] examples of voids with different sublattice imbalance in the middle of a graphene ribbon . from left to right the associated imbalance charge is @xmath29 ( vacancy ) , and 2.,title="fig : " ] let us consider now two voids , characterized _ locally _ by @xmath30 and @xmath31 , sufficiently separated so that they do not affect each other . according to the result of inui et al.@xcite , the single particle spectrum has , at least , @xmath32 midgap states . the important question is what happens when the distance between them is such that they do affect each other . if @xmath30 and @xmath31 have the same sign , there are @xmath33 midgap states , regardless of the distance . if they have different signs , e.g. , @xmath34 , eq . [ ita ] apparently warrants the annihilation of midgap states . within the non - interacting model midgap states are 100 percent sublattice polarized . the non - interacting hamiltonian has finite matrix elements between states that have weight on different sublattices . hence , the mechanism for midgap state annihilation is hybridization of midgap states localized in different sublattices . this annihilation occurs as bonding - antibonding pairs of midgap states form , resulting in a shift in their energy and in a loss of the sublattice polarization . for large distances , however , this annihilation does not occur . a well understood related example occurs in zigzag ribbons@xcite . the edge of a zigzag ribbon has a local sublattice imbalance . if the top edge belongs to the a sublattice , the bottom edge belongs to the b sublattice . states fully localized in the edge have zero energy and are localized in a single sublattice . to the extent that states mostly localized on the top edge penetrate into the ribbon , they hybridize with states mostly localized on the bottom edge . this mixing results in a bonding - antibonding splitting that takes these states away from the dirac energy . in the case of zigzag ribbons , the degree of localization in the unit cell depends on the wavevector . states close to the zone boundary are very localized in the edges and have energy very close to zero@xcite . the localization decreases as the wavevector departs from the boundary , resulting in the hybridization and the departure from zero energy@xcite . in the general case one can conclude that the minimum number of zero - energy states will be given by @xmath35 where the integer indeces @xmath36 and @xmath37 run over voids with the same imbalance sign , respectively . in practice , within an arbitrarily small energy interval @xmath38 , the number of zero - energy states can be as large as @xmath39 in the general case , @xmath40 will be a number between @xmath41 and @xmath42 . when electron - electron interactions are turned on , at least locally in the form of a hubbard - type interaction , lieb s theorem guarantees that @xmath43 for neutral graphene . the theorem , however , does not exclude the possibility of spin - symmetry broken local magnetic order when @xmath1 or small . for instance , two or more voids with local sublattice imbalances that cancel out the total imbalance can still retain their local magnetic order when they are not in proximity . when the imbalance of the void is zero but the size is large an internal ferrimagnetic order can not be discarded either . in general , calculations will be necessary to ascertain the spin texture in these situations . a few conclusions , however , can be reached without actually performing any calculations . one can distinguish four cases : 1 . @xmath44 : in this case all the voids are of the same sign . the coupling between them is always ferromagnetic and the spin of the ground state is @xmath45 . the splitting with smaller spin states will depend on the inter - void coupling . @xmath46 : in this case all the voids of different sign are in proximity and interact , yielding a @xmath47 state . calculations will be necessary to ascertain the spin texture in these situations . @xmath48 : in this case all the voids of different type are separated and uncoupled . the ground state has @xmath47 , but the spin - flip gap is negligible since there are uncoupled magnetic moments . @xmath49 : in this case there are voids of different sign , but some of them are uncoupled and some not . this is the most general case . the ground state has @xmath47 , but , as in the previous case , the spin - flip gap is negligible since there are uncoupled magnetic moments . calculations will be necessary to ascertain the spin texture in these situations as well . in this section we study the electronic structure of defective graphene ribbons within the non - interacting tight - binding model . the results are obtained using the cluster embedded method described in ref . . in this methodology a finite portion of the ribbon containing the defects is attached to two semi - infinite perfect ribbons of the same width and compute the dos of the defective region . in the defect free case we obtain a gap in the dos . we consider armchair graphene ribbons of width @xmath50 , where @xmath51 is an integer number and @xmath52 is the graphene lattice parameter . we only consider values of @xmath53 such that , within the first - neighbor tight - binding model , the ribbon is semiconducting@xcite . this happens if @xmath54 is not a multiple of three . more realistic calculations@xcite predict that , because of lattice distorsion of the edge atoms , even ribbons with @xmath55 , where @xmath56 is an integer , are semiconductors . semiconducting graphene ribbons attract interest due to possible applications in nanoelectronics@xcite . as in the case of si based semiconductors , their electronic structure might be strongly influenced by impurities . here we study the effect of vacancies and voids , which are expected to act as neutral impurities . the simplest defective structure is a perfect semiconducting graphene ribbon from which a single atom , a or b , is removed . in agreement to eq . [ ita ] , a zero energy state appears in the dos . for neutral graphene , this state is half filled . in other words , an spin unpaired electron occupies the midgap state . the local density of states at zero energy , which is nothing but the modulus square of the wave function associated with the zero energy state , is shown in the inset of fig . [ figsingle](a ) . the state is localized in the neighborhood of the vacancy . the shape of the midgap state is also peculiar : it has a clear directionality . monoatomic vacancies have the shape of a triangle . two vertices of the triangle point towards the edges of the ribbon , whereas the lateral vertex can point downstream or upstream along the ribbon . the midgap state is peaked around the lateral vertex . hence , midgap states have a strong directional character . importantly , the _ integrated _ charge , including both mid - gap and band states below the fermi energy , yields a homogeneous charge distribution : there is one electron per atom in every atom , even in the presence of the vacancy . hence , the localized midgap state does not imply charge localization , yet there is a finite spin density . the total spin @xmath0 of the neutral graphene with one vacancy is @xmath57 and the spin density does show a non - homogeneous texture , as shown in fig . [ figsingle](a ) . hence , the region of the material around the defect has no local charge but has local spin . [ hbt ] ( color online ) ( a ) magnetic moments on lattice sites around a single vacancy . inset : probability density of the zero - enery state built with gaussian functions on lattice sites . ( b ) same as in ( a ) but for a triangular void with @xmath58 . inset : same as in ( b ) but summing over the two zero - energy states.,title="fig : " ] ( color online ) ( a ) magnetic moments on lattice sites around a single vacancy . inset : probability density of the zero - enery state built with gaussian functions on lattice sites . ( b ) same as in ( a ) but for a triangular void with @xmath58 . inset : same as in ( b ) but summing over the two zero - energy states.,title="fig : " ] our next step is to consider larger voids . in fig . [ figsingle](b ) we show the results for a void with @xmath58 , obtained by removing four atoms ( a@xmath59b@xmath60 ) . in agreement [ ita ] , there are two midgap states ( per spin ) . their local density of states is shown in the inset of fig . [ figsingle](b ) . as in the case of monoatomic vacancy , two electrons occupy the midgap states . the _ integrated _ local charge is also homogeneous : one electron per site . within the framework of the non - interacting model we can not discriminate between the @xmath1 or the @xmath61 . as we discuss below , when hubbard repulsion is turned on lieb s theorem warrants that the spin of this structure is @xmath61 . in fig . [ figsingle](b ) we show the magnetization density , calculated including the interactions , as discussed below . as in the case of a sigle missing atom , there is a magnetic texture with @xmath61 which is localized in a region without localization of extra charge . triangular voids maximize @xmath10 while removing the minimum number of atoms . larger ones ( @xmath62 ) exhibit similar features to the ones already discussed . more complicated voids with zig - zag edges such as hexagons or rombhi , which have @xmath9 , can still present quasi zero - energy states if they are sufficiently big and therefore might exhibit spin textures as discussed below . as a step towards understanding the electronic structure of graphene with a finite density of defects , we first consider the electronic structure of two voids with the same absolute value of @xmath27 . each void has a well defined sublattice imbalance number @xmath27 when apart , which can be positive or negative . if the sublattice imbalance of the two voids has the same sign , the global structure has twice as many zero energy states as the separated defects . the non - interacting hamiltonian does not couple sites on the same sub - lattices so that the zero - energy states associated with two vacancies with sublattice imbalance of the same sign can not interact , regardless of the distance separating them , i.e. , the non - interacting dos will always present a delta function at zero energy that can accomodate @xmath63 electrons per spin channel . when the imbalance numbers are of different sign they cancel out each other . when the defects are far away from each other their local electronic structure is expected to be the same as that of a single defect : midgap states localized in a single sublattice around the missing atoms . as the defects become closer , the single particle hamiltonian , which couples atoms of different sublattices , will hybridize the otherwise sublattice polarized midgap states , which will result in bonding and antibonding pairs away from zero energy . the localization length of the single defect states sets the length scale at which this hybridization occurs . [ hbt ] ( color online ) density of states near the dirac point for an armchair ribbon of @xmath64 with two vacancies presenting charge of different sign and same modulus ( @xmath65 ) . solid lines correspond to the b+a case ( left lower inset ) and dashed lines correspond to the a+b case ( left upper inset ) . a finite broadening has been added for visibility s sake of the delta functions . the finite , but small , energy splitting in the former case is not visible for this broadening . right inset : bonding - antibonding energy splitting as a function of the distance between vacancies for the two different spatial orderings.,title="fig : " ] our numerical calculations confirm this scenario . for a given width , the hybridization depends on the distance and , given the directional character of the midgap states in ribbons , on the relative orientation . in fig . [ a-b.b-a.dos ] we show the dos for a system with two monoatomic vacancies a and b ( @xmath65 , respectively ) . they are aligned along the ribbon axis and placed at a distance of @xmath66 away from each other for the two possible spatial orderings , a+b ( head to head ) and b+a ( tail to tail ) as shown in the left insets . due to the high directional character of the associated zero - energy states , the coupling is not invariant against the interchange of positions and the zero - energy states hybridize differently , depending on the spatial ordering . in one case the two - fold zero - energy peak clearly splits into two above and below the fermi energy . in the other the splitting is much smaller ( not visible in this scale ) . for one relative orientation the wave functions overlap and the degeneracy is strongly removed . for the other the wave functions do not couple at this distance and the degeneracy is practically unaffected . in the right inset of fig . [ a-b.b-a.dos ] we show a logarithmic plot of the energy splitting as a function of the distance for the two cases . the splitting decays exponentially in both , reflecting the localized character of the zero - energy states . we now consider the case of pairs of defects with larger sublattice imbalance . figure [ 2a-2b ] shows the dos for two triangular voids characterized by @xmath67 and @xmath68 ( a@xmath59b@xmath60 and a@xmath60b@xmath59 , respectively ) at different distances . we have selected only one possible ordering in this case ( tail to tail ) . according to these sublattice imbalances each void has associated two localized states . these two states also present a strong directional character , but is different for the two . this can be inferred from the two different bonding - antiboding splitting energies for a given distance seen in fig . [ 2a-2b ] . we note that , even when the voids approach each other , the splitting associated with one of the localized states remains small , still being practically zero for small distances only in the extreme limit of zero distance when the two voids merge into a single one with sublattice imbalance @xmath9 ( a@xmath69b@xmath69 ) there are no zero - energy states . [ hbt ] ( color online ) dos projected on the vicinity of two triangular voids placed along the axis of a semiconducting ribbon for different relative distances . the imbalance charges are the same , but differ in sign ( @xmath70 ) . , title="fig : " ] as the sublattice imbalance of the merging voids becomes bigger and these condense into even bigger @xmath9 voids , the number of states that appear in a vicinity of zero @xmath71 increases with the charge of these . since the appearance of magnetic order relies on the existence of zero - energy states , large voids with @xmath9 can still present ferrimagnetic order , the only condition being that they are formed out of voids with large sublattice imbalance . in other words , their contours must present sufficiently long zig - zag sections . this limits the possible shapes of these voids to , e.g. , rombohedral ( see fig . [ rombo ] ) or hexagonal forms . this conclusion is no different from that reached on graphene hexagonal islands@xcite or finite length ribbons@xcite , where calculations have revealed compensated ferrimagnetic order developing along the edge beyond a certain critical size . finally , in order to stress the fact that there is nothing in the previous discussion specific to voids in the bulk of the ribbon , we compute the non - interacting dos for an a@xmath72b@xmath69 void plus an ab@xmath4 void placed on the edges ( i.e. , notches ) with @xmath58 and @xmath73 , respectively . removing just one atom to create a notch with @xmath73 would have given the same charge as the ab@xmath4 defect , but it would be chemically very unstable and we ignore that possibility . the notches are located on opposite edges , although the results apply the same for notches on the same edge . a single doubly - degenerate state appears at zero energy for the @xmath58 notch . when the second notch is added in close proximity only a single zero - energy state remains , according to the total sublattice imbalance of the system @xmath74 . [ hbt ] ( color online ) dos for a single notch with imbalance charge @xmath58 ( blue dashed line , right inset ) . the same notch with an additional notch nearby of charge @xmath73 ( black solid line , left inset).,title="fig : " ] in summary , defective structures with sublattice imbalance result in half - filled midgap states that are expected to yield magnetic moments when interactions are turned on . structures with global sublattice balance can still present midgap states and be prone to developing local magnetic order , at least in two situations : distant defects with @xmath27 of opposite sign and large voids with sufficiently long zigzag edges . in this section we verify whether the physical picture anticipated from the non - interacting model remains true when the on - site hubbard repulsions are included . as shown in previous section , unpaired spins appear in sublattice imbalanced structures . when @xmath75 , the non - interacting model predicts that a shell of @xmath10 degenerate mid - gap states is half - filled . the maximization of the spin is expected when coulomb repulsions are turned on , in the spirit of hund s rule . at half - filling , the exact ground state of the hubbard model for a bipartite lattice such as that of graphene satisfies lieb s theorem@xcite which relates the sublattice imbalance and the ground state total spin : @xmath76 . for unbalanced structures this inmediately confirms the hund s rule scenario . in the case of balanced structures the ground state spin must be zero , but this could happen with local moments , as it happens on the edges of infinite graphene ribbons . the numerical calculations of this section are done with a unit cell of width @xmath53 , length @xmath77 , where @xmath78 is the number of carbon atoms along an armchair chain , and with periodic boundary conditions along the @xmath79 direction to avoid spureous zigzag edges . we consider unit cells as long as 25 nm and the typical number of atoms in a self - consistent calculation is 1000 . importantly , our mean field results have the same relation between the sublattice imbalance and ground state spin than the exact state , as predicted by lieb theorem . we first revisit the single void samples . the ground state of structures with single atom vacancies have one unpaired electron within the @xmath81 model and , according to lieb s theorem , spin one half in the finite @xmath82 model . our mean field calculation , for @xmath83ev agrees with the lieb theorem . there is a spin splitting of the midgap state @xmath84 and a smaller spin splitting @xmath85 of the conduction and valence band states , as shown in fig . the spin degeneracy is thus broken , with only one of the spin channels of the midgap state occupied , the other being empty . this results in a finite magnetization density , localized around the vacancy , as shown in fig . [ figsingle ] . although the magnetization resides mostly in the majority sublattice , interactions induce some reversed magnetization in the other sublattice . [ hbt ] ( color online ) ( a ) the spin resolved dos for a ribbon with @xmath64 and @xmath83ev with one vacancy . spin @xmath86 ( @xmath87 ) is plotted as a positive ( negative ) number as a function of energy ( we take the fermi energy as zero ) . ( b ) zoom of the spin - split midgap state . ( c ) zoom of the conduction band minima . for clarity , we substitute the delta functions composing the dos by gaussian functions with a finite broadening.,title="fig : " ] we can gain some insight by doing an analytical description of the mean field results that involves some approximations valid when @xmath82 is much smaller than the band - gap of the ideal ribbon @xmath88 . in this case , we assume that only the midgap state is spin polarized : @xmath89 where @xmath90 is the @xmath81 square modulus of the midgap wave function . notice that the normalization of @xmath91 ensures that the total spin of the ground state is consistent with lieb s theorem , @xmath92 . the corresponding exchange splitting is @xmath93 [ hbt ] ( color online ) ( a ) non - interacting ( @xmath81 ) energy gap ( circles ) and interacting ( @xmath94 ) midgap spin splitting ( squares ) as a function of the ribbon width @xmath53 . ( b ) midgap spin splitting as a function of @xmath82 for two ribbon widths @xmath64 and @xmath95 . ( c ) @xmath81 inverse participation ratio for the midgap state . ( d ) standar deviation of the magnetization @xmath96 , as defined in eq . [ sigmajfr ] , as a function of @xmath82 for two ribbon widths . , title="fig : " ] we see that , within the simplified analytical model , the spin splitting of the midgap state is proportional to the inverse participation ratio , @xmath97 this quantity measures the degree of localization of the zero - energy state . an extended state in which the wave function is equally shared by @xmath98 atoms . has @xmath99 . in the opposite limit where the state is localized in a single atom we would have @xmath100 . the inverse participation ratio shown in fig . [ gaps](c ) corresponds to a number of atoms in the range @xmath101 to @xmath102 . as discussed above , the localization of the midgap states plays an important role in the minimal distance at which they are effectively decoupled . in fig . [ gaps](a ) we plot the @xmath81 gap of the ideal ribbon @xmath88 and the @xmath83ev spin splitting @xmath84 of the mid - gap state as a function of the ribbon width @xmath53 , as obtained from the full numerical calculation . as discussed above , we exclude the widths that give @xmath103 . we see that the midgap spin splitting is a decreasing function of @xmath53 . this is related to the fact that , in the small @xmath82 limit , @xmath84 is proportional to the inverse participation ratio @xmath104 , which is also a decreasing function of the ribbon width , as shown in fig . [ gaps](c ) . the extension of the midgap state increases as the ribbon becomes wider , resulting in a reduction of the midgap spin splitting . as @xmath88 tends to zero ( bulk graphene ) the midgap state becomes non - normalizable@xcite and @xmath84 is expected to vanish ( see below ) . whereas the total magnetic moment @xmath105 is given by the sublattice imbalance , the degree of localization of the spin texture is not . in order to quantify it we define the standard deviation @xmath106 in this definition @xmath96 is not normalized as usual by @xmath98 , the total number of atoms of the sample , since @xmath96 characterizes a localized object . for sufficiently large simulation cells , doubling @xmath98 would imply a decrease of a normalized @xmath96 without changing the local properties of the localized magnetic texture . notice that , within the analytical model valid for @xmath107 , we have @xmath108 . hence , in the absence of staggered magnetization , both @xmath96 and @xmath104 would measure the localization of the magnetic moments . for instance , if @xmath109 at all sites and @xmath110 , the maximal @xmath96 would be 0.5 . however , the graphene lattice responds with a staggered magnetization to the presence of defects and @xmath96 _ also _ measures the magnitude of that response . in figs . [ gaps](b ) and ( d ) we plot the mid - gap spin splitting @xmath84 and @xmath96 for two ribbons with @xmath64 and @xmath95 as a function of @xmath82 . the midgap spin splittings can be fitted to @xmath111 and to @xmath112 . according to eq . [ lineargap ] the linear coefficients should be compared to @xmath113 , which is 0.016 for @xmath114 and @xmath115 for @xmath116 . the non - linear terms arise from the interaction - driven mixing between the midgap states and the conduction states . this is also consistent with the fact that @xmath96 increases as a function of @xmath82 as shown in the lower panel of fig . [ gaps](d ) . the staggered magnetization is an increasing function of @xmath82 . the coefficient of the cuadratic term decreases with the length of the sample , namely , with the distance between vacancies since we are using periodic boundary conditions . we will come back to this issue in the last section . we have verified that the ground state of structures with single atom vacancies are locally neutral also with @xmath117 : the integrated electronic occupation in every site is one . hence , a localized spin texture with total spin @xmath57 occurs in the absence of any charge localization . our numerical results show that the addition of an extra electron to single vacancy structures results in a many electron state with total spin @xmath1 , local magnetization which is zero everywhere , and local charge accumulated in the same atoms and with the same distribution than the magnetic texture of the charge neutral structure . these results are shown in fig . [ spin - charge - sep ] for a ribbon with @xmath64 and @xmath94 . hence , it is apparent that the monoatomic vacancy results in a multielectronic state with spin charge separation : the neutral ground state has a net electric charge @xmath118 , but a total spin @xmath110 localized in a non - homogeneous spin texture in locally neutral atoms . the charged ground state has a net charge @xmath119 , total spin @xmath1 , no local magnetic moments , and a charge texture localized at the same location than the spin texture of the neutral ground state . this phenomenon resembles that reported by su - schrieffer - heeger@xcite in polyacetylene . [ hbt ] ( color online ) spin charge separation in single atom vacancy . left column : neutral case . right column : charged case . upper panels : charge density @xmath120 . lower panels : @xmath121 . the local charge and local spin are zero everywhere for the neutral and charged cases respectively . the spin texture of the neutral case is identical to the charge texture of the charged case . , title="fig : " ] we have also calculated the mean field magnetic structure for sublattice imbalanced larger voids . in fig . [ figsingle](b ) we show the magnetization profile for a triangular void with @xmath58 . for the chosen value of @xmath83 ev the staggered magnetization is barely visible in this scale . in agreement with lieb s theorem , it has a spin @xmath61 , made out of local moments localized , mostly , on the triangle boundaries . this object is the somehow complementary of the triangular graphene islands considered recently by two of us@xcite . figure [ rombo ] shows the ferrimagnetic spin texture around a rhomboidal void with imbalance charge @xmath122 , i.e. , composed of two triangular voids with @xmath123 . local moments with @xmath124 , three times smaller than those formed in the edges of infinite length zigzag ribbons , are formed on opposite corners of the void . we have verified that , for @xmath83 ev , the smallest void of this shape which features local moments is the one of the figure . the rhomboidal void is similar to the hexagonal islands considered in ref . in the sense that both have @xmath1 and develop local moments if they are sufficiently large . [ hbt ] ( color online ) emerging ferrimagnetic order in a rhombohedral void with imbalance charge @xmath122 situated the middle of a ribbon with @xmath125i for @xmath83ev . the largest magnetic moment per atom is @xmath126 , title="fig : " ] we now study the interaction between two magnetic defects with local sublattice imbalance @xmath65 . the lieb s theorem warrants that , when the sign of the sublattice imbalance is the same for the two defects , the total spin of the ground state is the sum of the spin of the individual defects . hence , they are coupled _ ferromagnetically_@xcite . in contrast , if the two defects have opposite sublattice imbalance so that the global sublattice imbalance is zero , lieb s theorem warrants that the total spin is zero . our calculations show that this can happen in two different scenarios : the local magnetization might be zero everywhere or the two defects could be magnetized along opposite directions , i.e. , could be coupled _ antiferromagnetically_. when the defects are sufficiently far apart from each other their local electronic structure should be identical to that of single defects . hence , the spin interaction between two magnetic defects can be either ferromagnetic or antiferromagnetic , as in the case of indirect exchange interactions ( rkky ) between single site magnetic moments@xcite , but can also result in the annihilation of the local magnetic order , an scenario that goes beyond the rkky picture . in fig . [ pairs ] we plot the normalized standard deviation of the two magnetic moments @xmath127 for ribbons with @xmath64 and @xmath95 as a function of the defect separation . we normalize the computed @xmath127 to the one corresponding to two independent single - defect magnetic textures , @xmath128 . here @xmath129 is the computed single vacancy standard deviation in the same ribbon . when the defects are sufficiently far away @xmath127 must tend to @xmath128 , i.e. , the normalized @xmath127 must tend to @xmath130 . we consider the effect of the ribbon width @xmath53 , interaction strength @xmath82 , and sublattice imbalance upon the magnetic interactions between the two defects . in the case of @xmath64 we show both monoatomic vacancies lying on the same sublattice ( a+a , open circles ) , whose ground state total spin is @xmath61 , and on different sublattices ( a+b , full circles ) , whose ground state spin is @xmath1 . the two curves for @xmath64 are calculated with @xmath83ev . at large distances , the two defects become decoupled , as expected . at short distances , the behaviour of the magnetic texture is radically different for both a+a and a+b structures . in the former case @xmath96 is enhanced , indicating the localization of the magnetic texture in a smaller region . since the total spin is @xmath130 , local moments survive even when the two defects are very close . as the separation between defects increases they become independent from each other and @xmath131 . when this happens , the energy gap between @xmath61 and @xmath1 should vanish . this is an example of rule number 1 . in contrast to the a+a case , the _ local _ magnetization of the a+b structure vanishes below a _ minimal distance _ this is an important result . in other words , there is a _ maximal density _ of defects above which zero energy states hybridize and local magnetic moments vanish . the critical density depends on the energy scales of the problem , the single particle gap @xmath88 , controlled by the ribbon width , and the on - site repulsion @xmath82 . for fixed @xmath82 the decoupling distance is definitely shorter for @xmath64 than for @xmath95 . hence , the critical ( linear ) density becomes _ smaller _ , as the ribbon width _ increases_. this is consistent with the fact that both the @xmath81 inverse participation ratio and the @xmath80 @xmath129 are decreasing functions of the ribbon @xmath53 . the wider the ribbon , the larger the delocalization of the zero energy state . hence , the hybridization between the midgap states associated with each vacancy survives at a larger distance for wider ribbons . finally , in fig . [ pairs ] we also show @xmath127 for @xmath95 and two values of @xmath82 , 2 and 4 ev . the decoupling distance ( critical density ) decreases ( increases ) as a function of @xmath82 . in other words , interactions drive the system magnetic , as expected . the a+b case is the simplest example that exemplifies rules number 2 and 3 . [ hbt ] ( color online ) normalized magnetization as a function of the distance between vacancies for four cases . ribbon with @xmath64 , @xmath83ev , vacancies in the same sublattice ( open circles ) , same ribbon , holes in different sublattices ( full circles ) , ribbont with @xmath95 , holes in different sublattices , @xmath83ev ( full squares ) and @xmath133ev ( stars),title="fig : " ] according to the two - vacancy calculation shown in the figure , the magnetic ( low density , large inter - vacancy distance ) and non - magnetic ( high density , low intervacancy distance ) phases are separated by a crossover region . if we take as an estimate of the critical distance below which local moments are quenched the distance for which @xmath134 , we find that @xmath135 for @xmath64 and @xmath136 for @xmath64 , both for @xmath83ev . the corresponding critical _ linear _ densities @xmath137 are @xmath138@xmath139 and @xmath140@xmath139 , respectively . the corresponding areal densities , @xmath141 are @xmath142@xmath143 and @xmath144@xmath143 respectively . these numbers should be taken as order of magnitude estimates of the real critical density . in the case of a+b pairs , the crossover from the locally magnetic to the non - magnetic state is similar to the one described in compensated graphene nanoislands@xcite : small islands are non - magnetic and larger islands have magnetic edges . the critical density depends on the extension of the magnetization , which in turn , depends on the ribbon single particle gap @xmath88 ( which controls the extension of the @xmath81 midgap states ) and on the on - site repulsion @xmath82 . the quenching of the local moments in the a+b structures is definitely related to the hybridization of the mid - gap states described in the non - interacting model . this phenomenon has an analog in zigzag ribbons . the midgap bands are linear combinations of top and bottom edge states . the hybridization is negligible in the brilloin zone boundary , and is much larger in the middle . as a result , the exchange interaction strongly renormalize the zone - boundary states , opening a magnetic gap , but they barely change in the middle of the zone@xcite . the physical picture that emerges from the previous discussion leads to an interesting conclusion : a semiconducting graphene ribbon with a density of vacancies that induce magnetism will behave like a diluted ( para-)magnetic semiconductor ( dms)@xcite provided that the density of defects is smaller than the critical density defined above ( this would be an example of rule number 3 ) . charged excitations will present a gap and spin excitations will not . the long range ferromagnetic order found by pisani _ ( see ref . ) only occurs when the vacancies are all in the same sublattice . it remains an open issue whether or not such a sublattice imbalance might occur in reality . unless this can be shown , one should not expect long range ferromagnetic order in real samples . interestingly , the conduction and valence bands depend on the magnetic order of the local moments , which might be induced by application of an external magnetic field . in the dms case , the conduction and valence bands are exchanged coupled to the local moments , provided by mn atoms . at zero field the mn spins are randomly oriented and the average spin splitting of the bands is zero . application of an external field orders the mn spins , resulting in a finite average exchange induced spin splitting of the bands which is much larger than the standard zeeman splitting . this is known as giant zeeman splitting . the same scenario might occur in semiconducting graphene ribbons with magnetic vacancies . when the inter - defect distance is larger than the critical spacing , @xmath132 , the local moments are independent . application of a magnetic field aligns them and induces an exchange - induced splitting of the conduction and valence bands much larger than the intrinsic zeeman splitting@xcite . in the case of graphene ribbons with vacancies we have computed the spin splitting of the bottom of the conduction band @xmath145 where @xmath146 is the first level above the mid - gap state [ see fig . [ dos](c ) ] . notice that the shift of the top of the valence band and the bottom of the conduction band is such that the gap , ignoring the midgap states , is _ spin - independent_. since we consider independent defects , the calculation is done with a single defect per unit cell . in fig . [ figure - dms ] we plot @xmath85 for a ribbon with @xmath64 . in the left panel we plot @xmath85 as a function of the inter - defect distance considering only values bigger than @xmath132 for which the defects are decoupled . for a @xmath83ev we find that @xmath85 ranges between 5 and 15 mev . this splitting could be obtained with an applied magnetic field such that @xmath147 , yet , @xmath148 . as in the case of real dms , this giant zeeman splitting scales linearly with the defect density , as shown in the inset of the right panel of fig . [ figure - dms ] . since there is a maximal density above which the local moments are coupled and eventually they vanish , the splitting @xmath85 can not be increased indefinitely . this phenomenon also has an analog in dms : direct antiferromagnetic coupling between mn spins eventually blocks the paramagnetic coupling to the external field . [ hbt ] ( color online ) left panel : bottom of the conduction band spin splitting , @xmath85 , as a function of the defect distance and as a function of the linear defect density ( inset ) for a ribbon with @xmath64 and @xmath1492ev . right panel : @xmath85 for the same ribbon for a fixed defect density as a function of @xmath82 . , title="fig : " ] in the right panel of fig . [ figure - dms ] we plot @xmath85 as a function of @xmath82 . for small @xmath82 we find that @xmath85 is almost linear with @xmath82 . this can be understood in the framework of the analytical model discussed above . if we consider that , to lowest order , the magnetization only comes from the mid - gap states with wave function @xmath91 and we compute the splitting of the bottom of the conduction band states to first order perturbation theory , we obtain : @xmath150 where @xmath151 is the @xmath81 single particle state of the bottom of the conduction band . if we approximate @xmath152 , where @xmath98 is the number of atoms in the unit cell , and we use the normallization condition of the midgap states , @xmath153 then we have : @xmath154 where @xmath155 is the number of magnetic vacancies per unit cell . this equation accounts also for the fact that @xmath85 scales linearly with the defect density . deviations from the linear behaviour arise due to the magnetization that arises from states other than mid - gap states . the strong sensitivity of the conduction states of the defective armchair ribbon on the application of a moderate magnetic field should give rise to strong spin - dependent magnetotransport and magnetooptic effects , in analogy with dms spintronic devices . notice that , in contrast to standard mn doped ii - vi semiconductors , for which electrical injection of carriers results in a new carrier mediated coupling@xcite , the addition of carriers in this system would results in the compensation of the mid - gap states and the disappearance of the local moments , as shown in fig . [ spin - charge - sep ] . so far we have considered the electronic structure of semiconducting graphene ribbons with vacancies and voids . as shown in fig . [ pairs ] the critical distance for the quenching of the magnetic moments increases with the ribbon width . an important question is whether or not this critical distance converges to a finite value in the two - dimensional limit . we have also seen in fig . [ gaps](d ) how the ( standard deviation of the ) magnetization @xmath96 associated with vacancies decreases as the gap of the ribbon decreases . in this section we address the question of what happens to these and other results obtained above in the limit of infinitely wide ribbons where the gap goes to zero , i.e. , bulk graphene . the extrapolation to the two - dimensional case is not straightforward . we thus consider a new strategy . here we consider unit cells with periodic boundary conditions in both directions . an infinite graphene crystal with a unit cell formed by @xmath51 parallel armchair - like chains , each of them containing @xmath78 carbon atoms . the dimension of this unit cell is @xmath156 being @xmath157 the graphene lattice parameter . we are interested in square unit cells and therefore we consider units cells @xmath158 of sizes ( 24,10 ) , ( 32,13),(40,16 ) , ( 48,20 ) , ( 60,26 ) , and ( 72,31 ) . we locate one or more vacancies in the unit cell considered and , as for ribbons , we obtain the eigenvalues , eigenfunctions , and the magnetization at each place of the system , @xmath159 , by solving self - consistently the hamiltonian . we locate a vacancy at the center of the unit cell in such a way that our system describes a square lattice of vacancies all located on the same type of atoms , e.g. , b in our notation . we obtain that , in agreement with lieb s theorem , the ground state of the system has a magnetic moment @xmath110 per vacancy . we also obtain the spin gap of the system @xmath84 , the wavefunction @xmath160 of the first empty state , and the local magnetization . in the limit of a large unit cell @xmath161 and @xmath162 should be the energy and the wavefunction , respectively , of an isolated vacancy . in two dimensions we can characterize the linear size of the vacancy wavefunction by its first moment @xmath163 where @xmath164 is the position of the vacancy and the sum is over all atoms of the unit cell . as discussed above , the bipartite character of the graphene lattice produces an antiferromagnetic coupling between the magnetization of the two sublattices of the system@xcite . we define the linear size of the magnetization in each sublattice as @xmath165 @xmath166 and @xmath167 have opposite signs , and their sum , @xmath168 indicates the extension of the net magnetization . , of the net magnetization , @xmath169 , and of magnetization in the sublattice a , @xmath170 , as function of the distance between vacancies . from top to bottom the panels correspond to hubbard constant , @xmath81 , @xmath171 and @xmath172 ev . the vacancies are located in sites type b. , width=340 ] in fig . [ size ] we plot the linear size of the vacancy wavefunction and of the magnetization as function of the distance between vacancies for different values of @xmath82 . the first thing to note is that the size of the wavefunction increases linearly with the size of the unit cell , practically independent of the value of @xmath82 . this result indicates that the electron - electron interaction almost does not affect @xmath173 which , when @xmath81 , becomes a quasilocalized state with weight in only one sublattice ( a ) and decays as @xmath174 , in agreeement with analytical results@xcite . when @xmath81 only the sublattice a is magnetized and @xmath169=@xmath175 . however , @xmath169 is considerable smaller than @xmath176 , indicating that the presence of the vacancy does not just creates a quasilocalized state , but also modifies strongly the wavefunctions of the states in the continuum . notice the difference with the ribbons where the magnetization follows the wavefunction for sizable confinement gaps . as we increase @xmath82 , the magnetic texture evolves in such a way that their size @xmath169 , as defined in eq . ( [ sizemag ] ) , decreases . this is accompained by an increase of the staggered magnetization , reflecting the antiferromagnetic tendency of the bipartite lattice , which polarizes the sublattice b in the opposite direction that sublattice a with @xmath177 . this effect can be rather dramatic for moderate values of @xmath82 , in fig . [ size ] we show that for @xmath178 ev the extension of the magnetization in sublattice a is considerable larger than @xmath179 and @xmath169 . note that @xmath180 ev is still below the critical value @xmath181 ev for the occurrence of an antiferromagnetic instability in perfect graphene@xcite . thus , a network of vacancies in the same sublattice would have a magnetic ground state , in agreement with lieb theorem , and enhanced staggered magnetization , compared to perfect graphene . we now consider the midgap spin splitting @xmath84 in two dimensional graphene with a finite density of vacancies in the same sublattice . in the previous case of semiconducting ribbons there was a strong indication that @xmath182 for any ribbon width in the single vacancy limit . hence , we might expect that @xmath84 vanishes in two dimensional graphene . when we have a finite density of defects , @xmath84 has also an inter - defect contribution arising from the hopping term . this mechansim is possible only if the midgap states have weight on the two sublattices . midgap states associated with defects in the same sublattice can only be coupled through interaction driven sublattice mixing . as the gap is a product of the coupling between the magnetizations induced by the hubbard interaction , the value of the gap increases quadratically with @xmath82 to lowest order . our calculations in fig . [ gap](a ) show that for @xmath183 , the midgap spin splitting @xmath84 goes to zero as the density of vacancies goes to zero . in fig . [ gap](b ) we plot the weight of @xmath162 on the sublattice where the vacancy is located . this quantity also tends to zero as the density of defects decreases . thus , in the single impurity limit the spin gap goes to zero and the interacting @xmath162 lives only on one sublattice . , width=340 ] in principle , one could expect that in real graphene samples the number of vacancies on sublattices a and b are roughly equal . in this situation lieb s theorem requires that the total spin of the system should be essentially zero . from the results in the case of ribbons one should expect an antiferromagnetic coupling between the magnetic moments for small concentrations of vacancies . for large concentrations the local magnetic moments should disappear and the sample should turn non - magnetic . in order to study the interaction between vacancies located in different sublattices and the local magnetization in a compensated system , we locate two vacancies with total sublattice imbalance equal to zero in the unit cell . the a and b impurities form two interpenetrated square lattices in such a way that the distance between impurities is maximum . we have checked that in agreement with lieb s theorem the @xmath0=0 solution is the ground state of the system . and @xmath184 forming two square lattice interprenetrated lattices . the inset indicates the variation of the standard deviation of the local magnetizations as function of the hubbard constant for a system of dimension ( 26,60 ) . the error bar is an estimation of the numerical error in the calculations . , width=340 ] we quantify the local magnetization studying the standard deviation of @xmath185 , @xmath186 where the sum is over all carbon atoms and @xmath98=@xmath187 is the number of atoms in the unit cell . in the inset of fig . [ u_cri ] we plot @xmath188 as a function of @xmath82 for a unit cell of size @xmath189 , which corresponds to a density of vacancies of @xmath190 cm @xmath191 . we obtain that for small values of @xmath82 the magnetization is zero everywhere and that there is a critical value of the hubbard coupling , @xmath192 , for which a local magnetization near the vacancies appears@xcite . this critical @xmath82 depends on the density of vacancies , and in fig . [ u_cri ] we plot @xmath192 as function of the density of vacancies . we obtain that @xmath192 decreases with the density of vacancies and from our results we conclude that the in the limit of zero density @xmath192 tends to zero . in that limit , each vacancy hosts a spin one half texture , which is decoupled from the others . for high density of vacancies and moderate values of @xmath82 , the kinetic energy coupling between the vacancies is stronger than the electronic repulsion , and in order to minimize the energy the system makes the local magnetization zero everywhere . when the density of impurities decreases , the kinetic energy coupling between vacancies decreases and eventually the impurities become uncoupled and each of them gets a total spin @xmath193 and the system behaves as a diluted antiferromagnetic system@xcite . unless a reason for global lattice imbalance exists ( unknown to us ) , for realistic values of @xmath82 in the range @xmath194 ev@xcite we expect that in highly irradiated graphene samples the local magnetization should be zero across the sample . some of the possible limitations of our approach have already been mentioned in the beginning . we have left aside the issue of the structural stability of passivated vacancies and voids . away from the edges , a single monoatomic vacancy might not result in a local atomic configuration that can be described with our model . on the other side , the effect of a hydrogen atom atop a carbon atom on graphene is very similar to the one described by our model@xcite . hence , the anomalous magnetic behaviour of irradiated graphite might be related to @xmath195 pairs , rather than to missing atoms . it is also important to signal that the mean field approximation is known to overestimate the appearence of magnetic order and yield critical values of @xmath196 smaller than those obtained with methods that include quantum fluctuations@xcite . finally , we have neglected both second neighbor hopping and interatomic coulomb repulsions . interestingly , both dft and mean field hubbard model yield a very similar description of the magnetic behaviour of graphene islands@xcite and zigzag graphene ribbons@xcite . this indicates that the couplings neglected in the simple hubbard model have a small effect on the low energy electronic structure that dominates the physical properties . we now summarize the main conclusions of this work . in the context of our model , the main results are : 1 . the electronic properties of the defects arising from the removal of atoms from graphene depend dramatically on an integer number , the sublattice imbalance ( or imbalance charge ) @xmath197 , which counts the difference in the total number of atoms per sublattice removed from a perfectly balanced graphene lattice . @xmath27 can take values @xmath198 etc . 2 . it can be rigorously shown@xcite that the single - particle spectrum of a structure with sublattice imbalance @xmath27 has , at least , @xmath10 midgap states per spin channel , occupied by @xmath10 electrons in neutral graphene . 3 . repulsive coulomb interactions will result in a many - body ground state with @xmath199 . this is an exact result in the case of the hubbard model@xcite . whereas the global electronic structure of a given graphene system is given by the lieb , the local structure is not . by assigning _ sublattice imbalance numbers to defects , provided that they are sufficiently apart , a set of rules to predict basic features of the magnetic structure has been proposed . 5 . we find that single voids with @xmath200 give rise to states with spin - charge separation , in the sense that a localized magnetic texture does not entail charge localization . the addition of a single electron to the system results in a many - body state with @xmath1 and the disappearance of the magnetic texture , which is substituted by a charge texture , as seen in fig . [ spin - charge - sep ] . in this sense , the properties of these states are very similar to su - schrieffer - heeger midgap states@xcite . the addition rules for two voids with a givel local sublattice imbalance or imbalance charge present similarities with those of vortices , e.g. , in superconductors . when sufficiently apart , two voids with local imbalance @xmath201 and @xmath202 behave like two independent objects with local spin @xmath199 . below a certain distance they annihilate each other and the local magnetization vanishes ( fig . [ pairs ] ) . when two voids with the same sign are brought together , they result in a region with enhanced local magnetization and spin @xmath203 , as seen in fig . [ figsingle ] . 7 . in analogy with graphene nanoislands@xcite , sufficiently large voids with @xmath9 can still have local magnetic moments . these can interpreted as if the large void with @xmath9 was the sum of two voids with @xmath204 . an example of this is the rhomboid of fig . [ rombo ] , obtained from merging two triangular voids with @xmath205 back to back . 8 . our results show that spin interactions between two magnetic defects of same @xmath10 can be of three types : ferromagnetic , antiferromagnetic or annihilating . in the first case , the ground state spin is the sum of the spins of the magnetic defects when infinitely apart , in the second case the spin is the difference between those two . in the third case both the total and the local spins are zero . antiferromagnetic and annihilating couplings occur in lattices without global sublattice imbalance , whereas ferromagnetic coupling requires global sublattice imbalance . our simulations show that , in balanced defective structures , there is a maximal density of monoatomic vacancies that can sustain local moments . in the case of ribbons , this critical density depends on the ribbon width . the critical density also depends on @xmath82 . a phase diagram for bulk graphene is provided in fig [ u_cri ] . 10 . depending on the density of vacancies , distributed randomly in the two sublattices , we distinguish three phases . in the very dilute limit the system is paramagnetic , with some common properties with ii - vi diluted magnetic semiconductors . in the opposite high - density limit the local moments are annihilated . the intermediate phase features antiferromagnetically coupled local moments . this would be a realization of the so - called diluted antiferromagnet@xcite 11 . we predict giant zeeman splitting in the case of semiconducting ribbons in the dilute limit . uppon application of an external magnetic field such that @xmath206 , the magnetic moment of all the defects would point parallel to applied field . this would result in a interaction induced splitting in the band states , much larger than the ordinary zeeman splitting , as seen in fig . [ figure - dms ] . 12 . a ferromagnetic phase is not expected for defective graphene , unless vacancies occur predominantly in one of the two sublattices . such an unbalanced situation would require further explanation . 13 . in the case of zero - gap graphene , we find that midgap states survive , even in the interacting case , in the very dilute limit ( fig . [ gap ] ) . since ideal graphene is a semimetal , the thermodynamics properties of graphene might be dominated by this type of defects . upon completion of this manuscript , a related work by o. v. yazyev ( axiv:0802.1735 ) has been reported . we acknowledge fruitful discussions with f. guinea , b. korgel , and p. lpez - sancho . this work has been financially supported by mec - spain under grants nos . mat2007 - 65487 , mat2006 - 03741 , and consolider csd2007 - 00010 , and by generalitat valenciana under grant no . acomp07/054 . h. ohldag , t. tyliszczak , r. hohne , d. spemann , p. esquinazi , m. ungureanu , and t. butz , phys . lett . * 98 * , 187204 ( 2007 ) . j. fernndez - rossier and j. j. palacios phys . . lett . * 99 * , 177204 ( 2007 ) .

we explore routes to realize electrically driven bose - einstein condensation of magnons in insulating antiferromagnets . even in insulating antiferromagnets , the localized spins can strongly couple to itinerant spins in adjacent metals via spin - transfer torque and spin pumping . we describe the formation of steady - state magnon condensates controlled by a spin accumulation polarized along the staggered field in an adjacent normal metal . two types of magnons exist in antiferromagnets , which carry opposite magnetic moments . consequently , and in contrast to ferromagnets , bose - einstein condensation can occur for either sign of the spin accumulation . this condensation may occur even at room temperature when the interaction with the normal metal is fast compared to the relaxation processes within the antiferromagnet . in antiferromagnets , the operating frequencies of the condensate are orders of magnitude faster than in ferromagnets .

1612.00366

bose - einstein condensation ( bec ) occurs in a wide variety of systems@xcite . at a sufficient density , magnons condense into a single bose quantum state . spectroscopically generated magnon condensates have been observed in ferrimagnetic insulators at room temperature@xcite . these findings imply that it is feasible to demonstrate coherent quantum phenomena using magnons . these effects could potentially be used in devices without the need for complicated cooling equipment . magnon bec manifests itself through a phase - coherent precession of the magnetization and an accompanying peak in the population of the magnons at the lowest energy spin - wave mode . associated with magnon bec is the possibility of realizing and controlling spin superfluidity@xcite . the superfluid properties could enable long - range dissipationless spin transport . in antiferromagnets , there are also reports of condensation and superfluidity induced by nuclear magnetic resonance@xcite . in this work , we explore an electrical route for controlling the bose - einstein condensation of magnons in antiferromagnetic insulators ( afis ) . spin pumping may be as operative from afis as from ferromagnetic insulators@xcite , in apparent contrast to nave intuition . this means that the dynamical precession of spins in afis may pump pure spin currents into adjacent normal metals that are as large as those of ferromagnetic insulators . the effectiveness of spin pumping from antiferromagnetic insulators to metals implies , via onsager reciprocity relations , that there is a considerable spin - transfer torque on the afis via spin accumulations in neighboring conductors@xcite . a spin accumulation can be generated via the spin hall effect or from other ferromagnets . the combination of significant spin - transfer torques and spin pumping enables terahertz antiferromagnetic spin hall nano - oscillators@xcite . for these reasons , antiferromagnetic insulators may be as effective as ferromagnetic insulators in spintronics devices . there is currently considerable interest in coupling the electronic properties of normal metals and ferromagnetic insulators . although there is increasing attention on antiferromagnetic insulators@xcite , they remain much less explored for spintronics purposes than their ferromagnetic counterparts . whereas there are predictions of electrically driven magnon condensation in ferromagnets@xcite , the case of antiferromagnets is entirely unexplored . in antiferromagnetic materials , the magnetic moments of the atoms exhibit a staggered ( nel ) order , which gives rise to long - range correlations between the moments . however , the order is such that the net magnetization vanishes in each unit cell . an interesting aspect of antiferromagnets is that the spin dynamics can be a thousand times faster than the magnetization dynamics in ferromagnetic systems . combining an antiferromagnetic insulator with a normal metal paves the way toward technological magnetic devices that operate at terahertz frequencies . to determine the feasibility of bose - einstein condensation of magnons in antiferromagnetic insulators , we generalize the theories of ( staggered ) spin transfer and spin pumping in normal metal - antiferromagnetic systems into the quantum domain with quantized spin wave excitations . to this end , we employ a quantum - mechanical model to describe both the localized electrons in the antiferromagnet and the conduction electrons in the metal . in the metal , a spin accumulation is assumed to exist via either the spin hall effect or spin injection from additional ferromagnets . we use a description in which the spins in the afi are exchange coupled to the itinerant spins in the metal . by computing the rates of change of the occupation of the magnons , we will determine the conditions for bose - einstein condensation driven by the spin accumulation . we will consider a quasi - equilibrated system , where the magnon lifetime is long compared to the thermalization time scale . we consider easy - axis antiferromagnets . in such antiferromagnets , there are two types of magnons . the difference between these magnons is that they carry magnetic moments in opposite directions along the easy axis . the two types of magnons have identical dispersions and will therefore be equally occupied at equilibrium . out - of - equilibrium , the two types of magnons are affected differently by the spin accumulation . consequently , we will show that magnon condensation occurs for either polarity of the spin accumulation . this distinct feature in antiferromagnets is because the spin - transfer torque dampens the excitations of magnons of one type but can dramatically change the state of the other type@xcite . the remainder of this paper is organized as follows . we introduce the hamiltonian describing the magnons in the antiferromagnet , the itinerant electrons in the normal metal , and the electron - magnon interaction across the afi - normal metal interface in sec . [ sec : dynamics ] . next , we compute the transport rate for the transfer of spin angular momentum across the interface in sec . [ sec : rates ] . in sec . [ sec : fixedt ] , we compute the conditions for bose - einstein condensation in systems where the magnon temperature is held fixed by another reservoir . we conclude our paper in sec . [ sec : conclusions ] . the appendices [ sec : scattering.magnons ] , [ sec : scattering.conduction ] , and [ sec : scattering.coefficients ] contain a substantial part of our work , including the microscopic calculation of the electron - magnon scattering amplitudes . the system consists of a normal metal , an antiferromagnetic insulator , and a metal - insulator interface . across the interface , the itinerant spins in the normal metal are exchange coupled to the localized spins in the antiferromagnetic insulator . the system is illustrated in fig . [ fig : intro.system ] . in the normal metal , the electrons are driven out of equilibrium , thereby resulting in a spin accumulation . we consider a scenario in which there is a spin accumulation that is polarized along the @xmath0 axis , which is the easy axis of the antiferromagnet . the spin accumulation is induced by the spin hall effect or by spin injection from ferromagnets@xcite . the antiferromagnet is shaped as a rectangular cuboid aligned with the simple cubic lattice , and it contains @xmath1 sites . there is one spin at each site . the axes of the lattice are aligned with those of the coordinate system defined in fig . [ fig : intro.system ] . we define two sublattices @xmath2 and @xmath3 . each sublattice contains half of the sites . in the classical ground state , all spins in sublattice @xmath2 are aligned along @xmath4 , and all spins in sublattice @xmath3 point in the opposite direction . both sublattices are face - centered cubic , and all nearest neighbors of a node in sublattice @xmath2 reside in sublattice @xmath3 and vice versa . we include spin - spin interactions between nearest neighbor atoms . furthermore , we assume that the interaction between any pair of nearest neighbors is the same . including an easy - axis anisotropy along the @xmath0 axis , the antiferromagnet hamiltonian becomes @xmath5 where @xmath6 is a pair of nearest neighbor sites . the indices @xmath7 and @xmath8 uniquely identify lattice sites , and @xmath9 indicates that site @xmath7 is in the antiferromagnet . we refer to the @xmath10-plane , where there is translation symmetry , as the transverse plane . spin transport takes place along the @xmath11 direction . the afi has left and right boundaries : the left one is at the normal metal - afi interface , and the right one is at the antiferromagnet - vacuum interface . the distance between these boundaries , @xmath12 , is the length of the afi . here , @xmath13 is the lattice constant . we assume that the afi is longer than the spin coherence length such that the magnon properties at the left and right boundaries of the afi are independent . similarly , the extent of the afi in the transverse directions is assumed to be sufficiently large such that we can disregard the details of the transverse boundaries . the @xmath14 and @xmath0 directions have widths @xmath15 and @xmath16 , respectively . we use periodic boundary conditions in the transverse plane . in this model , the two sublattices have the same reciprocal lattice and brillouin zone . the magnon momenta are defined within this brillouin zone . the brillouin zone of the full lattice is twice as large and contains twice as many possible momenta . using the holstein - primakoff transformation , we define two sets of magnon creation operators @xmath17 and @xmath18 . each operator corresponds to the excitation of a single spin at site @xmath7 ; @xmath19 and @xmath20 denote the corresponding sublattices . we expand the hamiltonian in powers of the magnon operators and disregard terms involving four or more magnons . the hamiltonian of eq . ( [ eq : dynamics.magnethamiltonian ] ) only results in terms involving an even number of magnons . we derive the energy eigenstates in appendix [ sec : scattering.magnons ] . most of the energy eigenstates are delocalized and extend throughout the magnet . additionally , there are surface states localized toward the interface . the decay length of the surface states , @xmath21 , is introduced in appendix [ sec : scattering.magnons ] . let us first consider surface states with energies , @xmath22 , that are of the same order as the gap , @xmath23 . in this case , the decay length , @xmath21 , is of the same order as the decay length of the magnon ground state @xmath24 . the latter is of the order @xmath25 . we assume that the anisotropy energy , @xmath26 , is sufficiently small compared to the exchange energy , @xmath27 , that @xmath25 is considerably longer than the length of the afi , @xmath12 . for the antiferromagnetic material rbmn@xmath28 , @xmath25 is on the order of @xmath29 mm@xcite . when @xmath25 is considerably longer than @xmath12 , the magnon surface states with energies on the order of the gap are approximately uniform throughout the afi . in appendix [ sec : scattering.magnons ] , we show that the uniform surface states are approximately equal to delocalized states with zero longitudinal momentum , @xmath30 . consequently , we model the behavior of the surface states using the uniform delocalized states . when the magnon energy , @xmath22 , is substantially higher than the gap , the surface state decay length , @xmath21 , is of the same order as the length of the afi . in appendices [ sec : scattering.magnons ] and [ sec : scattering.coefficients ] , we show that the magnon - electron interaction at the interface involving a surface state is comparable to that involving a delocalized state . for the surface states , the longitudinal momentum @xmath31 is determined by the transverse momentum , @xmath32 , via the boundary conditions , whereas @xmath31 is a free parameter for the delocalized states . there are therefore many more delocalized states than surface states . the latter non - uniform surface states only constitute a small fraction of the high - energy magnons and will be disregarded . with this information , we can model the behavior of all energy eigenstates using delocalized states , where the surface states are modeled as uniform delocalized states . we define two sets of delocalized states in terms of the annihilation operators @xmath33 and @xmath34 , respectively . here , @xmath35 is the wavevector of the incoming part of the delocalized state , and the superscripts @xmath36 and @xmath37 refer to the type of magnon . the @xmath36 ( @xmath37 ) magnons carry a spin angular momentum of @xmath38 pointing in the @xmath39 ( @xmath40 ) direction . the amplitude of a magnon in a delocalized state @xmath41 at node @xmath7 is described in terms of a wave function . we express the wave function using two continuous functions in real space ; each function provides the amplitudes associated with one sublattice . by interchanging those functions in the state @xmath42 between the two sublattices , we obtain the wave function associated with the state @xmath43 . the interchange corresponds to a reflection of all the spins of our system in the @xmath44-plane while simultaneously shifting the lattice by one lattice spacing in the transverse plane . because @xmath45 of eq . ( [ eq : dynamics.magnethamiltonian ] ) is symmetric both under the reflection and under the shift , the two types of magnons have the same dispersion relation . therefore , the magnon ground state is degenerate ; one degenerate state is of type @xmath36 , and the other is of type @xmath37 . we focus on the low - energy , long - wavelength spin - wave excitations , @xmath46 . the magnon dispersion for long wavelengths is derived in appendix [ sec : scattering.magnons ] . the result is well known , @xmath47 where the ( high - energy ) spin - wave velocity is @xmath48 and the spin - wave gap is@xcite @xmath49 we consider a normal metal with the same cubic lattice structure as the antiferromagnet . deviations from such an ideal system cause a renormalization of the electron - magnon scattering rates but do not change the main physics ; thus , they are not considered further . in a similar way as for the afi , we disregard the details of the boundaries in the transverse plane and use periodic boundary conditions . we assume that the scattering processes at the left and right boundaries of the normal metal are independent due to inelastic scattering in the bulk of the metal . we use a tight - binding model for the conduction electrons : @xmath50 where @xmath51 annihilates ( creates ) an electron with spin @xmath52 at node @xmath7 . here , @xmath53 indicates that site @xmath7 is in the normal metal . we assume half filling , and the fermi energy is then @xmath54 . we consider a local exchange interaction , with energy scale @xmath55 , between the spins in the antiferromagnetic insulator and the itinerant spins in the normal metal along the interface : @xmath56 where @xmath57 is a vector of pauli matrices . when the antiferromagnet is in its classical ground state , the exchange interaction of eq . ( [ eq : dynamics.interactionhamiltonian ] ) induces a static spin - dependent potential , @xmath58 , seen by the itinerant electrons at the interface . we determine the energy eigenstates of the electrons in the normal metal , including the interface potential @xmath58 . in appendix [ sec : scattering.conduction ] , we express the eigenstates as scattering states in a similar way as for the magnons in the antiferromagnet detailed in appendix [ sec : scattering.magnons ] . the annihilation operators of the scattering states are denoted as @xmath59 , where @xmath52 is the spin @xmath0 component of the electron spin and @xmath60 is a wavevector that is incoming to the afi - n interface . the conduction electron momenta are assumed to be near the fermi surface , and they are defined within the brillouin zone of the normal metal lattice . by expanding the interface exchange interaction of eq . ( [ eq : dynamics.interactionhamiltonian ] ) to the first order in the magnon operators , we find in appendix [ sec : scattering.coefficients ] an interaction of the form @xmath61 the sum over magnon and conduction electron momenta includes only the incoming wavevectors with respect to the interface . we separate the wavevectors @xmath60 into a part that is parallel to the direction of transport , @xmath62 , and a transverse part , @xmath63 . we assume that the exchange energy , @xmath64 , and the fermi energy are large compared to the other energy scales of the problem . in appendix [ sec : scattering.coefficients ] , we calculate the coefficients @xmath65 for the modes that are uniform in the direction of transport and the modes with finite longitudinal momenta @xmath66 separately , and we find that they differ by a factor@xcite of @xmath67 , @xmath68 the difference in the interface electron - magnon coupling of eq . ( [ eq : interface.vcoefficientfactor ] ) is the reason why the enhanced gilbert damping for the longitudinal finite wavelength modes is twice that of the longitudinal homogeneous mode@xcite . in the limit of a large exchange energy and a large fermi energy , we can show that the dominant contributions to the coefficients are @xmath69 as shown in appendix [ sec : scattering.coefficients ] . here , @xmath70 are kroenecker deltas for the transverse vector components , and @xmath71 . the parameter @xmath72 is introduced in appendix [ sec : scattering.magnons ] and is equal to the bogoliubov coefficient of the bulk model . furthermore , we have defined the umklapp - scattered momentum @xmath73 such that @xmath74 . we define @xmath75 in terms of energy conservation , @xmath76 . the amplitudes , @xmath77 , of eq . ( [ eq : interface.vcoefficients ] ) are directly proportional to the bogoliubov coefficient @xmath78 . because the exchange energy is considerably larger than the relevant magnon energies , @xmath79 thus , the inclusion of this factor strongly enhances the total transport rates . the amplitudes @xmath80 describe processes where a conduction electron creates or annihilates a magnon . the delta functions in the expression for @xmath81 ensure that the momentum of the conduction electrons is conserved in the transverse plane , but they also allow for a shift by a large momentum @xmath82 . we refer to the processes where transverse momentum is conserved as normal scattering processes and the processes where the transverse momentum is shifted as umklapp scattering processes . the terms proportional to @xmath83 in eq . ( [ eq : interface.vcoefficients ] ) are caused by the proximity effect . when @xmath84 , the proximity effect is negligible and the total rate of scattering involving magnons is low . when @xmath84 and electron scattering with a magnon does occur , the umklapp scattering process dominates the normal scattering process . we assume that the electron - magnon interaction at the interface is weak , and we treat it as a perturbation with respect to the decoupled systems of electrons and magnons . we now calculate the rate of change of the number of magnons of each type , @xmath85 , caused by the electron - magnon coupling at the interface . using fermi s golden rule , we find that @xmath86 \!\delta \!\left ( e _ { { f } } { - } e _ { { i } } \right ) \!. \label{eq : rates.fermicurrent}\end{aligned}\ ] ] in eq . ( [ eq : rates.fermicurrent ] ) , @xmath87 is the density matrix of the decoupled systems of electrons and magnons . the trace involves a sum over all quantum states . the first line of eq . ( [ eq : rates.fermicurrent ] ) represents the creation of magnons from an initial state , and the second line describes the annihilation of magnons . the difference between the creation and annihilation rates determines the magnon number rate of change @xmath85 . the creation and annihilation of magnons causes a change in the spin angular momentum of the itinerant electrons . @xmath88 and @xmath89 are the final and initial state energies of the non - interacting model , respectively . the magnons and conduction electrons are also affected by other interactions in the bulk of the materials . these interactions may be between the magnons or between the electrons , but there could also be interactions with other degrees of freedom , such as phonons . we consider bulk interactions that conserve the number of magnons and electrons . in other words , we assume that the magnon relaxation time is sufficiently long such that the magnon distributions can be experimentally observed . we use the phrase magnon number conserving process to describe interactions where the numbers of magnons of type @xmath36 and @xmath37 are conserved separately . treating the electron - magnon interaction at the interface as a perturbation , we assume that the bulk interactions are considerably faster than the interface interactions . furthermore , we assume that any bulk interactions that create or destroy magnons are slower than the interface interactions . the assumption that the magnon number conserving interactions are dominant is valid for the magnons in the ferrimagnet yttrium - iron - garnet ( yig)@xcite . the symmetry of the hamiltonian of eq . ( [ eq : dynamics.magnethamiltonian ] ) under rotations in spin space around the @xmath0 axis implies conservation of the total spin angular momentum along the @xmath0 direction . when the total spin angular momentum of the magnons is conserved , interactions conserving the total number of magnons also conserve the number of magnons of types @xmath36 and @xmath37 separately . the terms in the hamiltonian of eq . ( [ eq : dynamics.magnethamiltonian ] ) , where the magnon number is not conserved , simultaneously create ( annihilate ) both a @xmath36 and a @xmath37 magnon . this does not violate the conservation of angular momentum since the combined spin angular momenta of the two magnons in such pairs vanish . we disregard the processes where pairs of magnons are created or annihilated from our model . the remaining magnon - magnon interactions are examples of magnon number preserving bulk interactions . similarly , we assume that the magnon number is conserved in the magnon - phonon interactions . the bulk interactions can drive the magnons into a quasi - equilibrated distribution in a normal phase or a condensate phase . for both the normal phase and the condensate phase , there are two thermal baths of magnons with a bose - einstein distribution : one for each type of magnon . the electrons are assumed to be quasi - equilibrated with a fermi - dirac distribution , but we allow for a difference in the chemical potentials @xmath90 and @xmath91 of the spin - up and spin - down electrons . we define the spin accumulation as @xmath92 . the temperature of the conduction electrons is assumed to be independent of the spin . we will show that the two types of magnons are affected differently by the spin accumulation . the interactions across the interface are slow compared to the decoherence time of each subsystem . therefore , the density matrix of the coupled system is well approximated by a decoupled density matrix of the form @xmath93 . here , @xmath94 is the density matrix of the antiferromagnet , and @xmath95 is the density matrix of the normal metal . the quasi - equilibrated thermal states of the magnons and conduction electrons are described by the density matrices @xmath94 and @xmath95 , where @xmath96 here , @xmath52 is either spin up ( @xmath97 ) or spin down ( @xmath98 ) , and @xmath99 is the conduction electron energy . the presence of a bose - einstein condensate implies a macroscopic number of magnons in one or both of the magnon ground states , @xmath100 . in eq . ( [ eq : rates.thermalstates ] ) , @xmath101 ( @xmath102 ) is the chemical potential of the magnons of type @xmath36 ( @xmath37 ) , and @xmath103 , where @xmath104 ( @xmath105 ) is the effective temperature of the magnons of type @xmath36 ( @xmath37 ) . finally , @xmath106 , where @xmath107 is the temperature of the conduction electrons . we group the magnons into four contributions : the thermal magnons of types @xmath36 and @xmath37 , and the two possible condensates . the number of magnons in each group is denoted by @xmath108 and @xmath109 , respectively . with this information , we will now compute the rate of change of the magnon numbers @xmath108 due to electron - magnon scattering . to this end , we use eq . ( [ eq : rates.fermicurrent ] ) for the magnon current . we assume that the fermi energy of the conduction electrons and the exchange energy of the antiferromagnet are both considerably larger than the other relevant energy scales , such as the spin accumulation and the magnon gap . the dominant contribution to @xmath110 is @xmath111 in this expression , the coefficient @xmath112 that appears is defined in the following way . we define @xmath113 as the length of a wavevector @xmath35 , where @xmath114 and @xmath115 is the value of @xmath78 when @xmath116 . to the leading order , the bogoliubov parameter @xmath78 and the magnon energy @xmath117 only depend on @xmath35 through its magnitude . in eq . ( [ eq : rates.thermalcurrent ] ) , we have introduced @xmath118 here , @xmath119 is the volume of the antiferromagnet and @xmath120 is the volume of the normal metal , @xmath121 @xmath122 is the density of states of the magnons of type @xmath36 @xmath123 , and @xmath124 is the conduction electron density of states at the fermi surface . similarly , the current into the condensate is @xmath125 where @xmath126 the energy integral in eq . ( [ eq : rates.thermalcurrent ] ) runs over the energy of the magnons that are created or annihilated at the interface . only magnons with energies that are considerably smaller than the exchange energy and the fermi energy contribute to the current @xmath110 . therefore , it is sufficient to evaluate @xmath127 in the limit where @xmath128 is substantially smaller than the exchange energy and the fermi energy . in this limit , we find that @xmath129 where @xmath130 is the number of lattice nodes in one transverse layer . note that the currents of eq . ( [ eq : rates.condensatecurrent ] ) and eq . ( [ eq : rates.thermalcurrent ] ) are proportional to the large factor @xmath131 . these factors do not occur in the corresponding expressions for a ferromagnetic system@xcite . the implication is that the electron - magnon coupling at the interface is considerably stronger in antiferromagnets than in ferromagnets . consistent with this finding , a similar enhancement occurs in the heat transfer across afi - n interfaces@xcite . the density of states for the conduction electrons is approximately constant because the dynamics involve only electrons near the fermi surface . the magnon density of states , however , vanishes near the ground state energy . the densities of states of the conduction electrons and the magnons are respectively @xmath132 we assume periodic boundary conditions when estimating the densities of states of the afi . in appendix [ sec : scattering.magnons ] , we introduce a reflection angle , @xmath133 , that determines the magnon amplitude near the interface . we find that the reflection angle is much smaller than @xmath29 for all delocalized states with energy @xmath128 on the order of the gap . for magnons with energies that are much larger than the gap , we can disregard the gap in the dispersion relation , eq . ( [ eq : scattering.magnons.dispersion ] ) . using the approximate dispersion , we find that the reflection angle is of the order @xmath134 . typically , since the magnon wavelength is considerably longer than the lattice spacing , the reflection angle is also much smaller than @xmath29 in this regime . however , there are exceptions when the momentum vector @xmath35 is almost parallel to the interface , e.g. , when @xmath31 is relatively small . nevertheless , only a small portion of the delocalized states in the thermal bath have momenta parallel to the interface since the magnon dispersion is isotropic . because the delocalized states with finite reflection angles only constitute a small portion of the thermal cloud , and interact slower with the conduction electrons than the rest of the thermal bath , their contributions to the magnon currents in eq . ( [ eq : rates.venergy ] ) are negligible . the two types of magnons can both condense , and they also form separate thermal clouds , which could have different distributions . we assume that the temperatures of both clouds remain fixed by an external reservoir , such as phonons . we denote the common magnon temperature as @xmath135 . first , we consider the steady - state _ normal phase _ where no condensates are present . in this regime , the magnon chemical potentials for the two types together with the temperatures determine the magnon distribution . the magnon chemical potentials are smaller than the magnon gap , @xmath136 . the vanishing of the net currents of magnons into the antiferromagnet of eq . ( [ eq : rates.thermalcurrent ] ) determines the chemical potentials . we directly observe from eq . ( [ eq : rates.thermalcurrent ] ) that when there is no thermal bias , @xmath137 , the magnitudes of the chemical potentials equal the spin accumulation , @xmath138 . a finite thermal bias causes the magnitudes of the chemical potentials to deviate from the spin accumulation . we consider the case where @xmath135 is large compared to the magnon gap @xmath23 and the relative difference between @xmath107 and @xmath135 is small . in this limit , we find that the chemical potentials are @xmath139 the riemann zeta function is @xmath140 for integral numbers @xmath141 . when the magnon chemical potential for one type approaches the magnon gap , the number of magnons in the associated ground state becomes vast and macroscopic , creating a condensate . however , the chemical potential never increases beyond the magnon gap . in the _ condensate phase _ , one of the magnon chemical potentials , @xmath101 or @xmath142 , is equal to the gap , @xmath23 . in the condensate phase , we also need to determine the number of condensate magnons , @xmath143 and @xmath144 . a spin accumulation in the normal metal increases the number of magnons of one type and decreases the number of magnons of the other type . this asymmetry causes the creation of only one condensate . without a loss of generality , we assume that the spin accumulation @xmath145 is negative , which implies that only the condensate of type + is present . the dynamics of the system can now be described by @xmath146 [ eq : fixedt.numbers ] where @xmath147 is the net magnon transfer rate from the condensate into the thermal cloud . @xmath147 is caused by magnon number conserving bulk interactions . next , we will find steady - state solutions of the dynamical equations ( [ eq : fixedt.numbers ] ) and the threshold values of the spin accumulation and temperature for inducing a condensate . when the spin accumulation is larger than the gap , swasing can occur . the electron - magnon coupling - induced current of eq . ( [ eq : rates.condensatecurrent ] ) is proportional to the number of magnons already present in the condensate . this feature can lead to an exponential increase in the number of magnons in the condensate . it is this phenomenon that is a swasing instability@xcite . the direction of the current of eq . ( [ eq : rates.condensatecurrent ] ) changes when the spin accumulation increases beyond the gap . beyond the swasing instability , when the spin accumulation is larger than the gap , the number of magnons in the condensate rapidly increases , forming a large condensate . the growth of the condensate will continue until higher - order interactions between the magnons prevent a further population build - up . however , a large spin accumulation @xmath148 does not necessarily lead to swasing . to study this further , we need to take into account interactions between the thermal bath of magnons and the condensate , given by @xmath147 . such bulk interactions are fast and ensure that the magnons in the cloud remain in a non - equilibrium thermal distribution . even when the spin accumulation is larger than the gap , there could be stable steady - state solutions where @xmath149 . when @xmath149 , the system is in the normal phase . the normal phase is stable when the thermal cloud current of eq . ( [ eq : rates.thermalcurrent ] ) vanishes for a magnon chemical potential that is smaller than the gap , @xmath150 . this condition for the stability of the normal phase state does not depend on the relative magnitude of the spin accumulation and the gap . therefore , eq . ( [ eq : fixedt.chemicalpotential ] ) is valid even for large spin accumulations as long as both @xmath101 and @xmath142 are below the threshold , @xmath23 , when we maintain the assumption that @xmath135 is much larger than @xmath151 and @xmath152 . as shown in eq . ( [ eq : fixedt.chemicalpotential ] ) , the normal phase remains stable even when the spin accumulation is larger than the gap provided that the conduction electron temperature @xmath107 is smaller than the magnon temperature @xmath135 . however , when @xmath107 is larger than @xmath135 and @xmath148 , the normal phase is unstable and swasing will occur . the cases when @xmath148 and a normal phase state is a possible steady state solution are hysteretic@xcite . whether swasing or a normal phase occurs depends on the past history . we now further consider the case in which the spin accumulation is smaller than the gap , @xmath153 . if the conduction electron temperature @xmath107 is larger than that of the magnons , @xmath135 , the normal phase may become unstable even though the spin accumulation is small . in other words , a condensate can form for spin accumulations that are too small to induce swasing . when no stable normal phase state exists , the current @xmath154 of eq . ( [ eq : rates.thermalcurrent ] ) is always positive , and the chemical potential @xmath155 will increase until it approaches @xmath156 . when @xmath156 , @xmath154 can be fully determined using eq . ( [ eq : rates.thermalcurrent ] ) . we will denote this current by @xmath157 . in the limit where @xmath135 is considerably larger than @xmath151 and @xmath152 , we find that @xmath158 furthermore , the number of thermal magnons is fixed . in this case , using the conservation of the number of thermal magnons @xmath159 , eq . ( [ eq : fixedt.numbers ] ) and eq . ( [ eq : rates.condensatecurrent ] ) , we find that @xmath160 we define a time scale @xmath161 , and we use this time scale to express the solution of eq . ( [ eq : fixedt.condensateeqofmotion ] ) , @xmath162 from eq . ( [ eq : fixedt.condensatesolution ] ) , we observe that the steady - state number of condensed magnons is @xmath163 and independent of the initial conditions . in this steady state , @xmath164 is negative , which indicates that magnons are leaving the condensate across the interface . however , the condensate is maintained by the transfer from the thermal cloud . this is caused by the bulk interactions that redistribute the magnons from the cloud and into the condensate . in turn , the spin accumulation supplies magnons to the thermal cloud . this steady state is a magnon bose - einstein condensate@xcite . we have explored the prospect of bose - einstein condensation of magnons in insulating antiferromagnets coupled to normal metals . condensation occurs when large numbers of magnons are created at the interface . the creation of magnons at the interface can be stimulated by a spin accumulation in the normal metal and a temperature difference across the interface . starting from a quantum - mechanical model , we describe the dynamics of the antiferromagnet in terms of two types of magnons , which carry the same energy and opposite spin angular momentum . both types of magnons strongly interact with the conduction electrons in an adjacent normal metal compared to magnons in ferromagnets . the spin accumulation , which can be induced via the spin hall effect , causes an imbalance of the magnon distribution between the two types of magnons . for large imbalances , a coherent magnon condensate is created in one of the two degenerate uniform magnon ground states . in contrast to the ferromagnetic model , a condensate can form for both signs of the spin accumulation . the research leading to these results has received funding from the european research council via advanced grant number 669442 `` insulatronics '' . we will now calculate the magnon eigenstates of the semi - infinite heisenberg antiferromagnet . in a similar way , appendix [ sec : scattering.conduction ] presents the computation of the electron eigenstates in the normal metal . finally , we use the magnon and electron eigenstates to compute the interface electron - magnon coupling in appendix [ sec : scattering.coefficients ] . we are only interested in magnons with energies that are considerably lower than the exchange energy . however , to ensure that we find the correct solutions for small but finite magnon energies , we first find all magnon eigenstates and then take the low - energy limit . we define the sublattices @xmath2 and @xmath3 as described in sec . [ sec : dynamics ] . we represent the spin operators at each lattice site in terms of boson operators by using the holstein - primakoff transformation , @xmath165 [ eq : scattering.magnons.holsteinprimakoff ] where @xmath166 annihilates a magnon on site @xmath7 , which belongs to sublattice @xmath2 , and @xmath167 is the corresponding annihilation operator for a magnon at site @xmath7 in the @xmath3 sublattice . we assume that all the localized spins in the antiferromagnet have quantum number @xmath168 . other quantum numbers can be accounted for by redefining the exchange and anisotropy energies . the index @xmath7 can be decomposed into three integer indices @xmath169 corresponding to the location of a lattice site along the three axes @xmath11 , @xmath14 and @xmath0 . the sites where @xmath170 are in the normal metal , and the sites in the afi satisfy @xmath171 . in real space , the lattice site @xmath7 is at coordinate @xmath172 . we choose the coordinate system such that @xmath173 in the afi and @xmath174 in the normal metal , where @xmath175 is the thickness of the interface . the nodes in the antiferromagnet that are adjacent to the interface are located in the transverse plane @xmath176 . similarly , nodes , @xmath7 , in the normal metal that are adjacent to the interface satisfy @xmath177 . furthermore , we define transverse coordinates as @xmath178 . analogously , for wavevectors @xmath179 , we define the transverse wavevector @xmath180 . we also choose the indices such that the node @xmath181 is on sublattice @xmath2 . to second order in the boson operators , the hamiltonian of eq . ( [ eq : dynamics.magnethamiltonian ] ) becomes @xmath182 as discussed in sec . [ sec : dynamics ] , we use periodic boundary conditions in the transverse plane . we assume that the longitudinal length is larger than the inelastic scattering length and disregard the boundary conditions at the right boundary opposite to the interface . in this section , we will determine the energy eigenstates of the magnons . we find the magnon eigenstates with the ansatz @xmath183 [ eq : scattering.magnons.def.creationannihilation ] we have to determine the longitudinal ( @xmath11 ) dependence of the energy eigenstates represented by the functions @xmath184 and @xmath185 . the hamiltonian of eq . ( [ eq : scattering.magnons.hamiltoniannode ] ) relates the functions @xmath186 and @xmath187 by @xmath188 u_{{\ensuremath\boldsymbol{q}}}(x_i ) + \\ & \phantom{{}=1 } \frac { j \hbar^2}{2 } \left ( 2 \gamma^{\perp}_{{\ensuremath\boldsymbol{q}}_{\perp } } v_{{\ensuremath\boldsymbol{q}}}^{*}(x_i ) + v_{{\ensuremath\boldsymbol{q}}}^{*}(x_i-{d } ) + v_{{\ensuremath\boldsymbol{q}}}^{*}(x_i+{d } ) \right ) , \notag \\ 0 & = \left[(3j + k_z)\hbar^2 + e_{{\ensuremath\boldsymbol{q}}}\right ] v_{{\ensuremath\boldsymbol{q}}}(x_i ) + \\ & \phantom{{}=1 } \frac{j\hbar^2}{2 } \left ( 2 \gamma^{\perp}_{{\ensuremath\boldsymbol{q}}_{\perp } } u_{{\ensuremath\boldsymbol{q}}}^{*}(x_i ) + u_{{\ensuremath\boldsymbol{q}}}^{*}(x_i-{d } ) + u_{{\ensuremath\boldsymbol{q}}}^{*}(x_i+{d } ) \right ) \notag\end{aligned}\ ] ] [ eq : scattering.magnons.bulkequation ] for all nodes , @xmath7 , in the bulk of the afi . we have defined @xmath189 [ eq : scattering.magnons.def.gamma ] at the normal metal - afi boundary , we find the conditions @xmath190 u_{{\ensuremath\boldsymbol{q}}}({x_{i } } ) + \\ & \phantom{{}=1 } \frac { j \hbar^2}{2 } \left ( 2 \gamma^{\perp } _ { { \ensuremath\boldsymbol{q}}_{\perp } } v_{{\ensuremath\boldsymbol{q}}}^*({x_{i } } ) + v_{{\ensuremath\boldsymbol{q}}}^*({x_{i}}+{d } ) \right ) , \notag \\ 0 & = \left[(\frac{5}{2}j + k_z)\hbar^2 + e_{{\ensuremath\boldsymbol{q}}}\right ] v_{{\ensuremath\boldsymbol{q}}}({x_{i } } ) + \\ & \phantom{{}=1 } \frac{j\hbar^2}{2 } \left ( 2 \gamma^{\perp } _ { { \ensuremath\boldsymbol{q}}_{\perp } } u_{{\ensuremath\boldsymbol{q}}}^{*}({x_{i } } ) + u_{{\ensuremath\boldsymbol{q}}}^{*}({x_{i}}+{d } ) \right ) . \notag \end{aligned}\ ] ] [ eq : scattering.magnons.interfaceconditionstart ] in the remainder of this appendix , we will obtain the solutions to eq . ( [ eq : scattering.magnons.bulkequation ] ) and eq . ( [ eq : scattering.magnons.interfaceconditionstart ] ) . we first determine bulk solutions that solve eq . ( [ eq : scattering.magnons.bulkequation ] ) only , and then we combine the bulk solutions to fulfill the boundary condition of eq . ( [ eq : scattering.magnons.interfaceconditionstart ] ) . in bulk antiferromagnets , the system is translationally invariant in all directions . in this case , momentum is a good quantum number . the bulk equations , eq . ( [ eq : scattering.magnons.bulkequation ] ) , are solved by plane waves , @xmath191 and @xmath192 . this solution provides the momentum eigenstates @xmath193 [ eq : scattering.magnons.def.momentumcreationannihilation ] the bogoliubov parameters @xmath194 and @xmath195 satisfy the normalization condition @xmath196 . from eq . ( [ eq : scattering.magnons.bulkequation ] ) , we find the dispersion relation for the momentum eigenstates , @xmath197 in an infinite system , the momenta @xmath179 must be real valued since there are no boundaries where evanescent states can originate . in a semi - infinite system , there are momentum eigenstates both for complex and real - valued quantum numbers @xmath179 . when @xmath179 is real , the states represent propagating magnons with momentum @xmath179 . the momentum eigenstates with complex @xmath179 are evanescent states . the broken symmetry at the interface implies that the energy eigenstates are linear combinations of left- and right - going propagating states as well as evanescent states . explicitly , this follows from the boundary condition of eq . ( [ eq : scattering.magnons.interfaceconditionstart ] ) . all parts of the energy eigenstates have identical energy in the dispersion of eq . ( [ eq : scattering.magnons.dispersion ] ) and the same transverse momentum , @xmath198 . we classify the energy eigenstates according to the momentum of the incoming propagating wave , @xmath179 . the momentum eigenstate with incoming momentum @xmath179 can elastically scatter into three other momentum eigenstates . the reflected momentum of one of these states is @xmath199 . the other two states have momenta @xmath200 , which satisfy @xmath201 , whereas the transverse momentum remains @xmath202 . in total , the energy eigenstates are linear combinations of momentum eigenstates with momenta @xmath179 , @xmath203 , @xmath200 and @xmath204 . for evanescent momentum eigenstates , we only consider states that are localized near the interface . three types of energy eigenstates exist . there are i ) eigenstates containing only plane waves , ii ) purely evanescent states , and iii ) combinations of plane waves and evanescent waves . it can be shown that the eigenstates of type i ) have energies on the order of the exchange energy . such high - energy magnons are not relevant to our discussion of the low - energy physics . therefore , all the states of type i ) are disregarded from this point . type ii ) and type iii ) states have a wide range of energies from the magnon gap and up to the order of the exchange energy . we will first write expressions for the magnon eigenstates of type ii ) and type iii ) that are valid for all energies . later , we will consider the limiting case where the magnon energy is considerably smaller than the exchange energy . the evanescent surface states of type ii ) are described by@xcite @xmath205 [ eq : scattering.magnons.surfacestate ] where the longitudinal wave number @xmath31 is purely imaginary . the imaginary wave number @xmath206 and the real number @xmath207 are determined by fulfilling both the dispersion relation of eq . ( [ eq : scattering.magnons.dispersion ] ) and @xmath208 which follows from the boundary conditions of eq . ( [ eq : scattering.magnons.interfaceconditionstart ] ) . the coefficients @xmath209 , @xmath210 , @xmath211 , and @xmath212 appearing in the surface states of eq . ( [ eq : scattering.magnons.surfacestate ] ) should be determined from the bulk and boundary conditions of eq . ( [ eq : scattering.magnons.bulkequation ] ) and eq . ( [ eq : scattering.magnons.interfaceconditionstart ] ) and the normalization condition . for each value of the transverse momenta @xmath213 and @xmath214 , there are two surface states ; one of type @xmath36 and one of type @xmath37 . these states have the same energy and carry opposite angular momentum . the delocalized states of type iii ) are given by the functions @xmath215 [ eq : scattering.magnons.bulkstates ] the real - valued parameter @xmath216 is governed by @xmath35 via the dispersion relation of eq . ( [ eq : scattering.magnons.dispersion ] ) . the incoming longitudinal momentum @xmath31 is a free parameter , and the boundary conditions of eq . ( [ eq : scattering.magnons.interfaceconditionstart ] ) determine the angle @xmath133 , @xmath217 where we have introduced @xmath218 . the coefficients @xmath219 , @xmath220 , @xmath211 , and @xmath212 of the delocalized states are governed by the bulk and boundary conditions of eq . ( [ eq : scattering.magnons.bulkequation ] ) and eq . ( [ eq : scattering.magnons.interfaceconditionstart ] ) and the normalization condition . the magnon wavefunctions given in sec . [ sec : scattering.magnons.semi ] are eigenstates of the hamiltonian of eq . ( [ eq : scattering.magnons.hamiltoniannode ] ) , but the expressions for the magnon currents are complicated . in this section , we consider the low - energy limit , where the magnon wavefunctions are simpler . we consider a magnon temperature @xmath135 that is considerably smaller than the nel temperature , @xmath221 . in this low - energy limit , the magnon energy is much smaller than the exchange energy , @xmath64 . in this section , we study the momentum eigenstates of sec . [ sec : scattering.magnons.infinite ] in the low - energy limit . when the energy @xmath222 is considerably smaller than the exchange energy , we see from the dispersion of eq . ( [ eq : scattering.magnons.dispersion ] ) that @xmath179 satisfies @xmath223 . for low - energy propagating plane waves with momentum @xmath179 , @xmath224 . the low - energy momentum eigenstates with @xmath225 are staggered evanescent states . we consider the limit of long wavelength and small anisotropy . in other words , we perform expansions in @xmath226 and @xmath227 . the leading order terms in this expansion for the bogoliubov parameters , @xmath194 and @xmath195 , are given by @xmath228^{1/2}$ ] . the wavefunctions of the staggered evanescent states decay exponentially and oscillate rapidly along the @xmath11 axis within each sublattice . using the dispersion relation of eq . ( [ eq : scattering.magnons.dispersion ] ) , we expand @xmath229 in powers of @xmath226 and @xmath227 . we determine the wavefunctions of the staggered evanescent states in the long wavelength and small anisotropy limit using this expression for @xmath229 . to the leading order in the long wavelength and small anisotropy expansion , the decay length , @xmath230 , of the staggered states is @xmath231 . we now discuss the surface states of type ii ) , as given in eq . ( [ eq : scattering.magnons.surfacestate ] ) . we expand in the small parameters @xmath232 and @xmath227 . the non - staggered part of the wavefunction decays exponentially in the longitudinal ( @xmath11 ) direction with a decay length @xmath233 . the staggered part has a decay length @xmath230 . expanding in powers of @xmath234 and @xmath227 , we find to leading orders @xmath235 [ eq : scattering.magnons.limksurface ] substituting the expression eq . ( [ eq : scattering.magnons.limksurface.imq ] ) into eq . ( [ eq : scattering.magnons.dispersion ] ) , we find that to the leading order @xmath236 . here , we have defined @xmath237 . this means that the low - energy surface state has approximately the same energy as a delocalized state that is uniform along the @xmath11 direction . we will show in this section that we can approximate the low - energy surface states using uniform delocalized states . the decay length @xmath238 of the surface state is given by eq . ( [ eq : scattering.magnons.limksurface.imq ] ) . in the limit of long wavelength and small anisotropy , @xmath239 [ eq : scattering.magnons.limsolutionsurface ] the results in eq . ( [ eq : scattering.magnons.limksurface ] ) and eq . ( [ eq : scattering.magnons.limsolutionsurface ] ) determine the surface state wavefunction completely except for the normalization factor , @xmath209 . the condition for normalization , @xmath240 follows from the commutation relations between the magnon operators , @xmath241 . the staggered part of the wavefunction does not contribute to eq . ( [ eq : scattering.magnons.surfacenormalization ] ) because the normalization condition is dominated by contributions from the bulk of the afi and the staggered part of the wavefunction is localized near the interface . in eq . ( [ eq : scattering.magnons.surfacenormalization ] ) , @xmath242 is a bogoliubov parameter of the momentum eigenstate with momentum @xmath243 . when @xmath244 , the surface state is approximately uniform along the @xmath11 direction , and the normalization factor @xmath209 is approximately equal to the bogoliubov parameter @xmath242 . in sec . [ sec : dynamics ] , we showed that the relevant surface states to our problem are approximately uniform along the @xmath11 direction . these states are approximately equal to uniform delocalized states , meaning that their wavefunctions are approximately equal everywhere in the afi . we use the uniform delocalized states to model the behavior of the surface states . the uniform delocalized states are defined by eq . ( [ eq : scattering.magnons.bulkstates ] ) for @xmath30 and @xmath245 and satisfy @xmath246 . like the surface states , the uniform delocalized states fulfill eq . ( [ eq : scattering.magnons.limsolutionsurface ] ) . we now consider the delocalized states of type iii ) in the low - energy limit . in general , these states are given by eq . ( [ eq : scattering.magnons.bulkstates ] ) . the reflection angle , @xmath133 , is determined from eq . ( [ eq : scattering.magnons.phicondition ] ) . in the long - wavelength and low - anisotropy limit , we find that the reflection angle @xmath133 is @xmath247 . the smallest possible value of the momentum @xmath31 is on the order of the inverse of the length of the afi . this lower bound provides an upper boundary for the reflection angle , @xmath133 , of the order @xmath248 . we have shown in sec . [ sec : dynamics ] that @xmath249 is considerably longer than the afi length , @xmath12 , for magnon energies , @xmath117 , on the order of the gap , @xmath23 . therefore , the reflection angle @xmath133 is small for magnons with energies on the order of the gap . we have shown in sec . [ sec : rates ] that the reflection angle @xmath133 is also small for thermal magnons with energies that are considerably higher than the gap . consequently , we model the behavior of all the states of type iii ) using delocalized states with a vanishing reflection angle . we consider the long - wavelength and small - anisotropy limit . for small reflection angles , the delocalized states are described in terms of coefficients that , as the surface states , fulfill eq . ( [ eq : scattering.magnons.limsolutionsurface ] ) . however , to the leading order in the long - wavelength and small - anisotropy expansion , the normalization factor differs and is @xmath250 the normalization factor for a state of type iii ) is larger than that of a uniform delocalized state by a factor of @xmath67 . the enhancement is caused by constructive interference between the incoming and reflected waves . next , we aim to determine the scattering states in the normal metal . in the bulk , the hamiltonian is governed by eq . ( [ eq : dynamics.metalhamiltonian ] ) . we will find the energy eigenstates of the conduction electrons when the antiferromagnet is in the classical ground state . in sec . [ sec : rates ] , we included the coupling of conduction electrons and magnons as a perturbation . the exchange coupling of eq . ( [ eq : dynamics.interactionhamiltonian ] ) results in a spin - dependent potential at the interface that represents the proximity effect . the total hamiltonian for the conduction electrons is @xmath251 , where the interface proximity effect is captured by @xmath252 we consider a half - filled system . transport is governed by the states near the fermi surface . in this case , we assume that the scattering states are of the form @xmath253 the coefficient @xmath254 represents specular reflection , and the coefficient @xmath255 describes umklapp reflection . these coefficients are elements of the @xmath256 matrix . the @xmath256 matrix is unitary , which ensures the conservation of particle number . because the @xmath256 matrix is unitary , the coefficients @xmath254 and @xmath255 satisfy @xmath257 and @xmath258 . these conditions can be used to determine @xmath254 and @xmath255 . the normal metal has the same lengths @xmath259 and @xmath260 in the @xmath14 and @xmath0 directions as the afi , @xmath261 and @xmath262 . the normal metal extends a finite length @xmath263 in the negative @xmath11 direction from @xmath264 . the number of sites in the normal metal is @xmath265 . the plane wave states , @xmath266 , are eigenstates of @xmath267 except at the interface . the plane wave states and the scattering states both satisfy the dispersion relation @xmath268 + 6 t . \label{eq : scattering.conduction.dispersion}\ ] ] the scattering state @xmath269 consists of one incoming plane wave state with respect to the interface and two reflected ( normal and umklapp ) plane wave states . the wavevector @xmath60 satisfies @xmath270 and is therefore incoming . we require the scattering states of eq . ( [ eq : scattering.conduction.creationannihilationwavenormal ] ) to be eigenstates of @xmath271 . from the hamiltonian at the sites along the interface , we find the boundary conditions @xmath272 defining @xmath273 , the solutions to the boundary and unitarity conditions are @xcite @xmath274 in the limit of weak or strong exchange interaction at the interface , @xmath275 or @xmath276 , the umklapp scattering amplitude @xmath255 and hence the proximity effect are negligible . the influence of the antiferromagnet attains its maximum when the exchange interaction is comparable to the next - neighbor hopping parameter . the umklapp scattering amplitude can be related to the spin - mixing conductance@xcite . in appendices [ sec : scattering.magnons ] and [ sec : scattering.conduction ] , we have found the energy eigenstates of the noninteracting system of magnons and conduction electrons . in this section , we determine the dominant interaction term , starting from the interface interaction of eq . ( [ eq : dynamics.interactionhamiltonian ] ) . we will arrive at eqs . ( [ eq : dynamics.hamiltonianinterface ] ) and ( [ eq : interface.vcoefficients ] ) . we perform the holstein - primakoff transformation as described in sec . [ sec : dynamics ] , and we expand eq . ( [ eq : dynamics.interactionhamiltonian ] ) in powers of magnon operators . the zero - order terms have already been accounted for in appendix [ sec : scattering.conduction ] , and therefore , we disregard them here . disregarding higher - order terms in the magnon operators , we find that @xmath277 we now substitute the delocalized states derived in appendices [ sec : scattering.magnons ] and [ sec : scattering.conduction ] into eq . ( [ eq : scattering.coefficients.interactionhamiltonian ] ) . after this substitution , we arrive at eq . ( [ eq : dynamics.hamiltonianinterface ] ) , which defines the interaction amplitudes @xmath80 . @xmath80 is determined from the amplitude of the delocalized states at the interface . the amplitude of a magnon state @xmath42 ( @xmath43 ) along the interface is given by @xmath278 on the @xmath2 ( @xmath3 ) sublattice and @xmath279 on sublattice @xmath3 ( @xmath2 ) . to the leading order in the magnon exchange energy @xmath280 , we find that @xmath281 the wavefunction of the conduction electron scattering state @xmath282 at the interface has two components . one component oscillates with momentum @xmath283 and the other oscillates with momentum @xmath284 as we move along the interface . this mixing is caused by the proximity effect . the absolute value of the amplitude of each component is constant along the interface . this absolute value is @xmath285 for the component with momentum @xmath283 and @xmath286 for the component with momentum @xmath284 . the delta functions @xmath288 ensure conservation of momentum in the transverse plane . four different combinations of delta functions appear because each of the conduction electron states @xmath282 has two components . these delta functions in turn give rise to two different momentum conditions in eq . ( [ eq : interface.vcoefficients ] ) . 42ifxundefined [ 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 http://link.aps.org/abstract/rmp/v71/p463 [ * * , ( ) ] http://link.aps.org/doi/10.1103/physrevlett.84.5868 [ * * , ( ) ] http://dx.doi.org/10.1038/nature01617 [ * * , ( ) ] http://link.aps.org/doi/10.1103/physrevlett.95.127202 [ * * , ( ) ] http://dx.doi.org/10.1038/nature05117 [ * * , ( ) ] \doibase http://www.nature.com/nature/journal/v443/n7110/suppinfo/nature05131_s1.html [ * * , ( ) ] link:\doibase 10.1126/science.1140990 [ * * , ( ) ] http://dx.doi.org/10.1038/nature09567 [ * * , ( ) ] link:\doibase 10.1134/s0021364011130066 [ * * , ( ) ] http://www.sciencedirect.com/science/article/pii/0038109878902259 [ * * , ( ) ] http://dx.doi.org/10.1080/00018731003739943 [ * * , ( ) ] http://link.aps.org/doi/10.1103/physrevlett.108.177002 [ * * , ( ) ] http://link.aps.org/doi/10.1103/physrevb.89.024511 [ * * , ( ) ] http://link.aps.org/doi/10.1103/physrevlett.112.227201 [ * * , ( ) ] http://link.aps.org/doi/10.1103/physrevb.90.094408 [ * * , ( ) ] http://link.aps.org/doi/10.1103/physrevlett.115.237201 [ * * , ( ) ] http://link.aps.org/doi/10.1103/physrevlett.116.117201 [ * * , ( ) ] http://link.aps.org/doi/10.1103/physrevlett.115.156604 [ * * , ( ) ] http://link.aps.org/doi/10.1103/physrevlett.116.257205 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/physrevlett.108.246601 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/revmodphys.87.1213 [ * * , ( ) ] @noop * * , ( ) @noop ( ) @noop * * ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) _ _ , @noop ph.d . thesis , ( )

we examine whether essence and quantitative aspects of electronic excitation spectra are correctly captured by an effective low - energy model constructed from an _ ab initio _ downfolding scheme . a global electronic structure is first calculated by _ ab initio _ density - functional calculations with the generalized gradient approximation . with the help of constrained density functional theory , the low - energy effective hamiltonian for bands near the fermi level is constructed by the downfolding procedure in the basis of maximally localized wannier functions . the excited states of this low - energy effective hamiltonian ascribed to an extended hubbard model are calculated by using a low - energy solver . as the solver , we employ the hartree - fock approximation supplemented by the single - excitation configuration - interaction method considering electron - hole interactions . the present three - stage method is applied to gaas , where eight bands are retained in the effective model after the downfolding . the resulting spectra well reproduce the experimental results , indicating that our downfolding scheme offers a satisfactory framework of the electronic structure calculation , particularly for the excitations and dynamics as well as for the ground state .

0710.4371

first - principles electronic - structure calculations based on density - functional theory @xcite(dft ) within the local density approximation ( lda ) or the generalized gradient approximation ( gga ) for the exchange - correlation ( xc ) functional have opened a way to predict ground - state properties of various materials without introducing _ ad hoc _ parameters . however , there exist serious problems in which the dft fails even qualitatively . typical examples are found in strongly - correlated electron systems such as the genuine mott insulator , where an insulating gap opens in partially filled bands solely owing to the strong local electron - electron repulsion.@xcite the dft with lda / gga often predicts metals for these systems,@xcite indicating the fact that the xc functionals based on lda / gga do not correctly capture the local correlation in real space . other typical example is found in dynamics and excitation spectra of electrons , in which many - body correlation effects are also essential.@xcite even semiconductors , being supposed to belong to weakly - correlated electron systems in the ground state , may have highly - degenerate excited states arising from the local electron correlation effects , and thus the single - particle approximations such as the kohn - sham @xcite and hartree - fock @xcite schemes break down in general . it is well known that incorporating two - particle interactions between electrons and holes generated by the excitation is crucial in describing the electronic structure at low - energy levels . a typical example is found in excitonic excitations.@xcite to treat these excitations properly , we clearly need to go beyond the single - particle theory , while a full _ ab initio _ calculation taking into account the many - body correlation effects is practically intractable . to go beyond the lda / hartree - fock levels , we are required to develop a sufficiently accurate but efficient and practically feasible method . this challenge , so - called `` beyond lda / hartree - fock '' problem has attracted growing interest.@xcite the gw method @xcite has been developed to incorporate self - energy effects basically on the level of the random phase approximation ( rpa ) while strong correlation and fluctuation effects beyond the rpa level require a more accurate and reliable treatment . especially , an _ ab initio _ three - stage scheme has been rapidly developed by combining two procedures , namely , lda / hartree - fock framework and accurate low - energy solvers.@xcite the global electronic structure is first obtained by the lda / hartree - fock scheme . in the next stage , one performs a bridging treatment , that is _ downfolding _ , @xcite by eliminating the high - energy degrees of freedom leaving the low - energy effective model ( hamiltonian or lagrangian ) for local bases like wannier functions.@xcite the downfolding determines parameters for the effective low - energy model via first - principles calculations . the resulting low - energy model is , in the final stage , solved by low - energy reliable solvers such as dynamical mean field theory,@xcite path - integral renormalization group , @xcite and/or various monte - carlo methods @xcite developed for treating the correlation effects . such a hierarchical three - stage scheme instead of a full _ ab initio _ calculation allows us to perform a first - principles and parameter - free prediction of the electronic structure of the strongly - correlated electron system within the present feasibility of computer . in this paper , we present theoretical studies on the _ ab initio _ downfolding scheme to assess the reliability for treating dynamical properties . in our scheme , maximally localized wannier functions are introduced as a basis function for representing the model hamiltonian . this basis offers computationally convenient choice , because this wannier function can be computed with any basis functions ( plane wave,@xcite linearized muffin - tin orbital,@xcite linearized augmented plane wave,@xcite etc ) . transfer parameters are evaluated by calculating kohn - sham matrix elements in this basis , and onsite / offsite interaction parameters including screening effects are determined via constrained calculations.@xcite in the three - stage scheme , the reliability of the downfolding procedure and the accuracy of the resulting model parameters are crucially important . in particular , the reliability in describing dynamics and excitation spectra has to be critically tested . for this purpose , we make a critical comparison between experimental results and computational results for the generated model . in the present study , we focus on optical - absorption properties . it is widely recognized and accepted in the literature that the optical absorption of solids , in particular for semiconductors or insulators , is deeply affected by excitonic effects.@xcite this effect originates from an effective coulomb interaction between electrons and holes and therefore is sensitive to the magnitude of the interaction parameters in the model hamiltonian . to examine this effect through the present formalism , we choose gaas as a representative material exhibiting spectral enhancement due to the excitonic effect and calculate its optical spectra by taking account of the electron - hole interaction . there exist many experimental @xcite and highly - accurate _ ab initio _ @xcite spectral data for this material . therefore , our downfolding formalism and determined parameters can be critically tested by examining whether our model spectrum reproduces those data satisfactorily . this paper is organized as follows : in sec . ii we describe our downfolding procedure ; we introduce a complete - neglect - differential - overlap model which is used as our target model hamiltonian and describe computational details for determining the model parameters . in sec . iii , to take into account the electron - hole interaction , we introduce a single - excitation - configuration - interaction framework for calculating an optical absorption . efficient techniques to evaluate one - body velocity matrix elements needed in the spectral calculation , based on the wannier interpolation scheme , is described in appendix . the calculated optical spectra are compared with the experimental results . concluding remarks are given in sec . the first procedure derives the global electronic band structure by a conventional dft scheme . the present scheme is based on _ ab initio _ density functional calculations with _ tokyo ab initio program package _ @xcite developed by the condensed - matter - theory group in the university of tokyo . with this code , band calculations have been performed within the generalized gradient approximation @xcite to the exchange correlation functional , using a plane - wave basis set and the troullier - martins norm - conserving pseudopotentials @xcite in the kleinman - bylander representation.@xcite the energy cutoff is set to 25 ry , and a 15@xmath015@xmath015 @xmath1-point sampling is employed to represent electronic structures of the system . the resulting global band structure of gaas at an energy region [ @xmath215 ev:+ 30 ev ] is illustrated in the top panel of fig . [ fig : band : global ] ev:@xmath3 ev ] . ( bottom ) _ ab initio _ original ( solid line ) and interpolated ( dots ) bands . energy zero is set to the top of the valence bands . the energy window is set to [ @xmath4 ev:@xmath5 ev].,title="fig:",scaledwidth=40.0% ] + ev:@xmath3 ev ] . ( bottom ) _ ab initio _ original ( solid line ) and interpolated ( dots ) bands . energy zero is set to the top of the valence bands . the energy window is set to [ @xmath4 ev:@xmath5 ev].,title="fig:",scaledwidth=44.0% ] now we go onto the second stage and start the derivation of an effective low - energy hamiltonian by the downfolding procedure . before going to the downfolding itself , we first specify the form of the final effective low - energy hamiltonian . in the present downfolding procedure , we will make several approximations by simplifying the low - energy effective hamiltonian , in which we expect that the approximations do not alter the optical spectra seriously . the first approximation is to consider only the diagonal coulomb interaction and ignore the offdiagonal part ( exchange interaction ) in the low - energy effective model , which results in an extended hubbard model or , in other words , the complete - neglect - differential - overlap ( cndo ) model . the cndo hamiltonian has originally been introduced by pople @xcite to study electronic structures of small organic molecules and , to date , has been extended to study various properties of complicated systems ranging from transition - metal compounds @xcite to proteins @xcite and dna.@xcite a remarkable property of this hamiltonian is that it considers all the degrees of the freedom of valence electrons of the system , which allows describing the individual characters of the real material . the crystal cndo hamiltonian @xmath6 consists of a one - body part @xmath7 and an interaction part @xmath8 : @xmath9 the one - body part @xmath7 is given by @xmath10 where @xmath11 ( @xmath12 ) is a creation ( annihilation ) operator of a valence electron with spin @xmath13 in @xmath14-type localized basis centered at @xmath15th site in lattice @xmath16 . as mentioned in the introduction , we use the maximally localized wannier function ( mlwf ) as the basis function for representing the cndo hamiltonian . in the present study of gaas , there are eight wf s in the primitive cell , where the first four belong to a ga site , and the remaining four belong to an as site . thus , the index @xmath14 specifies four types of lobe directions ( band indices ) of the mlwf s , and the suffix @xmath15 specifies the ga or as sites . @xmath17 and @xmath18 are the ionization potential and the transfer integral , respectively . notice that the translational symmetry in the crystal is explicitly considered for matrix elements ; @xmath19 @xmath17 and @xmath20 @xmath21 for any @xmath22 and @xmath23 . the interaction part @xmath8 is written as @xmath24 here , @xmath25 , @xmath26 , and @xmath27 are the number operators , and @xmath28 is the core charge . @xmath29 and @xmath30 are the onsite intraorbital and interorbital coulomb repulsions , respectively . @xmath31 in the third term is an interatomic coulomb repulsion , and it is assumed to be independent of the lobe directions @xmath14 and @xmath32 . we now describe the downfolding procedure and parameterization for the cndo model of eq . ( [ eq : cndo ] ) . the downfolding consists of two parts . the first is the derivation of the kinetic energy part , where a tight - binding hamiltonian @xmath33 is derived . the second part is the derivation of the interaction part @xmath34 . the tight - binding hamiltonian @xmath33 given in eq . ( [ eq : h_t ] ) is derived from the global band structure after eliminating higher - energy bands . this downfolding may be performed by the perturbation scheme.@xcite the resultant band structure is normally very close to the low - energy part of the original band structure and the difference is not discernible when the low - energy retained part is isolated from the eliminated high - energy bands.@xcite this means that the self - energy of the retained bands caused by the higher - energy eliminated electrons is negligible . since such self - energy effects are smaller even in semiconductor systems , in this paper , we employ the low - energy part of the bands as the retained bands after the elimination of the higher - energy bands . now we retain eight bands near the fermi level and construct wannier orbitals from the retained band structure . to this end , _ ab initio _ mlwf s are constructed with the souza - marzari - vanderbilt algorithm.@xcite we set an energy window in the interval [ @xmath4 ev:@xmath5 ev ] , which includes four valence and four conduction bands of the system . the resulting ga and as mlwf s are displayed in the top and bottom panels of fig . [ fig : mlwf ] , respectively . we see that the wannier functions are almost localized at a single site and have an anisotropic character due to an @xmath35 hybridization . to show the accuracy of low - energy band structures represented by the resultant wf s , we compare in the bottom panel of fig . [ fig : band : global ] original bands ( solid line ) with interpolated bands ( dots ) obtained by diagonalizing @xmath1-space kohn - sham ( ks ) hamiltonian matrices represented by the wf s . we see a good agreement between the original and interpolated bands in the energy window . ionization potential @xmath17 and transfer integral @xmath36 are extracted from the matrix elements of the one - body ks hamiltonian @xmath37 in the basis of the mlwf s @xmath38as @xmath39 respectively . here @xmath40 with @xmath41 , and \{@xmath42 } are primitive lattice vectors . ( blue ) and @xmath43 ( red ) , where @xmath44 is the volume of the primitive cell . the shaded sheet represents a 3@xmath03@xmath03 fcc lattice and ga and as nuclei are illustrated as gray and blue dots.,title="fig:",scaledwidth=40.0% ] ( blue ) and @xmath43 ( red ) , where @xmath44 is the volume of the primitive cell . the shaded sheet represents a 3@xmath03@xmath03 fcc lattice and ga and as nuclei are illustrated as gray and blue dots.,title="fig:",scaledwidth=40.0% ] we next derive the interaction parameters @xmath29 , @xmath30 , and @xmath45 for the low - energy model . in the original cndo framework , the coulomb interaction between electrons has long - ranged tail as scaled by @xmath46 in the limit @xmath47 with @xmath48 being a distance between two electrons . we are , however , interested in the electronic interaction in condensed phase for which the coulomb interaction is effectively screened . in fact , the dielectric constant of gaas is rather high ; @xmath49 experimentally,@xcite indicating that the coulomb interaction decays as @xmath50 in the limit @xmath47 . we thus employ an approximation for this interaction ; we keep only onsite interactions @xmath29 and @xmath30 and the nearest - neighbor interaction @xmath51 . these parameters are determined with a constrained dft framework following a hopping - cutoff " treatment . @xcite as the basic strategy , one first kinetically decouple a specific site from the rest of the system , thus leaving this site isolated as the so - called atomic limit . this decoupling treatment is achieved by switching off the hoppings between the wannier orbitals at the specific site and all the other orbitals , where we identify the hoppings with the off - diagonal kohn - sham matrix elements in the representation of the wannier functions . with such a hopping cutoff , the standard constrained total energy calculations are performed to generate a potential energy surface with respect to constrained parameters such as occupancies of the wannier orbitals belonging to the decoupled site . the interaction parameters obtained with quadratic fitting to the resulting potential energy data include screening effects ascribed to the relaxation of the valence electron density around the decoupled site . in the present case , the procedure for determining the interaction parameters is somewhat complicated because of a large number of the interaction parameters to be derived . so , in the practical work , we divide the treatment into two steps ; the determination of @xmath52 and @xmath53 and the subsequent step for determining an offsite parameter @xmath51 . the basic strategy for obtaining @xmath52 and @xmath53 is to generate potential - energy - data sets with respect to two types of the constrained parameters ; ( i ) the first potential - energy data are obtained from the constrained calculations with respect to occupancy @xmath54 of a specific wannier orbital @xmath14 at the site @xmath55 , and ( ii ) another data are obtained from the constrained calculations for a site occupancy @xmath56 defined as the total amount of the orbital occupancies belonging to this site ; @xmath57 . the curvature of the first potential energy curve plotted as a function of the orbital occupation @xmath54 gives an estimate of the onsite intraorbital interaction for the orbital @xmath14 at the site @xmath55 , while the curvature of the second data represents the averaged value over the onsite intraorbital / interorbital interactions . from quadratic fitting to the mixed two data , we can determine the @xmath52 and @xmath53 parameters reasonably ( see below ) . though the types of the constraints are different in the two calculations , the method itself is the same , so , here , we describe only details for the constrained calculation with respect to the single orbital occupancy . the practical calculation proceeds as follows : we first consider a @xmath58 fcc supercell containing @xmath59 atoms , @xcite and choose the central ga site placed at the origin as the decoupled site . ( here , we describe only the ga case , but a parallel treatment can be applied to the @xmath52 and @xmath53 determination for the as site . ) we next introduce a cutting operator @xmath60 to switch off the hopping integrals connecting the four wannier orbitals of this site to the other wannier orbitals , @xmath61 where @xmath62 is a one - body ks hamiltonian , and @xmath63 is a projector onto the total wannier orbitals , @xmath64 here , @xmath65 is a lattice vector denoting a supercell , the suffix @xmath15 specifies the sites in the supercell , and @xmath14 stands for the band index of the wannier orbital . @xmath66 in eq . ( [ eq : cut ] ) is a projector onto the wannier orbitals belonging to the decoupled @xmath55 site in the home cell ( @xmath67 ) , @xmath68 with the cutting operator in eq . ( [ eq : cut ] ) , we define the constrained total energy as @xmath69 + \frac{1}{n } \sum_{\bm k } \sum_{\alpha } f_{\alpha { \bm k } } \bigl\langle \phi_{\alpha { \bm k } } | \hat{\lambda}_{cut } | \phi_{\alpha { \bm k } } \bigr\rangle + \lambda_{\mu i { \bm 0 } } \left [ \frac{1}{n } \sum_{\bm k } \sum_{\alpha } f_{\alpha { \bm k } } | \bigl\langle w_{\mu i { \bm 0 } } | \phi_{\alpha { \bm k } } \bigr\rangle |^{2 } - q_{\mu i { \bm 0 } } \right ] \biggr\}. \label{eq : e_ctot}\end{aligned}\ ] ] here , @xmath70 $ ] is a usual density functional with a total charge density @xmath71 with @xmath72 being the total number of @xmath1 points , the first term in the bracket @xmath73 $ ] is the definition itself for the orbital occupancy of the disconnected wannier orbital @xmath74 , and @xmath75 is the lagrange multiplier associated with the constraint to fix the orbital occupancy at @xmath76 . a functional derivative of @xmath77 with respect to the bloch orbital @xmath78 leads to the following constrained ks equation , @xmath79 | \phi_{\alpha { \bf k } } \rangle = \epsilon_{\alpha { \bm k } } | \phi_{\alpha { \bm k } } \rangle , \label{eq : kseq } \end{aligned}\ ] ] where the third term in the left hand side is an additional potential due to the occupancy constraint . for numerical details for solving the equation , readers are referred to ref . 29 . by solving the equation , we generate constrained potential energy data ( we refer to these data as data i ) , and plot them as functions of the orbital occupancy @xmath76 and the site occupancy @xmath80 . in parallel to this treatment , the constrained calculations for the site occupancy are performed , where there is a small modification in the constrained ks equation ( [ eq : kseq ] ) ; the additional potential is changed to @xmath81 . we again monitor the calculated constrained total energies as functions of @xmath76 and @xmath82 ( data ii ) . with the resulting data i and ii , we perform quadratic fitting of the following function around @xmath83 and @xmath84 : @xmath85 where @xmath86 and @xmath87 are equilibrium occupancies taken from the global band structure with no additional potential ( @xmath88 ) . the form of the fitting function in eq . ( [ eq : fitting ] ) is derived by exploiting the character of the fourfold degeneracy of the wannier orbitals @xmath89 ( appendix a ) . the @xmath90 and @xmath91 values thus obtained are 2.39 ev and 2.17 ev , respectively , which are largely reduced from the bare interaction values @xmath92 ev and @xmath93 ev . the same procedure is applied to the @xmath52 and @xmath53 determinations of the as site . it was found to be @xmath94 ev and @xmath95 ev . the corresponding bare values @xmath96 and @xmath97 are 11.45 ev and 9.80 ev , respectively . we next describe the determination of the offsite parameter @xmath51 . the interaction depends on the relative configuration between the wannier orbitals . for example , let us consider a configuration formed by a pair of the ga wannier orbital and the as wannier orbital , where these orbitals face along the covalent bond of the two atoms ( we call it _ facing configuration _ ) . an example for the facing configuration can be found in the two wannier orbitals displayed in the top and bottom panels of fig . [ fig : mlwf ] . obviously , the strength of the coulomb repulsion in the facing configuration is relatively large compared to that in the other configurations . it should be noted here that the @xmath51 parameter affects _ renormalized _ transfer integrals [ see eq . ( [ eq : fock_r ] ) in sec . ii d for the explicit form of the renormalized transfer integral ] . since bonding and anti - bonding orbitals are formed by the wannier orbitals in the facing configuration , the bonding and anti - bonding splitting , i.e. , band gap itself , is dominated by the magnitude of a renormalized transfer integral between the wannier orbitals in the facing configuration . thus , the @xmath51 value in this configuration is crucial for an accurate description of the low - energy band structure and optical excitations . the contributions from the intersite interaction in the other configurations are much smaller in optical response . the result would not change when we slightly overestimate them by @xmath51 in the facing configuration , because the renormalized part to the transfer integral appears as the product of @xmath51 and a density matrix [ see eq . ( [ eq : fock_r ] ) ] and it was found that the intersite density - matrix elements are almost zero except for that of the facing configuration in the present gaas case . therefore , we calculate the @xmath51 value in the facing configuration and employ it as the @xmath51 value of the cndo model . the actual determination of the @xmath51 parameter proceeds as follows : we first choose two decoupled sites ( the ga site placed at the origin and the neighboring as site being in the [ 111 ] direction ) . the similar cutting treatment to eq . ( [ eq : cut ] ) but extending the single - site - cutting formalism to the double - sites - cutting formalism is applied for this purpose . then , we perform constrained calculations by imposing a constraint that two occupancies of the wannier orbitals @xmath74 and @xmath98 in the facing configuration are kept at @xmath76 and @xmath99 , respectively . we then draw the two - dimensional potential energy surface with respect to the constrained parameters @xmath76 and @xmath99 , and perform a fitting of the quadratic function @xmath100 to the potential energy data . this fitting is performed by fixing @xmath101 and @xmath102 at predetermined values in the preceding @xmath52 and @xmath53 determination ; we treat only @xmath51 as a single fitting parameter to avoid fitting errors and uncertainties . the @xmath51 value thus determined is 0.71 ev , where we again see the large reduction from the bare interaction value @xmath103 ev . the resulting interaction parameters are summarized in table [ param ] . .interaction parameters determined in the present downfolding procedure . the energy unit is ev . [ cols="^,^,^,^,^,^",options="header " , ] [ param ] in general , the downfolded model contains an energy dependence in the interaction because the screening by the high - energy electrons necessarily causes a non - markoffian and retardation effect . such an energy dependence in the screened coulomb interaction @xmath104 is not described by the effective hamiltonian . however , in the low - energy region , the screened coulomb interaction is normally saturated to a constant and represented by the limiting value at @xmath105 . the constrained scheme is roughly regarded as the procedure to obtain this @xmath106 limit . the effect of the larger ( less screened ) interaction @xmath107 at larger @xmath108 as well as the effective interaction arising from virtual transition to eliminated bands can be accounted by the further consideration of the self - energy effect on the low - energy part.@xcite for a wide band system such as gaas , however , this effect may not be large @xcite and we thus ignore this effect . in the electronic structure calculation of a typical semiconductor , gaas , we expect that the strong correlation effect does not appear in the ground state and the hartree - fock ( hf ) approximation provides us a reasonable result , although the excitation spectra are reliably determined only through the account of the correlation effect more accurately . in this section , we consider how the hf solution for the ground state is calculated . the cndo hf hamiltonian in the basis representation of mlwf s is written as @xmath109 where @xmath110 is the fock matrix or the renormalized transfer matrix by the interaction part @xmath8 of eq . ( [ eq : h_c ] ) and the matrix element is given by @xmath111 u_{i } $ \\ $ \displaystyle + \sum _ { { \bm r } '' k } \left [ q_{k}({\mbox{\boldmath $ r$}}''-{\mbox{\boldmath $ r$ } } ) - z_{k } \right ] v_{ik}({\mbox{\boldmath $ r$}}''-{\mbox{\boldmath $ r$ } } ) $ } } & \mbox{(${\mbox{\boldmath $ r$}}={\mbox{\boldmath $ r$ } } ' , i = j , \mu = \nu$ ) , } \\[+10pt ] \displaystyle t_{\mu i \nu i}({\bm 0 } ) - \frac{1}{2 } q_{\mu i \nu i}({\bm 0 } ) u'_{i } & \mbox{(${\mbox{\boldmath $ r$}}= { \mbox{\boldmath $ r$ } } ' , i = j , \mu \ne \nu$ ) , } \\[+10pt ] \displaystyle t_{\mu i \nu j}({\bm 0 } ) - \frac{1}{2 } q_{\mu i \nu j}({\bm 0 } ) v_{ij}({\bm 0 } ) & \mbox{(${\mbox{\boldmath $ r$}}={\mbox{\boldmath $ r$ } } ' , i \ne j$ ) , } \\[+10pt ] \displaystyle t_{\mu i \nu j}({\mbox{\boldmath $ r$}}'-{\mbox{\boldmath $ r$ } } ) - \frac{1}{2 } q_{\mu i \nu j } ( { \mbox{\boldmath $ r$}}'-{\mbox{\boldmath $ r$ } } ) v_{ij } ( { \mbox{\boldmath $ r$}}'-{\mbox{\boldmath $ r$ } } ) & \mbox{(${\mbox{\boldmath $ r$}}\ne { \mbox{\boldmath $ r$}}'$ ) , } \end{array } \right . \label{eq : fock_r } \end{aligned}\ ] ] here @xmath112 is the density matrix and the matrix element is given by @xmath113 where @xmath114 is the hf ground state . the hf hamiltonian can be diagonalized with the bloch orbital , @xmath115 , where @xmath116 and @xmath117 are a band index and a wave vector , respectively , and @xmath72 is the total number of the unit cells in the system . the coefficients @xmath118 are determined by solving the following hartree - fock equation , @xmath119 with @xmath120 with the resulting @xmath121 , the real - space density matrix is calculated by @xmath122 with @xmath123 the cndo total energy with the hf approximation is given by @xmath124 , \label{eq : cndo_e } \end{aligned}\ ] ] where @xmath125 the matrix element of the core matrix @xmath126 is written as @xmath127 \displaystyle t_{\mu i \nu j}({\mbox{\boldmath $ r$ } } ) & \mbox{(otherwise ) . } \end{array } \right.\end{aligned}\ ] ] the actual cndo hf calculation proceeds along the schematic diagram shown in fig . [ fig : flowchart ] . the initial density matrix is made from diagonalizing the core matrix @xmath128 of eq . ( [ eq : h_core ] ) . since the fock matrix @xmath129 in eq . ( [ eq : fock_r ] ) depends on the density matrix @xmath130 of eq . ( [ eq : density_r ] ) , the hf equation ( [ eq : hfeq ] ) is solved self - consistently with an iterative procedure . to check the convergence , we monitor a density - matrix difference , @xmath131 } , \label{eq : delta_density } \end{aligned}\ ] ] and a total - energy difference , @xmath132 where an upper suffix @xmath15 stands for the number of the iteration step . the self - consistency condition we employ is satisfied when @xmath133 and @xmath134 ( a.u . ) . we show in fig . [ fig : band : model ] the self - consistent cndo band structures ( solid line ) , together with the _ ab initio _ interpolated band ( dots ) . we see a rigid band shift in the conduction band . this rigid band shift is attributed to the renormalization of the interaction part @xmath135 in eq . ( [ eq : h_c ] ) into the one - body part ; the so - called self - energy correction considered within the hf framework [ see eq . ( [ eq : fock_r ] ) ] . we note that the trend of the rigid band shift is basically the same as the quasiparticle band shift observed in the gw calculation for semiconductor.@xcite in the previous section , we have described the procedure for the downfolding . we now start calculating physical quantities using the downfolded model . the purpose of this sections is to examine the reliability of the model obtained by the downfolding . in particular , we highlight whether the model gives us a reliable excitation spectra and dynamical properties . for this purpose , we calculate optical - absorption spectra , based on a configuration - interaction ( ci ) treatment considering electron - hole interactions , and compare the computational result with the experiments . for electrons in solids , the number of configurations generated by the electronic excitations is in principle infinitely large , while the configurations capable of practical computations are limited . since we are interested in optical processes , we consider here only the single - excitation ( se ) configurations which play a primarily important role in the linear - absorption process , because the se configurations directly couple with the hf ground state via an electric dipole operator . the calculation at the seci level in fact takes into account electron - hole interactions ; the so - called excitonic effect in the spectrum . the seci many - body wave function with a wave vector @xmath136 is written as @xcite @xmath137 where @xmath138 is a spin - singlet se configuration given by @xmath139 here @xmath140 ( @xmath141 ) is a creation ( annihilation ) operator of the bloch electron in a virtual @xmath48 ( occupied @xmath142 ) band with spin @xmath13 and a wave vector @xmath117 . the ci coefficients \{@xmath143 } in eq . ( [ eq : seci ] ) and the excitation energy @xmath144 are obtained by solving the following ci equation , @xmath145 with @xmath146 where @xmath147 is the hf ground - state eigenenergy for the many - body hf hamiltonian @xmath148 in eq . ( [ eq : cndo - hf ] ) ; @xmath149 . for the cndo model , the second term in eq . ( [ eq : amat ] ) , called the exchange term , is calculated as @xmath150 and the last term in eq . ( [ eq : amat ] ) , refereed to as the direct term , is evaluated by @xmath151 with @xmath152 and @xmath153 \displaystyle u'_{i } & \mbox{(${\mbox{\boldmath $ r$}}={\bm 0 } , i = j , \mu \ne \nu$ ) , } \\[+5pt ] \displaystyle v_{ij}({\mbox{\boldmath $ r$ } } ) & \mbox{(otherwise ) . } \end{array } \right.\end{aligned}\ ] ] eqs . ( [ eq : ex ] ) and ( [ eq : dr ] ) imply the repulsive - exchange and attractive - coulomb interactions between an electron in the @xmath48 and @xmath154 bands and a positive hole in the @xmath142 and @xmath155 bands , with the total wave vector being kept constant [ @xmath156 . the structure of the present cndo - hf - seci equation ( [ eq : secieq ] ) is basically the same as that of the bethe - salpeter equation for two - particle green s functions.@xcite a difference is that the former electron - electron interaction in the two - electron integrals of eqs . ( [ eq : ex ] ) and ( [ eq : dr ] ) is represented by @xmath157 determined via the constrained scheme ( see sec . ii c ) , while the latter interaction is represented by the screened coulomb interaction evaluated with the random phase approximation . computationally , we note that four - center coulomb integrals in the original exchange / direct terms are reduced to two - center coulomb integrals including only the sites @xmath15 and @xmath158 because of the cndo approximation.@xcite therefore the computational cost for the matrix evaluation is much smaller in our ci calculation . the most time consuming step is the diagonalization of the ci matrix @xmath159 , which is scaled as the third power of the dimension of the @xmath159 ; @xmath160 . we next describe the expression for the seci optical absorption , which is given as the imaginary part of the macroscopic transverse dielectric function,@xcite @xmath161 where @xmath162 is a many - body position operator of electrons , and the normalization constant @xmath163 is determined via the sum rule @xcite @xmath164 with @xmath165 being the plasma frequency of the system . substituting eq . ( [ eq : seci ] ) into eq . ( [ eq : iseci ] ) and noting the commutation relation @xmath166 = \sum_{i } \left[\hat{h}_{{\rm hf}}(i ) , { \bm r}_{i } \right ] = - \sum_{i } \partial / \partial{\bm r}_{i}$ ] , we obtain @xmath167 where @xmath168 is the bloch state being the eigenstate of the fock operator @xmath169 and the matrix element of @xmath170 can be calculated with an interpolation scheme @xcite based on the mlwf s from first principles ( see appendix b ) . theoretically , @xmath171 contains an electron - hole - interaction effect due to the presence of the ci coefficients in eq . ( [ eq : iseci_2 ] ) . to see the electron - hole - interaction effect on the spectrum , it is convenient to compare with the spectrum obtained by the independent - particle approximation @xcite ( ipa ) where the ci coefficients are neglected in eq . ( [ eq : iseci_2 ] ) and the excitations are described just with optical transitions between independent hole and electron states , @xmath172 in the spectral calculations , we have chosen a @xmath1-grid different from the monkhorst - pack @xmath1-grid used in the band calculations.@xcite the present @xmath1-grid is generated as follows : we first make a uniform @xmath1-grid in an @xmath173 monkhorst - pack mesh and then slightly shift uniformly the sampling @xmath117 in the direction of @xmath174 with \{@xmath175 } being basic reciprocal lattice vectors . the resulting @xmath1-points are different from the high - symmetry directions of the crystal and therefore are not connected via rotational operations of the crystal with each other . this leads to a finer sampling for the spectral calculation . an unshifted grid corresponds to only 56 crystallographically different points , which are too few to achieve a good spectral resolution . on the other hand , the shifted grid leads to a grid of 1331 crystallographically different @xmath1-points , which gives a good spectral resolution . we show in fig . [ spectrum ] the calculated seci ( thick red line ) and ipa ( thin green line ) spectra . the closed blue and open blue circles denote experimental results.@xcite we see a clear contrast between the seci and ipa spectra ; by considering the electron - hole interaction with the seci method , the spectral intensity in the low - energy region ( @xmath176 ev ) is enhanced , thus reproducing the experimental results perfectly . the agreement is indeed somewhat surprising , when we consider the several simplified treatments employed here such as the reduction of the electron - electron interaction to the extended hubbard form . however , we emphasize that the nature such as the spectral enhancement observed in proceeding from ipa to seci is consistent with highly - accurate full _ ab initio _ results @xcite obtained from solving the bethe - salpeter equations . in the context of the downfolding , these results strongly support that our model construction by the downfolding described in sec . ii c offers a reliable description not only of the ground - state band structure of an insulator but also of the excitation spectra . for the completeness , we perform a more critical and elaborate assessment of the reliability of the downfolding ; we examine a sensitivity of the spectra to choices of hamiltonian parameters . in the present analysis , we focus on the check of the reliability of the offsite interaction parameter @xmath51 . in fact , the magnitude of this parameter is expected to crucially control the strength of the electron - hole interaction [ see eqs . ( [ eq : ex ] ) and ( [ eq : dr ] ) ] and thus directly affect the profile of the optical spectra . we may calculate the excitation spectra by using choices of interaction parameters different from the downfolded realistic values . thus , the reliability of the downfolding can be assessed by examining whether the spectrum obtained from the present downfolded hamiltonian gives the best agreement with the experiment among wider alternative choices of the interaction parameters . for this check , we introduce a scaling factor @xmath177 to scale @xmath51 to @xmath178 . with this definition , @xmath179 corresponds to the original _ ab initio _ @xmath51 value , while in the region @xmath180 1 , the nearest neighbor electron - electron repulsion is artificially overestimated . in the practical calculation , we monitor the values of @xmath177 at 0.8 , 1.0 , and 1.2 and then perform the seci calculations to obtain the optical spectra for each @xmath51 value . ( in the calculations , the onsite parameters are fixed at the _ ab initio _ determined values displayed in table [ param ] . ) we show in fig . [ fig : xdep ] ; ( a ) @xmath177 = 0.8 , ( b ) @xmath177 = 1.0 , and ( c ) @xmath177 = 1.2 . the closed blue and open blue circles represent experimental data taken from ref . 31 and ref . 32 , respectively.,scaledwidth=50.0% ] the resulting dependence of the seci spectra ( red lines ) on the scaling factor @xmath177 . the blue circles denote measured data . we see a notable change in the spectra due to the parameter increase [ ( a ) @xmath181 ( b ) @xmath181 ( c ) ] ; increasing @xmath177 makes a blue shift and an intensity decrease in the calculated spectra . we see that the spectrum at the downfolded choice ( i.e. , the case with @xmath177 = 1.0 ) exhibits the best agreement with the experiments among all the choices . the downfolded value offers the most realistic and accurate choices as the model parameters , and thus we conclude that the optical excitation spectrum is correctly captured by the downfolded hamiltonian . finally , we remark a recent development for _ ab initio _ evaluations for the offsite @xmath51 parameter . indeed , applications of the constrained schemes to the determinations of the offsite parameter is quite limited in the literature compared with those for the onsite parameters . recently , aryasetiawan _ et al_@xcite have proposed an rpa approach for calculating the interaction parameters . they first calculate a real - space screened coulomb interaction @xmath182 by excluding the polarization formed in target bands contained in the model hamiltonian and then evaluate the matrix element of @xmath52 in the localized basis such as linearized muffin - tin orbitals and/or maximally - localized wannier functions . this approach is able to derive all the offsite parameters . so , comparisons between the values obtained by the present constrained scheme and the values based on the rpa approach would contribute to making deeper understanding for the distant interaction parameters . we have examined whether the effective low - energy hamiltonian derived from the downfolding procedure is able to describe dynamics and excitation spectra in a proper way . the calculation is performed in the three - stage scheme . in the first stage , we calculate the global electronic structure from the density functional theory supplemented by the generalized gradient approximation . the high - energy degrees of the freedom in the global electronic band structure are , in the second stage , eliminated by the downfolding scheme , which leaves only the low - energy bands near the fermi level . in the present example of gaas , we retain up to 25 ry for the calculation of the global electronic bands , while the downfolded hamiltonian keeps only eight bands near the fermi level up to 15 ev ( @xmath183 1 ry ) . by the downfolding , kinetic and interaction energies are separately renormalized into the low - energy eight bands and the effective hamiltonian , where we employ the cndo model neglecting the offdiagonal part of the coulomb interaction , is constructed from first principles , with the help of the maximally localized wannier functions . this procedure , though several simplified treatments are employed , in principle , does not contain any _ ad hoc _ parameters . in the third stage , the hartree - fock method for the ground state supplemented by the single - excitation configuration - interaction treatment for considering the electron - hole interactions has been applied to obtain electronic excitation spectra of semiconductor gaas . the spectra thus obtained have quite well reproduced the experimental results ; the intensity and position for the excitonic peak are well reproduced at a quantitative level . we believe that the present model construction based on the downfolding offers a reliable _ ab initio _ scheme , where the downfolded effective hamiltonian is capable of not only the ground state properties but also the excitation spectra . the present result opens a way of treating excitations such as the optical spectra by the hybrid method combining the density functional approach and the accurate low - energy solver for the low - energy effective models . beyond the present application to semiconductors , it would be interesting to apply this approach to excitations in strongly correlated electron systems such as transition metal oxides including the cuprates . in the present paper , we have used the hartree - fock approximation for the ground state and the single - excitation configuration - interaction treatment for the excitations . optical excitation spectra of gaas have satisfactorily been treated by these approximations and the experimental results have been well reproduced . however , stronger electron correlation effects require more sophisticated low - energy solver than the hartree fock / single excitation configuration interaction treatment . for more different and challenging issues of the electron correlation , the low - energy effective hamiltonian may indeed be treated by much more reliable low - energy solver for electrons in solids , such as quantum monte carlo methods for lattice fermions,@xcite path - integral renormalization group method,@xcite and cluster extensions of the dynamical mean field theory.@xcite as is well known , there are many direct _ ab initio _ schemes aiming at considering correlation effects ; for example , the gw,@xcite transcorrelated , @xcite and quantum monte carlo methods.@xcite they are straightforward ways for approaching the problem compared to the present approach . however , the straightforward methods are faced with two serious problems : one is that the computational load becomes extremely heavy when all the electrons or even all the valence electrons are treated equally . the other problem is that the so far developed straightforward methods do not offer a sufficiently accurate framework if the electron correlation becomes strong such as in the genuine mott insulator . the crucial point is that we need to treat dynamical as well as spatial correlations and fluctuations near the fermi level in a controllable way . in the present stage of the computer power , such sufficient accuracies are undertaken only within simple models , which can be achieved in the low - energy effective model after downfolding . in fact , the high accuracy required from the temporal and spatial quantum fluctuations is important only in the low - energy region near the fermi level , which justifies to restrict the high - accuracy treatment only in the region of low - energy excitations and thus only within the downfolded hamiltonian . within the present computer power , this downfolding procedure opens an avenue of studying highly correlated electron systems as well as excitations without relying on _ ad hoc _ parameters . by explicitly considering the energy hierarchy in the electronic structure , the first principles calculations become tractable even when the electron correlation is essential . we would like to thank takashi miyake , atsushi yamasaki , youhei yamaji , takahiro misawa , and taichi kosugi for helpful discussions and comments . this work was supported by a grant - in - aid for scientific research in priority areas , `` development of new quantum simulators and quantum design '' ( no . 17064004 ) of the ministry of education , culture , sports , science and technology , japan . one of us ( k.n . ) acknowledges research fellowships of the japan society for the promotion of science for young scientists . all the calculations were performed on hitachi sr11000 system of the super computer center at the institute for solid state physics , the university of tokyo . here , we describe the details of the fitting function used in the onsite - parameter determination . in the atomic limit , the onsite hamiltonian for the decoupled site @xmath55 in the home cell is written as @xmath184 where @xmath185 and @xmath186 are number operators . @xmath187 and @xmath188 are the onsite intraorbital and interorbital coulomb repulsions , respectively . @xmath189 is a chemical potential . the onsite energy @xmath190 derived from eq . ( [ eq : h_onsite ] ) is expressed with the atomic - limit wave function @xmath191 as @xmath192 where we used @xmath193 . we introduce @xmath194 with @xmath195 defined as the orbital occupancy at the equilibrium state . at the equilibrium state , the term linear in @xmath196 should vanish , which results in cancellation of the chemical potential term with the linear term from the coulomb contribution . then the quadratic energy difference due to the charge fluctuations is derived as @xmath197 by noting @xmath198 and defining the site - charge fluctuation @xmath199 , we obtain the form @xmath200 one may specialize the charge fluctuation of one orbital because of the crystallographycal symmetry in the system . thus , the orbital index in the orbital - charge fluctuation is dropped and the cross term in eq . ( [ eq : e_site ] ) is rewritten in terms of @xmath201 and @xmath202 , @xmath203 inserting the above expression into eq . ( [ eq : e_site ] ) leads to the following expression for the onsite interaction energy : @xmath204 \left ( \delta q_{i { \bm 0 } } \right)^{2}.\end{aligned}\ ] ] here we describe details for the calculation of the matrix element @xmath205 in eq . ( [ eq : iseci_2 ] ) . we first rewrite it in terms of the maximally localized wannier function as @xmath206 with @xmath207 . in our calculation , the wannier function @xmath208 is stored as numerical data on the real - space grid , @xmath209 where @xmath210(@xmath211 ) is a cell vector defining a superlattice containing @xmath212 primitive cells , @xmath213 runs on the integer values : @xmath214 with @xmath215 being the total number of the grids in the @xmath15th direction ( in the present case , @xmath216 ) . since the wannier function satisfies the following periodic boundary condition , @xmath217 we can express @xmath218 in terms of the fourier transformation as @xmath219 where @xmath220 with @xmath221 , @xmath222 is the volume of the superlattice , and @xmath223 takes value from @xmath224 to @xmath225 . we note that @xmath226 is different from @xmath227 used in the _ ab initio _ band calculations ; the former is expressed in terms of the reciprocal - lattice vectors for the superlattice \{@xmath228 } , while the latter is written with the basic reciprocal lattice vectors \{@xmath175}. substituting eq . ( [ eq : ft ] ) into eq . ( [ eq : wdrw ] ) and using the translational symmetry for the matrix element leads to @xmath229 with @xmath230 the actual calculation proceeds as follows : we first transform the real - space wannier function @xmath218 into the reciprocal - space one @xmath231 in eq . ( [ eq : ft ] ) with the algorithm of the fast - fourier - transformation with radix-2 , 3 , and 5 . then , we calculate the @xmath170 matrix in the wannier basis [ @xmath232 in eq . ( [ eq : gij ] ) ] to obtain the desired quantity [ eq . ( [ eq : wdrw_2 ] ) ] . we note that this numerical procedure is a so - called wannier interpolation scheme;@xcite we construct the wannier functions with the _ ab initio _ bloch functions in the monkhorst - pack @xmath1-grid and then interpolate the matrix elements of @xmath170 at the slightly shifted @xmath1-grid used in the spectral calculation . n. f. mott , phys . london sect . a * 62 * , 416 ( 1949 ) ; m. c. gutzwiller , phys . * 10 * , 59 ( 1963 ) ; j. hubbard , proc . london , ser . a * 276 * , 238 ( 1963 ) ; j. kanamori , prog . theor . * 30 * , 275 ( 1963 ) . l. f. mattheiss , phys . b * 5 * , 290 ( 1972 ) ; _ ibid_. * 5 * , 306 ( 1972 ) ; k. terakura , t. oguchi , a. r. williams , and j. kubler , _ ibid_. * 30 * , 4734 ( 1984 ) ; g. a. sawatzky and j. w. allen , phys . lett . * 53 * , 2339 ( 1984 ) . s. albrecht , l. reining , r. del sole , and g. onida , phys . lett . * 80 * , 4510 ( 1998 ) ; e. l shirley , _ ibid_. * 80 * , 794 ( 1998 ) ; l. x. benedict , e. l shirley , r. b. bohn , _ ibid_. * 80 * , 4514 ( 1998 ) ; l. x. benedict and e. l shirley , phys . rev . b * 59 * , 5441 ( 1999 ) ; m. rohlfing and s. g. louie , phys . rev * 82 * , 1959 ( 1999 ) ; j. -w . van der horst , p. a. bobbert , m. a. j. michels , g. brocks , and p. j. kelly , _ ibid_. * 83 * , 4413 ( 1999 ) ; j. -w . van der horst , p. a. bobbert , p. h. l. de jong , m. a. j. michels , g. brocks , and p. j. kelly , phys . rev . b * 61 * , 15817 ( 2000 ) ; p. puschnig and c. ambrosch - draxl , _ ibid_. * 66 * , 165105 ( 2002 ) ; b. arnaud , s. lebegue , p. rabiller , and m. alouani , phys . lett . * 96 * , 026402 ( 2006 ) ; for a review , see g. onida , l. reining , and a. rubio , rev . * 74 * , 601 ( 2002 ) . for a time - dependent density - functional approach to the optical problem , see f. sottile , k. karlsson , l. reining , and f. aryasetiawan , phys . b * 68 * , 205112 ( 2003 ) ; a. marini , r. del sole , and a. rubio , phys . lett . * 91 * , 256402 ( 2003 ) . l. x. benedict , e. l shirley , r. b. bohn , phys . rev . b. * 57 * , r9385 ( 1998 ) ; b. arnaud and m. alouani , _ ibid_. * 63 * , 085208 ( 2001 ) ; p. h. hahn , k. seino , w. g. schmidt , j. furthmuller , and f. bechstedt , phys . stat , sol . ( b ) * 242 * , 2720 ( 2005 ) . l. hedin and s. lundqvist , solid state physics , * 23 * , 1 ( 1969 ) ; m. s. hybertsen and s. g. louie , phys . b * 34 * , 5390 ( 1986 ) ; _ ibid_. * 35 * , 5585 ( 1987 ) ; r. w. godby , m. schluter , and l. j. sham , _ ibid_. * 35 * , 4170 ( 1987 ) ; f. aryasetiawan , _ ibid_. * 46 * , 13051 ( 1992 ) ; f. aryasetiawan and o. gunnarsson , phys . * 74 * , 3221 ( 1995 ) ; f. aryasetiawan , rep . . phys . * 61 * 237 ( 1998 ) . v. i. anisimov , j. zaanen and o. k. andersen , phys . b * 44 * , 943 ( 1991 ) ; v. i. anisimov , f. aryasetiawan and a. i. lichtenstein , j. phys . cond . mat . , * 9 * , 767 ( 1997 ) ; a. i. lichtenstein and m. i. katsnelson , phys . rev . b * 57 * , 6884 ( 1998 ) . v. i. anisimov , a. i. poteryaev , m. a. korotin , a. o. anokhin , and g. kotliar , j. phys . : condens . matter * 9 * , 7359 ( 1997 ) ; g. kotliar , s. y. savrasov , k. haule , v. s. oudovenko , o. parcollet , and c. a. marianetti , rev . mod . phys . * 78 * , 865 ( 2006 ) . f. aryasetiawan , m. imada , a. georges , g. kotliar , s. biermann , and a. i. lichtenstein , phys . b * 70 * , 195104 ( 2004 ) ; f. aryasetiawan , k. karlsson , o. jepsen , and u. schonberger , _ ibid_. * 74 * , 125106 ( 2006 ) . e. pavarini , s. biermann , a. poteryaev , a. i. lichtenstein , a. georges , and o. k. andersen , phys . lett . * 92 * , 176403 ( 2004 ) ; v. i. anisimov , d. e. kondakov , a. v. kozhevnikov , i. a. nekrasov , z. v. pchelkina , j. w. allen , s .- k . kim , p. metcalf , s. suga , a. sekiyama , g. keller , i. leonov , x. ren , and d. vollhardt phys . b * 71 * , 125119 ( 2005 ) ; v. i. anisimov and a. v. kozhevnikov , _ ibid_. * 72 * , 075125 ( 2005 ) ; f. lechermann , a. georges , a. poteryaev , s. biermann , m. posternak , a. yamasaki , and o. k. andersen , _ ibid_. * 74 * , 125120 ( 2006 ) . for variational monte - carlo studies , see w. l. mcmillan , phys . rev . * 138 * , a442 ( 1965 ) ; d. ceperley , g. v. chester , and m. h. kalos , phys . b * 16 * , 3081 ( 1977 ) ; h. yokoyama and h. shiba , j. phys . * 56 * , 1490 ( 1986 ) ; as applications of a quantum monte - carlo method to lattice hubbard models , see r. blankenbecler , d. j. scalapino , and r. l. sugar , phys . d * 24 * , 2278 ( 1981 ) ; s. r. white , d. j. scalapino , r. l. sugar , e. y. loh , j. e. gubernatis , and r. t. scalettar , phys . b * 40 * , 506 ( 1989 ) ; m. imada and y. hatsugai , j. phys . * 58 * , 3752 ( 1989 ) ; n. furukawa and m. imada , j. phys . jpn . * 61 * , 3331 ( 1992 ) . p. h. dederichs , s. blugel , r. zeller , and h. akai , phys . 53 * , 2512 ( 1984 ) ; m. r. norman and a. j. freeman , phys . b * 33 * , 8896 ( 1986 ) ; o. gunnarsson , o. k. andersen , o. jepsen , and j. zaanen , _ ibid_. * 39 * , 1708 ( 1989 ) ; m. s. hybertsen , m. schlter , and n. e. christensen , _ ibid_. * 39 * , 9028 ( 1989 ) ; a. k. mcmahan , j. f. annett , and r. m. martin , _ ibid_. * 42 * , 6268 ( 1990 ) ; m. m. steiner , r. c. albers , and l. j. sham , _ ibid_. * 45 * , 13272 ( 1992 ) ; i. v. solovyev and p. h. dederichs , _ ibid_. * 49 * , 6736 ( 1994 ) ; i. v. solovyev , p. h. dederichs , and v. i. anisimov , _ ibid_. * 50 * , 16861 ( 1994 ) ; as a constrained calculation for complex molecules , see q. wu and t. vanvoorhis , phys . rev . a * 72 * , 024502 ( 2005 ) . for another approach for obtaining the onsite interaction parameter , see m. springer and f. aryasetiawan , phys . rev . b * 57 * , 4364 ( 1998 ) ; a. yamasaki and t. fujiwara , j. phys . 72 * , 607 ( 2003 ) . j. a. pople , d. p. santry , and g. a. segal , j. chem . phys . * 43 * , s129 ( 1965 ) ; j. a. pople and g. a. segal , _ ibid_. * 43 * , s136 ( 1965 ) ; j. a. pople and g. a. segal , _ ibid_. * 44 * , 3289 ( 1966 ) ; j. a. pople and d. l. beveridge , _ approximate molecular orbital theory _ , ( mcgrawhill , new york , london , 1970 ) . our constrained calculations were performed for a 3@xmath03@xmath03 fcc supercell containing 54 atoms , with 7@xmath07@xmath07 @xmath1-points sampling . it was found that the sizes of the supercell and sampling @xmath1-points are sufficient for obtaining a converged interaction parameters . for the detail of the hopping - cutoff treatment , see k. held , i. a. nekrasov , g. keller , v. eyert , n. blumer , a. k. mcmahan , r. t. scalettar , th . pruschke , v. i. anisimov , and d. vollhardt , phys . ( b ) * 243 * , 2599 ( 2006 ) .

i try to give an overview of future prospects of the experimental exploration of the unknowns in the neutrino mass pattern and the lepton flavor mixing . because of the nature of the lectures on which this manuscript is based , i give some pedagogical discussions to prepare for the presentation in the later part . i start from measuring by reactors and accelerators as a prerequisite for proceeding to search for leptonic cp violation . i then discuss how cp violation can be uncovered , and how the neutrino mass hierarchy can be determined . i do these by resolving so called the `` parameter degeneracy '' which is necessary anyway if one wants to seek precision measurement of the lepton mixing parameters . as a concrete setting for resolving the degeneracy i introduce the tokai - to - kamioka - korea two detector complex which receives neutrino superbeam from j - parc , sometimes dubbed as `` t2kk '' . it is shown that t2kk is able to resolve all the eight - fold parameter degeneracy in a wide range of the lepton mixing parameters . i also discuss an alternative way for lifting the octant degeneracy by reactor - accelerator combined method . finally , i discuss by taking some examples how some theoretically or phenomenologically motivated ideas can be tested experimentally . 1.4 cm

hep-ph0701070

neutrino physics , in particular its experimental part , has been extremely successful in the last 10 years . it would be worthwhile to look it back on this occasion as a prologue to our discussion on the future . in 1998 the super - kamiokande atmospheric neutrino observation confirmed @xcite the smoking gun evidence for atmospheric neutrino anomaly seen in the deficit of the rate and in the zenith angle distribution of @xmath2 induced events in the kamiokande experiment @xcite . it was the first evidence for mass - induced neutrino oscillation . oscillation in close analogy to @xmath3 oscillation @xcite . in 1962 maki , nakagawa , and sakata @xcite first pointed out the possibility of neutrino flavor transformation , the phenomenon established experimentally only recently as described here . ] the evidence for neutrino oscillation was readily confirmed by the first long - baseline ( lbl ) accelerator neutrino experiment k2k @xcite using man - made @xmath2 beam . in this sense , the first corner stone was placed in the 2 - 3 sector of the lepton flavor mixing matrix , the mns matrix @xcite . it established surprisingly that the mixing angle @xmath1 is large , which may even close to the maximal , refuting the prejudice of small flavor mixing angles deeply rooted among theorists at that time . on the other hand , there have been great amount of efforts in the solar neutrino observation pioneered by davis with his @xmath4cl experiment in sixties which was developed in close collaboration with the devoted theorist @xcite . in the last 20 years the field has been enriched by participation by kamiokande , ga , super - kamiokande , and sno experiments @xcite . in particular , the latter two experiments were united to the confirm the particle physics nature of the solar neutrino problem , the evidence for solar neutrino flavor transformation @xcite . later sno _ in situ _ confirmed the evidence @xcite . i would like to note here that the deficit of the @xmath5b flux obtained by davis in his @xmath4cl experiment @xcite , though suffered from stubborn skepticism for more than 30 years , was convincingly confirmed by the sno charged current ( cc ) measurement . the beautiful finale of the solar neutrino problem came with kamland @xcite which identified its cause as due to the mass - induced neutrino oscillation which clearly pinned down the large mixing angle ( lma ) msw solar neutrino solution @xcite . the particular significance of the kamland result in this context , so called the kamland massacre ( of non - standard scenarios ) , was emphasized by many people with detailed analysis for example in @xcite . the resultant mixing angle @xmath6 turned out to be large , but not maximal . finally , several experiments , super - kamiokande @xcite , kamland @xcite , k2k @xcite , and minos @xcite , observed the oscillatory behavior , thereby established the phenomenon of mass - induced neutrino oscillation . at this moment , it constitutes the first and the unique evidence for physics beyond the standard model . what is the next ? the most common answer , which i also share , is to explore the unknown 1 - 3 sector , for which the only knowledge we have is the upper bound on @xmath0 @xcite . discovery of leptonic cp violation must throw light on tantalizing mystery of interrelationship between quarks and leptons . the discussion of the quark - lepton correspondence which can be traced to early sixties @xcite , and in a modern context presented in a compelling form with the anomaly cancellation in standard model @xcite strongly suggests that they have common roots . it is also possible that the kobayashi - maskawa type cp violation @xcite in the lepton sector might be related to cp violation at high energies which is required for leptogenesis @xcite to work . see , e.g. , @xcite and the references cited therein for this point . despite the great progress in our understanding mentioned above we do have many important unanswered questions . the list includes , for example , the followings : what is the origin of neutrino masses and mixing ? what is the reason for disparity between small quark and large lepton mixings ? is there underlying quark - lepton symmetry , or quark - lepton complementarity ? is there flavor symmetry which includes quarks and leptons ? i am sure that many more questions exist . these points are discussed in depth in a recent review @xcite . the new stage of neutrino physics may also be characterized as beginning of the era of precision measurement of lepton mixing parameters . testing the various theoretical ideas proposed to understand the uncovered structure mentioned above requires accurate determination of mixing parameters . for example , to test the quark - lepton complementarity @xcite experimentally , one needs to improve accuracies for @xmath6 determination from the current one , @xmath7% for @xmath8 , to the one comparable to the cabibbo angle , @xmath9% @xcite . it will be discussed in sec . [ proceed ] . i must admit that the scope of my discussions is quite limited ; the crucially important issues such as absolute neutrino mass , nature of neutrinos ( majorana vs. dirac ) , majorana cp violation and leptogenesis are not covered . moreover , it covers only a part of the things that should be addressed for exploring unknowns done by the future lbl experiments , that is , concrete ways of how to determine the mixing parameters with the next generation conventional @xmath2 superbeam @xcite and reactor experiments . yet , conventional superbeam experiments are extremely interesting because , in principle , they can be done in the next 10 - 15 years without long - term r@xmath10d effort . here is a composition of this long report . first of all , i intend to be pedagogical in writing this report ; i met with many brilliant young people in `` world summit in galapagos '' @xcite , and a broad class of audiences who keenly interested in neutrino physics in `` heavy quarks and leptons '' @xcite , and it is a pity if this manuscript is entirely unreadable to them . in sec . [ theta23 ] , we review how the atmospheric parameters @xmath11 and @xmath1 are determined . in sec . [ theta13 ] , we explain how @xmath0 can be measured and briefly review the reactor and the accelerator methods . in sec . [ vac - matter ] , we provide a simple understanding of the interplay between the vacuum and the matter effect by introducing the bi - probability plot . in sec . [ delta ] , we mention two alternative strategies of how to measure cp violation and give some historical remarks on how the thoughts on measuring leptonic cp violation were evolved . in sec . [ pdege ] , we explain in a simple terms the cause of the parameter degeneracy by using the bi - probability plot . it is an important topics for precision measurement of the lepton mixing parameters because the degeneracy acts as a notorious obstacle to it . in sec . [ resolved ] , we discuss how the eight - fold parameter degeneracy can be resolved _ in situ _ by using `` t2kk '' , the tokai - to - kamioka - korea setting . in sec . [ reactor - accelerator ] , we describe an alternative method for solving a part of the degeneracy called the @xmath1 octant degeneracy by combining reactor and accelerator experiments . in sec . [ proceed ] , we discuss , by taking a concrete example , how theoretical / phenomenological hypothesis can be confronted to experiments . in sec . [ conclude ] , we give a concluding remark . `` bread and butter '' in the coming era of precision measurement of lepton mixing parameters is the accurate determination of the atmospheric parameters , @xmath12 and @xmath1 . it will be carried out by the accelerator disappearance experiments which measures energy spectrum modulation of muon neutrinos . ignoring terms proportional to @xmath13 and @xmath0 , the disappearance probability in vacuum can be written as @xmath14 in view of ( [ pvac ] ) , very roughly speaking , the position and the depth of the dip corresponding to the first oscillation maximum , @xmath15 , tell us @xmath12 and @xmath16 , respectively . the current limits on these parameters from the sk atmospheric neutrino observation are @xmath17 and @xmath18 at 90% cl @xcite . k2k , the first accelerator lbl experiment obtained the similar results , @xmath19 at 90% cl though the sensitivity to @xmath1 is much worse , @xmath20 @xcite . the currently running minos experiment @xcite aims at determining @xmath12 to @xmath21% level , and @xmath16 to @xmath22% level , both at 90% cl . the next generation lbl experiment t2k @xcite is expected to improve the sensitivity to @xmath23% level for @xmath12 excluding systematics , and @xmath24% level for @xmath16 including systematics , both at 90% cl @xcite . these numbers are cross checked in various occasions @xcite . the us project no@xmath25a @xcite will also have the similar sensitivities . these accuracies are quite essential to resolve the parameter degeneracy ( see sec . [ pdege ] ) to achieve the goal of precision determination of the lepton mixing parameters . to reach the goal of seeing leptonic cp violation , we have to clear the first hurdle , knowing the value of @xmath0 . what is the most appropriate way to measure the parameter ? to answer the question we consider the neutrino oscillation channel which involve @xmath26 , otherwise @xmath0 would not be contained in leading order . there are two candidate channels ; @xmath27 and @xmath28 ( or , @xmath29 ) . in our discussion that follows , @xmath30 is the same as @xmath29 . we note that @xmath31 probed at energy / baseline appropriate to atmospheric @xmath32 scale consists of interference terms between amplitudes @xmath33 and @xmath34 . then , obviously @xmath35 is involved in the disappearance probability . on the other hand , in the appearance channel @xmath28 , the oscillation probability contains interference terms between amplitudes @xmath36 and @xmath37 . then , the appearance channel looks to be advantageous because only a single power of @xmath38 is involved . but , it is untrue ; when there is a hierarchy in @xmath32 , @xmath39 , unitarity tells us that these two terms nearly cancel , leaving another power of @xmath38 . as a consequence , @xmath40 is also proportional to @xmath35 . hence , there are two comparably good ways to measure @xmath0 ; the reactor and the accelerator methods which measure @xmath31 and @xmath40 , respectively . let us describe them one by one . before getting into the task we give here the explicit expressions of oscillation probabilities . for @xmath41 disappearance channel it reads ( see e.g. , erratum in @xcite ) @xmath42 \sin^2 \left ( { \delta m^2_{21}l \over 4e } \right ) . \label{pvac_ee}\end{aligned}\ ] ] for the appearance channel , we use the @xmath43 ) oscillation probability with first - order matter effect @xcite @xmath44 & = & \sin^2{2\theta_{13 } } s^2_{23 } \left [ \sin^2 \left(\frac{\delta m^2_{31 } l}{4 e}\right ) -\frac { 1}{2 } s^2_{12 } \left(\frac{\delta m^2_{21 } l}{2 e}\right ) \sin \left(\frac{\delta m^2_{31 } l}{2 e}\right ) \right . \nonumber \\ & & \hspace*{24 mm } { } \pm \left . \left(\frac { 4 ea}{\delta m^2_{31}}\right ) \sin^2 { \left(\frac{\delta m^2_{31 } l}{4 e}\right ) } \mp \frac{al}{2 } \sin \left(\frac{\delta m^2_{31 } l}{2 e}\right ) \right ] \nonumber \\ & + & 2j_{r } \left(\frac{\delta m^2_{21 } l}{2 e } \right ) \left [ \cos{\delta } \sin \left(\frac{\delta m^2_{31 } l}{2 e}\right ) \mp 2 \sin{\delta } \sin^2 \left(\frac{\delta m^2_{31 } l}{4 e}\right ) \right ] \nonumber \\ & + & c^2_{23 } \sin^2{2\theta_{12 } } \left(\frac{\delta m^2_{21 } l}{4 e}\right)^2 , \label{pmue}\end{aligned}\ ] ] where the terms of order @xmath45 and @xmath46 are neglected . in eq . ( [ pmue ] ) , @xmath47 @xcite where @xmath48 is the fermi constant , @xmath49 denotes the averaged electron number density along the neutrino trajectory in the earth , @xmath50 @xmath51 denotes the reduced jarlskog factor , and the upper and the lower sign @xmath52 refer to the neutrino and anti - neutrino channels , respectively . in both of the oscillation probabilities , @xmath53 and @xmath40 , the leading atmospheric oscillation terms have the common factor @xmath54 , in agreement with the discussion given above . the last term in eq . ( [ pmue ] ) is the solar scale oscillation term , which will be important for resolving the @xmath1 degeneracy . it was proposed @xcite that by using identical near and far detectors which is placed close to and at around @xmath551 km from the reactor , respectively , one can search for non - zero @xmath0 to a region of @xmath56 . an advantage of the reactor @xmath0 experiments is their cost effectiveness which stems from that the beam is intense enough ( and furthermore free ! ) and low in energy to allows relatively compact detectors placed at baselines much shorter than those of accelerator experiments . intensive efforts over several years from these proposals entailed the various projects in world wide as described in @xcite . by now a few projects have already been approved , or are close to the status @xcite . scientific merit of the reactor measurement of @xmath0 is that it provides pure measurement of @xmath0 without being affected by other mixing parameters , as emphasized in @xcite . it implies , among other things , that it can help resolving the @xmath1 octant degeneracy as pointed out in @xcite , and recently demonstrated in detail in @xcite . on the other hand , the same property may be understood as `` shortcoming '' of the reactor experiment , if one want to search for leptonic cp violation . it is known that @xmath26 ( @xmath41 ) disappearance probability has no sensitivity to @xmath57 even in matter with arbitrary profile with negligible higher order correction @xcite . we note , however , that reactor @xmath0 experiment can be combined with accelerator appearance measurement to uncover cp violation @xcite . in contrast to the reactor experiments accelerator measurement of @xmath0 is `` contaminated '' ( or enriched ) by the other mixing parameters , in particular by @xmath57 in the case of low energy superbeam experiments . the sensitivity to @xmath0 therefore depends upon @xmath57 in a significant way . though it sounds like drawback of the accelerator method , it in turn means that the lbl @xmath0 experiments can be upgraded to search for leptonic cp violation . ( this is why and how the low - energy superbeam was originally motivated in @xcite . ) there exist an approved experiment t2k @xcite using the 0.75 mw neutrino beam from j - parc , and a competitive proposal of no@xmath25a @xcite which uses numi beam line in fermilab . the sensitivity to @xmath0 , is roughly speaking , up to @xmath56 . however , the better knowledge of background rejection and the systematic errors are required to make the number more solid . though less sensitive , minos @xcite and opera @xcite have some sensitivities to @xmath0 . if @xmath0 is really small , @xmath58 , probably we need new technology to explore the region of @xmath0 . the best candidates are neutrino factory @xcite or the beta beam @xcite . for them we refer @xcite for overview and for extensive references . to proceed further , we need some knowledges on neutrino oscillation in matter . there are several ways to simply understand the matter effect in neutrino oscillations . one is to use perturbative approach @xcite . the other is to rely on cervera _ et al . _ formula @xcite which applies to higher matter densities . the most important reason why we want to understand the feature of vacuum - matter interplay in neutrino oscillation is that they tend to mix and confuse with each other . for the early references which took into account the matter effect which inevitably comes in into lbl cp violation search , see e.g. , @xcite . probably , the simplest way to understand the matter effect as well as cp phase effect is to rely on the cp trajectory diagram in @xmath40 and @xmath59 space , for short , the bi - probability plot @xcite . it is given in fig . [ bi - p ] . by writing the bi - probability diagram one can easily understand the relative importance of cp and the matter effects in a pictorial way ; magnitude of effect of cp violating phase @xmath57 is represented as the size of the ellipses , while that of the matter effect can be read off as a separation between the two ellipses with positive ( blue ellipse ) and negative ( red ellipse ) sign of @xmath12 . the distance from the origin to the ellipse complex represents @xmath60 . we mention that the bi - probability plot can be extended to the one which includes t violation in which the charming relations between probabilities called the cp - cp and the t - cp relations are hidden @xcite . 0.3 cm - @xmath61 bi - probability plot with experimental parameters corresponding to numi off - axis project is presented for the purpose of exhibiting characteristic features of the neutrino oscillations relevant for low - energy superbeam experiments . namely , it can disply competing three effects , cp violating and cp conserving effects due to @xmath57 as well as the matter effects in a compact fashion . for more detailed description of its properties , see @xcite . the art work is done by adam para . , title="fig:",scaledwidth=60.0% ] if @xmath0 is not too small and is within reach by the next generation reactor or lbl experiments , the door is open to search for leptonic cp violation using conventional superbeam . when people started to think about the possibility of observing cp violation there were two alternative ways to approach the goal , high - energy vs. low - energy options , as described in @xcite . the high - energy option , the majority at that time , is based on the idea of neutrino factory @xcite which utilizes intense neutrino beam from muon storage ring . because background can be suppressed to a very small level due to clean detection of high - energy muons , the sensitivity to @xmath0 and @xmath57 can be extremely good . we do not quote the number here because its re - examination by taking into account the possibility of lowering the threshold is ongoing in the context of neutrino factory international scoping study @xcite , which should be available soon . the low - energy option is based on very simple fact that the effect of cp phase @xmath57 is large at low energies @xcite . what is good in the low - energy option is that it can be realized with conventional @xmath2 superbeam . it opens the possibility that the cp violation search can be pursued by relying on known beam technology with no need of an extensive r@xmath10d efforts , and is doable in the next 10 - 15 years if we can enjoy generous governmental support . on the other hand , @xmath26 appearance search with conventional @xmath2 beam inevitably has the intrinsic problem of background , not only of the beam origin but also due to the neutral current ( nc ) @xmath62 in the case of water cherenkov detectors . despite the potential difficulties , the possibility of experimental search for leptonic cp violation became the realistic option when loi of the t2k experiment with optimistic conclusion was submitted @xcite . unfortunately , the optimism in the early era was challenged by several potential obstacles . first of all , reducing the systematic error to a required level , a few % level , is a tremendous task . news is that several experiments are going on , or to be done , to measure hadron production @xcite and neutrino nucleus interaction cross sections @xcite . other difficulties include , for example : the possibility that cp violation could be masked by the unknown sign of @xmath12 , or in more general context the presence of parameter degeneracy @xcite which can obscure the cp violation , which will be the topics of the next section . since sometime ago people recognized that measurement of @xmath40 and @xmath59 at a particular energy , no matter how accurate they are , allows multiple solutions of @xmath0 and @xmath57 , the problem of parameter degeneracy . the nature of the degeneracy can be understood as the intrinsic degeneracy @xcite , which is duplicated by the unknown sign of @xmath32 @xcite , and the possible octant ambiguity of @xmath1 @xcite if it is not maximal . for an overview of the resultant eight - fold parameter degeneracy , see e.g. , @xcite . verses @xmath63 \equiv p(\bar{\nu}_\mu \rightarrow \bar{\nu}_e)$ ] plane . between the solid ( dashed ) lines is the allowed region for positive ( negative ) @xmath12 and the shaded region is where solution for both signs are allowed . the solid ( dashed ) ellipses are for positive ( negative ) @xmath12 and they all meet at a single point . this is the cp parameter degeneracy problem . we have used a fixed neutrino energy of 250 mev and a baseline of 130 km . the mixing parameters are fixed to be @xmath64 , @xmath65 , @xmath66 , @xmath67 and @xmath68 g @xmath69 . the figure is taken from @xcite . , scaledwidth=60.0% ] it is in fact easy to understand the cause of the parameter degeneracy if we use the bi - probability plot . look at fig . [ degeneracy ] . suppose that your experimentalist friend gives you the measurement point denoted as an open circle in fig . [ degeneracy ] . then , it is evident that you can draw two ellipses , as shown in blue solid lines in fig . [ degeneracy ] , that pass through the observed point , which implies the existence of two solutions of @xmath0 and @xmath57 . the two - fold ambiguity is usually called the intrinsic degeneracy . if we are ignorant of the neutrino mass hierarchy , i.e. , the sign of @xmath12 , the two more ellipses can be drawn , as shown by the red dashed line in fig . [ degeneracy ] ; duplication of the solution by the unknown mass hierarchy . altogether one has four - fold parameter degeneracy . unfortunately , it is not the end of the story . if @xmath1 is not maximal , we are enriched with another degeneracy , the @xmath1 octant degeneracy . the @xmath2 disappearance measurement of @xmath70 gives a value of @xmath16 . it then allows two solutions of @xmath1 if @xmath71 , @xmath72 $ ] , one in the first octant and the other in the second octant . since this is `` orthogonal '' to the intrinsic and the sign @xmath12 degeneracies with four solutions of @xmath0 and @xmath57 , the total eight - fold degeneracy results . prior to the systematic discussion of how to solve the degeneracy we want to mention about the simplest method of solving the @xmath73 degeneracy . by tuning the beam energy to the `` shrunk ellipse limit '' @xcite the degeneracy can be reduced to the two - fold one , @xmath74 . notice that there is no confusion between cp violation and cp conservation even in the presence of this degeneracy . it is known that degeneracy of neutrino oscillation parameters acts as a severe limiting factor to the precision determination of @xmath0 , @xmath1 , and @xmath57 . it is particularly true for the @xmath1 octant degeneracy @xcite . expecting the era of precision measurement in the next 10 - 30 years , it is the time that the formulation of the well defined strategy for exploring the whole structure of the lepton flavor mixing is of immense need . toward the goal , i explain in detail how the degeneracy can be resolved by using a concrete setting , which is called `` t2kk '' . it is an acronym for tokai - to - kamioka - korea two detector complex , an upgraded next project to t2k phase i for exploring the whole structure of lepton flavor mixing @xcite . it utilizes two half a megaton ( 0.27 mton fiducial volume ) water cherenkov detectors one in kamioka ( 295 km ) and the other in somewhere in korea ( @xmath551000 km ) which receive @xmath2 and @xmath75 superbeam of 4 mw from j - parc facility . we assume 4 years of running with each neutrino and antineutrino mode . for further details of t2kk , please consult to the original manuscripts @xcite . for a broader view of t2kk including wider class of detector options and locations , see the web page of the workshops which are focused on this project @xcite . though t2kk is _ not _ the unique way of resolving the eight - fold parameter degeneracy it is nice to have a concrete project to solve all the degeneracy _ in situ _ ; it provides the bottom line understanding on how it can be lifted , and the lesson may be useful to think about alternative ways . how does t2kk solve the 8-fold parameter degeneracy ? in a nutshell , the setting can resolve the three kind of degeneracies in the following ways : * the intrinsic degeneracy ; spectrum information solves the intrinsic degeneracy . * the sign-@xmath32 degeneracy ; difference in the earth matter effect between the intermediate ( kamioka ) and the far ( korea ) detectors solves the sign-@xmath32 degeneracy . * the @xmath1 octant degeneracy ; difference in solar @xmath32 oscillation effect ( which is proportional to @xmath76 ) between the intermediate and the far detectors solves the @xmath1 octant degeneracy . let me explain these points one by one . 0.3 cm space by 4 years of neutrino and antineutrino running in t2k ii ( left panel ) , and the kamioka - korea two detector setting ( right panel ) . they are taken from the supplementary figures behind the reference @xcite to which the readers are referred for details of the analysis . notice that the standard setting in t2k ii , 2 ( 6 ) years of neutrino ( antineutrino ) running , leads to a very similar results ( as given in @xcite ) to the one presented in the left panel of this figure . the true solutions are assumed to be located at ( @xmath54 and @xmath57 ) = ( 0.01 , @xmath77 ) with positive sign of @xmath78 , as indicated as the green star . the intrinsic and the @xmath78-sign clones appear in the same and the opposite sign @xmath78 panels , respectively . three contours in each figure correspond to the 68% ( blue line ) , 90% ( black line ) and 99% ( red line ) c.l . sensitivities , respectively . , title="fig:",scaledwidth=46.0% ] space by 4 years of neutrino and antineutrino running in t2k ii ( left panel ) , and the kamioka - korea two detector setting ( right panel ) . they are taken from the supplementary figures behind the reference @xcite to which the readers are referred for details of the analysis . notice that the standard setting in t2k ii , 2 ( 6 ) years of neutrino ( antineutrino ) running , leads to a very similar results ( as given in @xcite ) to the one presented in the left panel of this figure . the true solutions are assumed to be located at ( @xmath54 and @xmath57 ) = ( 0.01 , @xmath77 ) with positive sign of @xmath78 , as indicated as the green star . the intrinsic and the @xmath78-sign clones appear in the same and the opposite sign @xmath78 panels , respectively . three contours in each figure correspond to the 68% ( blue line ) , 90% ( black line ) and 99% ( red line ) c.l . sensitivities , respectively . , title="fig:",scaledwidth=46.0% ] first look at fig . [ intrinsickamkorea ] in which the sensitivities for resolving the intrinsic degeneracy by the tokai - to - kamioka phase ii ( t2k ii ) setting @xcite ( left panel ) and the kamioka - korea two detector setting ( right panel ) are presented . figure [ intrinsickamkorea ] is taken from supplementary figures prepared for the study reported in @xcite , in which the details of the analysis are described . in the left panel of fig . [ intrinsickamkorea ] it is shown that the intrinsic degeneracy in ( assumed ) each mass hierarchy is almost resolved by the t2k ii setting at the particular set of values of the mixing parameters as indicated in the caption . since the matter effect plays minor role in the t2k ii setting it is likely that the spectral information is mainly responsible for lifting the intrinsic degeneracy . in the right panel of the same figure it is exhibited that the intrinsic degeneracy is completely resolved by the t2kk setting at the same values of the mixing parameters , indicating power of the two detector method @xcite . namely , the comparison between the intermediate and the far detectors placed at the first and the second oscillation maxima , respectively , supersedes a single detector measurement in kamioka with the same total volume in spite of much less statistics in the korean detector . 0.3 cm . left panel is for t2kk and the right panel for a single 0.54 megaton detector placed in korea . , title="fig:",scaledwidth=46.0% ] . left panel is for t2kk and the right panel for a single 0.54 megaton detector placed in korea . , title="fig:",scaledwidth=46.0% ] it should be noted that the sign-@xmath32 degeneracy is also lifted though incompletely by the kamioka - korea setting as indicated in the right - lower panel of fig . [ intrinsickamkorea ] . it is well known that the interference effect between the vacuum and the matter effects depends upon the mass hierarchy , i.e. , the sign of @xmath12 , and many people have been proposed to utilize it to resolve the mass hierarchy . but , it is not the whole story here . to indicate this point the sensitivities for the two settings are compared in fig . [ signdm2kamkorea ] . one is t2kk ( left panel ) and the other is the case of single 0.54 megaton detector placed in korea ( right panel ) . it should be noticed that a single detector in korea with the same total volume fails to resolve the sign-@xmath32 degeneracy which is completely lifted by t2kk at the particular values of the mixing parameters . again , the sensitivity is enhanced by comparing the yields of the two identical detectors . the fact that the t2kk setting can resolve the four - fold degeneracy by the spectrum analysis and comparison between the two detectors is explained by plotting the energy dependences of the appearance probabilities in fig . [ spectral ] . -0.3 cm the @xmath1 octant degeneracy arise because accelerator disappearance and appearance measurement determine @xmath79 and the combination @xmath80 , respectively , but not @xmath81 itself . therefore , it is hard to resolve in the accelerator experiments using conventional @xmath2 beam . see @xcite for an explicit analytic treatment of this point . appearance oscillation probabilities with various values of cp phase @xmath57 ; @xmath82 ( dotted line ) , @xmath83 ( dashed line ) , @xmath84 ( dash - dotted line ) , and @xmath85 ( double - dash - dotted line ) . the neutrino mass hierarchy is assumed to be the inverted one . for the corresponding figure of the normal mass hierarchy , see @xcite . , scaledwidth=70.0% ] one way to solve the @xmath1 octant degeneracy is to utilize the solar @xmath32 oscillation term . this is the principle used by the atmospheric neutrino method for resolving the octant degeneracy @xcite . since the solar term , the last term in eq . ( [ pmue ] ) , depend upon @xmath76 ( not @xmath81 ) the degeneracy can be lifted . the next question is if it can be distinguished from the rest of the atmospheric oscillation terms . fortunately , the answer seems to be yes because of the clear difference in energy dependence , as shown in fig . [ solar - vs - atm ] . note that the figure is the inverted hierarchy version of fig . 2 in @xcite , and behavior of the solar term compared to the atmospheric ones is very similar to in the case of the normal hierarchy . in passing , we briefly comment on the problem of decoupling between the degeneracies . for a fuller treatment , see @xcite . the question is as follows : people sometimes discuss how to solve the degeneracy a without worrying about the degeneracy b , and vise versa . is this a legitimate procedure ? we want to answer to this question in the positive under the environment that the matter effect can be treated as a perturbation . to resolve the degeneracy one has to distinguish between the values of the oscillation probabilities with the two different solutions corresponding to the degeneracy . we define the probability difference @xmath86 as a measure for it where the superscripts @xmath87 and @xmath88 label the degenerate solutions . suppose that we are discussing the degeneracy a. the decoupling between the degeneracies a and b holds if @xmath89 defined in ( [ deltapdef ] ) for the degeneracy a is invariant under the replacement of the mixing parameters corresponding to the degeneracy b , and vice versa . the best example of the decoupling is given by the one between the @xmath1 octant and the sign-@xmath32 degeneracies . therefore , let us describe it here , leaving discussions on other cases to @xcite . one can easily compute @xmath90 for the @xmath1 octant degeneracy by using ( [ pmue ] ) . it consists of the solar and the solar - atmospheric interference terms , with over - all factor of @xmath91 because of the property @xmath92 in leading order in @xmath91 . the remarkable feature of @xmath90 is that the leading - order matter effect terms drops out completely . now , we notice the key feature of @xmath90 ; it is invariant under the transformations @xmath93 and @xmath94 , which exchanges the two sign-@xmath32 degenerate solutions , the invariance which holds in the presence of the solar term . it means that resolution of the @xmath1 degeneracy can be executed without knowing the mass hierarchy in experimental set up which allows perturbative treatment of matter effect . next , we examine the inverse problem ; does the determination of mass hierarchy decouple from resolution of the @xmath1 degeneracy ? one can compute in the similar way @xmath95 for the sign-@xmath32 degeneracy . because the exchange of two sign-@xmath32 degenerate solutions is the approximate symmetry of the vacuum oscillation probability @xcite , most of the vacuum terms drops out . we observe that @xmath96 is invariant under transformation @xmath97 and @xmath98 , because its @xmath0 and @xmath1 dependences are through the combination @xmath99 . therefore , resolution of the mass hierarchy can be carried out independently of which solution of the @xmath1 degeneracy is realized in nature . we mention here that the decoupling argument can be generalized to include the other pair of degeneracies as done in @xcite . 0.3 cm ( red dotted , yellow long - dashed , black solid , green dash - dotted , and blue short - dashed lines show the results for @xmath100 0.40 , 0.45 , 0.50 , 0.55 and 0.60 , respectively ) . the sensitivity is defined in the plane of @xmath101 versus cp phase @xmath57 . the top and bottom panels show the cases for positive and negative mass hierarchies , respectively . taken from @xcite . , scaledwidth=64.0% ] 0.3 cm . the meaning of the lines and colors are identical to that in fig . [ sensitivity - mass - hierarchy ] . taken from @xcite . , scaledwidth=64.0% ] -0.2 cm since the space is quite limited , we directly go to the results of our analysis . the original analysis in @xcite has been re - examined with an improved code which takes into account a difference between beam profiles in the intermediate and the far detectors , and the inclusion of the muon disappearance channel @xcite . in fig . [ sensitivity - mass - hierarchy ] , and in fig . [ sensitivity - cp ] , the results of re - analysis for the mass hierarchy resolution and cp violation , respectively , are presented . figure [ sensitivity - mass - hierarchy ] shows that the sensitivity to the mass hierarchy depends very weakly to @xmath1 , as expected by the decoupling argument given in @xcite . the same argument suggests that they obey the scaling behavior ; the curves falls to a single curve if plotted by @xmath102 . the sensitivity greatly improves the one possessed by the original t2k ii setup and is competitive to other similar projects . see @xcite for comparison between the performances of t2kk and t2k ii setting . the @xmath1-independence of the cp sensitivity is even more prominent , as shown in fig . [ sensitivity - cp ] . this feature is again consistent with the decoupling argument . the sensitivity to cp violation is similar to that of the t2k ii setting except for at large @xmath0 region where the t2kk sensitivity surpasses that of the t2k ii . it is due to the fact that the identical two - detector setting solves the degeneracies . we emphasize that the cp sensitivity of t2kk setting at the large @xmath0 region seems to be the largest among the similar proposals including neutrino factory . 0.3 cm octant degeneracy for 0.27 mton detectors both in kamioka and korea . 4 years running with neutrino beam and another 4 years with anti - neutrino beam are assumed . in ( a ) , the sensitivity is defined so that the experiment is able to identify the octant of @xmath1 for any values of the cp phase @xmath57 . in ( b ) , it is defined so that the experiment is able to identify the octant of @xmath1 for half of the cp @xmath57 phase space . taken from @xcite . , scaledwidth=78.0% ] -0.2 cm in fig . [ sensitivity - theta23-octant ] , the sensitivity to the @xmath1 octant degeneracy is presented . from this figure , we conclude that the experiment we consider here is able to solve the octant ambiguity , if @xmath103 or @xmath104 at 3 ( 2 ) standard deviation cl . roughly speaking , the sensitivity is independent of @xmath0 and the mass hierarchy . the dependence of this sensitivity on the cp phase @xmath57 is a mild one as one can see by comparing the left and the right panels of fig . [ sensitivity - theta23-octant ] , providing another evidence for decoupling . as discussed in detail in @xcite , the @xmath1 degeneracy is the difficult one to solve only by the accelerator experiments . though the argument is still true , t2kk circumvents it because it has sensitivity to the solar term . yet , the sensitivity is quite limited if plotted in @xmath81 plane , as one can observe in fig . [ sensitivity - theta23-octant ] . nonetheless , we stress that it is not easy to supersede the sensitivity presented in fig . [ sensitivity - theta23-octant ] . for example , t2kk s sensitivity is slightly better than the one by the atmospheric neutrino method based on 3 years observation in hyper - kamiokande reported in @xcite . we emphasize that our estimates of sensitivities for the mass hierarchy resolution , cp violation , and the @xmath1 octant degeneracy are based on the known technology for rejecting nc induced background in water cherenkov detectors . moreover , we have used a conservative value of 5% for most of the systematic errors @xcite . therefore , our results can be regarded as robust bottom - line sensitivities achievable by conventional superbeam experiments . of course , there may be ways to improve the sensitivities over the current t2kk design . at the end of this section , we should mention that the method explored in this article is by no means unique . with regard to the sign-@xmath32 ( mass hierarchy ) degeneracy we note that other methods include the one which utilize atmospheric neutrinos @xcite , supernova neutrinos @xcite , neutrino - less double beta decay @xcite , and @xmath26 and @xmath2 disappearance channels @xcite . we have already mentioned about the @xmath1 octant degeneracy , and a further comment follows immediately below . 0.8 cm octant degeneracy can be resolved for the case where the true value of @xmath105 = 0.458 ( 0.542 ) , corresponding to @xmath106 . the other input mixing parameters are given as @xmath107 ev@xmath108 , @xmath109 = 0.1 and @xmath110 , @xmath111 ev@xmath108 , @xmath8 = 0.31 ( the input values of @xmath109 and @xmath105 are indicated by the symbol of star in the plot ) . ( a ) the regions enclosed by the solid and the dashed curves are allowed regions only by the results of appearance and disappearance accelerator measurement , respectively . ( b ) the regions that remain allowed when results of appearance and disappearance measurement are combined . ( c ) the regions allowed by reactor measurement . ( d ) the regions allowed after combining the results of appearance and disappearance accelerator experiments with the reactor measurement . the exposures for accelerator are assumed to be 2 ( 6 ) years of neutrino ( anti - neutrino ) running with 4 mw beam power with hyper - kamionande whose fiducial volume is 0.54 mt , whereas for the reactor we assume an exposure of 10 gw@xmath112kt@xmath112yr . the case of optimistic systematic error is taken . the figure is taken from @xcite . , title="fig:",scaledwidth=86.0% ] -1.5 cm -2.0 cm space where the @xmath1 octant degeneracy can be resolved at 90% ( thin green ) and 99% ( thick red ) cl . the solid ( dashed ) curve is for the case of taking the normal ( inverted ) hierarchy to perform the fit , assuming the normal hierarchy as input . conservative systematic errors , as indicated in the figure , are considered here . taken from @xcite . , title="fig:",scaledwidth=92.0% ] -5.4 cm space where the @xmath1 octant degeneracy can be resolved at 90% ( thin green ) and 99% ( thick red ) cl . the solid ( dashed ) curve is for the case of taking the normal ( inverted ) hierarchy to perform the fit , assuming the normal hierarchy as input . conservative systematic errors , as indicated in the figure , are considered here . taken from @xcite . , title="fig:",scaledwidth=92.0% ] -3.4 cm detecting the solar oscillation effect is not the unique way of resolving the @xmath1 octant degeneracy . the alternative methods proposed include , in addition to the already mentioned atmospheric neutrino method , the reactor accelerator combined method @xcite , and the atmospheric accelerator combined method @xcite . here , we explain the reactor - accelerator combined method . the principle is again very simple ; the reactor measurement can pick up one of the solutions of @xmath0 because it is a pure measurement of @xmath0 , the possibility first explored in @xcite . this principle is explained in fig . [ 23fig1 ] and in the caption further , which are taken from @xcite . this reference gives a detailed quantitative analysis of the sensitivity achievable by the accelerator - reactor combined method . in fig . [ region_resolved ] presented is the region in @xmath113 space where the @xmath1 octant degeneracy can be resolved . the upper and the lower figures in fig . [ region_resolved ] are with a relatively pessimistic and an optimistic systematic errors , respectively , as indicated in the figures . for definition of the errors and details of the analysis procedure , see @xcite . by comparing fig . [ region_resolved ] with fig . [ sensitivity - theta23-octant ] , we observe that the sensitivity achievable by the reactor - accelerator combined method surpasses that of t2kk in large @xmath0 region , @xmath114 , the critical value very dependent of the systematic errors . in the bottom - up approach to the origin of neutrino mass and the mixing it is important to test various phenomenologically motivated ideas experimentally . in this article we discuss only one example , the quark - lepton complementarity ( qlc ) @xcite . and briefly mention about the @xmath115 exchange symmetry . an extensive list of the relevant references for the @xmath115 symmetry , which is too long to quote in this manuscript , may be found in @xcite . the empirically suggested relation @xmath116 with @xmath117 being the cabibbo angle is under active investigation @xcite and is dubbed as the qlc relation . if not accidental , it may suggest a new way of thinking on how quarks and leptons are unified . it may have extension to the 2 - 3 sector , @xmath118 . we now discuss how the relation ( [ qlc ] ) can be tested experimentally . since the cabibbo angle is measured in an enormous precision as emphasized earlier , the real problem is to what accuracy the solar angle @xmath6 can be measured experimentally . at this moment there exist two approaches to measure @xmath6 accurately . the first one is a natural extension of the method by which @xmath6 is determined today , namely , combining the solar and the kamland experiments . the other one is to create a dedicated new reactor experiment with detector at around the first oscillation maximum of reactor neutrino oscillation , `` sado '' ( see below ) . let me briefly explain about the basic ideas behind them one by one . combining the solar and the kamland experiments is powerful , assuming cpt invariance , because solar neutrino measurement is good at constraining @xmath6 and kamland determines with high precision the other parameter @xmath13 . the feature makes the analysis of the solar neutrino parameter determination essentially 1-dimensional . the former characteristics is particularly clear from the fact that the ratio of cc to nc rates in sno directly measures @xmath119 in the lma solution . the current data allows accuracy of determination of @xmath119 of about @xmath120% ( 2 dof ) ( the last reference in @xcite . ) further progress in measurement in sno and kamland may improve the accuracy by a factor of @xmath55 2 but not too much beyond that . 0.2 cm c|cc experiments & @xmath121 at 68.27% cl & @xmath121 at 99.73% cl + solar+ kl ( present ) & @xmath122 % & @xmath123 % + solar+ kl ( 3 yr ) & @xmath124 % & @xmath125 % + solar+ kl ( 3 yr ) + pp ( 1% ) & @xmath126 % & @xmath127 % + + sado for 10 @xmath128kt@xmath112yr & 4.6 % ( 5.0 % ) & 12.2 % ( 12.9 % ) + sado for 20 @xmath129kt@xmath112yr & 3.4 % ( 3.8 % ) & 8.8 % ( 9.5 % ) + sado for 60 @xmath129kt@xmath112yr & 2.1 % ( 2.4 % ) & 5.5 % ( 6.2 % ) + 0.4 cm 0.2 cm [ solar - kl ] 0.4 cm ( upper panel ) and @xmath13 ( lower panel ) reachable by kamland and sado ( both 1 dof ) are compared with the same systematic error of 4% . the geo - neutrino contribution was switched off . taken from @xcite . , title="fig:",scaledwidth=64.0% ] -0.2 cm if one wants to improve substantially the accuracy of @xmath6 determination , the existing solar neutrino experiments are not quite enough . measurement of low - energy pp and @xmath130be neutrinos is particularly useful by exploring vacuum oscillation regime . fortunately , varying proposal for such low energy solar neutrino measurement are available in the world @xcite . measurement of @xmath130be neutrinos is attempted in borexino @xcite and in kamland @xcite . the improvement that is made possible by these additional measurement is thoroughly discussed by bahcall and pea - garay @xcite . since the vacuum oscillation is the dominant mechanism at low energies measuring pp neutrino rate gives nothing but measurement of @xmath131 . on the other hand , @xmath130be neutrino may carry unique informations of oscillation parameters due to its characteristic feature of monochromatic energy . the solar - kamland method will allow us to determine @xmath119 to 4% level at 1@xmath132 cl @xcite . in the upper panels of table [ solar - kl ] , we tabulate the sensitivities ( 1 dof ) currently obtained and expected by the future measurement . -2.0 cm -@xmath13 space and are overlaid on fig.6 of the roadmap paper @xcite , in which the sensitivities of solar - kamland combined method are presented . the errors are defined both with 2 dof . taken from @xcite . , title="fig:",scaledwidth=66.0% ] -4.8 cm though natural and profitable as a dual - purpose experiment for both @xmath6 and solar flux measurement the solar - kamland method is not the unique possibility for reaching the region of the highest sensitivity for @xmath6 . the most traditional way of measuring mixing angles at the highest possible sensitivities is either to tune beam energy to the oscillation maximum ( for example @xcite which is for @xmath133 ) , or to set up a detector at baseline corresponding to it as employed by various reactor experiments to measure @xmath0 @xcite . it is also notable that the first proposal of prototype superbeam experiment for detecting cp violation @xcite entailed in a setup at around the first oscillation maximum . for @xmath6 the latter method can be applied to reactor neutrinos and in fact a concrete idea for possible experimental setup for dedicated reactor @xmath6 is worked out in detail @xcite . see also @xcite for the related proposals with reactor neutrinos . the type of experiment is dubbed in @xcite as `` sado '' , an acronym of _ several - tens of km antineutrino detector _ because of the range of baseline distance appropriate for the experiments . it is a very feasible experiment because it does not require extreme reduction of the systematic error to 1% level , as required in the @xmath0 measurement mentioned above . as is demonstrated in @xcite reduction of the systematic error to 4% level would be sufficient if no energy spectrum cut at @xmath134 mev is performed . it should be within reach in view of the current kamland error of 6.5% @xcite . the effect of geo - neutrino background , which then has to be worried about without spectrum cut , is shown to be tolerable even for most conservative choice of geo - neutrino model , the fully radiogenic model @xcite . the accuracy achievable by the dedicated reactor @xmath6 measurement is quite remarkable . it will reach to 2% level at 1@xmath132 cl ( 1 dof ) for 60 gw@xmath135@xmath136 exposure as shown in table [ solar - kl ] . with kashiwazaki - kariwa nuclear reactor complex , it corresponds to about 6 years operation for kamland size detector . it is notable that possible uncertainties due to the surrounding reactors are also modest , as one can see in table [ solar - kl ] . in fig . [ solar - sado ] we show in the two - dimensional space spanned by @xmath137 and @xmath13 the contours of sensitivities achievable by the solar - kamland method and by the dedicated reactor experiment . notice that the measurement is not yet systematics dominated and therefore further improvement of the sensitivity is possible by gaining more statistics . if sado can run long enough it can go beyond the solar - kamland method . the @xmath115 exchange symmetry is attractive because it predicts @xmath138 and @xmath139 in the symmetry limit . for extensive references on this symmetry , see e.g. , @xcite . but , since the symmetry is badly broken ( note that @xmath140 ) , the predictions @xmath138 and @xmath139 can not be exact . it is important to try to compute deviations from the results obtained in the symmetry limit . now , the question is how can we pick up the right one out of the vast majority of the proposed symmetries ? one way to proceed is to make clear how symmetry breaking affects the predictions . for example , it is shown in @xcite that breaking of the @xmath141 symmetry tends to prefer larger deviation from the maximal @xmath1 than vanishing @xmath0 . such study has to be performed in an extensive way including various symmetries . if one can make definitive prediction in a class of models on which octant of @xmath1 is chosen when the @xmath115 symmetry is broken , one can test such a class of models by resolving the @xmath1 octant degeneracy . we have discussed in sec . [ resolved ] and in sec . [ reactor - accelerator ] the ways of how it can be carried out . we have explained in sec . [ theta23 ] how the atmospheric @xmath11 , which may be better characterized as @xmath142 @xcite , can be determined . for various reasons one may want to improve the accuracy of determining @xmath11 to a sub - percent level . here , we want to remark that unfortunately there is a serious obstacle against it ; the problem of absolute energy scale error . see appendix of @xcite for an explicit demonstration of this fact . it comes from the limitation of the accuracy of calibrating the absolute energy of muons in the case of @xmath2 disappearance measurement . the current value for the error in super - kamiokande is about 2% at gev region ( second reference in @xcite ) and apparently no concrete idea has been emerged to improve it . it is believed to be a limiting factor in @xmath11 determination in much higher statistics region enabled by t2k ii with hyper - kamiokande . we note that there is the unique case which is free from the problem of energy scale error ; the recently proposed resonant @xmath41 absorption reaction enhanced by mssbauer effect @xcite . this method utilizes the recoilless resonant absorption reaction , @xmath143 , with monochromatic @xmath41 beam from the t - conjugate bound state beta decay , @xmath144 . when the source atoms are embedded into a solid the energy width of the beam is estimated to be @xmath145 ev , which is utterly negligible . if feasible experimentally , the monochromatic nature of the beam may allow accurate measurement of @xmath12 to @xmath146 at 1@xmath132 cl @xcite . we emphasize that it gives a very rare chance of achieving a sub - percent level determination of @xmath12 . i have tried to give an overview of neutrino physics emphasizing the experimental activities in the near future . i must admit that this is a personal overview , not mentioning very many important subjects and projects in all over the world , and i have to apologize for that . but , i tried to give a coherent view which largely come from the works i did in the last several years . i feel it appropriate to emphasize that we have learned a lot during the golden era of neutrino physics . but , it seems obvious to me that we have done only a half and new surprises are waiting for us in the future . i would like to thank all of my collaborators , in particular takaaki kajita , hiroshi nunokawa , masaki ishitsuka , shoei nakayama , renata zukanovich funchal , stephen parke , hiroaki sugiyama , and shoichi uchinami for fruitful collaborations and useful discussions . this work was supported in part by the grant - in - aid for scientific research , no . 16340078 , japan society for the promotion of science . y. fukuda _ et al . _ [ super - kamiokande collaboration ] , phys . rev . * 81 * , 1562 ( 1998 ) [ arxiv : hep - ex/9807003 ] ; y. ashie _ et al . _ [ super - kamiokande collaboration ] , phys . d * 71 * , 112005 ( 2005 ) [ arxiv : hep - ex/0501064 ] . b. t. cleveland _ et al . _ , astrophys . j. * 496 * , 505 ( 1998 ) ; y. fukuda _ et al . _ [ kamiokande collaboration ] , phys . rev . lett . * 77 * , 1683 ( 1996 ) ; j. n. abdurashitov _ et al . _ [ sage collaboration ] , phys . c * 60 * , 055801 ( 1999 ) [ arxiv : astro - ph/9907113 ] ; w. hampel _ et al . _ [ gallex collaboration ] , phys . b * 447 * , 127 ( 1999 ) ; m. altmann _ et al . _ [ gno collaboration ] , phys . b * 616 * , 174 ( 2005 ) [ arxiv : hep - ex/0504037 ] ; j. hosaka _ et al . _ [ super - kamkiokande collaboration ] , phys . d * 73 * , 112001 ( 2006 ) [ arxiv : hep - ex/0508053 ] ; b. aharmim _ et al . _ [ sno collaboration ] , phys . rev . c * 72 * , 055502 ( 2005 ) [ arxiv : nucl - ex/0502021 ] . q. r. ahmad _ et al . _ [ sno collaboration ] , phys . lett . * 87 * , 071301 ( 2001 ) [ arxiv : nucl - ex/0106015 ] ; s. fukuda _ et al . _ [ super - kamiokande collaboration ] , phys . lett . * 86 * , 5651 ( 2001 ) [ arxiv : hep - ex/0103032 ] . q. r. ahmad _ et al . _ [ sno collaboration ] , phys . rev . * 89 * , 011301 ( 2002 ) [ arxiv : nucl - ex/0204008 ] . s. n. ahmed _ et al . _ [ sno collaboration ] , phys . lett . * 92 * , 181301 ( 2004 ) [ arxiv : nucl - ex/0309004 ] . m. apollonio _ et al . _ [ chooz collaboration ] , phys . b * 420 * , 397 ( 1998 ) [ arxiv : hep - ex/9711002 ] ; _ ibid . _ b * 466 * , 415 ( 1999 ) [ arxiv : hep - ex/9907037 ] . see also , the palo verde collaboration , f. boehm _ et al . _ , phys . d * 64 * ( 2001 ) 112001 [ arxiv : hep - ex/0107009 ] . s. nakayama _ et al . _ [ k2k collaboration ] , phys . lett . b * 619 * , 255 ( 2005 ) [ arxiv : hep - ex/0408134 ] ; m. hasegawa _ et al . _ [ k2k collaboration ] , phys . rev . * 95 * , 252301 ( 2005 ) [ arxiv : hep - ex/0506008 ] ; m. o. wascko [ miniboone collaboration ] , arxiv : hep - ex/0602050 ; k. s. mcfarland [ minerva collaboration ] , arxiv : physics/0605088 ; a. a. aguilar - arevalo _ et al . _ [ sciboone collaboration ] , arxiv : hep - ex/0601022 . o. l. g. peres and a. y. smirnov , phys . b * 456 * , 204 ( 1999 ) [ arxiv : hep - ph/9902312 ] ; nucl . b * 680 * , 479 ( 2004 ) [ arxiv : hep - ph/0309312 ] ; m. c. gonzalez - garcia , m. maltoni and a. y. smirnov , phys . d * 70 * , 093005 ( 2004 ) [ arxiv : hep - ph/0408170 ] . m. shiozawa , t. kajita , s. nakayama , y. obayashi , and k. okumura , in proceedings of the rccn international workshop on sub - dominant oscillation effects in atmospheric neutrino experiments , kashiwa , japan , dec . 2004 , p.57 ; t. kajita , nucl . . suppl . * 155 * , 87 ( 2006 ) . t. kajita , talk given at next generation of nucleon decay and neutrino detectors ( nnn05 ) , aussois , savoie , france , april 7 - 9 , 2005 . http://nnn05.in2p3.fr/ ; s. palomares - ruiz and s. t. petcov , nucl . b * 712 * , 392 ( 2005 ) [ arxiv : hep - ph/0406096 ] ; s. t. petcov and t. schwetz , nucl . b * 740 * , 1 ( 2006 ) [ arxiv : hep - ph/0511277 ] ; r. gandhi , p. ghoshal , s. goswami , p. mehta and s. uma sankar , arxiv : hep - ph/0506145 . a. s. dighe and a. y. smirnov , phys . d * 62 * , 033007 ( 2000 ) [ arxiv : hep - ph/9907423 ] ; h. minakata and h. nunokawa , phys . b * 504 * , 301 ( 2001 ) [ arxiv : hep - ph/0010240 ] ; c. lunardini and a. y. smirnov , jcap * 0306 * , 009 ( 2003 ) [ arxiv : hep - ph/0302033 ] . v. barger , p. huber and d. marfatia , phys . b * 617 * , 167 ( 2005 ) [ arxiv : hep - ph/0501184 ] . a. s. dighe , m. t. keil and g. g. raffelt , jcap * 0306 * , 005 ( 2003 ) [ arxiv : hep - ph/0303210 ] ; _ ibid . _ * 0306 * , 006 ( 2003 ) [ arxiv : hep - ph/0304150 ] ; r. tomas , m. kachelriess , g. raffelt , a. dighe , h. t. janka and l. scheck , jcap * 0409 * , 015 ( 2004 ) [ arxiv : astro - ph/0407132 ] . h. minakata , h. nunokawa , w. j. c. teves and r. zukanovich funchal , phys . d * 71 * , 013005 ( 2005 ) [ arxiv : hep - ph/0407326 ] ; h. minakata , h. nunokawa , w. j. c. teves and r. zukanovich funchal , nucl . suppl . * 145 * , 45 ( 2005 ) [ arxiv : hep - ph/0501250 ] . a. bandyopadhyay , s. choubey , s. goswami and s. t. petcov , phys . d * 72 * , 033013 ( 2005 ) [ arxiv : hep - ph/0410283 ] ; j. f. kopp , m. lindner , a. merle and m. rolinec , arxiv : hep - ph/0606151 ; s. t. petcov and t. schwetz , arxiv : hep - ph/0607155 .

we present an approach that combines the local density approximation ( lda ) and the dynamical mean - field theory ( dmft ) in the framework of the full - potential linear augmented plane waves ( flapw ) method . wannier - like functions for the correlated shell are constructed by projecting local orbitals onto a set of bloch eigenstates located within a certain energy window . the screened coulomb interaction and hund s coupling are calculated from a first - principle constrained rpa scheme . we apply this lda+dmft implementation , in conjunction with a continuous - time quantum monte - carlo algorithm , to the study of electronic correlations in lafeaso . our findings support the physical picture of a metal with intermediate correlations . the average value of the mass renormalization of the fe 3 bands is about , in reasonable agreement with the picture inferred from photoemission experiments . the discrepancies between different lda+dmft calculations ( all technically correct ) which have been reported in the literature are shown to have two causes : i ) the specific value of the interaction parameters used in these calculations and ii ) the degree of localization of the wannier orbitals chosen to represent the fe 3 states , to which many - body terms are applied . the latter is a fundamental issue in the application of many - body calculations , such as dmft , in a realistic setting . we provide strong evidence that the dmft approximation is more accurate and more straightforward to implement when well - localized orbitals are constructed from a large energy window encompassing fe-3 , as-4 and o-2 , and point out several difficulties associated with the use of extended wannier functions associated with the low - energy iron bands . some of these issues have important physical consequences , regarding in particular the sensitivity to the hund s coupling .

0906.3735

this article has two purposes . the first one is to present a new implementation of dynamical mean - field theory ( dmft ) within electronic structure calculation methods . this implementation is based on a highly precise full - potential linear augmented plane wave method ( flapw ) , as implemented in the wien2k electronic structure code @xcite . the second purpose of this article is to report on dmft calculations for the iron oxypnictide lafeaso , the parent compound of the ` @xmath3'-family of recently discovered iron - based superconductors . the strength of electronic correlations in these materials is an important issue , which has been a subject of debate in the literature @xcite . the combination of dynamical mean - field theory with density - functional theory in the local density approximation ( lda+dmft ) provides a powerful framework for the quantitative description of electronic correlations in a realistic setting . a number of materials have been investigated in this framework over the past decade , such as transition metals and transition - metal oxides , rare - earth and actinide compounds , and organic conductors . these examples testify to the progress in our understanding of the key physical phenomena associated with the competition between the localized and itinerant characters of electrons belonging to different orbitals ( see e.g. refs . [ ] for reviews ) . in the past few years , a new generation of lda+dmft implementations have been put forward @xcite , which emphasize the use of wannier functions as a natural bridge between the band - structure and the real - space description of the solid in terms of orbitals . these functions span the subset of orbitals which are treated within the many - body dmft framework . in this article , we present an implementation of lda+dmft within the full potential linear augmented plane wave ( flapw ) framework , using atomic orbitals that are promoted to wannier functions by a truncated expansion over bloch functions followed by an orthonormalization procedure . this is a simpler alternative to the previous implementation of dmft within flapw @xcite , which constructed the wannier functions following the prescription of maximal localisation @xcite . the choice of flapw is motivated by the high level of accuracy of this all - electron , full - potential method . in the present work , we use the wien2k electronic structure package @xcite , and we have constructed an interface to it that allows for the construction of wannier - like functions used in dmft . our implementation is described in detail in sec . [ sect : theory ] . as a benchmark , we perform calculations on a test material , srvo@xmath4 , which are presented in appendix [ sect : srvo3 ] and compared to previously published results for this material @xcite . throughout this article , many - body effects are treated in the dmft framework using the recently developed continuous - time strong - coupling quantum monte carlo algorithm of p.werner and coworkers @xcite . because very low temperatures can be reached , and very high accuracy can be obtained at low - frequency , this algorithm represents a major computational advance in the field . in sec . [ sect : results ] , we address the issue of electronic correlations in lafeaso . the dmft calculations which have been published soon after the experimental discovery of superconductivity in the iron oxypnictides have provided seemingly contradictory answers to this question . in refs . [ ] , k. haule and g. kotliar proposed that lafeaso is a strongly correlated metal , rather close to the mott metal - insulator transition , and characterized by a reduced value of the quasiparticle coherence scale , resulting in bad metallic behavior . in contrast , in refs . [ ] , v.anisimov and coworkers proposed that these materials are in a weak to intermediate regime of correlations . our lda+dmft calculations for lafeaso support the physical picture of a metal with intermediate correlations . the average value of the mass renormalization of the fe 3@xmath0 bands is about @xmath1 , in reasonable agreement with the picture inferred from photoemission experiments . we also find that there is no technical inconsistency between different dmft results reported for lafeaso before . we show that the discrepancies in the literature are due to two causes : i ) the specific value of the interaction parameters used in these calculations and ii ) the degree of localization of the wannier orbitals chosen to represent the fe 3@xmath0 states , to which many - body terms are applied . in sec . [ sect : results ] , we perform detailed comparisons between lda+dmft calculations performed with different degree of localization of the correlated orbitals , associated with different choices of energy windows for the wannier construction ( and accordingly , different degrees of screening of the interaction parameters ) . we point out several difficulties associated with the use of more extended wannier functions associated with the low - energy iron bands only . some of these issues have important physical consequences , in particular regarding the sensitivity to the hund s coupling . this article ends with several appendices , reporting on more detailed aspects or technical issues . appendix a is devoted to a benchmark of our implementation on a `` classical '' test compound , srvo@xmath4 . appendix b details some technical issues associated with the projection scheme used to display partial spectral functions with a given orbital character . appendix c discusses the influence of spin - flip and pair - hopping terms on the degree of correlations , on the basis of model calculations . we conclude that while these terms are indeed important close to the mott transition , they can safely be neglected in the regime of correlations relevant to lafeaso . to make this article self - contained and in order to define the main notations , this subsection begins by briefly reviewing some essential aspects of the lda+dmft framework . the presentation is close to that of refs . [ ] , where additional details can be found . dynamical mean - field theory is a quantitative method for handling electron correlations , which can be described as an `` effective atom '' approach . the self - energy in the solid is approximated by that of a local model , a generalized anderson impurity model describing a specific set of atomic - like orbitals coupled to a self - consistent environment . the self - consistency requirement is that the local on - site green s function of the solid , calculated using this local self - energy , must coincide with the green s function of the effective impurity model . in order to formulate the local effective atom problem , a set of ( orthonormal ) local orbitals @xmath5 , and corresponding wannier - like functions @xmath6 , must be constructed . these wannier functions span the `` correlated '' subspace @xmath7 of the full hilbert space , in which many - body correlations ( beyond lda ) are taken into account . this set of orbitals spanning the correlated subspace must be clearly distinguished from the full basis set of the problem , in which the green s function of the solid can be expressed . obviously , the basis set spans a much larger hilbert space , involving all relevant electronic shells . below , we discuss in details how the @xmath6 are constructed from the local orbitals @xmath5 . the index @xmath8 is an orbital index within the correlated subset , @xmath9 denotes the atom in the unit cell , and @xmath10 is the spin degree of freedom . projections of quantities of interest on the subset @xmath7 are done using the projection operator @xmath11 the effective impurity model is then constructed for the correlated subset @xmath7 . it is defined by the green s function of the effective environment , @xmath12 and hubbard - kanamori interaction parameters @xmath13 . by solving this model in a suitably chosen manner one obtains the impurity green s function @xmath14 as well as the impurity self - energy @xmath15 for the formulation of the self - consistency condition relating the lattice green s function of the solid to the impurity model , it is convenient to choose the bloch basis @xmath16 as the complete basis set of the problem , since it is a natural output of any electronic structure calculation . the ( inverse ) green s function of the solid expressed in this basis set is given by : @xmath17 where @xmath18 are the kohn sham eigenvalues and @xmath19 is the approximation to the self - energy obtained by the solution of the dmft impurity problem . it is obtained by `` upfolding '' the impurity local self - energy as @xmath20 where @xmath21 are the matrix elements of the projection operator , eq . ( [ eq : projop ] ) and @xmath22 here , @xmath23 is the impurity self - energy , eq . ( [ eq : sigmaimp ] ) , expressed in the local orbitals , and @xmath24 is a double - counting correction , which will be discussed in sect . [ sect : impl_compmeth ] . the local green s function is obtained by projecting the lattice green s function to the set of correlated orbitals @xmath8 of the correlated atom @xmath9 and summing over the full brillouin zone , @xmath25 note that the local quantities @xmath26 and @xmath27 carry also an index @xmath9 , which we suppressed for better readability . the self - consistency condition of dmft imposes that the _ local _ green s function , eq . ( [ eq : local - g ] ) must coincide with the one obtained from the effective impurity problem , @xmath28 this equation implies that the green s function of the effective environment , @xmath29 , must be self - consistently related to the self - energy of the impurity model through : @xmath30 where the dependence of @xmath31 on @xmath32 is specified by eqs . ( [ eq : latt - g],[eq : local - g ] ) . in practice , the dmft equations are solved iteratively : starting from an initial @xmath29 , the impurity model is solved for @xmath32 , and a new @xmath29 is constructed from ( [ eq : selfconsist - sigma ] ) . the cycle is repeated until convergence is reached . in order to construct the set of wannier functions , we start from a set of local atomic - like orbitals @xmath5 defined in the unit cell . these orbitals can be expanded over the full bloch basis - set as : @xmath33 , and restricting the sum to those bloch states with kohn - sham energies @xmath34 within @xmath35 . the number of bands included in @xmath35 will in general depend on @xmath36 and @xmath10 . we thus define the modified orbitals ( which do not form an orthonormal set because of the truncation ) : @xmath37 the matrix @xmath38 is not unitary , except when the sum in ( [ eq : correl - orb-1 ] ) is carried over all bloch bands . it is also important to note that the matrices @xmath39 are in general non - square matrices . they reduce to square matrices only in the case when the number of kohn - sham bands contained in the chosen window equals at every @xmath40-point the number of correlated local orbitals to be constructed . the orbitals @xmath41 can be orthonormalized , giving a set of wannier - like functions : @xmath42 where @xmath43 and @xmath44 the overlap matrix elements . the overlap @xmath45 finally reads @xmath46 while the orthonormalized projectors are then written as @xmath47^{-1/2}\right\}_{m , m'}^{\alpha,\alpha ' } \widetilde{p}_{m'\nu}^{\alpha',\sigma}(\mathbf{k } ) . \label{eq : wannier - proj}\ ] ] in this work , the bloch basis @xmath16 are expanded in augmented plane waves ( apw / lapw ) , which will be briefly described in the following . as was first pointed out by slater @xcite , near atomic nuclei the crystalline potential in a solid is similar to that of a single atom , while in the region between nuclei ( in the interstitial ) the potential is rather smooth and weakly - varying . hence , one may introduce a set of basis functions , _ augmented plane waves _ , adapted to this general shape of the potential . first the crystal space is divided into non - overlapping muffin - tin ( mt ) spheres centered at the atomic sites and the interstitial region in between . in the interstitial region(@xmath48 ) the apw @xmath49 is simply the corresponding plane wave for given reciprocal lattice vector @xmath50 and crystal momentum @xmath40 : @xmath51 where @xmath52 is the unit cell volume . this plane wave is augmented inside each of the mt - spheres by a combination of the radial solutions of the schrdinger equation in such way that the resulting apw is continuous at the sphere boundary . the apw are then employed to expand the kohn - sham ( ks ) eigenstates @xmath53 for the full kohn - sham ( ks ) potential , ( without any shape approximation ) . in the original formulation of the apw method the radial solutions expanding a ks eigenstate inside mt - spheres had to be evaluated at the corresponding eigenenergy leading to an energy - dependent basis set and , hence , to a non - linear secular problem . in order to avoid this complication , linearized versions of the apw method have been proposed . there are two widely - used schemes for the apw linearization . in the first , the linear apw ( lapw ) method @xcite , the plane wave is augmented within mt - spheres by a combination of the radial solutions , evaluated at chosen linearization energies @xmath54 , and their energy derivatives . the resulting linear augmented plane wave then reads : @xmath55y_{m}^{l}(\hat{r } ) & \qquad\mathbf{r}\in r_{mt}^{\alpha}\end{array}\right.\label{eq : lapw}\ ] ] where the index @xmath56 runs over all @xmath57 atomic sites in the unit cell , the coefficients @xmath58 and @xmath59 are determined from the requirement for the linear apw to be continuous and differentiable at the sphere boundary , @xmath60 and @xmath61 are the radial and angular parts of the position vector , respectively . the energy - independent basis set ( [ eq : lapw ] ) leads to a linear secular problem , however , compared to the energy - dependent apw , a larger number of the lapw in the basis set is generally required to attain the same accuracy . in order to decrease the requirement for the number of apw another linearization scheme , apw+@xmath62 @xcite has been proposed . the apw+@xmath62 basis set consists of the augmented plane waves evaluated at a fixed energy @xmath54 : @xmath63 where the coefficient @xmath58 are determined from the requirement for @xmath49 to be continuous at the sphere boundary . to increase the variational freedom of the apw+@xmath62 basis set the fixed - energy apw ( [ eq : apw ] ) are supplemented for the physically important orbitals ( with @xmath64 ) by the local orbitals ( @xmath62 ) that are not matched to any plane wave in the interstitial and are defined only within the muffin tin spheres ( @xmath65 ) @xmath66y_{m}^{l}(\hat{r } ) \label{eq : lo - apw}\ ] ] with the coefficients @xmath58 and @xmath59 chosen from the requirement of zero value and slope for the local orbital at the sphere boundary . additional local orbitals ( usually abbreviated with the capital letters as @xmath67 , @xmath68 ) can be introduced to account for semicore states . they have a similar form as eq . ( [ eq : lo - apw ] ) with the redefined @xmath69 and a second term with a coefficient @xmath70 and the radial function evaluated at a corresponding energy @xmath71 for the semicore band . the coefficient @xmath59 is set to @xmath72 in the apw+@xmath62 framework . generally , in the full potential augmented plane wave method the lapw , apw+@xmath62 and @xmath67 types of orbitals can be employed simultaneously . the kohn - sham eigenstate is expanded in this mixed basis as : @xmath73 where @xmath74 is the number of the orbitals in the basis set . the lda+dmft framework introduced in the present work can also be used in conjunction with any mixed apw+@xmath62/lapw/@xmath67 basis set . having defined the basis set we may now write down the expression for the bloch eigenstate expanded in the general apw basis ( [ eq : ks - eigen ] ) . for @xmath75 in the interstitial region , it reads : @xmath76 while for the region within the mt - spheres @xmath77 , we have : @xmath78y_{m}^{l}(\hat{r})+\\ & + \overset{n_{lo}}{\underset{n_{lo}=1}{\sum}}c_{lo}^{\nu,\sigma } \left[a_{lm}^{\alpha , lo}u_{l}^{\alpha,\sigma}(r , e_{1l}^{\alpha})+c_{lm}^{\alpha , lo}u_{l}^{\alpha,\sigma}(r , e_{2l}^{\alpha})\right]y_{m}^{l}(\hat{r } ) , \end{split } \label{eq : ks - eigen-2}\ ] ] where @xmath79 is the total number of plane waves considered in the interstitial which in turn is augmented inside each mt - sphere , @xmath80 is the number of @xmath81 orbitals of eq . ( [ eq : lo - apw ] ) and @xmath82 the corresponding number of auxiliary orbitals for semicore states @xmath83 . in the framework of the apw method one has several choices for the ` initial ' correlated orbitals @xmath5 . any suitable combination of the radial solution of the schrdinger equation and its energy derivative for a given correlated shell @xmath84 can be employed , for example , the @xmath62-orbital ( [ eq : lo - apw ] ) . in the present paper we simply chose the @xmath5 s as the solutions of the schrdinger equation within the mt - sphere @xmath85 at the corresponding linearization energy @xmath54 . inserting @xmath86 and the expansion ( [ eq : ks - eigen-2 ] ) of the bloch eigenstate in terms of apws into eqs . ( [ eq : correl - orb-1],[eq : proj-1 ] ) , and making use of the orthonormality of the radial solutions and their energy derivatives : @xmath87 one obtains the following expression for the projection operator matrix element : @xmath88 in this expression , the first term in the right hand side of ( [ eq : proj-2 ] ) is due to the contribution from the lapw and/or apw+@xmath62 orbitals @xmath89 and the contribution due to the @xmath67 ( semicore ) orbitals that arises due to mutual non - orthogonality of the radial solutions of the schrdinger equation for different energies @xmath90 where @xmath91 is the corresponding overlap : @xmath92 then we orthonormalize the obtained local orbitals to form a set of wannier - like functions , eq . ( [ eq : wannier ] ) . the corresponding projection operator matrix elements ( [ eq : proj-2 ] ) are orthonormalized accordingly using eq . ( [ eq : wannier - proj ] ) . for the electronic structure calculation we use the full potential apw+@xmath62/lapw code as implemented in the wien2k package @xcite . we have built an interface that constructs the projectors to the correlated orbitals ( @xmath93 ) out of the eigenstates produced by the wien2k code , as described in sect . [ sect : locorbsinapw ] . in order to obtain the local green s function , the summation over momenta , eq . ( [ eq : local - g ] ) , is done in the irreducible brillouin zone ( bz ) only , supplemented by a symmetrization procedure which is standard in electronic structure calculations , @xmath94 where @xmath95 is any @xmath36-dependent matrix , @xmath96 the number of symmetry operations and @xmath97 the symmetrization matrices . furthermore , we construct the local orbitals in the local coordinate system of the corresponding atom . this means that the equivalent atoms in the unit cell for which the dmft should be applied , e.g. the two fe atoms in the oxypnictides , are exactly the same and the impurity problem has to be solved only once . afterward , the green s function and self - energies are put back to the global coordinate system of the crystal in which the bloch green s function , eq . ( [ eq : latt - g ] ) is formulated . for the solution of the impurity problem we use the strong - coupling version of the continuous - time quantum monte carlo method ( ctqmc ) @xcite . it is based on a hybridization expansion and has proved to be a very efficient solver for quantum impurity models in the weak and strong correlation regime . it allows us to address room temperature ( @xmath98ev@xmath99 ) without problems . in our calculations , we used typically around @xmath100 monte - carlo sweeps and 1000 @xmath36-points in the irreducible bz . since the ctqmc solver computes the green s function on the imaginary - time axis , an analytic continuation is needed in order to obtain results on the real - frequency axis . here , we choose to perform a continuation of the impurity self - energy using a stochastic version of the maximum entropy method,@xcite yielding real an imaginary parts of the retarded self - energy @xmath101 which can be inserted into eq . ( [ eq : latt - g ] ) in order to obtain the lattice spectral function and density of states . the ctqmc strong - coupling algorithm can deal with the full rotationally invariant form of the interaction hamiltonian @xcite . however , most calculations presented in this article will consider only the ising terms of the hund s coupling , yielding the interaction hamiltonian : @xmath102 with @xmath103 and @xmath104 the reduced interaction matrices for equal and opposite spins , respectively . this enables us to take advantage of a maximal amount of conserved quantum numbers and , hence , perform the ctqmc calculation without any truncation of the local basis . the effects of spin - flip and ` pair - hopping ' terms in the hund s interaction are discussed in appendix [ sect : hundsrule ] . in our approach , the interaction matrices are expressed in terms of the slater integrals @xmath105 , @xmath106 , and @xmath107 , where for @xmath0-electrons these parameters are related to the coulomb and hund s coupling via @xmath108 , @xmath109 , and @xmath110.@xcite using standard techniques the four - index @xmath111-matrix is calculated , and the reduced interaction matrices are then given by @xmath112 and @xmath113 . with the above definitions , the coulomb parameters @xmath111 and @xmath114 are related to the matrices via @xmath115 as mentioned above , the lda+dmft scheme ( as the lda+u one ) involves a double - counting correction @xmath24 in eqs . ( [ eq : latt - self],[sigma_mm ] ) . indeed , on - site coulomb interactions are already treated on mean - field level in lda . several forms of the double - counting correction term have been proposed and investigated.@xcite in this work we follow ref . and use the following double - counting correction : @xmath116,\label{eq : dc}\ ] ] where @xmath111 is the average coulomb interaction , @xmath114 the hund s rule coupling , @xmath117 the spin - resolved occupancy of the correlated orbitals , and @xmath118 . we compared the results obtained with eq . ( [ eq : dc ] ) also with the double - counting correction given in ref . which gave very similar results . the wannier functions defining the correlated subspace of orbitals for which a dmft treatment is performed , are constructed by truncating the expansion of the initial atomic - like local orbitals to a restricted energy window @xmath119 , as described above . the choice of this energy window is an important issue , which deserves further discussion . indeed , it will determine the shape and the degree of localization of the resulting wannier - like functions . let us consider first the case of a rather small energy window , containing only those bands that have dominantly an orbital character which qualifies them as `` correlated '' ( e.g. , the fe-@xmath120 orbitals in lafeaso or the v-@xmath120-t@xmath121 orbitals in srvo@xmath4 ) . in that case , the dimension of the kohn - sham hamiltonian used in eq . ( [ eq : latt - g ] ) coincides with that of the correlated subspace @xmath122 ( i.e. with the number of orbitals involved in the effective impurity model , for a single correlated atom per cell ) . the wannier - like functions are then quite extended in real space , and resemble strongly the wannier orbitals constructed within other schemes , such as the maximally localized wannier construction of ref . , or the @xmath123-th order muffin - tin basis set downfolded to that set of bands.@xcite in such a situation , hybridization of the correlated orbitals with states that lie outside the energy window is neglected at the dmft level . some information about the hybridization , e.g. , of the d - states with ligand orbitals is of course taken into account through the leakage of the wannier orbitals on neighboring ligand sites ( see , e.g. , ref . ) . in contrast , if a larger energy window is chosen , it will in general contain states treated as correlated as well as states on which no hubbard interactions are imposed . in this case , the dimension of the kohn - sham hamiltonian used in the lda+dmft calculation of the local green s function ( [ eq : latt - g],[eq : local - g ] ) exceeds the number of correlated orbitals involved in the effective impurity model . the wannier functions are more localized in space , and the information about the hybridization of the correlated orbitals with other states within this larger window is carried by the off - diagonal blocks of the hamiltonian between correlated and uncorrelated states . an instructive case occurs when the correlated bands are well separated from the uncorrelated bands at all @xmath40-points , but the bands overlap in energy . this situation is realized for instance for the t@xmath121 bands in srvo@xmath4 that extend into the energy region of the e@xmath124 bands . in order to strictly pick the three correlated t@xmath121 bands at each @xmath40-point , one would in that case have to introduce a @xmath40-dependent energy window . for a @xmath40-independent energy window , one will in general have more than three bands at some @xmath40-points , corresponding to some @xmath125 contribution in the chosen window . this can then be expected to result in slightly more localized orbitals . an example is given in appendix [ sect : benchmark ] . the choice of the energy window influences the value of the interaction parameters @xmath13 in a crucial manner , which can be traced back to two main reasons . first , the interaction parameters are related to matrix elements of a screened interaction between the chosen wannier functions . the more bands are included in their construction , the more localized they become and , hence , the matrix elements increase . second , screening effects themselves affect the value of @xmath111 . the more states are excluded from the screening process , the larger @xmath111 becomes . in what follows we distinguish carefully between these two effects . in the present work , we apply the present lda+dmft implementation to one of the new high - t@xmath126 superconductors , lafeaso . for constructing the wannier functions , we focus on an energy window that contains the 10 bands around the fermi level with dominantly fe-@xmath120 character and also the bands coming from the @xmath2-bands of o and as , which are mainly located in the energy region @xmath127$]ev , resulting in a `` @xmath128-hamiltonian '' . in addition , we performed calculations also for a smaller energy window containing only the fe-@xmath120 bands , yielding a `` @xmath0-hamiltonian '' , as well as for a very large window including around 60 bloch bands . the values of the coulomb interactions @xmath111 and @xmath114 are calculated from the constrained random phase approximation ( crpa)@xcite , using the recently developed scheme for entangled band structures@xcite . crpa calculations for lafeaso have been performed before in refs . . in ref . the screened coulomb parameters are obtained for three different situations : ( i ) by constructing wannier functions from an energy window comprising the fe-@xmath0 bands only , and screening calculated excluding the fe-@xmath0 channels only , ( ii ) by considering a larger window which also includes the as and o-@xmath2 states , so that the screening processes , for instance , from the as-@xmath2 states to the fe-@xmath0 ones are also excluded , ( iii ) and finally , a hybrid situation ( dubbed ` @xmath0-@xmath128 ' in ref . ) , in which the wannier functions are calculated from an energy window including fe-@xmath0 , o-@xmath2 , and as-@xmath2 , but only the fe-@xmath0 states are excluded from the screening . in other words , the screening is calculated as in ( i ) and the wannier functions as in ( ii ) . as discussed in ref . , this third option should be appropriate in a situation in which a full @xmath128-hamiltonian is used but hubbard interactions are only applied to the @xmath0 states . here , we follow the same approach as in the `` @xmath0-@xmath128 '' case of ref . , but using the new disentanglement scheme of ref . . first , a partially screened coulomb interaction @xmath129 is constructed as follows : wannier functions for the fe-@xmath0 states are calculated , and a basis for the complementary subspace ( containing in particular the ligands , but also higher lying f - states ) is constructed . based on the interpolating @xmath0-band structure , the @xmath130 polarization is computed , and the partially screened coulomb interaction @xmath129 is obtained by screening the bare coulomb interaction by all rpa screening processes _ except @xmath130_. finally , according to the crpa procedure the hubbard @xmath111 matrix is composed of the matrix elements of @xmath129 in the basis of @xmath128-wannier functions . as argued in ref . this procedure is suitable for calculations that deal with the full @xmath128-hamiltonian in the many - body calculations while explicitly retaining coulomb interactions on the @xmath0-submanifold only . in particular , we stress that to the extent that our projection method produces wannier functions for the @xmath128-window , the hubbard @xmath111 parameters are expressed in the same basis as the impurity quantities . it is important to note that we keep the screening channels in crpa , i.e. @xmath130 , unchanged when the energy window in our calculation is varied . thus , the different energy windows affect only the localization of the wannier functions , but not the screening process of the bare coulomb interaction , and therefore the effective interactions are increasing with increasing energy window . keeping the screening channels fixed is fully consistent with the fact that correlations are only included for the @xmath0 electrons , but not for the ligand states . for our purposes , we calculate the average coulomb interaction @xmath111 and hund parameter @xmath114 from the matrices calculated by crpa . with this @xmath111 and @xmath114 , the interaction matrices in the spherical symmetric approximation used in our calculation are obtained as discussed above . as we will discuss in more detail below , the comparison of the resulting @xmath103 and @xmath131 with the crpa matrices shows that for the @xmath128-hamiltonian , the approximation using atomic values for the ratios of slater integrals @xmath132 is well justified , whereas for the @xmath0-hamiltonian the crpa matrices show strong orbital anisotropies . let us start the discussion of correlation effects in lafeaso with our results for the @xmath128-hamiltonian , for which wannier functions are constructed from the energy window @xmath133$]ev . these wannier functions are quite well localised . the corresponding kohn - sham hamiltonian contains @xmath134 bloch bands , corresponding to the @xmath135 fe-@xmath120 bands , the @xmath136 as-@xmath2 and the @xmath136 o-@xmath2 bands . the local many - body interactions corresponding to this choice of wannier functions are obtained from crpa , as described in the previous section . they read : @xmath137 the ordering of orbitals in those matrices is @xmath138 , @xmath139 , @xmath140 , @xmath141 , @xmath142 . according to the conventions of the formulae eqs . ( [ eq : uaverage ] ) , ( [ eq : javerage ] ) , these matrices correspond to the values : @xmath143ev and @xmath144ev . using these values of @xmath111 and @xmath114 , we construct the spherically symmetric interaction matrices , @xmath145 it is obvious , that the approximation of the crpa matrices by using spherical symmetrisation is well justified in this case , with the largest absolute deviation being @xmath146ev , corresponding to a relative error of around 0.09 . the reason for this good agreement is that in the present case the wannier functions are already very close to atomic - like orbitals . we also checked that crpa yields a significantly larger value of @xmath111 for iron - oxide ( feo ) , as expected physically . ( color online ) total dos for laofeas , @xmath128 hamiltonian . black line : lda dos . red line : lda+dmft dos . ] ( color online ) real ( top ) and imaginary ( bottom ) part of the orbital dependent self - energy in the @xmath128 hamiltonian for @xmath143 , @xmath144 . ] ( color online ) dos of fe orbitals in laofeas , @xmath128 hamiltonian . color coding as in fig . [ fig : laofeas_dpp_tot ] . ] we carried out lda+dmft calculations for the @xmath128-hamiltonian using the above matrices at an inverse temperature @xmath147ev@xmath99 ( room temperature @xmath148k ) , using the experimental crystal structure of lafeaso . in fig . [ fig : laofeas_dpp_tot ] we display the resulting total densities of states ( dos ) together with the corresponding lda dos . the total densities of states were computed from the lattice green s function , eq . ( [ eq : latt - g ] ) , traced over all @xmath149 and integrated over bz . in order to obtain the corresponding lda densities of states @xmath150 in eq . ( [ eq : latt - g ] ) was set to zero . one sees in fig . [ fig : laofeas_dpp_tot ] that the lda+dmft dos near the fermi level displays characteristic features of a metal in an intermediate range of correlations . both occupied and empty states are shifted towards the fermi level due to the fermi - liquid renormalizations . no high - energy features that would correspond to lower or upper hubbard bands are present in the lda+dmft electronic structure . the fermi - liquid behavior is clear from the self - energy on the real - frequency axis , which we plot in fig . [ fig : sigma ] for the @xmath128-hamiltonian . although it shows a quite rich structure as a function of energy , the real part displays clear linear behavior at low - frequency . the imaginary part is small around @xmath151 and has a quadratic frequency dependence at low frequency . it does increase to rather large values at higher frequencies , however , especially for occupied states . hence , our results are in general agreement with the previous calculations of anisimov _ et al._@xcite and with the experimental photoemission ( pes)@xcite and x - ray absorption ( xas)@xcite spectra of lafeaso , which report a moderately - correlated system with mass renormalisation around @xmath152 . in order to analyze the strength of correlations for different fe 3@xmath0 orbitals we calculated the corresponding quasiparticle residues @xmath153\right]^{-1}$ ] from the self - energy eq . ( [ sigma_mm ] ) on the matsubara grid ( hence , avoiding all uncertainties related to the analytical continuation ) . the values are 0.609 , 0.663 , 0.609 , and 0.596 for the @xmath138 , @xmath139 , @xmath140 , and degenerate @xmath141/@xmath142 orbitals , resp . in this @xmath128 energy window the wannier functions become quite localized and their spread is expected to be isotropic . indeed , within the @xmath128-hamiltonian the difference in @xmath154 between the orbitals is rather small . the resulting value for the average mass renormalization ( between 1.5 and 1.7 ) is in reasonable agreement with the experimental estimate of 1.8 extracted in ref . from experimental pes the smaller mass renormalization found in our calculation compared to the experimental value can be attributed to the single - site approximation of dmft . spatial spin fluctuations , which are completely neglected in this approach , can eventually increase the effective mass of the quasiparticles . the partial densities of states for all fe 3@xmath0 orbitals computed within the @xmath128-model are displayed in fig . [ fig : laofeas_dpp_part ] . the partial lda+dmft dos for the @xmath155 and @xmath156 orbitals are shifted upwards relative to the @xmath157 . indeed , we found that the crystal field ( cf ) splitting between the fe 3@xmath0 orbitals is somewhat affected by correlations . the splitting between the lowest @xmath157 and highest @xmath158 orbitals remains unchanged ( @xmath159ev ) , while the @xmath155 and @xmath156 cf levels are shifted upwards by 0.15 and 0.08ev relative to their positions in lda . in lda+dmft they are located at 0.25 and 0.18ev , respectively , above the @xmath157 orbital . momentum - resolved spectral function of laofeas , @xmath128 hamiltonian . dark areas mark large spectral weight . ] comparison of the momentum - resolved spectral function of laofeas at low energies . upper panel : lda . lower panel : lda+dmft @xmath128 hamiltonian.,title="fig : " ] + comparison of the momentum - resolved spectral function of laofeas at low energies . upper panel : lda . lower panel : lda+dmft @xmath128 hamiltonian.,title="fig : " ] it is also instructive to look at the momentum - resolved spectral function @xmath160 of the crystal . it is obtained from the lattice green s function , eq . ( [ eq : latt - g ] ) , using the real - frequency self - energy and tracing over the orbital degrees of freedom . the result for the @xmath128-model is shown in fig . [ fig : laofeas_dpp_akw ] for an energy range including fe-@xmath0 , as-@xmath2 , and o-@xmath2 states . in agreement with the fermi - liquid picture of moderately - correlated quasiparticles discussed above , one can see well - defined excitations around the fermi level , which get more diffuse at higher binding energies . the bands above the fermi level are less affected , since the self - energies are quite asymmetric and smaller for positive frequencies , see fig . [ fig : sigma ] . additionally , it is easy to see that the as-@xmath2 states , dominantly in the energy range @xmath161$]ev , hybridize stronger with the fe-@xmath0 states and get , thus , affected by correlations . this effect is almost absent for the o-@xmath2 states , since they hybridize much less with fe-@xmath0 . in fig . [ fig : laofeas_dpp_akwzoom ] we show a comparison between the lda band structure and the lda+dmft @xmath162-resolved electronic structure in a low - energy range around the fermi level . this again reveals the coherent quasiparticles at the fermi level , as well as more diffuse bands at higher energies . the crossover between long - lived quasiparticles and more diffuse states with a shorter lifetime is around @xmath163ev , in qualitative agreement with existing arpes data@xcite . a point to mention here is the effect of the cf splitting on the band structure . for example , a difference between the lda and dmft results can be seen for the excitation with predominantly @xmath157 character . in lda it forms a hole - pocket with an excitation energy of @xmath164ev at the @xmath165 point . due to correlations , however , this band is shifted down significantly to the fermi level , and the third hole pocket stemming from the @xmath140 orbital could eventually vanish upon electron doping . one has to keep in mind that a direct comparison to experimental data is difficult for this compound , since ( i ) the experiments where done at low temperatures in the sdw phase , whereas our calculations are done at room - temperature using the tetragonal crystal structure , and ( ii ) arpes experiments on the @xmath3 family of pnictide superconductors are difficult to perform because of difficulties with single - crystal synthesis . nevertheless , on a qualitative level , there is a satisfactory agreement between lda+dmft and experiments . .quasiparticle weights for different interaction parameters , with wannier orbitals constructed from @xmath133$]ev ( @xmath128 hamiltonian ) . the values in boldface correspond to the interaction parameters obtained from crpa . [ cols="^,^,^,^,^",options="header " , ] [ tabl : z2 ] we also studied the dependence of the results on the values of the interaction parameters @xmath111 and @xmath114 . the resulting quasi - particle renormalizations @xmath166 are listed in table [ tabl : z ] . comparing the first two rows , one can see that a smaller value of @xmath114 decreases the degree of correlations . the third line correspond to values similar to the ones used in ref . , giving very similar results . we also increased @xmath111 to the ( unphysically ) large value of @xmath167ev , and the system still displays metallic behavior , although more correlated . hence , our calculations strongly suggest that lafeaso is not close to a mott metal - insulator transition . in order to check the robustness of our results , we also investigated the effect of increasing even further the spatial localization of the wannier functions , corresponding to a very large energy window @xmath168$]ev . we did several calculations for different parameter sets , and the resulting quasi - particle renormalizations @xmath166 of all these calculations are listed in table [ tabl : z2 ] . for this case , no crpa calculations for the interaction matrices were performed , but it is expected that @xmath111 and @xmath114 will slightly increase with more localised wannier orbitals . in that sense , the first row of table [ tabl : z2 ] corresponds to interaction parameters that could be realised for these wannier functions . it is very satisfying to see the calculations gave almost identical quasi - particle renormalizations . also the dependence on @xmath111 and @xmath114 is very similar to the one we found for the @xmath128 hamiltonian . in that sense we consider our calculations to be converged in terms of the number of bloch bands that are included for the construction of the wannier functions and the local hamiltonian . in this section , we address the lda+dmft calculations performed with the so - called @xmath0-hamiltonian , where only the @xmath135 fe-@xmath0 bands around the fermi level are used for the construction of the wannier orbitals . in doing so , we shall shed light on the discussion which has appeared in the literature@xcite regarding the results of lda+dmft calculations by different authors , and the degree of correlations of the @xmath3 family of pnictide superconductors . ( color online ) impurity spectral function within the @xmath0 hamiltonian using interaction parameters @xmath169ev and @xmath170ev as have been used in ref . . for those values a very correlated metal is obtained . ] the first thing to note is that the wannier functions constructed from a small energy window encompassing only the fe-@xmath0 bands are quite extended and very anisotropic , as discussed in details in ref . . this is directly reflected in the interaction matrices calculated by crpa in this restricted energy window : @xmath171 which display a strong orbital dependence . for instance , the intraorbital ( hubbard ) interaction spans from 2.17ev to 3.36ev . the interaction matrices in the spherical symmetric approximation using , the averages @xmath172 and @xmath173 , are @xmath174 the largest deviation in this case is @xmath175ev , corresponding to a relative error of about 0.26 . this shows clearly that the spherical approximation is highly questionable when using only the @xmath0-bands for the wannier construction . of course , the full anisotropic interaction matrices can in principle be used in the lda+dmft calculation , but this raises the very delicate issue of a reliable _ orbital - dependent _ double - counting correction . another consequence of using delocalized wannier functions is that they lead to significant non - local interactions @xmath176 , which we found to be ( from crpa ) of order @xmath177 to @xmath178 . these interactions are completely neglected in the single - site local dmft approach , suggesting the need for a cluster extension in that case . for these various reasons , we have reservations against using a @xmath0-only hamiltonian with extended wannier functions for dmft calculations on lafeaso , as also previously emphasized ref . . nevertheless , in order to clarify apparent discrepancies between previously published lda+dmft results@xcite , we performed calculations within the @xmath0-hamiltonian , for several interaction parameters reported in the literature . for the values @xmath169ev and @xmath170ev used in ref . , we do confirm that the results then display very strong correlations with quasiparticle renormalizations ranging from @xmath179 ( @xmath157 orbital ) to @xmath180 ( @xmath155 orbital ) . one may note that within the @xmath0-model there is a substantial orbital dependence of @xmath154 , with a stronger renormalization predicted for the @xmath157 , @xmath181 and @xmath182 orbitals . this is a clear consequence of the wannier functions being much more delocalized and anisotropic . the scattering rate at this inverse temperature of @xmath147ev@xmath99 is quite sizable ( @xmath183 , depending on the orbital ) , showing that the system is on the verge of a coherence - incoherence crossover and a bad metal . the impurity spectral function is plotted in fig . [ fig : laofeas_d ] . it resembles very much the one shown in fig . 3 of ref . , showing clear signatures of lower and upper hubbard bands . there are , though , some discrepancies with the total weight and the positions of the hubbard bands , but given the differences in the calculation ( underlying electronic structure method , temperature , interaction vertex which here is only density - density ) , this agreement with ref . is quite satisfactory . furthermore , using the parameters @xmath184ev and @xmath185ev from ref . , we find renormalizations in the range @xmath186 . this is somewhat smaller than reported in ref . , although not in drastic disagreement . finally , we investigated the dependence on the hund s rule coupling of calculations performed with the @xmath0-only hamiltonian . decreasing @xmath114 to the much lower value @xmath187ev but keeping @xmath188ev , we find the system to be much less correlated ( @xmath189 between 0.63 and 0.73 ) . we thus confirm , for those calculations , the great sensitivity to the hund s coupling reported in ref . . we note however that , although reducing @xmath114 does make the system somewhat less correlated in this case too , this sensitivity is much weaker when calculations are performed with the full @xmath128 hamiltonian , as reported above . in order to understand the origin of the remarkable sensitivity of the correlation strength to the value of @xmath114 observed with the @xmath0-hamiltonian we have studied the evolution of the ground state of the fe 3@xmath0 `` atomic shell '' as function of @xmath114 . we obtained the 3@xmath0 level positions corresponding to two different choices of the energy window : the `` small '' one corresponding to the @xmath0-hamiltonian and comprising 10 fe 3@xmath0 bands and the `` very large '' one comprising all as 4@xmath2 , o 2@xmath2 and fe 3@xmath0 bands as well as all unoccupied bands up to 13 ev above @xmath190 . the non - interacting level positions @xmath191 are then obtained as @xmath192 where the double counting term @xmath193 is calculated in accordance with eq . ( [ eq : dc ] ) but with the `` atomic '' occupancy @xmath194 of the fe 3@xmath0 shell . we used the same values of @xmath172 and 2.69ev for the `` small '' and `` very large '' window choices , respectively , while the value of @xmath114 was varied from 0.1 to 0.5ev . with @xmath191 corresponding to the @xmath0-hamiltonian we observed a level crossing at @xmath195 ev with the atomic ground state changing from the one with spin moment @xmath196 to the one with @xmath197 . in the case of the `` very large '' window the ground state always corresponds to @xmath197 , and the splitting between the ground state and first excited level is constant . it is obvious that a drastically different behavior of those two `` atomic '' models is related to the corresponding level positions @xmath198 , which are computed using different choices for the wannier orbitals . the observed change of the fe 3@xmath0 atomic ground state , induced by increasing @xmath114 , hints on a possible strong dependence of correlation strength on the hund s rule coupling for lafeaso , which is indeed observed in our lda+dmft calculations with the @xmath0-hamiltonian . however , this sensitivity stems from a particular choice of delocalized and anisotropic wannier functions and is much less pronounced when the energy window for the wannier function construction is increased . the bottom - line of this investigation is that all previously published calculations seem to be technically correct . however , as discussed above , one introduces several severe approximations when dealing with the @xmath0-hamiltonian only , and the justification of these approximations ( restriction to local interactions , single - site dmft , etc ... ) is questionable . this is especially true in this compound , due to the strong covalency between iron and arsenic states . in the first part of this work , we present an implementation of lda+dmft in the framework of the full - potential linearized augmented plane waves method . we formulate the dmft local impurity problem in the basis of wannier orbitals , while the full lattice green s function is written in the basis of bloch eigenstates of the kohn - sham problem . in order to construct the wannier orbitals for a given correlated shell we choose a set of local orbitals , which are then expanded onto the ks eigenstates lying within a certain energy window . in practice , we employ the radial solutions of the schrdinger equation for a given shell evaluated at the corresponding linearization energy as local orbitals . by orthonormalizing the obtained set of basis functions we construct a set of true wannier orbitals as well as projector operator matrices relating the bloch and wannier basis sets . we derive explicit formulas for the projected operator matrices in a general flapw framework , which may include different types of augmented plain waves , @xmath62 and @xmath67 orbitals . our new implementation is benchmarked using the test case of srvo@xmath4 , for which we have obtained spectral and electronic properties in very good agreement with results of previous lda+dmft calculations . in the second part of this paper we apply this lda+dmft technique to lafeaso in order to assess the degree of electronic correlations in this compound and clarify the ongoing controversy about this issue in the literature . we solved the dmft quantum impurity problem using a continuous - time quantum monte carlo approach . the wannier functions are constructed using an energy window comprising fe 3@xmath0 , as 4@xmath2 and o 2@xmath2 . the resulting wannier orbitals are rather well localized and isotropic . we take the average values of @xmath1992.69 ev and @xmath2000.79 from constrained rpa calculations , where the wannier functions and screening channels are consistent with our setting of the lda+dmft scheme . we have checked the robustness of these results by increasing the size of the energy window , what resulted in a very similar physical picture . our lda+dmft results indicate that lafeaso is a moderately correlated metal with an average value for the mass renormalization of the fe 3@xmath0 bands about 1.6 . this value is in reasonable agreement with estimates from photoemission experiments . we also consider a smaller energy window that includes fe-@xmath0 states only . the resulting wannier functions in this case are quite extended , leading to anisotropic and non - local coulomb interactions . we take different values for @xmath111 and @xmath114 , including the ones used in previous theoretical lda+dmft approaches . we demonstrate that different physical pictures ranging from a strongly correlated compound on the verge of the metal insulator transition to a moderately to weakly correlated one can emerge depending , in particular , on the choice of the hund s rule coupling @xmath114 as observed in ref . . however , there are conceptual difficulties when constructing a local hamiltonian from rather delocalized wannier orbitals . the interactions are very anisotropic and orbital dependent , and non - local interactions could also become important . in summary , we demonstrate that the discrepancies in the results of several recent theoretical works employing the lda+dmft approach stem from two main causes : i ) the choice of parameters of the local coulomb interaction on the fe 3@xmath0 shell and ii ) the degree of localization of the wannier orbitals chosen to represent the fe 3@xmath0 states , to which many - body terms are applied . regarding the first point , the calculated interaction parameters employed in the present work are significantly smaller than the values hypothesized in refs . [ ] . regarding the second point , we provide strong evidence that the dmft approximation is more accurate and more straightforward to implement when well - localized orbitals are constructed from a large energy window encompassing fe-3@xmath0 , as-4@xmath2 and o-2@xmath2 . this issue has fundamental implications for many - body calculations , such as dmft , in a realistic setting . for benchmarking purposes , we present in this appendix lda+dmft results for an oxide that has become a classical test compound for correlated electronic structure calculations , namely the cubic perovskite 3 . as a paramagnetic correlated metal with intermediate electron - electron interactions , it is in a regime that is neither well described by pure lda calculations nor by approaches such as lda+u that are geared at ordered insulating materials . from the experimental side , 3 has been characterized by different techniques ( angle - resolved and angle - integrated photoemission spectroscopy , optics , transport , thermodynamical measurements etc)@xcite . lda+dmft calculations have been performed both for an effective low energy model that comprises the three degenerate bands of mainly 2 g character that are located around the fermi energy taking advantage of the cubic crystal field that singles out this group of bands and for a bigger energy window comprising also the oxygen @xmath2-states . @xcite . in the low - energy effective 2 g model a quasi - particle renormalization of @xmath201 , compatible with experiments , is obtained for @xmath111 values around 4ev . the remaining spectral weight is shifted toward lower and upper hubbard bands . the lower hubbard band , located around @xmath202ev binding energy has indeed been observed in photoemission ; the high energy satellite of the 2 g model is located around 2.5ev . concerning calculations taking into account also the ligand states , it should be noted that possible lda errors on the separation of @xmath2- and @xmath0-states are not corrected by dmft , since only the @xmath0-states are treated as correlated . ( colore online ) dos for 3 , @xmath0-only model ( small energy window ) . top panel ( a ) : total dos of lda ( black ) and lda+dmft ( red ) . bottom panel ( b ) : lda local orbitals ( black ) , impurity spectral function @xmath203 ( red ) , and vanadium 2 g partial dos ( green ) . coulomb parameters for these calculations are given as inset . ] ( color online ) total dos for 3 , @xmath204 hamiltonian ( vanadium 2 g and oxygen @xmath2 ) . ] in the present work , we use 3 as a benchmark for our projector orbitals implementation of lda+dmft , with results very similar to previous theoretical studies . we performed two kinds of calculations : ( i ) we used as an energy window the range from @xmath205ev to 1.90ev which comprises the 2 g bands , and at some k - points one or both of the bands . this is closest in spirit to a 2 g model within a wannier function formalism , though not exactly the same due to the inclusion of some contribution . to recover a wannier prescription one would in fact have to choose a k - dependent window , such as to include exactly three bands _ at each k - point _ , corresponding to the three - fold degenerate manifold of dominantly 2 g bands . ( ii ) we used an energy window of @xmath206ev to 1.90ev , spanning both , the bands used in ( i ) and the oxygen @xmath2 dominated bands located between @xmath207 and @xmath208ev . please note that , in order to be consistent with existing literature , we use a different parametrisation of the interaction matrix compared to sect . [ sect : results ] . here we define @xmath111 to be the onsite intraorbital coulomb interaction , @xmath209 to be the inter - orbital interaction for electrons with opposite spin , and @xmath210 the inter - orbital interaction between electrons with equal spin . the results for the first case are shown in fig . [ fig : srvo3d ] . the upper panel , fig . [ fig : srvo3d ] ( a ) , displays the total spectral function of the thus defined model within lda and lda+dmft . our results recover previously published results , with a quasiparticle renormalisation of around @xmath211 for a values @xmath169ev and @xmath212ev . the contribution of the bands to the total dos can easily be identified from the lda+dmft spectra , where an additional hump between the quasiparticle peak and the upper hubbard band appears . [ fig : srvo3d ] ( b ) shows the local orbitals used for the dmft calculations , together with the corresponding impurity spectral function @xmath203 and the vanadium 2 g partial dos . the latter one is obtained bu projecting the lattice green s function to 2 g character using the partial projectors to be introduced in app . [ sect : pc_proj ] . the main difference to panel ( a ) is the absence of the additional character . finally , fig . [ fig : srvo3dp ] shows the lda+dmft spectral function compared to the lda density of states , as calculated within the larger energy defined in ( ii ) above . since the wannier function are more localised , as compared to case ( i ) , the value for the coulomb interactions has to be adjusted accordingly , and we chose a value of @xmath213ev . as can be seen in the figure , ligand states are barely modified by the correlations , and the results for the 2g - derived bands are very close to what is seen in the effective low energy model . the quasiparticle renormalisation is @xmath214 , in good agreement with the pure 2 g treatment discussed before . the results of these calculations correspond to what can be expected on the basis of previously published work , and thus validate our new implementation . [ sect : srvo3 ] in order to calculate the partial density of states for a given atomic site and particular orbital character ( correlated or not ) we construct a different type of projectors , which we call @xmath215 . the wannier operators of eq . [ eq : wannier - proj ] project onto a given wannier - like orbital . on the other hand , the new set @xmath215 , as we will show , project onto a given orbital of certain character for which we do not apply any orthonormalization process as in the first . unlike the wannier projectors , the @xmath215 s can also project to other orbitals atoms apart from the correlated set . a given orbital character contributes in the eigenstates through the solutions of the schrdinger equation inside the sphere @xmath216 , @xmath217 and @xmath218 which do not form an orthonormalized basis set . it is more convenient to construct these projectors if the wave function is rewritten in an orthonormal basis set . in a general form , inside a given sphere we can express @xmath53 as @xmath219 where we simplify the notation by omitting the angular and spin parts and defining @xmath220 , @xmath221 and @xmath222 as combined coefficients which are generally k - dependent and contain the sum over the plane waves and local orbitals . we also define @xmath223@xmath224 , @xmath225 and @xmath226 . we then rewrite @xmath53 as a function of a set of orthogonal orbitals @xmath227 , _ _ j__=1,2,3 as follows @xmath228 the coefficients @xmath229 are the matrix elements of the square root of the corresponding overlap matrix @xmath230 in this way , rewriting eq . [ eq : wave - func2 ] as @xmath231 the matrix elements of the projector to a given atom with _ character finally reads , @xmath232 the spectral function of a given atom _ i _ with orbital character _ m _ , is obtained as @xmath233\ ] ] quasi - particle renormalisation @xmath189 in a 3-orbital hubbard model . calculations have been done using a semicircular dos with bandwidth @xmath234 . top panel : @xmath235 ( half filling ) . bottom panel : @xmath236 . ] in our dmft calculations using ctqmc as impurity solver , we restricted the hund s rule interaction to ising - type interactions only , although there is no conceptual limitation of the algorithm to this type of interactions . the reason for doing this is of purely technical nature , since in this case one can diagonalize the local problem very efficiently , and furthermore , it enables us to use the so - called segment - picture update scheme,@xcite which increases the efficiency of the ctqmc method a lot . one may now ask how results change if the fully rotational - invariant hund s rule exchange is taken into account . for this purpose , we study a multiband model hamiltonian , assuming degenerate bands , no interband hybridizations , and a semicircular density of states . applying the self - energy functional theory ( sft),@xcite we can study the quasiparticle renormalization @xmath189 as function of interactions @xmath111 and @xmath114 . in this study , we choose the convention of setting the intraorbital coulomb repulsion to @xmath111 and the interorbital to @xmath237 , and give all energies in units of the single - particle hopping amplitude @xmath238 , i.e. , the band with of the dos is @xmath234 . in addition to the density - density interactions , we consider also the additional spin - flip and pair - hopping terms of the local hamiltonian , @xmath239 we do calculations at @xmath240 , and choose the reference system for the sft framework to consist of one bath degree of freedom for each correlated orbital . hence , going up to @xmath241 orbitals , we have to diagonalize a local problem consisting of at most 10 orbitals . the upper panel of fig . [ figz ] shows @xmath189 for a 3-orbital model at half - filling , @xmath235 , for @xmath242 . a tremendous reduction of the critical @xmath243 of the metal - to - insulator transition ( mit ) is observed , already for ising - like interactions . this is a well known fact that for multi - orbital systems at or close to half - filling , the effect of @xmath114 should be strongest.@xcite the inclusion of spin - flip and pair - hopping terms gives raise to two effects . ( i ) for moderate correlations , @xmath244 , these terms lead to a slight reduction of @xmath189 , but ( ii ) the critical @xmath111 for the mit is shifted upwards . this qualitatively holds also away from half - filling , which can be seen in the lower panel of fig . [ figz ] , where we plotted @xmath189 for @xmath236 . although the transition is not of first order any more , one can again identify two regimes . for moderate correlations , @xmath189 _ decreases _ , whereas close to the transition the spin - flip and pair - hopping terms _ increase _ the renormalization @xmath189 and the critical @xmath243 is pushed to higher values . this is consistent with a numerical renormalization group study for the 2-orbital hubbard model.@xcite we also considered the case relevant for pnicitde materials , i.e. , @xmath241 , @xmath245 , and @xmath189 around 0.5 . this regime could be realize by setting ( i ) @xmath246 and @xmath247 , which shows a reduction of @xmath189 from 0.52 to 0.47 due to spin - flip and pair - hopping , or ( ii ) @xmath248 , @xmath249 , giving a reduction from 0.61 to 0.57 . in conclusion , this analysis shows that the picture of a moderately - correlated metal as argued in sect . [ sect : results ] holds also when a fully rotational - invariant hund s exchange is considered . for systems close to a mit , this is no longer true and the spin - flip and pair - hopping terms become crucial . we are grateful to vladimir anisimov , ryotaro arita , gabriel kotliar , igor mazin , frank lechermann , alexander lichtenstein and , especially , kristjan haule for useful discussions and correspondence . is grateful to f. assaad for enlightening discussions on the stochastic maximum entropy method . we acknowledge the support of the agence nationale de la recherche ( under project correlmat ) , and of genci and idris ( under project 091393 ) for supercomputer time . m.a . acknowledges financial support from the austrian science fund ( fwf ) , grant j2760-n16 .

the similarity between observed velocity structures of al iii and singly ionized species in damped ly systems ( dlas ) suggests the presence of ionized gas in the regions where most metal absorption lines are formed . a simplified model consisting of _ region 1 ) _ a plane - parallel ionization bounded region illuminated by an internal radiation field , and _ region 2 ) _ a neutral region with a negligible metal content is considered . we calculate photoionization equilibrium models for region 1 , and constrain the ionization parameter by the observed(al iii)/(si ii ) column density ratio . under these conditions we find that ionization effects are important . if these effects are taken into account , the element abundance ratios in dlas are quite consistent with those observed in milky way stars and in h ii regions of local low - metallicity blue compact dwarf galaxies . in particular we can not exclude the same primary n origin in both dlas and metal - poor galaxies . no depletion of heavy elements on dust grains needs to be invoked , although our models do not exclude the presence of little depletion . although highly simplified and relying on the strong assumption of a significantly lower metal content in region 2 , our model appears to be supported by recent data on a local dla and it is not in contradiction with the current knowledge on high redshift dlas . if correct , it offers a clear simplification in the understanding of heavy element abundance ratios in dlas and their comparison with the local universe . = cmr10 scaled 1000 _ accepted for publication in apj ( scheduled in march 10 , 2001 , issue ) _

astro-ph0010643

damped lyman @xmath0 systems ( dlas ) are the class of qso absorption systems showing the largest column density , with @xmath1(h i ) @xmath2 2@xmath310@xmath4 @xmath5 . they are considered to be the high - redshift progenitors of present - day luminous galaxies ( e.g. wolfe 1990 ) and dominate the neutral gas content of the universe ( wolfe et al . 1995 ) , a mass comparable with that of the stars in present day spiral galaxies . this may suggest that dlas are the source of most of material available for star formation at high redshift ( lanzetta et al . later studies have shown that probable optical counterparts of damped ly@xmath0 systems at @xmath6 are galaxies of different morphological types ( le brun et al . 1997 ) . understanding the chemical evolutionary history of galaxies is fundamental to the study of galaxy formation . pettini et al.(1994 , 1997 ) have performed extensive studies on the metallicity of the damped ly@xmath0 systems at high redshift . their results indicate a mean metallicity zn / h @xmath7 1/10 of the solar abundance with a large dispersion . it was also noted that the element cr is generally less abundant than zn relative to their solar proportion . since zn and cr are thought to be made in solar proportions throughout the milky way based on their abundances in galactic disk and halo stars , the overabundance of zn to cr found in damped ly@xmath0 systems was taken as evidence that some cr was depleted from the gas phase by dust grains . the hires w.m.keck telescope observations have opened the opportunity to study ly@xmath0 absorption lines for a large number of elements including n , o , mg , al , si , s , cr , mn , fe , ni and zn . lu et al . ( 1996a ) , prochaska & wolfe ( 1999 ) , outram et al . ( 1999 ) and pettini et al . ( 2000 ) have found that the relative abundance patterns indicate that the bulk of heavy elements in these high - redshift galaxies was produced by type ii supernovae ; there is little evidence for significant contribution from low- to intermediate - mass stars or from type ia supernovae . however , the current interpretation combining the effects of dust depletion and type ii supernovae enrichment ( lu et al . 1996a ; prochaska & wolfe 1999 ; vladilo 1998 ) leads to several inconsistencies . in this picture the overabundance of zn relative to cr or fe found in dlas is attributed to selective depletion of cr and fe by dust grains . however , it then turns out that such an interpretation is inconsistent with some of the other elemental abundance ratios seen in these galaxies , in particular n / o , s / fe , mn / fe and ti / fe . this led lu et al . ( 1996a ) and prochaska & wolfe ( 1999 ) to conclude that the overabundance of zn relative to cr in damped ly@xmath0 galaxies may be intrinsic to their stellar nucleosynthesis . they emphasize the need for another physical process to explain the damped ly@xmath0 abundance patterns . furthermore , comparisons of the distribution of fe / h versus redshift ( lu et al . 1996a ; prochaska & wolfe 2000 ) for the sample of damped ly@xmath0 galaxies with the corresponding relation for the milky way disk indicate that the damped ly@xmath0 galaxies are much less metal - enriched than the galactic disk in its past . since there is evidence that depletion of fe by dust grains is relatively unimportant , the difference in the enrichment level between the sample of damped ly@xmath0 galaxies and milky way disk suggests that dlas are probably not high - redshift spiral disks . a similar conclusion has been reached by pettini et al . they argue that the relative abundances of heavy elements in dla systems are consistent with a moderate degree of dust depletion that , once accounted for , leaves no room for the enhancement of @xmath0 elements over iron seen in metal - poor stars in the milky way . recently centurin et al . ( 2000 ) have found the s / zn abundance ratio in the relatively high - metallicity dlas to be lower than solar ratio . this is contrary to previous assertions that dlas have been enriched solely by type ii supernovae , but it can be understood if star formation in these systems proceeded at a lower rate than in the early history of our galaxy ( pettini et al . 1999 ; centurin et al . 2000 ) . it is generally believed that the gas in dlas is mostly in the neutral stage due to the very high optical depth beyond the hydrogen ionization limit ( e.g. , viegas 1994 ; prochaska & wolfe 1996 ) . therefore , it is assumed that all species with ionization potentials lower than hydrogen are in a singly ionized stage , while other species are neutral . with this assumption correction factors for ionization are small and the ratios of column densities of ions are equal to the ratios of element abundances . however , lu et al . ( 1996a ) , prochaska & wolfe ( 1999 ) and wolfe & prochaska ( 2000 ) find a good correlation between the velocity structure of al iii and singly ionized species . the ionization potential of al@xmath8 is 18.8 ev , i.e. greater than that of hydrogen . therefore , al@xmath9 is likely present in ionized , not in neutral gas . to explain the similarity of al iii and other low ionization species line profiles , howk & sembach ( 1999 ) and izotov & thuan ( 1999 ) proposed that these lines originate in the same ionized region or in a mix of neutral and ionized clouds , and stressed the importance of abundance correction for ionization effects . indirect arguments to consider ionized gas in dlas come from studies of the warm ionized medium in the milky way and other nearby galaxies ( howk & sembach 1999 ; howk , savage & fabian 1999 ; sembach et al . 2000 ; jenkins et al . 2000 ) . higher ionization species of c iv and si iv are also observed in the spectra of dlas . however , in the majority of cases , the c iv and si iv profiles differ from those of lower ionization species ; few cases exist where some components of the c iv and si iv absorption lines coincide with e.g. si ii , al ii and al iii ( see e.g. , lu et al . 1996a ; prochaska & wolfe 1999 ; wolfe & prochaska 2000 ) . the high ionization species are thus apparently formed in regions with different conditions and will not be considered here . in rare cases absorption lines from neutral species with ionization potentials below that of hydrogen , such as c i and mg i , are observed in dlas . the existence of such lines is a priori not in contradiction with the presence of al iii , since neutral species can also be present in small quantities in ionized gas ( cf . sect . 3 ) . motivated by the close similarity between the velocity structures of al iii and singly ionized species in dlas , which suggests the presence of ionized gas in the regions where these absorption lines are formed , we here consider a simple model that takes these effects to first order into account . we examine a model of dlas in which absorption lines of neutral and low - ionization heavy element species originate in ionized rather than in neutral gas , and we explore the possible implications on abundance determinations in such systems . as indicated by the presence of high ionization species and the complex structure of the absorption line profiles seen in the high resolution spectra , the conditions of the absorbing gas must in reality be highly variable and of complex nature ( see e.g. lauroesch et al . 1996 ) . in the future , both more advanced analysis of the detailed line profiles and sub - components as well as more sophisticated models allowing for complex geometries , various ism phases etc . will be necessary to obtain a more accurate description of the absorbing systems . it is the hope that the present simplified model reflects the main trends due to ionization effects . the model assumptions are described in section 2 . the results from the model calculations and the implications on heavy element abundance ratios in dlas are discussed in section 3 . in section 4 we briefly consider the application of our model to other objects and summarise the main results from this work . the above discussion shows that the properties of the damped ly@xmath0 systems are yet poorly understood . while a significant amount of neutral gas is observed in these systems , it is not clear how the regions of ionized gas are related to those of neutral gas . what is the main source of ionization in dlas - internal stellar radiation or harder background intergalactic radiation ? where are absorption lines of heavy elements formed - in ionized or neutral gas ? is the metallicity of ionized and neutral gas similar or not ? is heavy element depletion by dust grains important ? in the present paper we explore the following `` multi - component '' picture for dla systems : _ 1 ) _ an ionization bounded region illuminated by an internal radiation field complemented by _ 2 ) _ a neutral region with a lower metal content . these working hypotheses are adopted to study the main effects intrinsic stellar ionizing radiation may have on observed heavy element patterns in such conditions . to explore the consequences of such a models we use the cloudy ionization equilibrium code ( version c90.05 ; ferland 1996 ; ferland et al . 1998 ) to calculate the ionization state of region 1 . for simplicity , we assume internal stellar radiation and consider a plane - parallel geometry . the outer boundary of the slab is determined by the condition that a low electron temperature of 2000 k is reached . for the cases discussed here ( low metallicity and low dust content ) this boundary condition is a good approximation for the outer boundary of ionization - bounded h ii regions . for the ionizing spectrum , we adopt the spectrum of a typical o star . we use an atlas line - blanketed model atmosphere ( kurucz 1991 ) with effective temperatures @xmath10 = 33000 k , 40000 k and 45000k and log @xmath11 = 4.0 . the lower temperature is likely more appropriate for the conditions in the interstellar medium of metal - rich spiral galaxies and the local ism in our galaxy ( cf . sembach et al . 2000 ) . in low - metallicity environments the temperature of the ionizing radiation tend to be higher as evidenced from studies of nearby extragalactic giant h ii regions ( e.g. stasiska & leitherer 1996 ) . the ionization parameter which is proportional to the ratio of the number density of the ionizing photons to that of the electrons is varied in the range @xmath12 = 10@xmath13 10@xmath14 . we consider the metal abundance of 0.1 solar , with relative element abundances equivalent to those observed in the low - metallicity bcds ( e.g. , izotov & thuan 1999 ) . in particular , c / o and n / o abundance ratios are reduced by factors of 2 and 5 respectively as compared to solar abundance ratios . we also adopt for @xmath15he the mass fraction @xmath16 = 0.252 ( or log he / h = 1.07 ) , which is higher than the primordial mass fraction @xmath17 = 0.245 ( izotov et al . 1999 ) , but is typical for the bcds with the heavy element abundance 1/10 solar ( izotov & thuan 1998 ) . dust grains with a dust - to - gas ratio of 1/10 that in milky way are taken into account in heating and cooling processes , although the differences in ionic fractions derived in the models with and without dust grains are small . a similar model has recently been considered by howk & sembach ( 1999 ) . the major difference with their approach is that they consider ionized density - bounded regions surrounding the neutral , h i bearing clouds . the boundary of the ionized region is defined where the local fraction of neutral hydrogen climbs above 10% . with these assumptions the fraction of neutral species like o@xmath18 , n@xmath18 or ar@xmath18 is significantly reduced because they always reside near the edges of h ii region . additionally , the variation of the local ionization fraction of neutral hydrogen at the edge of ionized density - bounded region results in large variations of some low - ionization species , in particular , of cr@xmath8 and fe@xmath8 . further resulting differences between the model by howk & sembach ( 1999 ) and our ionization - bounded model will be discussed below . an important feature of both models is that the heavy element abundances in the ionized and neutral regions are a priori unrelated . this implies in particular that relative abundance determinations with respect to hydrogen , as derived from the observed column densities , may not be meaningful or are at least uncertain . in table 1 we show the ionization potentials and the ionization fractions @xmath19(x@xmath20 ) is the ratio of the radially integrated number density of ion x@xmath20 to the radially integrated total number density of the element x in all ionization stages , i.e. @xmath21 , the relative column density of the ion x@xmath20 . ] of the most important ions which are observed or can potentially be observed in dlas , as a function of the ionization parameter @xmath12 for two temperatures @xmath10 = 33000 k and 40000 k of ionizing radiation and the heavy element mass fraction @xmath22 = 0.1@xmath23 . on the first line we also show @xmath1(h@xmath18+h@xmath8 ) , the total hydrogen column density across the ionized slab . over the entire range of @xmath12 , except for the highest value shown here ( @xmath24 ) , @xmath1(h@xmath18+h@xmath8 ) remains significantly smaller than the column density of neutral hydrogen observed in damped ly@xmath0 system = 40000 k and the `` typical '' value of @xmath25 3.25 the column density is @xmath26 40 % of the lower limit which defines damped systems . the baryonic mass of low column density dlas could thus be underestimated by up to this amount . ] . for an ionization parameter @xmath12 @xmath26 10@xmath13 hydrogen is essentially neutral , while at @xmath12 @xmath2 10@xmath27 it is almostly ionized as it is evidenced by small neutral hydrogen fraction @xmath19(h@xmath18 ) ( table 1 ) . calculations for lower heavy element abundances show that the ionization fractions are not strongly changed . therefore , the data from table 1 can be used for objects with metallicity [ fe / h ] @xmath28 1 . the distribution of the ionic fractions @xmath19(x@xmath29 ) of selected ions as a function of ionization parameter @xmath12 is shown in figure 1 . while the ionization fractions of all neutral and singly ionized species decrease with increasing ionization parameter @xmath12 , the ionization fraction of al@xmath9 is increased . therefore , the ratio @xmath19(al@xmath9)/@xmath19(al@xmath8 ) = @xmath1(al iii ) / @xmath1(al ii ) is a very sensitive indicator of ionization conditions and can be used to determine @xmath12 . unfortunately , although line profiles of al ii and al iii in damped ly@xmath0 systems are essentially similar ( lu et al . 1996a ; prochaska & wolfe 1999 ) , in the many cases the al ii lines are saturated or marginally saturated . this circumstance complicates direct estimates of the ionization parameter @xmath12 . however , very often , unsaturated lines of si ii , fe ii and al iii are observed in the same spectra and they can be used to estimate @xmath12 . for this , we assume that the observed heavy element abundance patterns at low metallicities are similar to that observed in low - metallicity halo stars , i.e. [ si / fe ] @xmath26 0.3 , [ al / fe ] @xmath26 0.4 ( timmes et al . 1995 ; samland 1998 ) , and adopt that log ( al / si)@xmath30 @xmath7 log ( al / fe)@xmath30 @xmath7 1.0 ( anders & grevesse 1989 ) . then , the expected ratio of column densities is equal to @xmath31 @xmath32 the typical values of log [ @xmath1(al iii)/@xmath1(si ii ) ] and log [ @xmath1(al iii)/@xmath1(fe ii ) ] in damped ly@xmath0 systems are @xmath26 2.0 to 2.5 and @xmath26 1.5 to 2.0 , respectively , which translates to log [ @xmath19(al@xmath9)/@xmath19(si@xmath8 ) ] @xmath26 0.3 to 0.8 , [ @xmath19(al@xmath9)/@xmath19(fe@xmath8 ) ] @xmath26 0.1 to 0.6 , or log @xmath12 @xmath26 3 ( fig . this estimated value of ionization parameter is dependent on the uncertainties in abundance determination and those for the adopted photoionized model and possible depletion of elements into grains . in principle , other single ionized species , e.g. zn ii , can be used instead of si ii or fe ii . however , si ii and fe ii absorption features are more often seen and measured in the dlas , and the atomic data for these ions are known with better precision ( howk & sembach 1999 ) . the above estimate of @xmath12 is based on the assumption that heavy elements are not present in the neutral gas surrounding the region of ionized gas . if the neutral gas is significantly enriched in heavy elements then an additional contribution to the column densities of si ii and fe ii comes from neutral gas and the ionization parameter should be increased to account for observed @xmath1(al iii)/@xmath1(si ii ) and @xmath1(al iii)/@xmath1(fe ii ) ratios . before we proceed to discuss the implications of this model when applied to observations of damped ly@xmath0 systems it is useful to recall that some neutral species with ionization potentials below that of hydrogen can be present in ionized gas as well . for instance lu et al . ( 1996a ) have measured @xmath33(mg ii ) @xmath34 14.38 and @xmath33(mg i ) = 12.44@xmath350.04 in the @xmath36 = 0.8598 system toward q0454 + 0356 which are consistent with the predictions of our simplified models ( table 1 ) . their measurements for another @xmath36 = 1.2667 system toward q04491326 provide only lower limits @xmath33(mg ii ) @xmath34 14.1 and @xmath33(mg i ) @xmath34 12.62 , also not in contradiction with the ionization model . prochaska & wolfe ( 1999 ) have derived @xmath33(c ii ) @xmath34 14.6 and @xmath33(c i ) @xmath37 12.8 in the @xmath36 = 2.140 toward q0149 + 33 which are consistent with predictions of ionization models . hence , the presence of species like c i or mg i does not necessarily mean that they reside in neutral gas . a useful validity check of our model is the comparison of the predictions of photoionization models with the observed values of [ o i / si ii ] . in the case of the neutral gas model it is expected that [ o i / si ii ] @xmath26 0 , while in the case of our ionization models [ o i / si ii ] = 0.5 - 0.8 depending on @xmath10 and for the range log @xmath12 = 3.5 - 3.0 ( table 1 ) . in the majority of cases observations of dla systems give only lower limits of o i abundance because of saturation of the o i @xmath381302 absorption line . the available data in the literature shows that , in general , the lower limits of [ o i / si ii ] are consistent with values expected from the ionization models . however , in some dlas the lower limits of [ o i / si ii ] are larger than those expected from the ionization models . ( 1996b ) find [ o i / si ii ] @xmath34 0.27 for the @xmath39 = 4.38 absorber towards 12020275 , while prochaska & wolfe ( 1999 ) find [ o i / si ii ] = 0.17 in the absorber towards q095104 . these values seems difficult to reconcile with ionization models unless a low ionization parameter is appropriate . however , no measurements of al iii have been done for these absorbers which would allow to constrain the ionization parameter @xmath12 . recently molaro et al . ( 2000 ) have derived a high value of [ o i / si ii ] = 0.0 - 0.1 from the unsaturated o i and si ii lines in the @xmath39 = 3.39 absorber towards the qso 00002620 . for this absorber @xmath1(al iii ) / @xmath1(si ii ) = 2.4 is rather high , suggesting an ionization parameter @xmath25 3.0 to 3.5 . however , inspection of fig . 1 in prochaska & wolfe ( 1999 ) indicates different profiles of si ii @xmath381808 and al iii @xmath381854 in this absorber , at variance with other absorbers rendering the estimate of the ionization parameter @xmath12 rather uncertain . the above discussion suggests that large ionized regions and abundance discontinuities might be present in some dlas , while other dlas might be essentially neutral . it is likely that dlas may represent a mixed bag of systems with varying ionization conditions . while the ionization model does not pretend to explain abundance pattern in all damped ly@xmath0 systems , we conclude that in many cases the derived lower limits of [ o i / si ii ] do not rule out this model . however , we point out that better determinations of [ o i / h i ] are needed to test the importance of the ionization effects in the damped ly@xmath0 systems . a very extensive literature exists on the interpretation of the metallicity of dla as a function of redshift ( e.g. pettini et al . 1994 ; lu et al . 1996a ; lindner , fritze - von alvensleben & fricke 1999 ; prantzos & boissier 2000 ) . among the main issues discussed are the question of apparent evolution or non - evolution of metallicity with @xmath39 , the origin of the scatter , and possible observational biases affecting this relation . a detailed comparison with such studies is beyond the scope of the present paper . here , we shall only indicate the typical metallicity shift expected in the framework of our model and mention the main implication in terms of age - metallicity relations . pettini et al . ( 1994 , 1999 ) , lu et al . ( 1996a ) , prochaska & wolfe ( 2000 ) noted that the chemical enrichment processes in damped ly@xmath0 galaxies appear to be quite inhomogeneous . the [ zn / h ] and [ fe / h ] distributions show a scatter by a factor of @xmath26 20 with the mean value of [ zn / h ] which is 0.4 larger than that of [ fe / h ] for the dlas at @xmath39 @xmath7 2 ( prochaska & wolfe 2000 ) . furthermore , lu et al . ( 1996a ) and pettini et al . ( 1999 ) noted that the degree of chemical enrichment in the sample of damped ly@xmath0 galaxies is , on average , considerably lower than the milky way disk at any given time in the past . this lead them to conclude that the chemical enrichment of damped ly@xmath0 galaxies is different from that in the disk of milky way . a possible explanation for the difference in chemical enrichment is simply that dlas are in general not necessarily spiral galaxies ( le brun et al . 1997 ) . alternatively , if in the framework of our model the heavy element lines are produced only in the ionized region ( i.e. the neutral gas has a very low metal content ) , the real fe / h abundance in the ionized gas is several times higher than that derived from the @xmath1(fe ii)/@xmath1(h i ) ratio , because of the non negligible upward correction for iron ( table 1 ) . the upward correction for zinc is lower , and interestingly it nearly compensates the difference between [ fe / h ] and [ zn / h ] discussed by prochaska & wolfe ( 2000 ) . then , the age - metallicity relation for damped ly@xmath0 galaxies is more consistent with that for thick disk stars of our galaxy . the opposite situation is of course realised if the neutral gas has the same metallicity as the ionized gas . in this case the observed fe / h ratio is higher than real because in ionized gas @xmath19(fe@xmath8 ) @xmath34 @xmath19(h@xmath18 ) while in neutral gas @xmath19(fe@xmath8 ) = @xmath19(h@xmath18 ) = 1 . as we will show below , it is , however , difficult to adjust all observed heavy element abundance ratios in the frame of the model with similar metallicities of neutral and ionized gas . there is little doubt that some depletion of heavy elements by dust is present in dlas . however , the level of depletion in dlas is generally smaller than that in the disk and possibly in the halo of our galaxy . pettini et al . ( 2000 ) adopt the model of dla consisting only of neutral gas and propose for dlas with low level of depletion ( systems with typically [ zn / fe ] @xmath28 0.3 ) to correct abundances of refractory elements assuming constant depletion factors for all these elements . this approach obviously leaves the question open how to explain observed abundances in systems with more depletion . besides that , the similarity of the line profiles of singly ionized ions and al iii and the large @xmath1(al iii)/@xmath1(al ii ) ratio is not addressed . prochaska & wolfe ( 1999 ) have suggested that if the overabundance of si / fe relative to solar is indicative of dust depletion , then one might expect a correlation between [ si / fe ] and [ zn / fe ] with the most heavily depleted regions showing the largest si / fe and zn / fe ratios . indeed , they found such correlation and concluded that this appearance is consistent with that expected for dust depletion ( their fig . 23 ) . alternatively , as shown in fig . 2 , the regression between [ si / fe ] and [ zn / fe ] can be explained by ionization effects . dashed , solid and dotted lines in figure connect theoretical points [ @xmath19(si@xmath8)/@xmath19(fe@xmath8 ) , @xmath19(zn@xmath8)/@xmath19(fe@xmath8 ) ] in models with @xmath22 = 0.1 @xmath23 calculated for various @xmath12 values and for three values @xmath10 = 33000 k , 40000 k and 45000 k of the effective temperature characterising the hardness of the ionizing radiation . note that [ @xmath19(si@xmath8)/@xmath19(fe@xmath8)]@xmath40 and [ @xmath19(zn@xmath8)/@xmath19(fe@xmath8)]@xmath40 are the correction factors to convert respectively si@xmath8/fe@xmath8 to si / fe and zn@xmath8/fe@xmath8 to zn / fe and they are independent on the intrinsic abundances of si , fe and zn as far as these elements do not contribute significantly to the thermal balance of the ionized gas . however , the correction factors for the considered ions can depend on the helium and the total heavy element abundances which regulate the thermal balance and the ionization structure . as mentioned above , the values of the ionization fractions @xmath19 are essentially metallicity independent for heavy element mass fractions @xmath41 0.1 0.01 @xmath23 . we point out that the quantities commonly ( e.g. prochaska & wolfe 1999 ) denoted as [ si / fe ] and [ zn / fe ] are actually @xmath42 } = \log [ n{\rm ( si~ii)}/n{\rm ( fe~ii ) } ] - \log{\rm ( si / fe)}_\odot\ ] ] and @xmath43 } = \log [ n{\rm ( zn~ii)}/n{\rm ( fe~ii ) } ] - \log{\rm ( zn / fe)}_\odot,\ ] ] respectively . the total abundance ratios are then defined as @xmath42_{cor } } = { \rm [ si / fe ] } - \log[x({\rm si}^+)/x({\rm fe}^+)]\ ] ] and @xmath43_{cor } } = { \rm [ zn / fe ] } - \log[x({\rm zn}^+)/x({\rm fe}^+)].\ ] ] we find that both the observed range of @xmath19(al@xmath9 ) / @xmath19(si@xmath8 ) and [ zn / fe ] are consistent with same ionization parameter for the ionizing radiation with @xmath10 = 33000k and 40000k . figure 2 shows that if one applies the correction for ionization effects for the radiation effective temperature of 40000 k , assuming that after correction [ zn / fe]@xmath44 = 0 , the resulting [ si / fe]@xmath44 value remains on average above zero with a small spread of points ( see also fig . 4 ) , and no additional correction for dust depletion is required . however , models with harder radiation with @xmath10 = 45000 k do not reproduce the observed trend . if our approach is correct then we obtain important constraint on the hardness of the ionizing stellar radiation in dlas from figure 2 . note that with the hard intergalactic background radiation field considered by howk & sembach ( 1999 ) the observed [ zn / fe ] vs [ si / fe ] relation can not be reproduced because of their adopted boundary conditions . in this case ( see their fig . 1 ) the correction for ionization of [ zn / fe ] is positive , while that of [ si / fe ] is negative . even if an ionization bounded model with the hard intergalactic background radiation field is considered , the correction for ionization of [ zn / fe ] , though smaller , remains positive . in our case both corrections are negative . in summary , it is possible to reproduce the observed @xmath1(al iii)/@xmath1(si ii ) ratios and the trend seen in [ si / fe ] vs [ zn / fe ] diagram assuming ionization by a moderately hard stellar ionizing radiation field . in principle , no correction for dust depletion needs to be invoked , although it might be present in small amounts and can easily be taken into account in the ionization model without significant changing the main features of the [ si / fe ] vs [ zn / fe ] distribution . additional support of the ionization model comes from fig . 3 where we show the dependence of [ al iii / fe ii ] vs [ zn ii / fe ii ] . models with @xmath10 = 33000k ( dashed line ) and 40000k ( solid line ) fit quite well the observed points from prochaska & wolfe ( 1999 ) and lu et al . ( 1996a ) at the values of ionization parameter @xmath12 close to that obtained from fig . the above conclusion is broadly consistent with expectations for other heavy element abundance ratios . to illustrate this we apply a `` typical '' ionization correction factor to the observational data . to do so we adopt the characteristic correction factors of the model with gas ionized by the radiation with @xmath10 = 40000 k. we consider the metallicity @xmath22 = 0.1 @xmath23 as representative for the sample of dla systems and adopt log @xmath12 = 3.25 . then , our calculations yield the following mean ionization correction factors @xmath45 $ ] defined by @xmath46_{\rm corrected } = \log[{\rm x / h}]_{\rm observed } + \delta \log[{\rm x / h}]$ ] : @xmath47log[al / fe ] = 0.63 , @xmath47log[si / fe ] = 0.31 , @xmath47log[s / fe ] = 0.32 , @xmath47log[ti / fe ] = 0.12 , @xmath47log[cr / fe ] = + 0.16 , @xmath47log[mn / fe ] = 0.01 , @xmath47log[ni / fe ] = 0.38 , @xmath47log[zn / fe ] = 0.40 . note the small corrections for [ ti / fe ] and [ mn / fe ] which remain almost unchanged while the downward correction of [ zn / fe ] is high . interestingly , it is the same as the difference between mean [ fe / h ] and [ zn / h ] values for dlas at @xmath39 @xmath7 2 from prochaska & wolfe ( 2000 ) . the case of nitrogen is discussed separately below . the above `` typical '' correction factors have been applied to the sample of damped ly@xmath0 galaxies from lu et al . ( 1996a ) , prochaska & wolfe ( 1999 ) and pettini et al . ( 2000 ) . only galaxies with definitely derived column densities are included and we do not consider objects with lower or upper limits . the original ( open symbols ) and corrected ( filled symbols ) abundance ratios are shown in fig . 4 . as pointed out earlier , an upward correction of [ fe / h ] must also be applied because of the large column density of neutral hydrogen . however , this correction is uncertain and does not change the relative heavy element abundances . therefore , we simply use [ fe / h ] values derived by lu et al . ( 1996a ) , prochaska & wolfe ( 1999 ) and pettini et al . ( 2000 ) keeping in mind that correction might be up to of the order of + 1 dex . for 4 dlas of the sample the [ fe / h ] are not available . we adopt [ fe / h ] = [ cr / h ] for those systems . for [ ni / fe ] we have applied an additional upward correction which comes from the new measurements of oscillator strengths for ni ii from uv absorption lines ( fedchak & lawler 1999 ) ; older @xmath48-values were used by prochaska & wolfe ( 1999 ) and lu et al . ( 1996a ) . although the correction is different for different ni ii transitions , its mean value is @xmath26 0.3 dex , which we adopt for the sake of simplicity . we now discuss the behaviour of the element ratios of si , zn , s , mn , cr , ni , ti , and al with respect to fe . * silicon , zinc : * the case of si and zn has already been treated before ( sect . 3.3 ) . in short , it can be seen from fig . 4 that the corrected [ zn / fe ] values scatter around zero , while [ si / fe ] remains positive suggesting the production of this element in type ii supernovae . * sulphur : * [ s / fe ] shows the same snii like pattern as [ si / fe ] . we note that if the gas in dla galaxies is ionized by softer stellar radiation with @xmath10 = 33000 k , the downward correction of s / fe abundance ratio is 0.17 dex larger and in this case [ s / fe ] @xmath37 0 , in contradiction with the production of s in type ii supernovae . this difference in s correction factors also favours harder ionization radiation for the low - metallicity dlas as compared to that in the milky way . a new analysis of the s / zn abundance ratio in a sample of dlas was recently presented by centurin et al . they find [ s / zn ] @xmath37 0 in dlas with [ zn / h ] @xmath2 1.0 . both s and zn are not depleted into the dust grains and the correction for dust depletion does not change s / zn ratio . centurin et al . ( 2000 ) thus conclude that such unusual abundance ratios suggest that the dla galaxies are objects with low star formation rates such as low surface brightness galaxies or dwarf galaxies . based on our model , the correction for the ionization results in the increase of both [ s / zn ] ( table 1 ) and [ zn / h ] and in better agreement with [ s / fe ] for halo stars . if instead the model with the hard background intergalactic radiation is adopted , the correction of [ s / zn ] is negative and it increases the difference between [ s / zn ] in dlas and [ s / fe ] in halo stars . * manganese , chromium : * [ mn / fe ] remains nearly unchanged because of small correction factor for this ratio . the cr / fe ratio is overabundant by @xmath26 0.2 dex with respect to the solar value , which seems to be in contradiction with the stellar data . this may constitute a difficulty for our model . uncertainties in oscillator strengths for cr ii transitions are better than 10% ( bergeson & lawler ( 1993 ) . however uncertainties in the ionization and recombination cross - sections should be examined . adopting a higher effective temperature decreases the overabundance of cr relative to fe , as can be seen from table 1 . the mn / fe ratio remains underabundant supporting the origin of this odd - element in type ii supernovae . * nickel : * note that [ ni / fe ] calculated with the new oscillator strengths before the ionization correction ( open circles ) are already above zero . this poses a problem in the explanation of [ ni / fe ] when dust depletion is taken into account , since this correction shifts the points upward . although a large negative correction of 0.38 is predicted for [ ni / fe ] in the ionization model , it is almostly compensated by the upward shift of points introduced by the new oscillator strengths by fedchak & lawler ( 1999 ) . the points in [ ni / fe ] vs [ fe / h ] diagram scatter around the mean [ ni / fe ] value of 0.2 . * titanium : * prochaska & wolfe ( 1999 ) pointed out problems in explaining the ti / fe abundance ratio in the `` dust depletion '' model , since ti is expected to be more heavily incorporated into dust grains . hence , the correction for depletion shifts observed [ ti / fe ] to larger values . in our model , the correction of [ ti / fe ] for ionization effects is negative and small . therefore , its application to prochaska & wolfe ( 1999 ) data leaves ti moderately overabundant relative to iron ( filled squares in fig . 4 ) , in agreement with that observed in galactic halo stars . * aluminium : * our model predicts the largest downward correction factor of 0.63 dex for al / fe abundance ratio . in the prochaska & wolfe ( 1999 ) sample the [ al / fe ] scatter around zero value . this is likely at variance with what is observed in lowest - metallicity halo stars where [ al / fe ] scatters below zero with mean value of @xmath26 0.4 ( ryan et al . 1996 ; timmes et al . 1995 ; samland 1998 ) . our correction eliminates this difference and suggests that the odd - element al seen in damped ly@xmath0 galaxies is likely also produced by type ii supernovae . if instead the hard intergalactic ionization radiation is applied , the correction of [ al / fe ] is @xmath26 + 0.1 and shifts points upward in the direction opposite to the nucleosynthetic predictions . overall , we find that in the framework of our simplified model of photoionized gas and despite the uncertainties in the atomic data , the abundances in dla systems reproduce quite well observed halo star abundance patterns and lead to a consistent picture . several problems pertaining to neutral dla models are thus eliminated . in the new model , there is no need to take depletion of heavy elements into dust grains into account , although small depletion of heavy elements might be present . in the framework of the proposed model the abundance ratios of neutral and singly ionized species n i / o i , si ii/ s ii , mn ii / fe ii ( table 1 ) depend only weakly on ionization effects and can be used as a good measure of the total abundance ratios . we now turn to a discussion of the nitrogen abundance , which is of particular interest since it is one of the elements observable both in metal - poor hii regions and dlas , and because its primary and/or secondary origin remains debated ( see e.g. henry , edmunds & kppen 2000 ) . the heavy element abundance patterns seen in dla galaxies show several similarities with those found in local blue compact dwarf galaxies . izotov & thuan ( 1999 ) , based on the large sample of low - metallicity bcds , have demonstrated that all elements of @xmath0 process are overabundant relative to iron and hence they are produced in type ii supernovae . in this respect the n / o ratios seem , at first glance , to constitute the major difference . the n / o ratios in damped ly@xmath0 systems appear to be significantly lower than those measured in low - metallicity bcds ( cf . fig . 5 ) . for the most metal deficient of these objects with [ o / h ] @xmath28 1.3 thuan , izotov & lipovetsky ( 1995 ) and izotov & thuan ( 1999 ) have derived a mean value of log n / o = 1.60 with a very small dispersion . they argue that this value constitutes a lower limit to log n / o in bcds , which is set by primary n production in massive stars . pettini , lipman & hunstead ( 1995 ) have found an upper limit log n / o @xmath28 1.9 in one damped ly@xmath0 system , 0.3 dex below the value of izotov & thuan . however , within the errors the log n / o value derived by pettini et al . ( 1995 ) is still consistent with the mean value for bcds . furthermore , lu , sargent & barlow ( 1998 ) have compiled n / si abundance ratios in 15 damped ly@xmath0 galaxies , and have set upper limits for log n / o ranging from from 1.2 to 2.1 . whereas all these measurements constitute upper limits , an actual value of [ n / si ] = 1.70 has been derived in one system , 1946 + 7658 , by lu et al . ( 1998 ) . from these dla measurements of n / o and n / si , pettini et al . ( 1995 ) conclude that both primary and secondary nitrogen production are important over the whole range of metallicity measured in damped ly@xmath0 systems , while lu et al . ( 1998 ) favor the time - delayed primary nitrogen production by intermediate - mass stars . can these different scenarios be reconciled ? how well are these differences established , especially in view of the possible ionization corrections discussed above for other elements detected in dla systems ? indeed if our model holds , ionization effects are also important for the n / si abundance ratio . the inspection of table 1 and figure 1 shows that correction factors for the observed @xmath1(n@xmath18)/@xmath1(si@xmath8 ) ratio increase with increasing @xmath12 . for the typical value log @xmath12 = 3.25 adopted for the above analysis of other heavy element abundance ratios ( cf . 4 ) we find @xmath47log[@xmath19(n@xmath18)/@xmath19(si@xmath8 ) ] = 0.72.log[@xmath19(n@xmath18)/@xmath19(si@xmath8 ) ] is significantly larger ( 1.48 dex ) because of the assumption of density - bounded h ii region in their case . ] the observations of n / o in bcds ( izotov & thuan 1999 , filled circles ) and the dla data from lu et al . ( 1998 ) and outram et al . ( 1999 ) ( open circles ) assuming neutral systems are plotted in figure 5 . overplotted for illustration is the contribution of primary and secondary nitrogen production as well as the combined production by both mechanisms . the `` typical '' ionization correction for n / si and si / h is shown by the arrow in the lower left corner of figure 5 . taking into account such a correction , the vast majority of measurements for dlas are found at or above the n / o value from the bcds . this suggests that a significant part of the difference in [ n / si ] can be explained by ionization effects . the typical correction brings even the lowest value of [ n / si ] @xmath26 1.7 , measured in the @xmath36 = 2.8448 damped ly@xmath0 system toward q1946 + 7658 , to [ n / si ] @xmath26 0.98 after correction for ionization , fairly close to the [ n / o ] values for the bcds . the open square in fig . 5 shows the n / o abundance ratio derived recently by molaro et al . ( 2000 ) in a dla with @xmath49 = 3.3901 from the unsaturated o i and n i absorption lines . potentially , the n / o abundance ratio derived directly from n i and o i ( instead of using si as a proxy ) is the least model - dependent . also it has the advantage of being very little dependent on the correction for ionization ( see table 1 ) . most interestingly , the n / o value of molaro et al . ( 2000 ) is in very good agreement with n / o derived in the low - metallicity bcds . in any case , if the typical ionization corrections derived from our model apply , we have to conclude that the bulk of the present data does not show a significant difference in n / o abundance ratios between low - metallicity bcds and dla systems implying similar origin of these two elements . we shall now briefly summarise recent progress in stellar models regarding the nucleosynthetic origin of nitrogen . interestingly both intermediate - mass and massive stars are found to be potential producers of primary @xmath50n in the most recent stellar models . intermediate mass stars ( @xmath51 3.5 @xmath52 ) produce primary nitrogen when the cn cycle operates in their convective envelope during the agb phase ( e.g. renzini & voli 1981 ) . this production results from the combined effect of the 3rd dredge - up and envelope burning , and thus varies both with stellar mass and metallicity ( e.g. forestini & charbonnel ( 1997 ) and references therein ) . it is also highly dependent on the parameters which intervene in the description of agb stars . among them , the mixing length and the dredge - up parameters which are empirically calibrated to reproduce peculiar observed properties of different agb populations ( in particular the observed luminosity function of carbon stars in the lmc and smc ) , and which influence the efficiency of the burning inside the envelope . in addition , the mass loss prescription which remains uncertain ( e.g. vassiliadis & wood 1993 ) has a crucial impact on the theoretical nucleosynthesis by controlling the number of thermal pulses undergone by the agb star and the duration of the envelope burning . the most recent models find a @xmath50n production similar to that of renzini & voli ( 1981 , @xmath53 case with the largest n production ) at [ o / h ] @xmath26 0.7 but which decreases with increasing metallicity ( see marigo 1998 ; marigo et al . 1998 ; lattanzio , forestini & charbonnel 1999 ) . intermediate mass stars are thus potential primary @xmath50n producers . however , no calculations are yet available in the metallicity range of interest for our study . the behaviour at [ o / h ] @xmath37 0.7 remains to be investigated . regarding massive stars it is well known that standard stellar evolution models do not produce primary nitrogen ( e.g. woosley & weaver 1995 ) . however , considerable advances have been made recently to include rotation and its effects on the transport of elements and angular momentum in the description of stellar interiors and evolution ( see review of maeder & meynet 2000 ) . the corresponding new models for massive stars , which solve a number of difficulties encountered by `` standard '' models ( cf . maeder 1995 ) , also produce primary nitrogen ( maeder 1999 ; langer et al . in particular the models of maeder ( 1999 ) reproduce well the n excess in smc supergiants found by venn ( 1999 ) , and suggest that these stars could be responsible for primary n production at early epochs . increasing rotational velocities found at low o / h ( maeder , grebel & mermilliod 1999 ) could favor the primary n production in metal - poor environments . quantitative yields from these new stellar models have still to be awaited . last , but not least , such new yields from intermediate and massive stars will be included in chemical evolution models to fully exploit all the observational constraints regarding n and other elements . this should hopefully solve the question on the origin of primary n. independently of the outcome , the conclusion from our work is that in the light of possible ionization corrections for dlas one can not presently exclude the same primary origin of n in both dlas and nearby metal - poor galaxies . the above analysis shows the importance of consideration of the ionization effects in the dlas element abundance studies which was pointed out earlier by prochaska & wolfe ( 1999 ) , howk & sembach ( 1999 ) and izotov & thuan ( 1999 ) . however , at variance with many previous studies assuming that the heavy element absorption lines in dla galaxies are formed in the neutral gas , we consider the following simplified `` multi - component '' picture for dla systems : _ region 1 ) _ a plane - parallel ionization bounded region illuminated by an internal radiation field complemented by _ region 2 ) _ , a neutral region with a lower metal content . interestingly it appears that the structure of at least some dla systems might be similar to that observed in nearby low - metallicity blue compact dwarf galaxies . for example , the bcd galaxy sbs 0335052 showing supergiant h ii regions over a region of @xmath26 7 kpc with ongoing star formation embedded in much larger clouds of neutral gas with a column density @xmath1(h i ) as high as 7@xmath310@xmath54 @xmath5 , can be considered as a local damped ly@xmath0 system ( e.g. , thuan & izotov 1997 ) . another interesting parallel is the most metal - poor bcd galaxy i zw 18 , which shows a high hydrogen column density @xmath1(h i ) @xmath7 3.5@xmath310@xmath54 @xmath5 ( kunth et al . 1994 ) thus also qualifying as a local dla system , and for which recent _ fuse _ spectroscopic observations were obtained by vidal - madjar et al . in contrast with high - redshift dlas the structure of i zw 18 is known better . it consists of two central regions of active star formation ionizing the ambient gas . radio observations and the presence of strong ly@xmath0 absorption line indicate that the h ii region is embedded in the outer h i cloud . measurements of the column densities of n i , si ii , ar i and fe ii derived from the _ fuse _ spectrum of i zw 18 were kindly made available to us by alfred vidal - madjar ( 2000 , private communication ) . it is striking that the column density ratios given in table 2 ( column 2 ) are very different from those observed in the h ii regions ( nw and se components ) of i zw 18 and other low - metallicity bcds also given in the table 2 . we suspect that such large differences are due to ionization effects . to analyse the influence of the ionization on the observed column densities we consider a spherically - symmetric ionization - bounded h ii region model with the parameters derived from the observations ( see e.g. hunter & thronson 1995 ; de mello et al . 1998 ) . in particular , the heavy element mass fraction in h ii region is adopted to be 1/50 @xmath23 , while for the neutral gas we assume zero metallicity . the results of the calculations are shown in table 2 , where the observed ratio from _ fuse _ ( col . 2 ) and the ionization corrected column density ratio ( col . 3 ) are given . note the good agreement with the abundance ratios derived from the _ hst _ and ground - based observations of i zw 18 and other bcds , while solar ratios do not match corrected ratios . to obtain the total hydrogen column density a large amount of gas must now be added to that of the ionized gas ( cf . table 2 ) . however , since the heavy element ratios are already close to the observed values from the h ii regions in bcds , there is little room for additional metals in the neutral phase . hence , this analysis provides some support in favor of the basic assumption of our model that neutral gas in dlas systems might be more metal - poor as compared to the ionized gas . \1 . using the observed @xmath1(al iii)/@xmath1(si ii ) column density ratio to constrain the ionization parameter @xmath12 , one obtains almost complete ionization of hydrogen in regions with a column density @xmath1(h@xmath18+h@xmath8 ) @xmath26 10@xmath4 @xmath5 ( region 1 ) , several times smaller than the observed hi column in dla systems . in the ionized region fe is mainly in the fe@xmath9 form . both effects imply that the true fe / h abundance in the ionized gas must be larger ( by up to one order of magnitude ) than that derived from the @xmath1(fe ii)/@xmath1(h i ) ratio . compared to [ fe / h ] derived from a neutral dla model , such ionized systems have an age - metallicity relation in better agreement with that of the thick disk of the milky way . however , in general , due to the assumption of a significant metal - deficiency in the neutral region ( 2 ) , abundance ratios with respect to hydrogen are a priori not meaningful when the total absorbing column is taken into account . accounting for ionization effects improves the agreement between dlas and milky way stellar thick disk measurements . although not excluded , no significant depletion of elements by dust grains is required . the majority of heavy element abundance ratios are brought into agreement with those expected from the production by type ii supernovae . the apparent difference between [ n / o ] derived in local blue compact dwarf galaxies and [ n / si ] derived in dla systems can be fairly well explained by ionization effects in the latter . thus one can not currently exclude the same primary n origin in both dlas and metal - poor galaxies . the main sources of uncertainties potentially affecting our results are : poor atomic data ( oscillator strengths , recombination coefficients etc . for the photoionization models , affecting e.g. zn , cr , and ni ) , the unknown structure and geometry of the absorbing systems , the spatial distribution of ionizing sources and their radiation field , and uncertainties related to the measurement of column densities . future investigations should hopefully be able to improve on these issues and to assess the validity of our picture . for example , quantitative studies of the line profile structure of different ions , and detailed ionization properties of individual systems and large samples , should be extremely useful . it is the hope that the proposed simplified model reflects the main trends due to ionization effects in dlas . if correct , our picture of dla systems partly ionized by internal radiation offers a clear simplification in the understanding of heavy element abundance ratios in dlas and their comparison with the local universe . thanks the staff of the observatoire midi - pyrnes for their kind hospitality , and acknowledges an invitation by the universit paul sabatier in toulouse . alfred vidal - madjar kindly provided us with results of _ fuse _ observations . we thank richard f. green and an anonymous referee for useful comments on this paper . this international collaboration was possible thanks to the partial financial support of intas grant no . 97 - 0033 for which y.i.i . and d.s . are grateful . y.i.i . acknowledges the partial financial support of the joint research project between eastern europe and switzerland ( scope ) no . 7ukpj62178 . anders , s. , & grevesse , n. 1989 , geochim . acta , 53 , 197 bergeson , s. d. , & lawler , j. e. 1993 , , 408 , 382 centurin , m. , bonifacio , p. , molaro , p. , & vladilo , g. 2000 , , 536 , 540 de mello , d. f. , schaerer , d. , heldmann , j. , & leitherer , c. 1998 , , 507 , 199 fedchak , j. a. , & lawler , j. e. 1999 , , 523 , 734 ferland , g. j. 1996 , hazy , a brief introduction to cloudy 90 ( university of kentucky , department of physics and astronomy , internal report ) ferland , g. j. , korista , k. t. , verner , d. a. , ferguson , j. w. , kingdon , j. b. , & verner , e. m. 1998 , , 110 , 761 forestini , m. , & charbonnel , c. 1997 , , 123 , 241 henry , r. b. c. , edmunds , m. g. , & kppen , j. 2000 , , 541 , 660 howk , j. c. , & sembach , k. r. 1999 , , 523 , l141 howk , j. c. , savage , b. d. , & fabian , d. 1999 , , 525 , 253 hunter , d. a. , & thronson , h. a. , jr . 1995 , , 452 , 238 izotov , y. i. , chaffee , f. h. , foltz , c. b. , green , r. f. , guseva , n. g. , & thuan , t. x. 1999 , , 525 , 757 izotov , y. i. , & thuan , t. x. 1998 , , 500 , 188 . 1999 , , 511 , 639 jenkins , e. b. , oegerle , w. r. , gry , c. , vallerga , j. , sembach , k. r. , shelton , r. l. , ferlet , r. , vidal - madjar , a. , york , d. g. , linsky , j. l. , roth , k. c. , dupree , a. k. , & edelstein , j. 2000 , , 538 , l81 kunth , d. , lequeux , j. , sargent , w. l. w. , & viallefond , f. 1994 , , 282 , 709 kurucz , r. l. 1991 , in proc . workshop on precision photometry : astrophysics of the galaxy , ed . a. c. davis philip , a. r. upgren , & k. a. james ( schenectady : davis ) , 27 lanzetta , k. m. , wolfe , a. m. , & turnshek , d. a. 1995 , , 440 , 435 lattanzio , j. , forestini , m. , & charbonnel , c. , 1999 , in international workshop on `` the changes in abundances in agb stars '' , monteporzio , in press ( astro - ph/9912298 ) lauroesch , j. t. , truran , j. w. , welty , d. e. , & york , d. g. 1996 , , 108 , 641 le brun , v. , bergeron , g. , boiss , p. , & deharveng , j. m. 1997 , , 321 , 733 lindner , u. , fritze - von alvensleben , u. , & fricke , k. j. 1999 , , 341 , 709 lu , l. , sargent , w. l. w. , & barlow , t. a. 1998 , , 115 , 55 lu , l. , sargent , w. l. w. , barlow , t. a. , churchill , c. w. , & vogt , s. s. 1996a , , 107 , 475 lu , l. , sargent , w. l. w. , womble , d. s. , & barlow , t. a. 1996b , , 457 , l1 maeder , a. 1995 , in astrophysical applications of stellar pulsation , ed . stobie , p.a . whitelock , asp conf . series , 83 , 1 maeder , a. 1999 , in wolf - rayet phenomena in massive stars and starburst galaxies , ed . k. a. van der hucht , g. koenigsberger , & p. r. j. eenens ( sheridan books : michigan ) , 177 maeder , a. , & meynet , g. 2000 , , in press maeder , a. , grebel , e. k. , & mermilliod , j .- c . 1999 , , 346 , 459 marigo , p. 1998 , phd thesis , universita de padova marigo , p. , bressan , a. , & chiosi , c. 1998 , , 331 , 564 molaro , p. , bonifacio , p. , centurin , m. , dodorico , s. , vladilo , g. , santin , p. , & di marcantonio , p. 2000 , , 541 , 54 outram , p. j. , chaffee , f. h. , & carswell , r. f. 1999 , , 310 , 289 pettini , m. , ellison , s. l. , steidel , c. c. , & bowen , d. v. 1999 , , 510 , 576 pettini , m. , ellison , s. l. , steidel , c. c. , shapley , a. e. , & bowen , d. v. 2000 , , 532 , 65 pettini , m. , king , d. l. , smith , l. j. , & hunstead , r. w. 1997 , , 478 , 536 pettini , m. , lipman , k. , & hunstead , r. w. 1995 , , 451 , 100 pettini , m. , smith , l. j. , hunstead , r. w. , & king , d. l. 1994 , , 426 , 79 prantzos , n. , & boissier , s. 2000 , , 315 , 82 prochaska , j. , & wolfe , a. m. 1996 , , 470 , 403 . 1999 , , 121 , 369 . 2000 , , 533 , l5 renzini , a. , & voli , m. 1981 , , 94 , 175 ryan , s. g. , norris , j. e. , & beers , t. c. 1996 , , 471 , 254 samland , m. 1998 , , 496 , 155 sembach , k. r. , howk , j. c. , ryans , r. s. i. , & keenan , f. p. 2000 , , 528 , 310 stasiska , g. , & leitherer , c. 1996 , 107 , 661 thuan , t. x. , & izotov , y. i. 1997 , , 489 , 623 thuan , t. x. , izotov , y. i. , & lipovetsky , v. a. 1995 , , 445 , 108 timmes , f. x. , woosley , s. e. , & weaver , t. a. 1995 , , 98 , 617 vassiliadis , e. , & wood , p. r. 1993 , , 413 , 641 venn , k. a. 1999 , , 518 , 405 vidal - madjar , a. 2000 , private communication vidal - madjar , a. , kunth , d. , lecavelier des etangs , a. , lequeux , j. , andr , m. , benjaffel , l. , ferlet , r. , hbrard , g. , howk , j. c. , kruk , j. w. , lemoine , m. , moos , h. w. , roth , k. c. , sonneborn , g. , & york , d. g. 2000 , , 538 , l77 viegas , s. m. 1995 , , 276 , 268 vladilo , g. 1998 , , 493 , 583 wolfe , a. m. 1990 , in the interstellar medium in galaxies , ed . h. a. thronson & j. m. shull ( dordrecht : kluwer ) , 387 wolfe , a. m. , lanzetta , k. m. , foltz , c. b. , & chaffee , f. h. 1995 , , 454 , 698 wolfe , a. m. , & prochaska , j. x. 2000 , , in press ( preprint astro - ph/0009081 ) woosley , s. e. , & weaver , t. a. 1995 , , 101 , 181 lcccccc @xmath1(h ) , @xmath5 & 2.1@xmath310@xmath54&7.7@xmath310@xmath55&&&&@xmath56 , mpc & & 10&&&&log @xmath57 , s@xmath40 & & 51.6&&&&log @xmath58 , cm&&20.08&&&&@xmath10 , k&&50000&&&&@xmath1 , cm@xmath27&&100&&&&@xmath48&&0.1&&&&log n / si & 0.77&0.14&&0.09@xmath350.23&0.10@xmath350.06&@xmath60 log ar / si & 1.32&0.57&0.61@xmath350.22&0.74@xmath350.22&0.72@xmath350.09&@xmath61 log fe / si & 0.77&0.42&&0.32@xmath350.24&0.24@xmath350.12&@xmath62

we study the phase behavior of a symmetric binary polymer blend which is confined into a thin film . the film surfaces interact with the monomers via short range potentials . we calculate the phase behavior within the self - consistent field theory of gaussian chains . over a wide range of parameters we find strong first order wetting transitions for the semi infinite system , and the interplay between the wetting / prewetting behavior and the phase diagram in confined geometry is investigated . antisymmetric boundaries , where one surface attracts the component with the same strength than the opposite surface attracts the component , are applied . the phase transition does not occur close to the bulk critical temperature but in the vicinity of the wetting transition . for very thin films or weak surface fields one finds a single critical point at . for thicker films or stronger surface fields the phase diagram exhibits two critical points and two concomitant coexistence regions . only below a triple point there is a single two phase coexistence region . when we increase the film thickness the two coexistence regions become the prewetting lines of the semi infinite system , while the triple temperature converges towards the wetting transition temperature from above . the behavior close to the tricritical point , which separates phase diagrams with one and two critical points , is studied in the framework of a ginzburg landau ansatz . two - dimensional profiles of the interface between the laterally coexisting phases are calculated , and the interfacial and line tensions analyzed . the effect of fluctuations and corrections to the self - consistent field theory are discussed . + pacs : 05.70-h , 68.45-gd , 83.80-es epsf

cond-mat0007358

the phase behavior of fluid mixtures in confined geometry has attracted abiding interest over many decades.@xcite the preferential interactions at the surfaces give rise to an enrichment of one component at the surface . in a semi - infinite system at phase coexistence , the thickness of this enrichment layer diverges at the wetting transition.@xcite upon approaching the wetting transition temperature from below the thickness of the enrichment layer might increase continuously ( second order wetting ) or jump from a microscopically thin layer to a macroscopic layer at the ( first order ) transition . this latter case is by far the most common experimentally . if the transition is of first order a continuation of the singularity persists also slightly above the wetting transition temperature . at a chemical potential ( partial pressure ) of the preferred species , which is smaller than the coexistence value ( undersaturation ) , a thin and a thick enrichment layer coexist . upon following this coexistence line ( prewetting ) to higher temperatures we decrease the difference in the enrichment layers of the coexisting phases and encounter a prewetting critical point . if a symmetric binary mixture is confined into a film with antisymmetric boundaries , i.e. , the upper surface attracts one species with exactly the same strength than the lower surface attracts the other species , no phase transition will occur close to the bulk critical point . upon decreasing the temperature the enrichment layers at both surfaces gradually develop and stabilize an @xmath3 interface in the center of the film ( `` soft mode '' phase ) . it is only close to the wetting transition temperature that the symmetry is spontaneously broken and the @xmath3 interface is localized close to one surface . this interface localization delocalization transition@xcite and the anomalous fluctuations of the delocalized @xmath3 interface in the `` soft mode '' phase have attracted recent interest@xcite and experimental realizations in terms of polymeric systems have been investigated.@xcite the application of experimental techniques ( e.g. , nuclear reaction analysis or neutron reflectometry ) is facilitated by the large length scale of the enrichment layers , which is determined by the molecules end - to - end distance @xmath4 . the macromolecular architecture also allows a successful comparison to the results of the mean field theory . the free energy cost of an @xmath3 interface @xmath5 on the length scale @xmath4 increases with chain length ; a fact that reduces the effect of interface fluctuation on the phase diagram for very long chains . the extended fractal shape of the polymers leads to a strong interdigitation of different molecules . the large number of neighbors with which a molecule interacts strongly suppresses composition fluctuations and imparts mean field behavior to the phase diagram except for the ultimate vicinity of the critical point . in the following we consider a symmetric binary polymer mixture confined into a thin film with antisymmetric boundaries and study how the wetting transition in the semi - infinite geometry affects the phase stability in a film . we employ self - consistent field calculations@xcite to calculate the phase diagram as a function of the incompatibility , the short range surface interactions , and the film thickness . our paper is arranged as follows : in the next section we describe the self - consistent field technique.@xcite then we present the phase behavior in a thin film with antisymmetric boundaries . for thick films or strong surface fields the phase diagram contains two critical points , corresponding to the prewetting critical points of each surface . interfacial profiles between the coexisting , laterally segregated phases are discussed , and the interfacial and line tensions are analyzed . the paper closes with a summary and a discussion of fluctuation effects . we consider a binary polymer blend in a volume @xmath6 . the film contains @xmath7 polymers . @xmath8 denotes the film thickness , while @xmath9 is the lateral extension of the film . let @xmath10 be the monomer number density in the middle of the film . the density at the film surfaces deviates from the density in the middle and it is useful to introduce the thickness ( volume ) @xmath11 ( @xmath12 ) of an equivalent film with constant monomer density @xmath13 . the two surfaces of the film are impenetrable and hard . in a boundary region of width @xmath14 the total monomer density drops to zero at both walls . in our calculations we assume the monomer density profile @xmath15 to take the form@xcite @xmath16 a film with the same number of monomers but uniform density would have the thickness @xmath17 . we assume the width of the boundary region to be small compared to the characteristic length scale of the composition profile , i.e. , @xmath18 . in accord with previous studies@xcite we choose @xmath19 . this particular choice of the density profile is employed for computational convenience . if we chose a small value of the ratio @xmath20 the results would remain ( almost ) unaltered but the computational effort ( i.e. , the required number of basis functions ( cf . below ) ) would increase substantially . both polymer species of the blend denoted @xmath0 and @xmath1 contain the same number of monomeric units @xmath21 and are of the same architecture . we model them as gaussian chains of end - to - end distance @xmath4 . there is a short range repulsion between the two monomer species which can be parameterized by the flory - huggins parameter @xmath22 . the reduction of the total monomer density imparts also a lower segregation to the boundary regions ( `` missing neighbor '' effect ) . both walls interact with the monomer species via a short range potential . the monomer wall interaction @xmath23 in units of the thermal energy @xmath24 is modeled as@xcite : @xmath25 a positive value ( @xmath26 ) corresponds to an attraction for the @xmath0 monomers and a repulsion for the @xmath1 species . the range of the monomer wall interaction is assumed to be much smaller than the chain extension and , for convenience , we employ the same numerical value as for the width of the boundary region in the monomer density profile . the normalization of the surface fields @xmath27 and @xmath28 , which act on the monomers close to the left and the right wall , is chosen such that the integrated interaction energy between the wall and the monomers is independent of the width of the boundary region @xmath14 . in the following we consider antisymmetric surface fields , i.e. , @xmath29 . the microscopic @xmath0 monomer density @xmath30 can be expressed as a functional of the polymer conformations @xmath31 : @xmath32 where the sum runs over all @xmath33 @xmath0 polymers in the system and @xmath34 parameterizes the contour of the gaussian polymer . a similar expression holds for @xmath35 . with this definition the semi - grandcanonical partition function of a binary blend takes the form : @xmath36 { \cal p}_a[{\bf r } ] \int { \cal d}_b[{\bf r } ] { \cal p}_b[{\bf r } ] \nonumber \\ & & \;\times \;\exp\left ( - \rho \int { \rm d}^3{\bf r } \left\ { \chi \hat \phi_a \hat \phi_b - h({\bf r})(\hat \phi_a({\bf r})-\hat \phi_b({\bf r } ) ) \right\}\right ) \nonumber \\ & & \;\times\;\delta\left ( \phi_0({\bf r } ) - \hat \phi_a({\bf r } ) - \hat \phi_b({\bf r } ) \right)\end{aligned}\ ] ] where @xmath37 and @xmath38 represents the exchange potential between @xmath0 and @xmath1 polymers . the functional integral @xmath39 sums over all chain conformations of the gaussian polymers and @xmath40 \sim \exp \left(- \frac{3}{2r_e^2 } \int_0 ^ 1 { \rm d}\tau \;\left(\frac{{\rm d}{\bf r}}{{\rm d}\tau}\right)^2 \right)$ ] denotes the statistical weight of a non interacting gaussian polymer . this simple model neglects the coupling between the interaction energy and the chain conformations,@xcite and a finite stiffness of the polymers . hence , the chain extensions parallel to the walls remain always unperturbed . the boltzmann factor in the partition function incorporates the thermal repulsion between unlike monomers and the interactions between the monomers and the walls . the last factor represents the incompressibility of the melt in the center of the film and enforces the monomer density to decay according to eq.([eqn : dens ] ) at the walls . a finite compressibility of the polymeric fluid , which results in a reduction of the monomer density at an @xmath3 interface , is neglected . introducing auxiliary fields @xmath41 , @xmath42 , @xmath43 , @xmath44 and @xmath45 we rewrite the partition function of the multi chain system in terms of the partition function of a single chain @xmath46}{k_bt}\right ) \label{eqn : z}\ ] ] the free energy functional has the form : @xmath47}{n k_bt } & \equiv & + \ ; \ln\frac{n}{v_0 } \ - \;\ln \big\ { \exp(\delta \mu/2k_bt ) { \cal q}_a[w_a ] + \exp(-\delta \mu/2k_bt ) { \cal q}_b[w_b ] \big \ } \nonumber \\ & & + \ ; \frac{1}{v } \int { \rm d}^3{\bf r } \;\ ; \chi n \phi_a({\bf r } ) \phi_b({\bf r } ) - \ ; \frac{1}{v } \int { \rm d}^3{\bf r } \;\ ; h({\bf r } ) n \left\ { \phi_a({\bf r})-\phi_b({\bf r})\right\ } \nonumber \\ & & - \ ; \frac{1}{v } \int { \rm d}^3{\bf r } \;\ ; \left\ { w_a({\bf r } ) \phi_a({\bf r } ) + w_b({\bf r } ) \phi_b({\bf r})\right\ } - \ ; \frac{1}{v } \int { \rm d}^3{\bf r } \;\ ; \xi({\bf r } ) \left\{\phi_0({\bf r } ) - \phi_a({\bf r } ) - \phi_b({\bf r})\right\ } \label{eqn : f}\end{aligned}\ ] ] where @xmath48 denotes the single chain partition in the external field @xmath41 : @xmath49 = \frac{1}{v_0 } \int { \cal d}_1[{\bf r } ] { \cal p}_1[{\bf r } ] \;\ ; \exp\left ( - \int_0 ^ 1 { \rm d}\tau\ ; w_a({\bf r}(\tau ) ) \right)\ ] ] and a similar expression holds for @xmath50 the functional integration in eq.([eqn : z ] ) can not be carried out explicitly . therefore we employ a saddlepoint approximation , which replaces the integral by the largest value of the integrand . this maximum occurs at values of the fields and densities determined by extremizing @xmath51 with respect of each of its five arguments . these values are denoted by lower case letters and satisfy the self - consistent set of equations : @xmath52 and a similar expression for @xmath53 . the abbreviation @xmath54 denotes the semi - grandcanonical single chain partition function @xmath55 at this stage fluctuations around the most probable configuration are ignored . most notably , the @xmath3 interfaces in the self - consistent ( scf ) field calculations are ideally flat and there is no broadening by fluctuations of the local position of the @xmath3 interface ( capillary waves ) . to calculate the monomer density it is useful to define the end segment distribution @xmath56 @xmath57 { \cal p}_1[{\bf r } ] \delta({\bf r } - { \bf r } ( t ) ) \exp\left(- \int_0^t { \rm d}\tau\ ; w_a({\bf r}(\tau ) ) \right)\ ] ] and a similar equation holds for @xmath58 . the end segment distribution satisfies the diffusion equation : @xmath59 the @xmath0 monomer density can be expressed via the end segment distribution @xmath60 and the single chain partition function is given by : @xmath61 substituting the saddlepoint values of the densities and fields into the free energy functional ( [ eqn : f ] ) we calculate the free energy of the different phases . @xmath62 the free energy and the monomer densities are invariant under a change @xmath63 . hence , we adjust the constant @xmath64 such that the last term in the equation above vanishes . for an homogeneous bulk ( i.e. , @xmath65 ) , we obtain from eq.([eqn : scf ] ) @xmath66 and eq.([eqn : f2 ] ) yields @xmath67 the helmholtz free energy @xmath68 in the canonical ensemble is related to @xmath69 via the legendre transformation @xmath70 and for the homogeneous bulk system @xmath68 takes the flory huggins form : @xmath71 in inhomogeneous systems we expand the spatial dependence of the densities and fields in a set of orthonormal functions : @xmath72 this procedure results in a set of non linear equations which are solved by a newton - raphson like method . for the one dimensional profiles ( @xmath73 , @xmath74 ) we use up to 120 basis functions and achieve a relative accuracy @xmath75 in the free energy . in order to investigate the interface between laterally coexisting phases we employ two dimensional scf calculations in the canonical ensemble . in the canonical ensemble the relation between the fields @xmath76 and @xmath77 and the densities is still given by eq.([eqn : scf ] ) , but the densities are obtain via : @xmath78 @xmath79 denotes the average composition of the system . we use 320 basis functions for film thickness @xmath80 . the helmholtz free energy takes the form : @xmath81 where we have used @xmath82 . the temperature scale in the scf calculations is set by the incompatibility @xmath83 , the length scale is set by the molecule s end - to - end distance @xmath4 , and the strength of the surface fields appear only in the combination @xmath84 and @xmath85 . moreover , we employ the reduced chain length @xmath86 to measure the degree of mutual interdigitation . in the framework of the scft , systems with the same values of @xmath83 and @xmath4 but different @xmath87 exhibit identical behavior . as we shall discuss , the mean field approximation is appropriate in many aspects in the limit @xmath88 , whereas there are corrections to the scf calculations for finite @xmath87 . a different interesting behavior emerges at strong segregation @xmath89 ( ssl ) . in this regime , many properties of the scf calculations are describable by simple analytical expressions , and we shall denote these expressions by ssl in the following . coexistence between different phases occurs if the two phases have the same semi - grandcanonical energy at fixed temperature @xmath90 and exchange potential @xmath38 . since the bulk is symmetric with respect to exchanging @xmath91 , phase coexistence occurs at @xmath92 and the phase diagram is given implicitly by eq.([eqn : mu ] ) . the critical temperature is given by @xmath93 . of course , the scf theory yields a parabolic shape of the binodal close to the critical point , because it is a mean field theory . the location of the wetting transition can be determined via the young equation.@xcite the @xmath0 component wets the surface , if @xmath94 , where @xmath95 denotes the @xmath3 interface tension between the coexisting bulk phases , and @xmath96 and @xmath97 denote the excess surface free energies per unit area of a surface in contact with the @xmath0rich or @xmath1rich phase , respectively . in the strong segregation limit ( ssl ) , i.e. , @xmath98 , the @xmath3 interface tension takes the form:@xcite @xmath99 the excess surface free energy of a surface in contact with the @xmath0rich phase has two contributions . on the one hand the polymer conformations are restricted due to the presence of the surface and the decay of the density profile in the vicinity of the wall . since the @xmath0 and @xmath1 polymers have identical architecture this conformational entropy contribution to the excess surface free energy is , however , the same for the two species and does not enter into the difference @xmath100.@xcite on the other hand , the surface fields @xmath27 and @xmath28 give rise to a contribution to the excess surface free energy . if the surface is completely covered by the @xmath0 component , this contribution amounts to : @xmath101 the contribution of a surface covered by the @xmath1 component has the opposite sign . if the wetting transition occurs at high incompatibility it will be of first order . in this case the enrichment layer in the non wet state is negligible small . monte carlo simulations show that this is a good approximation.@xcite the young equation@xcite yields for the strength of the surface field at the wetting transition : @xmath102 if the integrated monomer - wall interaction @xmath103 which is the experimentally relevant quantity does not depend on the chain length @xmath21 , the left hand side of the equation will be large and the wetting transition in the binary polymer blend will occur in the strong segregation limit , i.e. , @xmath104 . this is in contrast to the behavior of mixtures of small molecules , where the cahn argument@xcite suggests that the wetting transition occurs close to the critical point . if the wetting transition is first order a prewetting lines emanates from the coexistence curve above the wetting transition temperature . along this line a thin and a thick enrichment layer coexist at undersaturation . for short range forces the prewetting line approaches the bulk coexistence curve linearly @xmath105.@xcite upon increasing the temperature , the difference in the thickness of the enrichment layers decreases and the prewetting line ends in a prewetting critical point . this prewetting behavior is pertinent to the phase behavior in thin films . only for very weak surface fields the wetting transition occurs close to the critical point . in this limit polymers exhibit a behavior similar to small molecules and the wetting behavior has been studied within the square gradient approximation.@xcite the latter assumes that the concentration varies slowly on the length scale @xmath4 . in this approximation the dependence of the bare surface free energy on the composition at the wall plays a central role . in our model , both the surface fields and the `` missing neighbor '' effect due to the decay of the monomer density at the wall give rise to a composition dependence of the bare surface free energy . @xmath106 where @xmath107 denotes the composition at the surface . other contributions to the bare surface free energy ( e.g. , terms proportional to the gradient of the composition at the surface ) are omitted . from this we identify the coefficients @xmath108 and @xmath109 . one central result of the square gradient theory is that wetting transitions close to the critical point are of second order and occur at @xmath110 ( a detailed derivation of this equation in the framework of the square gradient approximation can be found in ref@xcite ) . using the parameters of our model we rewrite this result in the form : @xmath111 for arbitrary strength of the surface fields we expect the variable @xmath112 to be a function of @xmath113 and the above equations describe the limit @xmath89 and @xmath114 . rather than focusing on the detailed composition profile across the enrichment layers at the surface much qualitative insight into the wetting behavior and the interface localization delocalization transition can be gained from characterizing the profile only by the distance @xmath115 between the wall and the @xmath3 interface . the dependence of the free energy per unit area the effective @xmath3 interface potential @xmath116 on the distance @xmath115 determines the wetting behavior . the short range surface fields distort the profile in the vicinity of the wall and give rise to an effective interaction which decays exponentially with the distance @xmath115 between the wall and the @xmath3 interface . qualitatively , the effective interface potential @xmath117 in the semi infinite system can be expanded in the form@xcite @xmath118 this expression retains only the lowest powers of @xmath119 , which are necessary to bring about the salient features of the wetting behavior . we neglect the temperature dependence of the coefficients @xmath120 and @xmath64 . @xmath121 denotes the length scale of the interaction between the surface and the @xmath3 interface ( cf . below ) . in principle , the numerical values of the coefficients can be obtained from fitting the results of our scf calculation to the equation above , but we can not offer analytical expression for the coefficients in the framework of the scft . in the following we discuss the qualitative behavior which arises from an effective interface potential of type ( [ eqn : eih ] ) . if the coefficient @xmath120 is negative the wetting transition is of second order and occurs when the coefficient @xmath122 changes its sign . following parry and evans@xcite we obtain the effective interface potential @xmath123 in a thin film by superimposing the interactions originating from each individual wall ; @xmath124 . the qualitative form of the potential is shown in the inset of fig.[fig : ginz](*a * ) , where the parameter @xmath125 is proportional to the distance of the interface from the center of the film . the values of @xmath126 correspond to various temperatures . a second order wetting transition gives rise to a second order interface localization delocalization transition . above the transition a single @xmath3 interface parallel to the surfaces is stable , the system is in the one phase region . below the transition , the system phase separates laterally into a phases where the @xmath3 interface is located close to the right or the left surface , respectively . the transition is of second order , i.e. , the composition difference between the coexisting phases increases continously . the transition temperature approaches rapidly the wetting transition temperature of the semi infinite system from below as we increase the film thickness . if the coefficient @xmath120 is positive , the form of @xmath117 leads to a first order wetting transition in the semi infinite system where a thin layer of thickness @xmath127 coexists with a macroscopically thick enrichment layer at @xmath128 . the prewetting critical point is located at @xmath129 and @xmath130 . the superposition of interactions between the @xmath3 interface and the opposing walls yields the effective interface potential in a thin film . the qualitative shape of the potential is shown in fig.[fig : ginz ] ( * a * ) schematically , while panel ( * b * ) presents the results of the scf calculations for @xmath80and @xmath131 . at low temperature we find phase coexistence between two laterally segregated phases . upon increasing the temperature we encounter a triple point at which these two phases , where the @xmath3 interface is located close to the right or the left surface , respectively , coexist with third phase , where the @xmath3 interface is delocalized at the center of the film . the location of the triple point ( @xmath132 and @xmath133 ) is given by the conditions @xmath134 and @xmath135 . for large film thickness this yields : @xmath136 ; i.e. , as the film thickness is increased the triple temperature converges exponentially fast to the temperature of the first order wetting transition of the semi infinite system from above . above this triple point there are two coexistence regions which each correspond to the prewetting coexistence of the semi infinite system . at @xmath137 we find the coexistence of a thick and a thin enrichment layer of the @xmath0 species at the @xmath0attracting surface and at @xmath138 a similar coexistence at the opposite surface . the two coexistence regions end in critical points close to the prewetting critical temperature of the semi infinite system . this first order interface localization delocalization behavior is the analogon of the first order wetting behavior of the semi infinite system . the different coexisting phases and their semi grandcanonical free energy @xmath69 are presented in fig.[fig : gfree ] for @xmath80 and @xmath139 . below the triple point @xmath140 the phases are well segregated . the monomer density profiles of the @xmath0component are depicted on the left side . upon following the coexistence curve to higher temperatures @xmath69 decreases . at the triple temperature , these two phases coexist with a third phase in which the interface is delocalized in the middle of the film . from there onwards , there are two coexistence regions at positive and negative values of the exchange potential @xmath38 . profiles of the two phases of the @xmath0poor coexisting region are presented on the right side . they consist of a thin ( upper right inset of fig.[fig : gfree ] ) and a thin ( lower right inset ) enrichment layer of the @xmath0 component at the surface that favors @xmath0 . for our strictly antisymmetric system , the concentration corresponding to the triple point always is exactly @xmath141 due to the symmetry . this has an interesting consequence if one cools a mixture at @xmath142 : while in the bulk this mixture would undergo a second order phase separation ( critical unmixing at @xmath143 , @xmath144 ) , one finds a single first order unmixing transition at @xmath145 . for asymmetric compositions , however , enrichment layers form gradually at a wall close to the bulk critical temperature . this stabilizes an @xmath3 interface , which runs parallel to the surfaces . the interface is located close to one surface ; its position is given by the composition of the system . close to the prewetting critical point , the enrichment layer may phase separate laterally into a thick and a thin enrichment layer . upon further cooling , we encounter a second phase transition where the thickness of the thick enrichment layer become comparable to the film thickness , i.e. , two almost completely segregated phases coexist . previous monte carlo simulations@xcite yield evidence that the interaction range @xmath146 in eq.([eqn : eih ] ) is determined by the bulk correlation length @xmath147 for large distances between the @xmath3 interface and the surface . this is in accord with the expectation that the @xmath3 interface profile in the outer wing is characterized by the length scale @xmath147 , which measures the decay of composition fluctuations in the bulk , rather than @xmath148 , which characterizes the slope of the @xmath3 interface profile at the center of the interface . this is further corroborated by our scf calculations . for two temperatures @xmath149 and @xmath150 above the critical temperature we have measured the free energy density @xmath151 as a function of the composition @xmath152 for various film thicknesses . around @xmath142 the free energy density can be expanded in the form @xmath153 . this yields for the effective @xmath3 interface potential @xmath154 . above the critical temperature we can estimate the effective range @xmath146 of the interaction according to @xmath155 . in fig.[fig : range ] we plot @xmath156 _ vs. _ the film thickness @xmath11 . for large film thicknesses the data exhibit an exponential dependence on the film thickness . upon increasing the temperature the interaction increases as the surfaces repel the @xmath3 interface stronger . for large @xmath11 the interaction range is compatible with @xmath157 , where we have used the behavior of the correlation length at strong segregation . for small @xmath11 , i.e. , distances between the @xmath3 interface and the surface which are not very much larger than the interfacial width @xmath158 , the interaction decays somewhat faster @xmath159 . at these intermediate distances a rather complicated interaction has been predicted.@xcite upon varying the sign of the coefficient @xmath120 we alter the order of the interface localization delocalization transition . at the tricritical point the order of the transition changes . for small values of @xmath120 we make a phenomenological ginzburg - landau ansatz for the effective interface potential @xmath160 in terms of the ( not normalized ) order parameter @xmath161 . for antisymmetric surface fields the effective interface potential is invariant under the transformations @xmath91 and must be an even function of @xmath162 . we assume the simplest ansatz which allows for three phase coexistence : @xmath163 the coefficients of this landau expansion ( [ eqn : landau ] ) can be derived from an effective interface hamiltonian ( [ eqn : eih ] ) : @xmath164 and @xmath165 , where @xmath166 denotes the @xmath7th derivative of the wall interface potential at the center of the film . this effective interface potential is depicted in fig.[fig : ginz](*a * ) . for @xmath167 ( inset ) the coefficient in front of the fourth order term @xmath168 is positive and we find a single second order phase transition at @xmath169 , @xmath170 , and @xmath171 . the case @xmath172 corresponds to a first order interface localization delocalization transition . for @xmath173 there is a three phase coexistence , at which the order parameter of the coexisting phases takes the values @xmath174 and @xmath175 . of course , the landau expansion is only appropriate for small @xmath176 . the parameter @xmath126 characterizes the temperature difference to this triple point . above the triple temperature , we find two coexistence regions and , eventually , we encounter two critical points at @xmath177 and order parameters @xmath178 . the critical chemical potential @xmath179 is given by @xmath180 . @xmath181 marks the tricritical transition ; the three coexisting phases collapse to a single one with order parameter @xmath182 . the critical and triple temperature coincide likewise . the fourth order coefficient in the ginzburg landau ansatz vanishes and the binodals close to the critical temperature open like @xmath183 with @xmath184 rather than with @xmath185 ( in mean field approximation ) . the qualitative features of this scenario are confirmed by the scf calculations . results for film thickness @xmath186 are presented in fig.[fig : d0.9a ] . panel ( * a * ) presents the phase diagram for various strength of the surface fields as a function of temperature and composition . for weak surface fields @xmath187 the interface localization delocalization transition is of second order and we obtain phase diagrams with one critical point at @xmath2 upon increasing the strength of the surface fields the critical point shifts to lower temperatures . around @xmath188 the binodal become flatter and are compatible with an exponent @xmath184 . in accordance with the ginzburg landau ansatz this marks the tricritical transition . at stronger surface fields we obtain phase diagrams with two critical points , which correspond to the prewetting critical points of the first order wetting transition in the semi infinite system . when we increase @xmath187 further , the two critical points , which are located symmetrically around @xmath142 , gradually move to lower temperatures , and higher or lower @xmath0 concentration , respectively . moreover , the temperature distance between the critical points and the triple point increases . fig.[fig : d0.9a](*b * ) depicts the behavior in terms of temperature and chemical potential difference . for second order interface localization delocalization transitions the coexistence chemical potential is @xmath92 by virtue of the symmetry with respect to exchanging @xmath91 . the same holds true for first order transitions below the triple point . at the triple point , however , the coexistence curve bifurcates into two symmetrical branches , which correspond to the prewetting lines of the semi infinite system . these lines end at critical points . upon increasing the strength of the surface fields the two critical points move to lower temperatures and larger absolute values of @xmath38 . to make closer connection to the ginzburg landau ansatz , we assume that the parameter @xmath176 , which drives the transition between the two types of phase diagrams , varies as a function of the surface field @xmath187 . then the ginzburg landau ansatz predicts that the quantities @xmath189 . this is tested in fig.[fig : crit ] . indeed , our scf calculations confirm that these quantities exhibit a very similar dependence on the surface field close to the transition . moreover , we estimate the critical value of the surface field to be @xmath190 . the corresponding power laws for the location of the critical points in the vicinity of tricriticality are also displayed in the figs.[fig : d0.9a](*a * ) and ( * b * ) . additionally , the inset of fig.[fig : crit ] shows that the binodals are characterized by an exponent @xmath184 at this tricritical value of the surface field . this provides strong evidence that the ginzburg landau ansatz captures the salient features of the tricritical transition . square gradient calculations@xcite and recent monte carlo simulations@xcite of the ising model indicate that the interface localization delocalization transition can be second order in thin films ( @xmath191 ) , even if the wetting transition is first order . within our ginzburg landau ansatz this finding can be rationalized as follows : close to the tricritical point the coefficient @xmath192 is small and the temperature of the triple point is given by the condition @xmath193 or @xmath194 . at the tricritical film thickness @xmath195 the coefficient @xmath196 changes its sign . neglecting terms of order @xmath197 we obtain : @xmath198 . if the semi infinite system exhibits a first order wetting transition the coefficients @xmath120 and @xmath64 are positive . hence , for large @xmath8 the coefficient @xmath176 is positive and leads to a first order interface localization delocalization transition . if the film width @xmath8 becomes comparable to the correlation length @xmath146 , however , the second term might drive the coefficient @xmath176 negative upon decreasing the film thickness . this is further explored in our scf calculations . in fig.[fig : thick](*a * ) we present the phase diagrams as a function of the film thickness at @xmath199 . for large film thickness @xmath200 , we find a first order interface localization delocalization transition . upon decreasing the film thickness the two critical points move to lower temperatures and closer to the symmetry axis . @xmath201 corresponds to the tricritical transition : there is only a single critical point but the binodals are describable by the exponent @xmath184 . upon further decreasing @xmath8 the critical temperature increases and the binodals assume parabolical shape . the corresponding coexistence curves are shown in fig.[fig : thick](*b * ) . the mechanism is most clearly visible in the effective @xmath3 interface potential @xmath202 , which is presented in fig.[fig : thick](*c * ) for @xmath203 as a function of the distance @xmath204 between the surface and the @xmath3 interface . for the largest film thickness @xmath205 @xmath123 is to a good approximation the effective potential of the nearest surface @xmath206 . @xmath203 corresponds to a temperature above the wetting ( triple ) transition but below the ( prewetting ) critical temperature of the thick film . each surface repels the @xmath3 interface and there is a shoulder in the effective interaction around @xmath207 . qualitatively , the scf calculations confirm that @xmath123 is the linear superposition of the interactions with each surface . when we decrease the film thickness larger values of @xmath115 become unfavorable , because the more the interface moves away from one surface the more it experiences the repulsion from the opposite surface . this results in a minimum of @xmath123 close to each surface . the shape of @xmath123 corresponds to a temperature below the triple point , i.e. , the interface is localized at one or the other surface . this shows that the triple temperature increases when we decrease the film thickness . the first order character of the transition is associated with the shoulder of @xmath206 around @xmath208 . obviously , this feature of @xmath123 disappears when the film thickness @xmath11 is of the order @xmath209 and we find second order transitions for smaller film thicknesses . this thickness dependence implies that the value @xmath210 obtained from the phase diagram of rather thin films @xmath186 is not a reliable approximation of the strength of the surface fields at which a tricritical wetting transition in the semi infinite system occurs . we have attempted to locate the strength of the surface fields at which the binodals are describable by the exponent @xmath184 as a function of the film thickness @xmath11 . the results of this procedure are collected in fig.[fig : thick](*d * ) . as we increase @xmath11 the surface fields at which the tricritical transition occurs decreases and the transition temperature approaches the critical point . for film thickness much larger than the range of interaction @xmath146 between the surface and the @xmath3 interface we expect a thin film to behave similar as a semi infinite system . close to the critical point , however , the range of interaction @xmath211 between the surface and the @xmath3 interface increases . therefore we anticipate very pronounced finite size effects even for film thicknesses which exceed the end to end distance @xmath4 by far . these difficulties prevent us from reliably estimating the tricritical wetting transition of the semi infinite system or comparing our calculations to the prediction ( [ eqn:2nd ] ) of the square gradient approximation.@xcite from the behavior at film thickness @xmath212 we conclude that critical wetting transitions in the semi infinite systems occur only for @xmath213 and @xmath214 in our model . qualitatively this is in agreement with scf calculations of carmesin and noolandi@xcite and monte carlo simulations@xcite which find only first order wetting transitions , except for the ultimate vicinity of the critical point which has not been investigated . the effective interface potential also determines the composition profiles across an interface between the laterally coexisting phases . at low temperatures the coexisting phases are almost completely segregated , i.e. , the thickness of the enrichment layers of the minority components are small . in this case the interface between the coexisting phases is planar and makes an angle @xmath215 with the surface . the contact angle @xmath215 is given by the young equation : @xmath216 the width @xmath158 of the interface between the coexisting phases is given by @xmath217 , where @xmath218 denotes the distance of the @xmath3 interface from the center of the film in each phase . at a second order interface localization delocalization transition the contact angle decreases linearly with the distance from the transition temperature @xmath126 and the composition difference between the coexisting phases vanishes like @xmath219 . hence , the width of the interface between the coexisting phases diverges like @xmath220 where @xmath221 is the mean field exponent for the correlation length upon approaching the critical point . close to a first order interface localization delocalization transition the effective interface potential exhibits more structure and this will modify the shape of the interface . within the mean field approximation the shape of the interface will minimize the effective interface free energy . approximating the @xmath3 interface profile as a sharp kink at the position @xmath115 we obtain for the free energy the effective interface hamiltonian:@xcite @xmath222}{k_bt } = \int { \rm d}y{\rm d}z\ ; \left\ { \sigma_{ab } \left(\sqrt{1 + \left(\frac{{\rm d}l}{{\rm d}y}\right)^2}-1 \right ) + g(l ) \right\ } \approx l \int { \rm d}y\ ; \left\ { \frac{\sigma_{ab}}{2 } \left(\frac{{\rm d}l}{{\rm d}y}\right)^2 + g(l ) \right\ } \label{eqn : h}\ ] ] where we have assumed that the position @xmath115 of the @xmath3 interface depends only on one lateral coordinate @xmath223 . the last approximation is valid if the angle between the interface and the surface is small . this is justified in the vicinity of the wetting transition , but the approximation breaks down at low temperatures , where the interface runs almost perpendicular to the surfaces . this effective interface hamiltonian yields the euler lagrange equation : @xmath224 which can be interpreted as the trajectory @xmath115 of a particle in the potential @xmath225 . to obtain a qualitative insight we extract the effective interface potential from the one dimensional scf calculations @xmath226 at @xmath227 . typical shapes of the interface between the coexisting phases for a film thickness @xmath80 are presented in fig.[fig : prof ] . far below the triple temperature the interface is planar and makes an angle @xmath215 with the surfaces . slightly below the triple temperature , however , the interfacial profile becomes @xmath228shaped , i.e. , the angle between the interface and the surface is larger in the vicinity of the surface than at the center of the film . note that the lateral interfacial width can exceed the film thickness by far in the vicinity of the triple temperature . the flatter portion in the center of the film is a consequence of the metastability of the third phase with composition @xmath142 or the additional local minima in @xmath123 , respectively . upon approaching the triple temperature the interfacial width becomes larger and the central portion of the profiles becomes flatter and more extended . this central portion might be conceived as a microscopic layer of the metastable delocalized phase ( @xmath142 ) , which completely wets the interface between the @xmath0-rich and @xmath1rich phase at the triple point . in the semi infinite system the interfacial tension varies smoothly upon rising the temperature through the wetting transition temperature . the excess free energy of the interface approaching the surface the line tension @xmath229 varies rapidly close the wetting transition . employing an effective interface hamiltonian of the form ( [ eqn : h ] ) , indekeu@xcite has obtained a simple expression for the line tension @xmath229 . @xmath230 where @xmath231 is the position of the minimum of @xmath117 close to the surface and @xmath232 . this formula has been applied to analyze recent experiments.@xcite the behavior of the line tension close to the wetting transition depends on the order of the transition and the range of the monomer wall interaction . for short range forces the line tension @xmath229 reaches a finite positive value at the wetting transition temperature , while it is negative far below the wetting temperature . the effective interface hamiltonian captures only the qualitative behavior . monte carlo simulations and scf calculations have shown that the properties of the @xmath3 interface depend on the distance @xmath115 from the surface . this gives rise to a position dependence of the tension@xcite and width of the @xmath3 interface.@xcite a more detailed description of the interface is provided by the two dimension composition profiles in fig.[fig : profs ] . due to the choice of basis functions the profiles are periodic in @xmath223 direction and only half the system is shown . in qualitative agreement with the considerations above , the interface between the @xmath0rich and @xmath1rich phases runs straight across the film at low temperatures ( @xmath233 and @xmath234 ) . the contact angle at @xmath235 is about @xmath236 . upon increasing the temperature the contact angle between the surface and the interface decreases and the interface becomes @xmath228shaped in the vicinity of the triple point . the scf calculations also reveal that the interface becomes broader when we increase the temperature . moreover , the width of the interface is broader in the vicinity of the surfaces than in the middle of the film . this effect is due to the reduction of the monomer density in the vicinity of the surface , which imparts a reduced effective incompatibility ( `` missing neighbor effect '' ) on the surface region . a similar effects has been observed in confined systems containing copolymers@xcite this effect gives rise to a negative contribution to the line tension when the interface approaches the surface . we decompose the free energy of systems containing two interfaces between an @xmath0rich and a @xmath1rich phase into bulk , surface and line contributions : @xmath237 note that the surface fields and the entropy loss of the chains at the surfaces give rise to a thickness dependence of the bulk free energy density @xmath238 . varying both @xmath9 and @xmath11 we have estimated the coefficients in our scf calculations and the results are displayed in fig.[fig : line ] . qualitatively similar to the behavior at a first order wetting transition the line tension changes its sign from negative to positive upon approaching the triple point temperature from below . the coefficient @xmath239 decreases as we approach the triple point . to a first approximation one would expect a behavior of the form @xmath240 . the data are also compared to indekeu s formula ( [ eqn : ind ] ) . in order to apply the formula to thin films , we extend the integration only to the middle of the film and we shift the constant @xmath241 accordingly to @xmath242 in the dimensionless units of the scf calculations we obtain : @xmath243 this approximation for a thin film gives a reasonable estimate for the temperature dependence and the order of magnitude of the line tension ; however , the value of the lines tension in a thin film is systematically underestimated . when we apply the above approximation to thicker films the value of the line tension increases . if we use the film thickness @xmath205 instead of @xmath80 the approximation yields @xmath244 instead of @xmath245 . unfortunately , we are unable to extend our scf calculations to larger film thicknesses . in addition to the finite film thickness there are other effects which might upset the comparison between the scf calculations and the effective hamiltonian description : it is unclear how accurate the identification of the effective interface potential via @xmath246 is . this identification of the interface position via the absorbed amount is a good approximation for large distances between the surface and the @xmath3 interface . in this case , the composition profile across the @xmath3 interface is well describable by the interfacial profile between the coexisting bulk phases . if the @xmath3 interface is close to the surface , however , the profile becomes strongly distorted ( cf . fig.[fig : gfree ] ) and the definition of the interface position @xmath115 is somewhat ambiguous , but this is exactly the region which gives the dominant contribution to the line tension . moreover there are non local contributions to the free energy , e.g. , due to the conformational entropy . the polymers change their conformation as to fill the wedge shaped volume between the surface and the @xmath3 interface . this differ from the behavior close to a surface or an @xmath3 interface and gives rise to a contribution to the line tension , which is only partially described by the effective interface hamiltonian . we have calculated the phase diagram of a symmetric polymer mixture confined to a thin film in mean field approximation . the left surface attracts the @xmath0 component with the same strength than the right surface the @xmath1 component of the mixture . the calculations reveal a rich interplay between the phase behavior in confined geometry and the wetting behavior of the semi infinite system . if the wetting transition of the semi infinite system is second order so is the interface localization delocalization transition in a thin film.@xcite at stronger surface fields the wetting transition in the semi infinite system is first order and this gives rise to a first order interface localization delocalization transition in a thick film . the phase diagram in a thin film exhibits two critical points symmetric around @xmath142 . these correspond to the prewetting critical points of the semi infinite system . at lower temperatures we encounter a triple point at which an @xmath0rich phase , a phase where the @xmath3 interface is located in the center of the film , and a @xmath1rich phase coexist . this triple temperature converges from above to the wetting temperature as we increase the film thickness . below the triple temperature there is a single coexistence region between an @xmath0rich and a @xmath1rich phase . the interplay between the prewetting behavior and the phase diagram in a thin film has been considered for symmetric surface fields ( capillary condensation),@xcite but to the best of our knowledge phase diagrams with two critical points far below the bulk critical temperatures in films with antisymmetric surface fields have neither been discussed analytically@xcite nor observed in experiments or simulations . as we shall discuss below , we do not expect corrections to the mean field calculations to alter our conclusions qualitatively and we hope our predictions to be confirmed by experiments or simulations . qualitatively , the interplay between the prewetting behavior and the phase diagram in a film with antisymmetric boundaries is not specific to polymer blends but is rather characteristic of all binary mixtures . symmetric polymer mixtures might , however , be especially suitable model systems for exploring these effects experimentally and we hope our detailed calculations to provide some guidance . the existence of the triple point also influences the shape of the interface between the laterally segregated , coexisting phases . at low temperatures the interfaces runs straight across the film and the angle between the interface and the surface is given by the macroscopic contact angle . upon approaching the triple point from below , however , the profiles becomes s shaped with a flatter portion at the center of the film . this signals the metastability of the third delocalized phase with composition @xmath142 . at the triple point the delocalized phase completely wets the interface between the @xmath0rich and @xmath1rich phase . upon approaching the triple temperature from below the line tension changes sign from negative to positive . the properties of the interface between the coexisting phases are in qualitative agreement with the results on an effective hamiltonian description . in thin films the interface localization delocalization transition might be of second order even though the wetting transition is of first order . this has been predicted in the framework of a square gradient approach by swift _ et al._,@xcite and is in accord with simulations of the ising model.@xcite a similar behavior is found in our self - consistent field calculations for polymer blends . for our model second order wetting transitions are restricted to the ultimate vicinity of the critical point of the bulk while second order interface localization delocalization transitions can be observed far below the bulk critical temperature for film thickness comparable to the end to end distance @xmath4 . this leads us to anticipate very strong finite film thickness effects close to a second order wetting transition even for film thicknesses which exceed @xmath4 by far . of course , our self - consistent field calculations neglect fluctuations . in the vicinity of the critical point we rather expect 2d ising critical behavior with much flatter binodals than the parabolic binodals of the mean field universality class . the ginzburg criterion ensures , however , that these composition fluctuations are only important in the ultimate vicinity of the critical point @xmath247 , where the reduced chain length @xmath248 measures the degree of interdigitation . as we increase @xmath249 the relative temperature distance from the critical point for which these composition fluctuations are important decreases . for small and intermediate chain lengths an interesting interplay between mean field , 3d ising and 2d ising critical behavior is anticipated . in the limit of large interdigitation , however , we expect composition fluctuations to be only of minor importance for most part of the phase diagram . moreover , the interface profiles in the self - consistent field calculations are ideally flat , i.e. , there are no capillary waves of the interfaces . the importance of these fluctuations is not restricted to the vicinity of critical points . on the one hand , capillary waves lead to a broadening of profiles across the interface . the `` internal '' @xmath3 interfaces run parallel to the surfaces and the effective interaction between the @xmath3 interface and the surface imparts a long wavelength cut off @xmath250 to the spectrum of capillary waves . hence , the interfacial width does not grow unbound as we increase the lateral system size , but still the self - consistent field calculations might severely underestimate the width of the `` internal '' @xmath3 interfaces.@xcite within a convolution approximation the apparent width @xmath251 of the @xmath3 interface , which is observed in experiments or simulations , is related to the intrinsic width @xmath252 in the scf calculations via : @xmath253 where we have used the temperature dependence of the interfacial tension and width for strong segregation to obtain the last expression.@xcite @xmath1 is a short length cut off for the capillary wave spectrum . analytical calculations and recent monte carlo simulations show that @xmath1 tends to @xmath254@xcite or @xmath255,@xcite respectively , in the strong segregation limit . even though the second term in the above equation is only of the order @xmath256 the increase of the apparent interfacial width due to capillary waves typically is of the order of the intrinsic width for experimentally relevant chain lengths.@xcite on the other hand , capillary waves renormalize the effective interaction between the surface and the @xmath3 interface . for instance , the effective interaction range @xmath146 is increased to @xmath257 , where the capillary parameter@xcite @xmath258 measures the strength of fluctuation effects and decreases like @xmath259 . at weak segregation @xmath260 and @xmath261 and the capillary parameter increases like @xmath262^{-1/2}$ ] . in the ultimate vicinity of the critical point the mean field theory breaks down and we expect a crossover to a constant value.@xcite the divergence of @xmath263 upon approaching the critical point is cut off around @xmath264 , in accord with the ginzburg criterium for the crossover from mean field to ising critical behavior . in the strong segregation limit the correlation length @xmath147 approaches the temperature independent limit @xmath265 and the capillary parameter scales like @xmath266 $ ] . conformational changes are incompletely described by our self - consistent field calculations . within the gaussian chain model the lateral extension @xmath267 of a molecule parallel to the surfaces remains always unperturbed , i.e. , it is independent of the local composition , the distance between the molecules and the surface , and the film thickness . experiments and monte carlo simulations , however , do reveal a dependence of the lateral chain extension on the parameters above . partially , some of these effects can be rationalized as follows : if the film thickness is very thin the lateral chain extension @xmath267 has to increase as to restore a constant monomer density . this increase of the lateral chain extensions occurs if @xmath268 or @xmath269 . under these conditions the chains are quasi two dimensional and the density of monomers belonging to the same chain inside the coil volume is not small . @xcite an @xmath0chain in a @xmath1rich environment shrinks as to exchange energetically unfavorable intermolecular contacts with contacts along the same chain . by reducing its size it increases the density of its own monomers inside the coil volume . the energy gain upon shrinking is counterbalanced by the loss of conformational entropy . scaling arguments , monte carlo simulations and scf calculations@xcite yield for the relative reduction of a minority chain in the strong segregation limit : @xmath270 . in the limit of infinite interdigitation @xmath271 the above correction to the scf calculations become small . however , even for experimentally relevant chain lengths finite @xmath87effects might give rise to sizable corrections to the scf calculations ( e.g. , broadening of the apparent width of @xmath3 interfaces by capillary waves ) . besides finite @xmath87effects there are other corrections which are not captured by the scf calculations and which remain important even in the limit of infinite interdigitation . the finite compressibility of the polymeric fluid , for instance , gives rise to packing effects at the surfaces . the monomer density profile in the vicinity of the wall is determined by an intricate interplay between equation of state effects , loss of conformational entropy , fluid like packing effects and surface fields . these effects are not included in the scf calculations of gaussian chains , but require a detailed consideration of the molecular architecture and fluid like packing structure . however , we do not expect these effects to change our conclusions qualitatively . likewise , it is difficult to find an experimental realization of a symmetric mixture confined into a film with antisymmetric boundaries . the effects of deviations from perfectly antisymmetric surfaces and the crossover between capillary condensation ( for strictly symmetric boundary fields ) to interface localization delocalization has been explored in the framework of our model.@xcite this study shows that phase diagrams with two critical points also occur for nearly antisymmetric surface fields . the stronger the order of the wetting transition ( the more extended the prewetting line ) the more stable is the topology of the phase diagram against small deviations from perfect antisymmetry . 99 r. evans , j.phys.cond.matter * 2 * , 8989 ( 1990 ) . a.o . parry , j.phys.condens.matter * 8 * , 10761 ( 1996 ) . cahn , j.chem.phys . * 66 * , 3667 ( 1977 ) . m. schick , liquids at interfaces , les houches , session xlviii , j. charvolin , j.f . joanny , and j. zinn - justin ( eds ) , elsevier , amsterdam ( 1990 ) . s. dietrich , phase transitions and critical phenomena , vol 12 , c. domb and j. lebowitz ( eds ) academic press , london ( 1988 ) . sullivan , m.m . telo da gamma , fluid interfacial phenomena , c.a . croxton , new york , wiley ( 1986 ) . de gennes , wetting statics and dynamics , rev.mod.phys . * 57 * , 827 ( 1985 ) . f. brochard wyart and p .- g . de gennes , acad.sci . paris * 297 * , 223 ( 1983 ) . a.o . parry and r. evans , phys.rev.lett . * 64 * , 439 ( 1990 ) , physica * a 181 * , 250 ( 1992 ) . swift , a.l . owczarek , and j.o . indekeu , europhys.lett . * 14 * , 475 ( 1991 ) . k. binder , d.p . landau , and a.m. ferrenberg , phys.rev.lett . * 74 * , 298 ( 1995 ) ; phys.rev . * e 51 * , 2823 ( 1995 ) . albano , k. binder , d.w . heermann , and w. paul , surf.sci . * 233 * , 151 ( 1989 ) . indekeu , a.l . owczarek , and m.r . swift , phys.rev.lett . * 66 * , 2174 ( 1991 ) . a.o . parry and r. evans , phys.rev.lett . * 66 * , 2175 ( 1991 ) . j. rogiers and j.o . indekeu , europhys.lett . * 24 * , 21 ( 1993 ) . e. carlon and a. drzewinski , phys.rev.lett . * 79 * , 1591 ( 1997 ) . t. kerle , j. klein , and k. binder , phys.rev.lett . * 77 * , 1318 ( 1996 ) , europ . j. b*7 * , 401 ( 1999 ) . m. sferrazza , m. heppenstall butler , r. cubitt , d. bucknall , j. webster , r.a.l . jones , phys.rev.lett . * 81 * , 5173 ( 1998 ) ; m. sferrazza , c. xiao , r.a.l . jones , g.d . bucknall , j. webster , j. penfold , phys.rev.lett . * 78 * , 3693 ( 1997 ) . a. werner , f. schmid , m. mller , and k. binder , j.chem.phys . * 107 * , 8175 ( 1997 ) . a. budkowski , adv . polymer sci . * 148 * , 1 ( 1999 ) . k. binder , adv . polymer sci . * 138 * , 1 ( 1999 ) . e. helfand and y. tagami , j.chem.phys . * 56 * , 3592 ( 1972 ) . j. noolandi and k.m . hong , macromolecules * 14 * , 727 ( 1981 ) . l. schlangen , f.a.m . leermakers , and l.k . koopal , j.chem.soc.faraday trans . * 92 * , 579 ( 1996 ) . matsen , j.chem.phys . * 106 * , 7781 ( 1997 ) . t. geisinger , m. mller , and k. binder , j.chem.phys . * 111 * , 5241 and 5251 ( 1999 ) . m. mller , k. binder , and e.v . albano , physica * a 279 * , 188 ( 2000 ) . m. mller , k. binder , and e.v . albano , europhys.lett . * 50 * ( jun , 15 ) ( 2000 ) . m. mller , macromolecules * 31 * , 9044 ( 1998 ) . m. mller , macromol . theory simul . ( feature article ) * 8 * , 343 ( 1999 ) . t. young , philos.trans.r.soc . ( london ) * 5 * , 65 ( 1805 ) . semenov , j.phys . ii ( france ) * 6 * , 1759 ( 1996 ) . m. mller and k. binder , macromolecules * 31 * , 8323 ( 1998 ) . e.h . hauge and m. schick , * b 27 * , 4788 ( 1983 ) . i. schmidt and k. binder , j.physique * 46 * , 1631 ( 1985 ) . h. nakanishi and p. pincus , j.chem.phys . * 79 * , 997 ( 1983 ) . a.m. ferrenberg , d.p . landau , and k. binder , phys.rev . * e 58 * , 3353 ( 1998 ) . i. carmesin and j. noolandi , macromolecules * 22 * , 1689 ( 1989 ) . huse , w. van saarloos , and j.d . weeks , phys.rev . * b 32 * , 233 ( 1985 ) . indekeu , physica * a 183 * , 439 ( 1992 ) ; j.o . indekeu , int.j.mod.phys . * b 8 * , 309 ( 1994 ) . wang , s. betelu , and b.m . law , phys.rev.lett . * 83 * , 3677 ( 1999 ) . fisher and h. nakanishi , j.chem.phys . * 75 * , 5857 ( 1981 ) . h. nakanishi and m.e . fisher , j.chem.phys . * 78 * , 3279 ( 1983 ) . d. nicolaides and r. evans , phys.rev . * b 39 * , 9336 ( 1989 ) . r. evans and u. marini bettolo marconi , phys . rev * a 32 * , 3817 ( 1985 ) . semenov , macromolecules * 26 * , 6617 ( 1993 ) ; _ ibid _ * 27 * , 2732 ( 1994 ) . a. werner , f. schmid , m. mller , and k. binder , phys.rev . * e 59 * , 728 ( 1999 ) . shull , a.m. mayes , and t.p . russell , macromolecules * 26 * , 3939 ( 1993 ) . d.s . fisher and d.a . huse , phys.rev . * b 32 * , 247 ( 1985 ) . fisher and h. wen , phys.rev.lett . * 68 * , 3645 ( 1992 ) . however , a non vanishing stiffness of the molecule gives rise to an additional change in the lateral extension @xmath267 close to the film surface . this orientation effects is not describable in the gaussian chain model and persists also for thick films . , , and as a function of the surface field . in agreement with the ginzburg landau ansatz all quantities show the same dependence on the parameter . the dashed line is only a guide to the eye . the location of the critical surface field is indicated . . the inset display the behavior of the binodals for @xmath273 . the curve corresponds to the scf calculations , while the dashed line indicates the with . ( * a * ) phase diagram for @xmath131 and various film thicknesses @xmath8 . for @xmath205 and @xmath274 the interface localization delocalization transition is first order , @xmath275 corresponds to a tricritical transition , while the transition is second order for @xmath276 . ( * b * ) phase diagram as a function of temperature and chemical potential for the same parameters than in ( * a * ) . ( * c * ) effective interface potential @xmath277 at @xmath278 as a function of the film thickness @xmath8 for @xmath203 . ( * d * ) surface field ( circles ) and temperature ( squares ) of the tricritical transition as a function of the film thickness @xmath8 .

we study holomorphic spheres in certain symplectic cobordisms and derive information about periodic reeb orbits in the concave end of these cobordisms from the non - compactness of the relevant moduli spaces . we use this to confirm the strong weinstein conjecture ( predicting the existence of null - homologous reeb links ) for various higher - dimensional contact manifolds , including contact type hypersurfaces in subcritical stein manifolds and in some cotangent bundles . the quantitative character of this result leads to the definition of a symplectic capacity .

1109.1389

in a previous paper @xcite we studied moduli spaces of holomorphic discs in certain @xmath0-dimensional symplectic cobordisms . as in the classical work of hofer @xcite on the weinstein conjecture in dimension @xmath1 , the non - compactness of these moduli spaces was used to detect periodic reeb orbits . moreover , this approach enabled us to give a unified view of many results in @xmath1-dimensional contact topology and @xmath0-dimensional symplectic topology . in the present paper we extend this work to higher dimensions . using an idea that can be traced back to mcduff @xcite , we modify our set - up by constructing a symplectic cap on the convex end of the symplectic cobordism . we can then work with moduli spaces of holomorphic spheres rather than discs , which allows us to invoke a compactness theorem from symplectic field theory @xcite . our main technical result ( theorem [ thm : main ] ) makes quantitative predictions about periodic reeb orbits in the concave end of the symplectic cobordisms under consideration . in corollary [ cor : main ] we rephrase this as a statement about the weinstein conjecture @xcite in the strong version proposed by abbas et al . @xcite , which will be recalled in section [ section : weinstein ] . we also recover a result of mcduff about a class of symplectic fillings whose boundary is necessarily connected ( theorem [ thm : mcduff ] ) . in section [ section : applications ] we explore various applications of these results . our main focus is on specific instances of the strong weinstein conjecture and on quantitative reeb dynamics . these examples include contact type hypersurfaces in subcritical stein manifolds ( corollary [ cor : stein ] ) and in cotangent bundles over split manifolds @xmath2 ( corollary [ cor : cotangent ] ) . as in our previous paper , the quantitative results give rise to the definition of a symplectic capacity via the periods of reeb orbits on contact type hypersurfaces . a typical application is the weinstein conjecture for subcritically stein fillable contact manifolds ( corollary [ cor : filling ] ) . this result is complementary to that of albers hofer @xcite on the weinstein conjecture for higher - dimensional contact manifolds that are overtwisted in the sense of niederkrger @xcite and hence , as shown there , do not admit semipositive strong symplectic fillings . hofer study holomorphic discs in trivial symplectic cobordisms only , but it seems likely that this can be extended to the more general cobordisms considered here . for other recent work on the higher - dimensional weinstein conjecture see @xcite . the final two sections contain the proof of the main technical result . in section [ section : completing ] we describe a completion of the symplectic cobordism . in section [ section : moduli ] we introduce certain moduli spaces of holomorphic spheres in this completed cobordism ; the non - compactness of these moduli spaces implies the main result . part of the motivation for this paper comes from the recent article by oancea viterbo @xcite on the topology of symplectic fillings . at the end of the present paper we briefly indicate the relation of our results with their work . let @xmath3 be a closed @xmath4-dimensional manifold carrying a ( cooriented ) contact structure @xmath5 , i.e. a tangent hyperplane field defined as @xmath6 for some @xmath7-form @xmath8 such that @xmath9 is a volume form on @xmath3 . we equip @xmath3 with the orientation induced by this volume form ; write @xmath10 for @xmath3 with the reversed orientation . the reeb vector field @xmath11 of @xmath8 is defined by the equations @xmath12 and @xmath13 . paraphrasing weinstein @xcite we say that @xmath14 _ satisfies the weinstein conjecture _ if for _ every _ contact form @xmath8 defining @xmath5 the corresponding reeb vector field @xmath15 has a closed orbit . when we speak of a contractible periodic orbit , the period is not required to be the minimal one . a * reeb link * for a contact form @xmath8 is a collection of periodic orbits of @xmath15 , not necessarily of minimal period . its * total action * is the sum of the periods . in other words , when we speak of a reeb link we allow the components of this link to be multiply covered . this convention is important for the following definition . following abbas et al . @xcite we say that @xmath14 _ satisfies the * strong * weinstein conjecture _ if for every @xmath8 defining @xmath5 there exists a _ nullhomologous _ reeb link . here is a simple example of a whole class of contact manifolds @xmath16 admitting nullhomologous reeb links . consider a closed manifold @xmath17 with a symplectic form @xmath18 such that @xmath19 represents an integral cohomology class . let @xmath20 be a principal circle bundle of ( real ) euler class @xmath21 $ ] . then the connection @xmath7-form @xmath8 on this bundle with curvature form @xmath18 , i.e. @xmath22 , is a contact form on @xmath3 whose reeb orbits are the fibres of the @xmath23-bundle . when @xmath17 is a surface , the euler class may be regarded as an integer , and the @xmath24-fold multiple of the fibre is nullhomologous in @xmath3 . unless @xmath17 is a @xmath25-sphere , no multiple of the fibre will be nullhomotopic . this follows from the homotopy exact sequence @xmath26 of the fibration , since surfaces of genus at least one are aspherical . in the case @xmath27 , choose a @xmath25-dimensional integral homology class of @xmath3 on which @xmath28 evaluates non - trivially . any such class can be represented by an embedded surface @xmath29 , see ( * ? ? ? * thorme ii.27 ) . then the @xmath30-fold multiple of the fibre will be homologically trivial in @xmath31 and , _ a fortiori _ , in @xmath3 . our main theorem answers the strong weinstein conjecture in the affirmative for contact manifolds @xmath16 that arise as the strong concave boundary of a suitable compact symplectic cobordism @xmath32 of dimension @xmath33 , which we equip with the orientation given by the volume form @xmath34 . throughout this paper it is understood that @xmath35 . the convex end of @xmath32 is required to contain one connected component ( or collection of components ) @xmath36 of the type we describe next . this class of manifolds includes spheres and ellipsoids with their standard contact form . let @xmath37 be a compact , connected @xmath38-dimensional symplectic manifold ( with boundary ) admitting a strictly plurisubharmonic potential , by which we mean the following . we require the existence of an almost complex structure @xmath39 tamed by @xmath40 , i.e.@xmath41 for all non - zero tangent vectors @xmath42 , and a smooth function @xmath43 having the boundary @xmath44 as a regular level set and with @xmath45 a straightforward calculation , cf . * lemma 4.11.3 ) , shows that the restriction of a ( strictly ) plurisubharmonic function to a non - singular holomorphic curve is ( strictly ) subharmonic , whence the name . this entails the maximum principle for such curves ( even in the non - strict case ) . given any point of an almost complex manifold and a holomorphic tangent vector at that point , one can find a local holomorphic curve passing through that point in the given tangent direction ( * ? ? ? * theorem iii ) . combined with the maximum principle this implies that @xmath46 attains its maximum on the boundary @xmath44 only . after changing @xmath46 by an additive constant we may assume @xmath47 . equip the product @xmath48 with the almost complex structure @xmath49 and the symplectic form @xmath50 , where @xmath51 denotes the fubini study form of total integral @xmath52 on @xmath53 . observe that the khler manifold @xmath54 may be interpreted as the one - point compactification of the open unit disc @xmath55 in @xmath56 with its standard area form . on @xmath57 we have the corresponding strictly plurisubharmonic function @xmath58 . we write @xmath59 and @xmath60 . the manifold @xmath36 is supposed to be a regular level set @xmath61 with @xmath62 smaller than both @xmath63 and @xmath64 , or a collection of connected components of such a level set . this choice of @xmath62 ensures that @xmath36 is a compact hypersurface in @xmath65 . it inherits the contact form @xmath66 the contact structure @xmath67 is given by the @xmath68-invariant sub - bundle of the tangent bundle @xmath69 . the liouville vector field @xmath70 for @xmath71 defined by @xmath72 satisfies @xmath73 , so the contact manifold @xmath74 is the strong convex boundary of the symplectic manifold @xmath75),\omega_c\bigr)$ ] . here are the properties we require of the symplectic cobordism @xmath32 : * @xmath32 is compact , connected and @xmath52-semipositive ( see below for the definition ) . * the oriented boundary of @xmath76 equals @xmath77 where @xmath78 is allowed to be empty . the manifolds @xmath3 , @xmath78 and @xmath36 are not required to be connected . * @xmath16 is the strong concave boundary of @xmath32 . by this we mean that there is a liouville vector field @xmath70 for @xmath18 defined near @xmath79 and pointing into @xmath76 along @xmath3 such that @xmath80 . * on a neighbourhood of @xmath81 there is an @xmath18-tame almost complex structure @xmath82 relative to which the boundary @xmath78 is @xmath82-convex , i.e. the @xmath82-invariant subbundle of @xmath83 is a contact structure . * @xmath74 is a strong convex boundary component of @xmath32 . for the definition of @xmath52-semipositivity , which is essentially the one given in ( * ? ? ? * definition 6.4.5 ) , recall that a homology class @xmath84 is said to be _ spherical _ if it lies in the image of the hurewicz homomorphism @xmath85 . write @xmath86 for the first chern class of the symplectic manifold @xmath32 , defined via any almost complex structure on @xmath76 in the contractible space of @xmath18-tame structures , and @xmath87 for the evaluation of this class on @xmath88 . by @xmath89 we denote the evaluation of the de rham cohomology class @xmath90\in h^2_{\operatorname{dr}}(w)$ ] on @xmath88 . let @xmath91 be a positive real number . a symplectic manifold @xmath32 of dimension @xmath33 is * @xmath91-semipositive * if any spherical class @xmath84 with @xmath92 and @xmath93 satisfies @xmath94 . note that this condition is automatically satisfied for symplectic manifolds of dimension at most @xmath95 . one particular case of interest to us is the one where , in addition to conditions ( c1 ) to ( c5 ) , the symplectic form @xmath18 is an exact form @xmath96 with @xmath97 . we shall refer to this as the _ liouville case _ , since the conditions are those of a concave boundary in a liouville cobordism , cf . @xcite . in this case , too , @xmath91-semipositivity is automatic . we can now formulate our main result . [ thm : main ] given a symplectic cobordism @xmath32 satisfying conditions ( c1 ) to ( c5 ) , there exists a nullhomologous reeb link in its concave end @xmath98 of total action smaller than @xmath52 . in the liouville case there is in fact a contractible reeb orbit of period smaller than @xmath52 . the choice @xmath99 in ( c1 ) is made purely for notational convenience . in the general case , one would have to replace the fubini study form @xmath51 by @xmath100 , and the open unit disc @xmath55 by a disc of radius @xmath101 . the action of the reeb link predicted by our theorem would then be smaller than @xmath91 . a neighbourhood of @xmath102 looks like a neighbourhood of @xmath103 in the half - symplectisation @xmath104 . this allows us to define a symplectic form @xmath105 on the manifold @xmath106\times m\cup_m w\ ] ] by @xmath107\times m$}. \end{cases}\ ] ] any contact form defining the cooriented contact structure @xmath108 can be written as @xmath109 for some smooth function @xmath110 . rescaling this by a constant function ( which does not change the reeb dynamics qualitatively ) we may assume that @xmath111 takes negative values only . replacing @xmath3 by @xmath112\times m\co x\in m\ } \subset ( -\infty,0]\times m\cup_m w\ ] ] we obtain a cobordism as in theorem [ thm : main ] , with concave boundary @xmath113 . in the liouville case we can define the collar of @xmath3 in @xmath76 via the liouville vector field @xmath114 given by @xmath115 . then both @xmath116 and @xmath117 are @xmath70-invariant and evaluate to zero on @xmath70 ; since they coincide on @xmath118 , they coincide near @xmath3 . in other words , @xmath117 defines an extension of the primitive @xmath116 to @xmath119\times m$ ] . these considerations lead to the following corollary . [ cor : main ] given a symplectic cobordism @xmath32 satisfying conditions ( c1 ) to ( c5 ) , the contact manifold @xmath14 satisfies the strong weinstein conjecture . in the liouville case , any contact form defining @xmath5 has a contractible periodic reeb orbit . \(1 ) let @xmath14 be a closed contact manifold of dimension @xmath1 or @xmath120 occurring as the concave end of a strong symplectic cobordism whose convex end is @xmath121 or @xmath122 , respectively , with its standard contact structure @xmath123 . then @xmath14 satisfies the strong weinstein conjecture . \(2 ) if @xmath14 with @xmath124 is liouville cobordant to @xmath125 , e.g. if @xmath125 can be obtained from @xmath14 by contact surgery , then any contact form defining @xmath5 has a contractible closed reeb orbit . related cobordism - theoretic arguments allow one to reprove the result of abbas et al . @xcite that planar contact structures on @xmath1-manifolds satisfy the strong weinstein conjecture . our proof of theorem [ thm : main ] rests on the construction of a symplectic cap , see section [ subsection : cap ] below . a cap in the case of planar contact structures has been constructed by etnyre @xcite . details of the ensuing holomorphic curves argument can be found in @xcite . the methods for proving theorem [ thm : main ] also yield the following result . on the face of it , this is stronger than a result of mcduff ( * ? ? ? * theorem 1.4 ) , but her proof actually yields the result we formulate here . in fact , the proof we give in section [ subsection : mcduff ] below is essentially hers , except that we paraphrase it in the more sophisticated language of @xcite . [ thm : mcduff ] let @xmath32 be a symplectic cobordism satisfying conditions ( c1 ) to ( c5 ) , but now with the additional assumption that @xmath3 be empty , i.e. there is no concave boundary component . then @xmath78 is likewise empty . before we turn to the proof of theorem [ thm : main ] , we describe a number of applications . some are parallel to the @xmath0-dimensional applications of the ` ball theorem ' proved in @xcite ; we shall be brief in the discussion of those . in @xcite viterbo proved the existence of closed characteristics on compact contact type hypersurfaces in @xmath126 with its standard symplectic structure . in ( * theorem 4.4 ) he extended this to contact type hypersurfaces in subcritical stein manifolds ; an alternative proof was given by frauenfelder schlenk ( * ? ? ? * corollary 3 ) . our main theorem allows us to prove the strong weinstein conjecture in these situations . first we recall the basic definitions . a hypersurface @xmath3 in a symplectic manifold @xmath127 is said to be of * contact type * if there is a liouville vector field @xmath70 for @xmath18 defined near and transverse to @xmath3 . the hypersurface is said to be of * restricted contact type * if @xmath70 is defined on all of @xmath128 . it will be understood that a contact type hypersurface @xmath3 is equipped with the contact form @xmath129 . if @xmath70 and hence the primitive @xmath130 for @xmath18 is globally defined , this places us in the liouville case of our main theorem . recall our convention from section [ section : weinstein ] : when we say that @xmath3 satisfies the weinstein conjecture , we mean that _ every _ contact form defining the contact structure @xmath131 has a closed reeb orbit . periodic reeb orbits of the specific contact form @xmath129 will be referred to as * closed characteristics*. a * stein manifold * is a complex manifold @xmath132 admitting a proper holomorphic embedding into some @xmath133 . then @xmath132 admits an exhausting ( i.e. bounded from below and proper ) strictly plurisubharmonic function @xmath134 , e.g. the restriction of the function @xmath135 on @xmath133 . the @xmath25-form @xmath136 is then a symplectic form on @xmath128 . by ( * theorem 1.4.a ) , any other exhausting strictly plurisubharmonic function on @xmath132 gives rise to a symplectomorphic copy of @xmath137 . a * stein domain * is a regular sub - level set @xmath138 ; this is also called a * stein filling * of the level set @xmath61 with contact structure given by the @xmath139-invariant sub - bundle of its tangent bundle . if @xmath134 is also a morse function , then the index of any of its critical points is at most equal to @xmath140 , cf . remark [ rem : morse ] below . a stein manifold or domain is called * subcritical * if @xmath134 is morse with all critical points of morse index strictly smaller than @xmath140 . stein fillability of a contact manifold @xmath14 can also be defined by requiring the existence of a compact complex manifold @xmath128 with boundary @xmath3 , admitting a strictly plurisubharmonic function @xmath134 for which @xmath3 is a regular level set @xmath61 . the interior @xmath141 then admits an exhausting strictly plurisubharmonic function ( with the same level sets as @xmath134 ) : simply replace @xmath134 by @xmath142 with @xmath143 convex and strictly increasing , with @xmath144 as @xmath145 . by grauert s famous theorem @xcite , @xmath141 is then again a stein manifold . moreover , thanks to gray stability , a level set sufficiently close to @xmath146 will be a contactomorphic copy of @xmath14 . [ cor : stein ] any smooth compact hypersurface of contact type ( with respect to some symplectic form @xmath147 ) in a subcritical stein manifold satisfies the strong weinstein conjecture . if the hypersurface is of restricted contact type , every contact form defining the induced contact structure has a contractible periodic reeb orbit . by a result of cieliebak @xcite , cf . @xcite , any subcritical stein manifold is symplectomorphic to a split one @xmath148 , where @xmath149 is some stein manifold . so we have a strictly plurisubharmonic function @xmath150 on @xmath128 , and @xmath151 is a strictly plurisubharmonic function on @xmath152 . thus , we are dealing with a compact hypersurface @xmath14 of ( restricted ) contact type in @xmath153 . without loss of generality we may take @xmath3 to be connected . since @xmath152 has trivial homology in codimension @xmath7 , @xmath3 separates @xmath152 into a bounded and an unbounded part . choose a regular level set @xmath61 containing @xmath3 in the interior , and write @xmath76 for the part between @xmath3 and @xmath61 . the liouville vector field @xmath70 near @xmath3 ( or on all of @xmath76 in the case of restricted contact type ) points into @xmath76 along @xmath3 , otherwise theorem [ thm : mcduff ] would be violated . hence corollary [ cor : main ] applies to this symplectic cobordism @xmath76 . ( for this qualitative result , it is irrelevant that we have replaced @xmath55 by @xmath56 . ) as this proof shows , the corollary is close in spirit to the work of floer et al . @xcite , where the existence of closed characteristics is proved in split symplectic manifolds @xmath154 with @xmath155 closed and @xmath156 . the following corollary extends earlier work of andenmatten ( * ? ? ? * theorem 1.4 ) and yau @xcite , who impose additional homological conditions on the stein filling . [ cor : filling ] if @xmath14 is a subcritically stein fillable contact manifold , any contact form defining @xmath5 has a contractible periodic reeb orbit . by assumption , @xmath14 is a level set @xmath61 of an exhausting strictly plurisubharmonic morse function @xmath134 on a stein manifold @xmath132 . a contact form defining @xmath5 , the @xmath139-invariant sub - bundle of @xmath118 , is given by the restriction of the global primitive @xmath157 of @xmath147 . so we are in the restricted contact type case of the preceding corollary . the floer - homological methods of viterbo @xcite and frauenfelder schlenk @xcite produce a closed reeb orbit contractible in the ambient manifold . in the situation of corollary [ cor : filling ] , @xmath3 is disjoint from the isotropic skeleton of the stein filling , and by general position a closed orbit contractible in the subcritical filling is also contractible in @xmath3 itself . so that last corollary can alternatively be derived from their result . given a closed manifold @xmath3 with contact form @xmath8 we write @xmath158 for the infimum of all positive periods of closed orbits of the reeb vector field @xmath15 . when there are no closed reeb orbits , we have @xmath159 , otherwise an arzel ascoli type argument as in @xcite shows that @xmath158 is a minimum , and in particular positive ; the latter is also a simple consequence of the flow box theorem . let @xmath127 be a @xmath33-dimensional symplectic manifold . the manifold @xmath128 is allowed to be non - compact or disconnected . it may also have non - empty boundary , in which case @xmath128 should be replaced by @xmath141 in the following definition of a symplectic invariant of @xmath127 : @xmath160 here the supremum is taken over all closed , but not necessarily connected , contact manifolds @xmath16 of dimension @xmath161 . by a contact type embedding @xmath162 we mean that there is a liouville vector field @xmath70 for @xmath18 defined near @xmath163 such that @xmath164 . in @xmath165 with its standard symplectic form @xmath166 write @xmath167 for the open @xmath33-ball of radius @xmath168 and @xmath169 for the cylinder over the open @xmath25-ball of radius @xmath168 . for @xmath170 we simply write @xmath17 and @xmath171 , respectively . [ thm : capacity ] the invariant @xmath172 is a symplectic capacity , i.e. it satisfies the following axioms : monotonicity : : if there exists a symplectic embedding @xmath173 , then @xmath174 . conformality : : for any @xmath175 we have @xmath176 . normalisation : : @xmath177 . monotonicity and conformality are obvious from the definition . the @xmath4-sphere of radius @xmath178 with its standard contact form has all reeb orbits closed of period @xmath179 , and it admits a contact type embedding into both @xmath17 and @xmath171 ; cf . @xcite , where this is computed explicitly in the @xmath0-dimensional case . this implies that @xmath180 and @xmath181 are bounded from below by @xmath52 . if @xmath182 is a contact type embedding of some @xmath4-dimensional contact manifold @xmath16 into @xmath17 or @xmath171 , then the image @xmath163 is contained in the interior of an ellipsoid @xmath183 for @xmath184 sufficiently large and @xmath185 sufficiently small . now our main theorem applies to the symplectic cobordism given by the region in @xmath186 between @xmath163 and @xmath187 . more generally , this argument shows that if @xmath37 is a connected but not necessarily compact @xmath38-dimensional symplectic manifold admitting a plurisubharmonic potential @xmath46 ( with the boundary as a regular level set in the compact case ) , then @xmath188 . it is not difficult to give examples where equality holds , e.g.if @xmath155 is a cotangent bundle or a stein manifold of finite type . as an immediate consequence of theorem [ thm : capacity ] , we have gromov s non - squeezing theorem @xcite : there is a symplectic embedding @xmath189 if and only if @xmath190 . our main theorem allows us to define other capacities in a similar fashion . one option is to take the infimum over the total action of null - homologous reeb links . a further possibility , for exact symplectic manifolds , is to work with restricted contact type embeddings and then to take the infimum over contractible reeb orbits , see @xcite . a comprehensive survey on symplectic capacities can be found in @xcite . in the case of exact symplectic manifolds @xmath191 , there are in fact two sensible ways to introduce a capacity . one , carried out in the @xmath0-dimensional setting in @xcite , is to consider restricted contact type embeddings for the given primitive @xmath116 , i.e. embeddings @xmath192 with @xmath193 . this leads to an invariant on the set of exact symplectic manifolds with given primitive . alternatively , and more in the spirit of an ekeland hofer capacity @xcite , one can define a capacity exclusively for subsets @xmath194 of a given exact symplectic manifold @xmath195 . this can be done via restricted contact type embeddings @xmath196 , i.e. it is only required that there be _ some _ global primitive @xmath197 for @xmath18 , depending on the embedding @xmath182 , with @xmath198 , but in addition the condition @xmath199 is imposed . all these capacities have the same normalisation constants , but we do not know if they are different , in general . by way of an example , we show that our capacity can be used to recover a result of frauenfelder et al . * remark 1.13.3 ) . we improve the constant in their result by appealing to classical geometry . let @xmath200 be a compact hypersurface of contact type . then @xmath201 . since the symplectic form @xmath202 is translation - invariant , we have a contact type embedding of @xmath16 into @xmath167 for any @xmath168 greater than the circumradius of @xmath3 , which by @xcite is at most equal to @xmath203 . this bound is optimal ; it is attained for the regular @xmath33-simplex . hence @xmath204 in the case of restricted contact type hypersurfaces , one obtains the same quantitative estimate on _ contractible _ closed reeb orbits , cf . @xcite . the weinstein conjecture for contact type hypersurfaces in a cotangent bundle @xmath205 ( with its canonical symplectic structure ) is of particular interest , since this includes the question of closed characteristics on energy surfaces in classical mechanical systems . the solution to the existence question in this classical case is described in ( * ? ? ? * chapter 4.4 ) . viterbo @xcite proved the existence of closed characteristics on contact type hypersurfaces in @xmath205 enclosing the zero section . viterbo ( * ? ? ? * theorem 3.1 ) , @xcite covers the case where the fundamental group of @xmath206 is finite . the following corollary provides new instances of the weinstein conjecture in cotangent bundles , and in fact gives the strong version . [ cor : cotangent ] the strong weinstein conjecture holds for closed contact type hypersurfaces in @xmath207 , where @xmath208 is any closed manifold . let @xmath209 be a closed hypersurface of contact type . we want to show that we can realise @xmath14 as a hypersurface of contact type in a symplectic manifold of the form @xmath210 with @xmath155 as in section [ subsection : s ] . the cobordism @xmath76 will then be defined by the part of @xmath210 between @xmath3 and a sufficiently high level set @xmath36 of the strictly plurisubharmonic potential on @xmath210 . then we choose @xmath211 large enough such that @xmath212 and apply theorem [ thm : main ] . for the qualitative statement of the corollary it is irrelevant that we need to replace @xmath55 in the original formulation of the main theorem by @xmath213 . the canonical symplectic form on @xmath214 is given by @xmath215 . the compact hypersurface @xmath3 lies in @xmath216 for @xmath217 large enough , and a symplectic embedding of @xmath218 into @xmath219 is given by @xmath220 . in order to finish the proof , we need to equip @xmath221 with an almost complex structure and an exhausting strictly plurisubharmonic function . in ( * appendix b ) it is explained how this can be done , starting from a riemannian metric on @xmath208 . since the symplectic embedding @xmath222 is not surjective , the image of a hypersurface of restricted contact type will only be of non - restricted contact type , in general . so our argument does not allow us to make any statement about contractible periodic reeb orbits . notice that a closed contact type hypersurface in a symplectic manifold of the form @xmath210 is displaceable . in this situation , a result of frauenfelder schlenk ( * ? ? ? * theorem 3 ) predicts the existence of a closed characteristic . in @xcite , albers et al . collect conditions on a @xmath1-dimensional contact manifold that prevent the existence of non - separating contact type embeddings into any @xmath0-dimensional symplectic manifold . our main theorem and its consequences for the quantitative reeb dynamics allow us to make such statements in higher dimensions , e.g.for spheres and ellipsoids . however , as we need to control the reeb dynamics , we need to fix the induced contact form on the hypersurface . here is a simple example , pointed out to us by max drner . consider the standard liouville form @xmath223 on @xmath126 , and write @xmath224 for its restriction to the tangent bundle of the sphere @xmath225 of radius @xmath168 . recall from section [ subsection : capacities ] that when we speak of a contact type embedding of @xmath226 into a symplectic manifold @xmath127 , the contact form @xmath224 being given _ a priori _ , we mean that there is a liouville vector field @xmath70 for @xmath18 defined near and transverse to @xmath227 with @xmath228 . any contact type embedding of @xmath226 into a closed @xmath91-semipositive symplectic manifold @xmath127 with @xmath229 is separating . suppose we have a non - separating contact type embedding of @xmath226 into @xmath127 . then a neighbourhood of @xmath230 looks like a neighbourhood of the sphere of radius @xmath168 in @xmath231 . remove an open tubular neighbourhood around @xmath227 corresponding to the shell between @xmath232 and @xmath233 in the euclidean model , where @xmath234 has been chosen sufficiently small . this defines a symplectic cobordism @xmath32 from @xmath235 to @xmath236 satisfying the assumptions of theorem [ thm : main ] , where in ( c1 ) we replace @xmath52-semipositivity by @xmath179-semipositivity . this tells us that there should be a reeb link in @xmath235 of total action less than @xmath179 . but all the simple reeb orbits of @xmath226 are closed of period @xmath179 , cf . this contradiction proves the proposition . the same argument applies to any ellipsoid whose minimal half - axis satisfies the inequality in the proposition . we now begin with the preparations for the proof of the main theorem . we define a ` completion ' of our symplectic cobordism @xmath76 which contains holomorphic spheres , and we describe some simple properties of these holomorphic spheres . the initial step in the proof of theorem [ thm : main ] is analogous to the arguments of mcduff in @xcite . we complete @xmath76 by attaching a negative half - symplectisation along @xmath3 , as in section [ subsection : main ] , and a ( perforated ) symplectic cap @xmath237 along @xmath36 , where @xmath238 is the closure of the component of @xmath239 containing @xmath240 ( for the notation cf . section [ subsection : s ] ) . by gluing the convex boundary @xmath74 of @xmath32 and the concave boundary @xmath74 of @xmath237 we obtain the symplectic manifold @xmath241\times m\cup_m w,\omega_-\bigr)\cup_s ( c_{\infty},\omega_c);\ ] ] see figure [ figure : open - cap ] . @xmath76 at 388 231 @xmath238 at 690 425 @xmath3 at 116 237 @xmath78 [ l ] at 663 90 @xmath36 [ b ] at 605 458 @xmath240 [ r ] at 747 359 @xmath242 [ l ] at 905 355 .,title="fig : " ] [ rem : morse ] notice that the morse index of a non - degenerate critical point of a plurisubharmonic function is at most half the dimension of the almost complex manifold . otherwise the negative definite subspace of the hessian at the critical point would contain a complex line , and by @xcite we could then find a local holomorphic curve tangent to that line . such a curve would violate the maximum principle . it follows that if @xmath46 is a morse function , then the @xmath38-dimensional manifold @xmath155 has the homotopy type of a complex of dimension at most @xmath243 . the homology exact sequence of the pair @xmath244 then shows that @xmath44 is connected when @xmath245 . by a @xmath246-small perturbation compactly supported in @xmath247 we can turn any given @xmath46 into a morse strictly plurisubharmonic function , so this topological conclusion about @xmath44 always holds in our set - up . the apparently disconnected @xmath44 in figure [ figure : open - cap ] is an artefact of the lack of dimensions . the same argument applied to the connected components of @xmath248)$ ] shows that all these components have a connected boundary . this means that each component of the level set @xmath61 , and hence in particular @xmath36 ( which is a collection of such components ) is separating . for more about plurisubharmonic functions on almost complex manifolds see @xcite . on the symplectic manifold @xmath250 we choose an almost complex structure @xmath139 tamed by @xmath251 , subject to the following conditions : * on @xmath252 , the almost complex structure @xmath139 equals the split structure @xmath253 . * on the cylindrical end @xmath119\times m$ ] , the almost complex structure @xmath139 is cylindrical and symmetric in the sense of @xcite , i.e. it preserves @xmath6 and satisfies @xmath254 . * on a neighbourhood of @xmath78 , the almost complex structure @xmath139 equals @xmath82 ( cf . ( c4 ) ) . * outside the regions described in ( j1 ) to ( j3 ) , the almost complex structure is chosen so as to satisfy certain genericity assumptions that will be described later . before defining and studying the moduli space of holomorphic spheres in @xmath255 representing a certain homology class , we collect some information about more general holomorphic spheres that will be relevant in the bubbling - off analysis . the @xmath139-convexity of @xmath78 allows one to write @xmath78 as a level set of a strictly plurisubharmonic function defined in some collar neighbourhood @xmath256 of @xmath78 , cf . * remark 4.3 ) . then the maximum principle holds for @xmath139-holomorphic curves in @xmath256 . recall from section [ subsection : s ] that @xmath36 is a collection of components of a regular level set @xmath61 . define a closed neighbourhood of @xmath257 by @xmath258 this neighbourhood is obviously foliated by holomorphic spheres @xmath259 . [ lem : spheres ] let @xmath260 be a smooth non - constant @xmath139-holomorphic sphere . * if @xmath261 , then @xmath262 . * if @xmath263 , then @xmath264 and @xmath265 is of the form @xmath266 with some holomorphic branched covering @xmath267 of @xmath53 by itself . * if @xmath268 , then @xmath265 is one of the spheres in ( ii ) . * @xmath269 . \(i ) suppose @xmath265 is a holomorphic sphere intersecting @xmath238 but not @xmath240 . then the strictly plurisubharmonic function @xmath134 is defined on @xmath270 and attains a maximum in the interior , forcing @xmath270 to lie in a level set of @xmath134 . so either @xmath271 is completely contained in the exact symplectic manifold @xmath272 , or @xmath265 is contained in @xmath273 and touches the convex boundary @xmath36 of that manifold . either alternative forces @xmath265 to be constant . \(ii ) on the preimage of @xmath274 we can write @xmath265 in the form @xmath275 with @xmath276 a @xmath39-holomorphic and @xmath277 a holomorphic function . thanks to the strictly plurisubharmonic function @xmath46 on @xmath155 , the maximum principle applies to @xmath276 , so as in ( i ) we argue that @xmath265 must be globally of the form @xmath278 . the non - constant holomorphic map @xmath277 is a branched covering @xmath279 . \(iii ) a sphere @xmath265 with image contained in @xmath238 can be globally written as in ( ii ) , that is , @xmath280 . if @xmath281 , then this is one of the spheres from ( ii ) . if @xmath282 , the point @xmath283 does not lie in @xmath238 , so @xmath277 is not surjective , and hence constant . \(iv ) if @xmath265 is a holomorphic sphere intersecting @xmath256 , then with the strictly plurisubharmonic function defined on @xmath256 we argue as in ( i ) that @xmath265 must be constant . if @xmath78 is only weakly @xmath139-convex , which means that it is the level set of a ( non - strictly ) plurisubharmonic function , then the maximum principle still applies , but there can be non - constant holomorphic spheres entirely contained in a level set . however , such spheres can not occur in a bubble tree arising as the limit of spheres in the moduli space considered in the next section , because one sphere in such a bubble tree , as we shall see , always intersects @xmath240 , so at least one sphere would touch a level set but not be entirely contained in it . so the results of the present paper remain valid under this weaker assumption on @xmath78 . in this section we define the relevant moduli spaces of holomorphic spheres in the completed symplectic cobordism . these moduli spaces will be shown to be non - compact , which then leads to a proof of theorems [ thm : main ] and [ thm : mcduff ] . fix a point @xmath284 . then @xmath285 is a holomorphic sphere in @xmath255 . write @xmath286 for the moduli space of smooth @xmath139-holomorphic spheres @xmath287 that represent the class @xmath288\in h_2({\widetilde{w}})$ ] . the intersection number of the classes @xmath288\in h_2({\widetilde{w}})$ ] and @xmath289\in h_{2n-2}({\widetilde{w}},\partial{\widetilde{w}})$ ] equals @xmath7 , so with lemma [ lem : spheres ] we see that any holomorphic sphere in the class @xmath288 $ ] that touches @xmath274 or is completely contained in @xmath238 must be of the form @xmath290 with @xmath281 and @xmath291 an automorphism of @xmath53 . the observation before the proposition tells us that near its boundary the moduli space @xmath286 looks like a neighbourhood of @xmath293 crossed with the @xmath95-dimensional automorphism group @xmath294 , i.e. like a manifold with boundary of the claimed dimension . so for all practical purposes we can apply transversality arguments as if @xmath286 had no boundary . a further consequence of the homological intersection of @xmath288 $ ] and @xmath289 $ ] being equal to @xmath7 is that all spheres in the class @xmath288 $ ] will be simple ( i.e. not multiply covered ) . the moduli space @xmath286 will be a manifold provided we can choose @xmath139 to be regular in the sense of ( * ? ? ? * definition 3.1.4 ) , i.e. such that the linearised cauchy riemann operator @xmath295 is surjective for each @xmath296 . in the regions where the choice of @xmath139 is prescribed by conditions ( j2 ) and ( j3 ) , the maximum principle applies , so no non - constant holomorphic sphere can lie entirely in one of these regions . by ( * ? ? ? * remark 3.2.3 ) , the freedom of choosing @xmath139 in the complementary region then suffices to achieve regularity for all holomorphic spheres in @xmath249 , cf . * remark 4.1.(2 ) ) . by ( * theorem 3.1.5 ) , the dimension of the manifold @xmath286 is given by @xmath297)$ ] . since the normal bundle of @xmath298 in the product manifold @xmath299 is trivial ( as a complex bundle ) , @xmath300)$ ] equals the euler characteristic of the sphere @xmath298 . the quotient space @xmath301 of @xmath302 under the diagonal action of the @xmath95-dimensional automorphism group @xmath303 , where @xmath304 acts by @xmath305 is then a @xmath33-dimensional manifold , since the spheres in @xmath286 being simple implies that this action is free . furthermore , there is a well - defined evaluation map @xmath306 & \longmapsto & u(z ) , \end{array}\ ] ] where @xmath307 $ ] denotes the class represented by @xmath308 . let @xmath309 be a proper embedding of the interval @xmath310 $ ] or @xmath311 into @xmath249 , with @xmath312 and no other image point of @xmath309 in @xmath257 . so in the case of the closed interval @xmath310 $ ] we must have @xmath313 ; in the case of the half - open interval @xmath311 we go to @xmath314 in the cylindrical end @xmath315\times m$ ] as we approach @xmath7 . set @xmath316 where by slight abuse of notation we identify @xmath309 with its image in @xmath249 . [ prop : mgamma ] given @xmath309 , a generic choice of @xmath139 can be made such that @xmath317 is a @xmath7-dimensional manifold including one component diffeomorphic to a half - open interval . in particular , @xmath317 is not compact . for @xmath317 to be a @xmath7-dimensional manifold ( away from potential boundary points ) , we need to ensure that the evaluation map @xmath320 be transverse to the submanifold @xmath309 of @xmath249 . by ( * theorem 3.4.1 and remark 3.4.8 ) , for generic @xmath139 transversality holds for all simple spheres not contained entirely in a region where @xmath139 has been fixed by one of the conditions ( j1 ) to ( j3 ) . this is the generic choice we want to make in ( j4 ) . the aforementioned theorem from @xcite only treats the case without boundary , but it can still be applied here . indeed , for the interval @xmath310 $ ] we have @xmath313 , where by lemma [ lem : spheres ] ( iv ) there are no non - constant holomorphic spheres . a neighbourhood of the boundary point @xmath321 in @xmath309 lies in @xmath274 , where the explicit description of holomorphic spheres in the class @xmath288 $ ] as maps of the form @xmath322 with @xmath304 shows that at any point @xmath307\in{\mathcal m}$ ] with @xmath323 the evaluation map @xmath324 is submersive . so near the preimage of @xmath321 , the moduli space @xmath317 looks like a single half - open interval , and there are no other boundary points in @xmath317 . we conclude that the corresponding component is a half - open interval . by proposition [ prop : mgamma ] we can find a sequence in @xmath317 without any convergent subsequence . we now want to show that such a sequence can not have any subsequence gromov - converging to a stable map in the sense of ( * ? ? ? * definition 5.1.1 ) . together with the compactness theorem from symplectic field theory @xcite this will imply , for a generic choice of contact form @xmath8 , that there has to be a gromov hofer - convergent subsequence where breaking occurs . this will lead to the existence of periodic reeb orbits . naively speaking , the non - existence of a gromov - convergent subsequence follows from the fact that bubbling is a phenomenon in codimension at least @xmath25 ( in the @xmath52-semipositive situation ) see the dimension formula for the moduli space @xmath330 in the proof of lemma [ lem : tree ] below and we are only considering a @xmath7-parameter family of holomorphic curves . for the formal argument , suppose @xmath331 $ ] is a sequence in @xmath317 with @xmath332 gromov - convergent to a stable map @xmath333 modelled on a tree @xmath334 . write @xmath335 for the number of edges of @xmath334 . we want to show @xmath336 , in which case the limit would be a classical one in the @xmath337-topology . by ( * proposition 6.1.2 ) we find a simple stable map @xmath338 , i.e. with each non - constant @xmath339 a simple map and with different non - constant spheres having distinct images , such that @xmath340 moreover , with @xmath341 denoting the homology class represented by the holomorphic sphere @xmath339 , and with suitable weights @xmath342 , we have @xmath343=\sum_{\beta\in t'}m_{\beta}a_{\beta}.\ ] ] there are two distinguished vertices in the bubble tree @xmath344 ; these may coincide . one is the bubble @xmath345 corresponding to the limit marked point @xmath346 , in particular @xmath347 . notice that @xmath345 may well be constant , a so - called ghost bubble . this is the case if the image of @xmath346 on @xmath309 happens to be the image of a nodal point joining two non - constant bubbles . the second is the bubble @xmath348 with @xmath349 . we claim that this property uniquely determines @xmath348 , and @xmath350 . the homological intersection of @xmath298 with @xmath240 equals @xmath7 , so there has to be at least one non - constant sphere @xmath339 intersecting @xmath240 . there are no non - constant holomorphic spheres contained in @xmath240 ( thanks to the strictly plurisubharmonic function @xmath46 and the maximum principle ) , so positivity of intersection with the holomorphic hypersurface @xmath240 , see ( * ? ? ? * proposition 7.1 ) , tells us that there is a unique @xmath339 intersecting @xmath240 , and this sphere has to be simple . from lemma [ lem : spheres ] it then follows that @xmath351 the symplectic energy @xmath353)$ ] of any sphere @xmath296 equals @xmath354)=\pi$ ] . assuming @xmath355 ( which implies that there are at least two non - constant bubbles ) , we have @xmath356 for every @xmath357 with @xmath358 . for @xmath359 we may regard @xmath360 for homological computations as a spherical class in @xmath361 . our choice ( j4 ) of almost complex structure was made so as to guarantee regularity for spheres . then the moduli space of simple holomorphic spheres in any class @xmath358 is a non - empty manifold of dimension @xmath362 , which implies @xmath363 . so the requirement that @xmath32 be @xmath52-semipositive implies @xmath364 as in the proof of proposition [ prop : mgamma ] , by possibly refining the generic choice of almost complex structure in ( j4 ) we can ensure that the moduli space @xmath330 of unparametrised simple stable maps modelled on @xmath344 ( with a single marked point ) is a smooth manifold . as shown in ( * ? ? ? * theorem 6.2.6 ) , the dimension of this moduli space equals @xmath365 now consider the evaluation map @xmath366 & [ \{v_{\beta}\},z ' ] & \longmapsto & v_{\beta_0}(z ' ) . \end{array}\ ] ] since transversality of the evaluation map is automatic at ghost spheres , we see as in proposition [ prop : mgamma ] that @xmath367 is a non - empty manifold of dimension @xmath368)-2 e(t')-3\\ & = & 1 - 2e(t').\end{aligned}\ ] ] this dimension must be non - negative , hence @xmath352 after all . one of the reasons why a stable map may fail to be simple is that it might contain several components having the same image . in this case @xmath335 would be larger than @xmath369 . in our situation , however , an intersection argument with the class @xmath289 $ ] as in the proof of proposition [ prop : mgamma ] allows us to conclude that @xmath334 , likewise , must consist of a single vertex @xmath370 , with @xmath371 a simple map . as shown in the final paragraph of @xcite , by invoking the arzel ascoli theorem one can reduce the proof of theorem [ thm : main ] to the case where in addition the contact form @xmath8 is assumed to be non - degenerate ( i.e. where the linearised poincar return map along closed orbits of the reeb vector field @xmath15 , including multiples , does not have an eigenvalue @xmath7 ) . under this assumption , the compactness theorem from symplectic field theory ( * ? ? ? * theorem 10.2 ) applies . this tells us that we can find a sequence in @xmath317 convergent ( in the sense of that paper ) to a holomorphic building of height @xmath372 with @xmath373 . the lowest level of this building consists of holomorphic curves in the symplectisation @xmath374 with positive punctures only and symplectic energy smaller than @xmath52 , for the total symplectic energy of the holomorphic building , which is a homological invariant , equals @xmath52 . so these positive punctures constitute a null - homologous reeb link in @xmath16 of total action smaller than @xmath52 . any holomorphic building of genus @xmath375 and height @xmath372 with @xmath373 contains at least two finite energy planes . by the argument in the proof of lemma [ lem : spheres ] , at most one of these can intersect @xmath240 ; any other finite energy plane in the top level @xmath376 of the building stays inside @xmath76 . in the liouville case there can be no finite energy plane in @xmath76 with a negative puncture , because by stokes s theorem its energy would be negative ; cf . * lemma 5.16 ) . hence , in the liouville case there has to be a finite energy plane with a positive puncture in one of the lower levels @xmath374 of the building , corresponding to a contractible reeb orbit of period smaller than @xmath52 . this concludes the proof of theorem [ thm : main ] . arguing by contradiction , we assume that @xmath78 is not empty . then we can choose a path @xmath309 in @xmath76 from @xmath298 to @xmath78 as in section [ subsection : arc ] . since the concave end @xmath3 of @xmath76 is assumed to be empty , the non - compactness of @xmath317 can only be caused by bubbling , which is precluded by the argument in section [ subsection : stable ] . in other words , @xmath317 would have to be compact , contradicting proposition [ prop : mgamma ] . the recent paper of oancea viterbo @xcite also grew from taking a fresh look at the work of mcduff @xcite . their main interest is the topology of symplectic fillings . although their approach seems quite different on a technical level , one can interpret their results in our set - up as statements about the special case where @xmath3 is empty ( and hence so is @xmath78 , by theorem [ thm : mcduff ] ) . in that special case , one can define the moduli space @xmath377 of holomorphic spheres @xmath296 with @xmath378 for @xmath379 . requiring the image of three marked points to lie on three respective hypersurfaces is essentially equivalent to taking the quotient under the action of the automorphism group @xmath294 . for generic @xmath139 subject to conditions ( j1 ) to ( j4 ) the moduli space @xmath377 turns out to be a _ compact _ oriented manifold ( with boundary ) of dimension @xmath380)-3\cdot 2=2n-2.\ ] ] since the holomorphic spheres near the boundary @xmath257 of @xmath249 are the obvious ones by lemma [ lem : spheres ] , one sees that the evaluation map @xmath381 which maps boundary to boundary , is of degree @xmath7 . the homological arguments in the proof of ( * ? ? ? * proposition 2.11 ) go through unchanged , even though in our set - up we work with manifolds with boundary . in this way one can derive one of their main theorems ( * ? ? ? * theorem 2.6 ) . the homological conditions imposed there are necessary , in general , to guarantee the compactness of the moduli space . let us point out one special case of their theorem where these homological conditions are superfluous . let @xmath36 be as in section [ subsection : s ] and @xmath32 a symplectically aspherical strong symplectic filling of @xmath382 . then the homomorphism @xmath383 on homology with coefficients in any field @xmath384 , induced by the inclusion @xmath385 , is surjective in all degrees . we are grateful to alexandru oancea for encouraging us with this project . we thank peter albers , frdric bourgeois , max drner and samuel lisi for useful conversations , and the referee for numerous perceptive comments and suggestions . 10 , the weinstein conjecture for planar contact structures in dimension three , _ comment . * 80 * ( 2005 ) , 771793 . , on nonseparating contact hypersurfaces in symplectic @xmath0-manifolds , _ algebr . _ * 10 * ( 2010 ) , 697737 . , on the weinstein conjecture in higher dimensions , _ comment . * 84 * ( 2009 ) , 429436 . , the weinstein conjecture and holomorphic curves in higher dimensions , ph.d . thesis , new york university ( 1999 ) . , compactness results in symplectic field theory , _ geom . _ * 7 * ( 2003 ) , 799888 . , subcritical stein manifolds are split , preprint ( 2002 ) , arxiv : math/0204351 . , _ from stein to weinstein and back symplectic geometry of affine complex manifolds _ , book in preparation . , quantitative symplectic geometry , in : _ dynamics , ergodic theory , and geometry _ , math . . publ . * 54 * , cambridge university press , cambridge ( 2007 ) , 144 . , symplectic hypersurfaces and transversality in gromov witten theory , _ j. symplectic geom . _ * 5 * ( 2007 ) , 281356 . , planar contact structures , giroux torsion and the weinstein conjecture , diplomarbeit , universitt zu kln ( 2011 ) . , capacits symplectiques , _ c. r. acad . paris sr . _ * 307 * ( 1988 ) , 3740 . , convex symplectic manifolds , in : _ several complex variables and complex geometry _ ( santa cruz , ca , 1989 ) , proc . sympos . pure math . * 52 * , part 2 , american mathematical society , providence , ri ( 1991 ) , 135162 . , planar open book decompositions and contact structures , _ int . res . not . _ * 2004 * , 42554267 . , the weinstein conjecture in @xmath154 , _ math . z. _ * 203 * ( 1990 ) , 469482 . , energy capacity inequalities via an action selector , in : _ geometry , spectral theory , groups , and dynamics _ , contemp . math . * 387 * , american mathematical society , providence , ri ( 2005 ) , 129152 . , hamiltonian dynamics on convex symplectic manifolds , _ israel j. math . _ * 159 * ( 2007 ) , 156 . , _ an introduction to contact topology _ , cambridge stud . adv . math . * 109 * , cambridge university press , cambridge ( 2008 ) . , how to recognise a @xmath0-ball when you see one , preprint ( 2011 ) , arxiv : 1104.1543 . , on levi s problem and the imbedding of real - analytic manifolds , _ ann . of math . ( 2 ) _ * 68 * ( 1958 ) , 460472 . , pseudoholomorphic curves in symplectic manifolds , _ invent . math . _ * 82 * ( 1985 ) , 307347 . , pseudoholomorphic curves in symplectizations with applications to the weinstein conjecture in dimension three , _ invent . * 114 * ( 1993 ) , 515563 . , the weinstein conjecture in cotangent bundles and related results , _ ann . scuola norm . ( 4 ) _ * 15 * ( 1988 ) , 411445 . , _ symplectic invariants and hamiltonian dynamics _ , birkhuser adv . texts - basler lehrbcher , birkhuser verlag , basel ( 1994 ) . , ueber die kleinste kugel , die eine rumliche figur einschliesst , _ j. reine angew . _ * 123 * ( 1901 ) , 241257 . , symplectic manifolds with contact type boundaries , _ invent _ * 103 * ( 1991 ) , 651671 . , _ @xmath139-holomorphic curves and symplectic topology _ , amer . . publ . * 52 * , american mathematical society , providence , ri ( 2004 ) . , the plastikstufe a generalization of the overtwisted disk to higher dimensions , _ algebr . geom . topol . _ * 6 * ( 2006 ) , 24732508 . , the weinstein conjecture in the presence of submanifolds having a legendrian foliation , preprint ( 2011 ) , arxiv : 1104.0250 . , some integration problems in almost - complex and complex manifolds , _ ann . of math . ( 2 ) _ * 77 * ( 1963 ) , 424489 . , on the topology of fillings of contact manifolds and applications , _ comment . math . _ , to appear . , quelques proprits globales des varits diffrentiables , _ comment . _ * 28 * ( 1954 ) , 1786 . , a proof of weinstein s conjecture in @xmath126 , _ ann . poincar anal . non linaire _ * 4 * ( 1987 ) , 337356 . , exact lagrange submanifolds , periodic orbits and the cohomology of free loop spaces , _ j. differential geom . _ * 47 * ( 1997 ) , 420468 . , functors and computations in floer homology with applications i , _ geom . funct . * 9 * ( 1999 ) , 9851033 . , on the hypotheses of rabinowitz periodic orbit theorems , _ j. differential equations _ * 33 * ( 1979 ) , 336352 . , cylindrical contact homology of subcritical stein - fillable contact manifolds , _ geom . _ * 8 * ( 2004 ) , 12431280 .

as a preparation for the dynamical investigations , this paper begins with a short review of the three - layer gravastar model with distinguished attention to the structure of the pertinent parameter space of gravastars in equilibrium . then the radial stability of these types of gravastars is studied by determining their response for the totally inelastic collision of their surface layer with a dust shell . it is assumed that the dominant energy condition holds and the speed of sound does not exceed that of the light in the matter of the surface layer . while in the analytic setup the equation of state is kept to be generic , in the numerical investigations three functionally distinct classes of equations of states are applied . in the corresponding particular cases the maximal mass of the dust shell that may fall onto a gravastar without converting it into a black hole is determined . for those configurations which remain stable the excursion of their radius is assigned . it is found that even the most compact gravastars can not get beyond the lower limit of the size of conventional stars , provided that the dominant energy condition holds in both cases . it is also shown independent of any assumption concerning the matter interbridging the internal de sitter and the external schwarzschild regions that the better is a gravastar in mimicking a black hole the easier is to get the system formed by a dust shell and the gravastar beyond the event horizon of the composite system . in addition , a generic description of the totally inelastic collision of spherical shells in spherically symmetric spacetimes is also provided in the appendix .

1008.0554

there are more and more astrophysical observations justifying the existence of extremely compact massive objects with size close to their schwarzschild radius @xcite . it is widely accepted that these observations also provide indirect justifications of the existence of black holes ( bhs ) . nevertheless , there are also alternative ideas about claiming that exotic states of matter may exist which could stabilize extremely compact stars suiting to the aforementioned astrophysical observations ( see , e.g. , @xcite for a recent review ) . one of the most popular among these types of bh mimicking objects is the gravitational vacuum star ( gravastar ) model which has received considerable attention not least because its relation to the concept of dark energy . in this model of mazur and mottola @xcite an interior de sitter spacetime region is connected via three intermediate layers to an outer schwarzschild solution such that the radius of the outermost layer is supposed to be slightly larger than the schwarzschild radius of the system . it is worth mentioning that in advance to the gravastar model there were several constructions in which the matching of a de sitter region to the schwarzschild spacetime was applied . for instance , to get rid of the @xmath0 singularity of the schwarzschild spacetime , frolov , markov and mukhanov in @xcite proposed a matching of a de sitter interior to it at a small radius of the planck scale ensuring thereby that the curvature remains bounded everywhere in the yielded spacetime . dynamical investigation of this model was already carried out in @xcite . note that up to certain extent the model used in @xcite could be considered as the precursor of the gravastar model although in the latter the matching was made in a more elaborated way and the outermost matching surface , at the boundary of the schwarzschild region , was supposed to be arranged such that its radius is slightly larger than the pertinent schwarzschild radius . once such a model is set up the following questions manifest themselves : * what type of physical process may produce such a gravastar ? * is a gravastar stable ? * if it is , does it provide a viable alternative to bhs ? while the first question has not been tackled yet even the second question turned out to be too complex within the original model of mazur and mottola although in @xcite an argument claiming for its thermodynamical stability was given as it is composed by making use of three different types of regions with unspecified matter . to reduce the related ambiguities visser and wiltshire @xcite introduced a simplified three - layer gravastar model where the interior de sitter region is matched to the exterior schwarzschild spacetime via a single matter shell . this model is simple enough to carry out various analytic investigations by making use of the thin - shell formalism of israel @xcite . visser and wiltshire besides deriving the basic relations determining the evolution of gravastars also carried out the first investigation of their radial stability . since then the stability of gravastars has been studied by several authors within this simplified model or within its continuum correspondence @xcite . results relevant for radial stability can be found in @xcite , and in case of electrically charged gravastars in @xcite . the stability has also been investigated with respect to axial perturbations @xcite . in all of these investigations attention was restricted to the space of gravastars in equilibrium , i.e. the radial stability was investigated by determining the response of a gravastar to a slight formal change of the underlying effective potential . for instance , in @xcite the excursions of gravastars was investigated in such a way that their evolution started with carefully prepared initial conditions . in all of the pertinent investigations it was demonstrated that by suitably adjusting the equation of state ( eos ) of the matter forming the surface of the gravastar the subspace of ` stable ' gravastars may always be ensured to be of non - zero measure . nevertheless , it was also found that whenever the measure of the subspaces of the configuration space representing stable and unstable gravastars is compared the former is always found to be negligible with respect to the latter . this observation was commonly interpreted that gravastars may not offer a viable alternative to bhs . the main purpose of the present paper is to determine the response of a gravastar in equilibrium to the arrival of a dust shell onto its surface . this is done not merely by considering some formal change of the effective potential determining the state of a gravastar but also by making use of the full dynamical setup . for the sake of definiteness , we assume that the surface of the gravastar and the dust shell collide in a totally inelastic manner . in addition , concrete eoss are chosen and it is assumed that the dominant energy condition ( dec ) holds and the speed of sound in the surface of the gravastar does not exceed that of the light . then the relevant non - linear problems the basic equations of which are based on the dynamics of spherical shells are solved by using analytical and numerical approaches . in this way we can not only study the excursion of particular gravastar models but also determine the maximal mass of the dust shell colliding with the surface of the gravastar without converting the latter into a bh . this paper is organized as follows . in section [ visser - model ] some of the basics of the visser and wiltshire three - layer dynamics gravastar model are recalled using dimensionless variables . as a preparation for the aforementioned dynamical investigations a short survey of the configuration space of stable gravastars is also provided . ( although there are no completely new results in this section , we believe that this review provides a good reference frame for the results of the succeeding dynamical investigations . ) in section [ collision - section ] the dynamics of the system composed by the spherically symmetric dust shell falling onto a stable gravastar , along with their collision , is described . section [ results ] reports about our analytical and numerical results concerning the dynamics of maximally loaded gravastars , while section [ conclusions ] contains our concluding remarks . finally , in the appendix , a generic description of the totally inelastic collision of spherical shells in spherically symmetric spacetimes is provided . throughout this paper the geometrized units , with @xmath1 , are applied . throughout this paper considerations will be restricted to the three - layer spherically symmetric gravastar model of visser and wiltshire @xcite . this simplified model consists of an external schwarzschild vacuum region with the mass parameter @xmath2 , representing the total adm mass of the gravastar , an interior de sitter region with energy density @xmath3 , and an infinitesimally thin shell at the mutual boundary of the aforementioned two regions at radius @xmath4 with surface energy density @xmath5 and surface tension @xmath6 . the line element representing the interior desitter and exterior schwarzschild spacetimes is given in the form of ( [ met1 ] ) and ( [ met2 ] ) , where the pertinent forms of @xmath7 are specified in the paragraph below ( [ met2 ] ) . note that the radius is allowed to vary in time . as the shell is assumed to coincide , at any moment , with one of the @xmath8 group orbits , the quantities characterizing the shell are only the functions of the radius . to have a closed system of equations governing the evolution of the shell we also need to assume the existence of an eos . as was noted above @xmath5 and @xmath6 must be functions of the radius exclusively . thereby , in virtue of the implicit function theorem , the eos , relating the tension , @xmath6 , of the surface to the mass density , @xmath5 , of the surface of the gravastar , may always be assumed to possess the form @xmath9 . this section , besides recalling the most important equations determining the dynamics , also provides a short comprehensive review of the basic properties of the above - outlined three - layer gravastar model . in studying the gravastar model the use of the following dimensionless parameters turned out to be advantageous . they are the dimensionless radius @xmath10 , surface density @xmath11 , tension @xmath12 and the dimensionless de sitter density parameter , @xmath13 , where @xmath14 . the systems degenerates to a shell moving in a minkowski - schwarzschild spacetime . ] note that in the chosen setup positive tangential pressure does correspond to negative surface tension . to avoid the appearance of the event horizon in the schwarzschild region the inequality @xmath15 is required to be satisfied . similarly , in order to exclude the presence of a ` cosmological ' horizon in the de sitter region the inequality @xmath16 is imposed . accordingly , in case of the considered gravastar models the value of @xmath17 is restricted to the interval @xmath18 . finally note that in dynamical situations the time dependence may conveniently be expressed via the dimensionless proper time measured by observers moving radially along the world - sheet of the shell , in @xmath19 units . the basic equations determining the dynamics of the shell of the gravastar can be derived in various ways . the most widely applied method uses the thin - shell formalism of israel @xcite based on the results given by sen , lanczos , darmois @xcite . in this approach the induced metrics , @xmath20 and @xmath21 , on the mutual boundary of the two spacetime regions the inner and the outer on the two sides of the shell are assumed to coincide , i.e. the metric on the shell can be given as @xmath22 , while the discontinuity of the pertinent extrinsic curvature tensors , @xmath23 and @xmath24 , are related to the surface energy density @xmath25 via the matching condition @xmath26 by making use of the pertinent results of @xcite , along with the above - introduced parameterization , these junction conditions can be seen to take the form @xmath27 where the over - dot denotes the derivative with respect to the dimensionless proper time along the shell . as the energy density of the shell is always assumed to be non - negative , the first square root on the left - hand side of ( [ sigma - dyn ] ) must be greater than the second one , which implies that @xmath28 . note that this restriction , viewed as a condition on the value of @xmath29 , is more demanding than the previous one excluding the presence of a cosmological horizon in the inner de sitter region . in virtue of ( [ sigma - dyn ] ) it is straightforward to see that whenever @xmath30 is non - negative it attains its maximum value for vanishing @xmath31 for any fixed value of @xmath29 . in describing the dynamics of gravastars it turned out to be advantageous to introduce an ` effective potential ' defined via the relation @xmath32 assuming that @xmath30 is positive , by a straightforward algebraic manipulation , one finds that @xmath33 can be given as a function of @xmath29 and @xmath34it also depends on the dimensionless de sitter density parameter @xmath17and it reads @xmath35 \ , , \ ] ] which , in virtue of ( [ eqv])despite its complicated analytic form has to be smaller than or equal to zero . once @xmath30 is known as a function of the dimensionless radius , @xmath29 , so is @xmath36 , and , in turn , ( [ eqv ] ) can be used to determine the motion of the shell . the next subsection shows that the eos of the shell , along with the conservation law , @xmath37where @xmath38 denotes the covariant derivative operator compatible with the metric , @xmath39 , on the shell can be used to determine the functional form of @xmath40 . recall first that the eos of the surface matter of the gravastar , if given in terms of the dimensionless variables , may be assumed to possess the form @xmath41 . admittedly , the entire concept of an infinitesimally thin shell may be considered to be somewhat artificial . nevertheless , in our investigations we need to apply some assumptions on the type of its matter . the use of a suitable energy condition is essential . in this respect the use of the dec seems to be the most appropriate which is supposed to guarantee that the concept of causality is properly adopted in describing the evolution of coupled gravity matter systems . note that the less stringent weak and null energy conditions are automatically satisfied whenever the dec is guaranteed to hold . for an infinitesimally thin shell the dec can be seen to hold whenever @xmath42 @xcite . throughout this paper the function @xmath41 will be assumed to be continuous and piecewise differentiable . to avoid dynamical instabilities we shall also assume that the square of the speed of sound , @xmath43 , is non - negative , and to be compatible with the concept of relativity that @xmath44 is required to be less than or equal to the square of the speed of light . accordingly , unless otherwise stated , it will be assumed that @xmath45 . it is well - known that the radial conservation equation is @xmath46 where @xmath47 denotes the dimensionless proper time along the radial trajectories generating the shell . ( [ conserv ] ) may be derived either by referring the conservation law @xmath37 or by making use of equations ( [ sigma - dyn ] ) and ( [ theta - dyn ] ) as in @xcite . by introducing the dimensionless radius @xmath29 , instead of @xmath47 , as an independent variable ( [ conserv ] ) takes the form @xmath48 this is the very point where the eos of the shell comes into play since by substituting the relation @xmath41 into ( [ dsigma ] ) the functional form of @xmath34 can be determined . note that as @xmath49 , in virtue of ( [ dsigma ] ) , @xmath40 has to be a monotonic decreasing function . since ( [ dsigma ] ) is a separable differential equation , an implicit solution of it can be given as @xmath50 where the integration constant @xmath51 is fixed such that @xmath52 . although latter @xmath53 will label equilibrium states , it should keep in mind that @xmath53 does not have such an interpretation yet . ] to have some specific examples in later sections we shall use the following three types of eos . the simplest eos that has been used in most of the former investigations of gravastars is linear homogeneous possessing the functional form @xmath54 , which , however , is not satisfied for a generic linear functional relation . in this case @xmath55 and @xmath56 clearly , it is of interest to consider more complex systems , as well . since @xmath57 is related to the speed of sound and its value at an equilibrium point will play an important role in the stability of the gravastar , it seems to be reasonable consider the eos with varying sound speed at the equilibrium point . the simplest generalization of a homogeneous linear eos would be the non - homogeneous linear one . however , we would also like to have the dec to hold which requires @xmath58 to vanish at @xmath59 . therefore we use instead the ` broken linear ' eos @xmath60 here the parameter @xmath61 signifies the value of the breaking point in the @xmath30 range , while for the values of the slopes , @xmath62 and @xmath63 , before and after the matching the inequalities @xmath64 hold . then ( [ general - eos ] ) can be evaluated explicitly . in doing so the dimensionless radius of the matching point @xmath65 , gets to be determined by the relation @xmath66 as @xmath67 then , whenever @xmath68 @xmath69 while for @xmath70 @xmath71 another obvious choice is a polytrop eos . note , however , that in general a polytrop eos is supposed to relate the rest energy density @xmath72 , as opposed to the total energy density @xmath5 , to the pressure in which case it takes the functional form @xmath73 , with parameters @xmath74 and @xmath75 . thereby , whenever we replace @xmath72 by @xmath5 the ` inverse form ' of the polytrop eos reads as ( see , e.g. , @xcite ) @xmath76 the solution of ( [ general - eos ] ) can be given explicitly as @xmath77 where @xmath78 is related to the value of @xmath51 by ( [ polytrop ] ) . it is straightforward to check that the dec is satisfied if and only if @xmath79 $ ] . assume , now that if we are given both of the values @xmath51 and @xmath78 and the eos is chosen to possess the form ( [ polytrop ] ) then @xmath75 has to take value in the interval @xmath80 $ ] , where @xmath81 . then the square of initial speed of sound at @xmath51 can be given as @xmath82 . in closing this subsection let us emphasize again that whenever an eos , @xmath41 , is chosen @xmath30 may be given as a function @xmath34 and , in virtue of ( [ v ] ) , the effective potential can be given as an explicit function @xmath83 depending exclusively on the dimensionless radius @xmath29 and the dimensionless de sitter density parameter @xmath17 . in this subsection considerations will be restricted to the case of gravastars in equilibrium . equilibrium , as usual , signifies those configurations in rest for which forces are also balanced , i.e. @xmath84 and @xmath85 , where @xmath53 stands for the value of @xmath29 in equilibrium , while prime , @xmath86 , denotes derivative with respect to @xmath29 . as is expected the vanishing of @xmath33 and @xmath87 is equivalent to the vanishing of @xmath31 and @xmath88 , which can be verified by making use of ( [ eqv ] ) . in depicting of the configuration space of gravastars in equilibrium we shall apply the values @xmath89 and @xmath90 , which , by making use of ( [ sigma - dyn ] ) and ( [ theta - dyn ] ) , along with the vanishing of @xmath31 and @xmath88 at @xmath91 , can be given as @xmath92 and @xmath93 using the algebraic equations of ( [ sigma - stat ] ) and ( [ theta - stat ] ) and the condition that @xmath94 , it can be shown that for a gravastar in equilibrium the surface tension , @xmath95 , is always negative . this then justifies that no exotic matter ( with negative pressure ) is required to form the shell of a gravastar in equilibrium . note that whenever an eos is given only one of the values of @xmath17 and @xmath53 can be chosen freely because , in virtue of ( [ sigma - stat ] ) and ( [ theta - stat ] ) , @xmath96 has to be consistent with the eos . on the other hand whenever no definite choice is made for an eos then the configuration space of gravastars in equilibrium may be parameterized by @xmath17 and @xmath53 . in this case the value of @xmath97 gets to be determined by relations ( [ sigma - stat ] ) and ( [ theta - stat ] ) . in depicting the configuration space of gravastars in equilibrium it is useful to identify the region on the @xmath98 plain where the dec holds ( see figure [ figure2 ] ) . in doing so recall first that whenever @xmath51 is non - negative for a gravastar in equilibrium @xmath78 has to be negative . this immediately implies that the dec reduces to @xmath99 ; furthermore the upper boundary of the subregion of the @xmath98 plain where the dec is satisfied is signified by the @xmath100 curve . this curve determined by the implicit relation @xmath101can be given as @xmath102 which intersects the @xmath53-axis at the value @xmath103 ( see figure [ figure2 ] ) . consequently , whenever the dec holds @xmath53 has to be greater than or equal to @xmath104 . this value of the minimal radius immediately imposes a strict limitation on the compactness , expressed by the @xmath105 ratio , of a gravastar in equilibrium . it also follows from ( [ dec ] ) that for small @xmath17 the approximation @xmath106 holds . for comparison it is useful to recall that the value of the ratio @xmath105 is also restricted in the case of conventional star models composed by matter satisfying the dec . in the case of spherically symmetric static configurations under various assumptions concerning the energy density and pressure of the star the ratio @xmath105 has been investigated . a comprehensive summary of the pertinent results can be found in @xcite . it is shown there that if pressure is non - negative ( although it may be anisotropic ) and the dec holds , then the numerical value of the sharp upper bound for @xmath105 is @xmath107 . it is interesting that the corresponding upper bound for a gravastar in equilibrium is a little bit smaller @xmath108 . one of the readings of this finding is that whenever the dec holds the radius of the most compressed gravastar in equilibrium is definitely larger than the radius of the most compressed conventional star . this indicates then that instead of asking whether these gravastars can be distinguished from bhs , by any sort of astrophysical observations , it is more appropriate to ask whether they can be distinguished from the most compact stars . in deciding whether a gravastar is in equilibrium there has been no need to choose a specific eos . nevertheless , whenever we want to study the radial stability of these configurations we shall need to know the eos , at least , in a sufficiently small neighborhood of @xmath51 as merely the knowledge of @xmath78 and @xmath51 will not suffice then . before turning to the radial stability of gravastars there are some conceptual issues to be mentioned . in spite of the familiarity of the dynamical equation @xmath109 it is better not applying automatically the analogous one - dimensional classical mechanical arguments . the main reason is related to the fact that @xmath83 is not an external potential . in the case of the classical mechanical problem the ` total energy ' need not to be zero so we may simply either increase the kinetic energy term or change the potential energy by moving out of our system from its equilibrium . in both cases , since the potential is supposed to be intact , the total energy is changed . as opposed to this in the case of a gravastar if the initial velocity is changed the potential has also to be changed since the formal ` total energy ' needs to remain zero . the change in the initial velocity @xmath31 affects the initial surface energy density , @xmath51 , therefore the solution @xmath40 of ( [ general - eos ] ) , and , in turn , @xmath110 is also affected . this verifies that it is better to avoid thinking of @xmath110 as an external potential . in carrying out the radial stability of gravastars we shall assume that @xmath33 smoothly depends on its indicated variables . this implies then that under a sufficiently small perturbation of a gravastar in equilibrium the sign of the second derivative , @xmath111 , will be intact . in particular , since the perturbation increases the kinetic term , the potential has to become negative in a sufficiently small neighborhood of @xmath53 whence the yielded motion of the gravastar will be bounded if the value of @xmath111 is guaranteed to be positive . accordingly , a gravastar in equilibrium , with radius @xmath53 , is considered to be stable if @xmath112 . as was already mentioned , the vanishing of @xmath113 and @xmath114in the classical mechanical analogy corresponds to vanishing speed , @xmath115 , and , in turn , the forces are balanced . the equilibrium configurations can be classified according to the convex - concave character of the potential , i.e. it is reasonable to distinguish the following three categories . * _ stable _ : a gravastar in equilibrium is considered to be stable whenever small perturbations of the system yield small local oscillations around the equilibrium state . in this case the potential has to be a convex function near @xmath53 . this may happen whenever @xmath116 . ( see , e.g. , the transition from potential 1 to potential 2 in figure [ figure1 ] ) . * _ locally unstable _ : this occurs if the potential is concave ( at least from one side ) near @xmath53 . this means that small perturbations may yield non - local motions . this locally unstable situation may arise , e.g , whenever @xmath117 . ( for an illustration see graph 3 in figure [ figure1 ] , where @xmath65 takes the role of @xmath53 ) . if @xmath118 is not continuous at @xmath53 but has well - defined left- and right - hand - sided limit values the equilibrium state will be called locally unstable if either of these limit values is negative . + the vanishing of @xmath111 may also occur . in this case the sign of the least non - zero @xmath29-derivative of @xmath33 determines the character of the potential . for instance , the graph of @xmath33 may change its character from concave into convex or vice versa through an inflexion point whenever @xmath119 is the first non - zero higher order derivative . this means that the equilibrium is locally unstable in one direction . however if the first non - zero @xmath29-derivative of @xmath33 is @xmath120 , then the graph of @xmath33 may have local minimum or maximum in accordance with the sign of @xmath120 . * _ neutral _ : it may also happen that the potential is identically zero , @xmath121 , in the neighborhood of @xmath53 . in this particular case the equilibrium is called neutral ( throughout this neighborhood ) . 5.5 mm in determining the value of @xmath111 recall first that @xmath44 can be given as @xmath122 the terms @xmath123 and @xmath124 can be expressed by using equations ( [ sigma - dyn ] ) and ( [ theta - dyn ] ) . for instance , in virtue of ( [ v ] ) , and since @xmath125 and @xmath126 , we find that @xmath124 can be given in terms of @xmath127 and @xmath128 exclusively . similarly , @xmath123 can be seen to depend on @xmath127 , @xmath128 and @xmath129 exclusively , i.e. no higher order @xmath29-derivatives of @xmath127 are needed in evaluating @xmath123 . therefore , in the case of a gravastar in equilibrium , i.e. whenever @xmath130 and @xmath131 , @xmath111 can be expressed as a function of the variables @xmath17 , @xmath53 and @xmath132 . this means as was mentioned before in carrying out the stability analysis of gravastars in equilibrium only the derivative @xmath133 of the eos at @xmath51 comes into play . it can be shown that for stable configurations the square of the speed of sound is bounded from below by @xmath134 where @xmath135 is nothing but the square of the speed of sound @xmath136 at @xmath53 for configurations with @xmath137 . this , in particular , means that @xmath138 whenever @xmath139 . to see that ( [ omegamin0 ] ) holds note first that @xmath30 and @xmath58given originally as functions @xmath140 and @xmath141in virtue of the relations @xmath142 and @xmath143 can be given as the functions of the form @xmath144 and @xmath145 . then ( [ omega ] ) , along with the relations @xmath146 and @xmath147 , and the vanishing of @xmath148 and @xmath149 in the case of a gravastar in equilibrium justify the first equality in ( [ omegamin0 ] ) . the second one holds because , in virtue of ( [ sigma - dyn ] ) and ( [ sigma - stat ] ) @xmath150 , and similarly , in virtue of ( [ theta - dyn ] ) and ( [ theta - stat ] ) @xmath151 . by taking into account ( [ sigma - stat ] ) and ( [ theta - stat ] ) the minimal value of @xmath136 can be given as @xmath152 in figure [ figure2 ] the configuration space of gravastars in equilibrium is illustrated with the help of the dimensionless parameters @xmath17 and @xmath53 . in explaining the meaning of some of the characteristic curves plotted in figure [ figure2 ] ) and it shows the upper boundary of the region where dec holds . this curve meets the horizontal axis at @xmath153 and it has its global maximum at the point @xmath2 with coordinates @xmath154 . the dashed thick line represents configurations with @xmath155 and with @xmath156 . stability holds , i.e. @xmath157 with @xmath158 , to the right of this curve , while instable configurations are to the left of this curve . the thin dashed line represents configurations with @xmath155 and @xmath159 . this curve also separates stable and unstable configurations when @xmath160 is restricted to zero . the dashed thin line enters to the dec region at the point @xmath161 with coordinates @xmath162 . on the thin dashed - dot line @xmath163 and @xmath164 holds . the dashed thick line , the thin dashed - dot line and the curve determining the boundary of the dec region all intersect at @xmath2 . two subregions are indicated by gray shading . in the shaded region stable gravastars can be found within the domain where dec holds and the speed of sound at @xmath53 does not exceed the speed of light . the dark gray region represents stable gravastar configurations irrespective of the value of @xmath136 until it is non - negative . , width=377 ] note that the thin dashed , dashed - dotted and thick dashed lines do correspond to configurations for which the square of the speed of sound @xmath136 at @xmath53 takes the minimal values @xmath165 , respectively . correspondingly , the thick dashed line represents the lower bound of the radius for stable configurations for which the sound speed does not exceed the speed of light . if the dec is also required to hold , then @xmath53 has to be larger than @xmath166 which is the @xmath53 value of the configuration represented by the point @xmath2 in figure [ figure1 ] . plane . an eos may be represented by a curve on this plane . in order to provide a better orientation the @xmath167 curves for values ranging from @xmath168 to @xmath169 , with separation @xmath170 , are indicated . the styles of the sheading of the distinguished subregions and that of the characteristic curves and points are the same as in figure [ figure1 ] . , width=415 ] [ figure3 ] in figure [ figure3 ] a different representation of the very same configuration space , as depicted in figure [ figure2 ] , is given on the @xmath171 plane . the points representing gravastars in equilibrium by the values @xmath172 in figure [ figure2 ] are mapped as @xmath173 . the styles of the sheading of the distinguished subregions and those of the characteristic curves and points are in correspondence in these two figures . an interesting feature of figure [ figure3 ] is that , in virtue of the above argument in connection with @xmath135 , the slope of the tangent at a point of an @xmath167 curve is exactly the value of @xmath135 relevant for the configuration represented by the pertinent point . the interaction of a gravastar with a spherical dust shell falling onto it will be studied in this section . in advance to the arrival of the dust shell onto the surface of the gravastar the history of the shell is completely determined by its relativistic equation of motion on a schwarzschild background . to make our investigations somewhat realistic we shall assume that the collision of the dust shell and the gravastar is totally inelastic , i.e. all the dust particles will be assumed to move together with the surface layer of the yielded dynamical gravastar . in determining its basic parameters the velocity , mass density , etc.the conservation of the @xmath174-momenta will be applied . finally , by inspecting the new potential relevant for the composed system the dynamics of the yielded gravastar can be determined . inside the radius @xmath175,schwarzschild exterior with mass parameter @xmath176 gravastar shell with rest mass @xmath177 and the moving dust shell with rest mass @xmath178 are shown . right : the moment of the inelastic collision is indicated.,title="fig:",height=291 ] inside the radius @xmath175,schwarzschild exterior with mass parameter @xmath176 gravastar shell with rest mass @xmath177 and the moving dust shell with rest mass @xmath178 are shown . right : the moment of the inelastic collision is indicated.,title="fig:",height=291 ] according to the above - outlined program we shall start with a gravastar in equilibrium the total mass of which @xmath176 will be used as our reference unit . the dimensionless radius and the mass ratio parameter of this gravastar are @xmath179 and @xmath180 , respectively . we assume that a dust shell , with gravitational mass @xmath181 , falls onto the surface of the gravastar and the collision between the dust shell and the surface of the gravastar is totally inelastic . this , in particular , means that once the collision occurs a new three - layer dynamical gravastar forms . figure [ coll ] provides a simple depiction of the underlying process . ( for another schematic spacetime diagram see figure [ space - time - diagram ] . ) denote the proper mass of the initial shell of the gravastar , of the dust shell and of the shell of the gravastar after the collision by @xmath177 , @xmath178 and @xmath182 , respectively . the total mass of the system is @xmath183 . to guarantee @xmath175 to be greater than the schwarzschild radius of the event horizon of the yielded gravastar , we require the inequality @xmath184 to hold . the dynamical characterization of the dust shell in advance to the collision can be given based on the results outlined in the appendix . concerning the velocity of the particles of the dust shell at the moment of the collision it will be assumed that either it is equal to the velocity that would be acquired by the shell if it was starting to move toward the gravastar from rest at infinity , i.e. @xmath185 , or , in the second case , the particles of the dust shell will be assumed to be simply attached to the surface of the gravastar with zero velocity , i.e. @xmath186 . note that in the @xmath185 case the gravitational mass of the dust shell equals to its rest mass , i.e. @xmath187 ( see the appendix for more details ) , and also that , in virtue of ( [ rdot ] ) , the velocity @xmath188 at the moment of the collision can be given as @xmath189 the conservation of the @xmath174-energy - momentum can be used to determine the radial energy and the radial momenta of the surface of the gravastar yielded by the collision . the relevant equations are derived in the appendix where a detailed and generic investigation of an inelastic collision of two concentric spherically symmetric matter shells is also carried out . because of spherical symmetry the radial component of the 4-momentum conservation does correspond to the relativistic 3-momentum conservation given by equation ( [ appendixv3 ] ) . in virtue of this relation the radial velocity , @xmath190 , of the boundary of the yielded gravastar if @xmath191 can be given as @xmath192 where @xmath193 here @xmath194 denotes the energy density of the original gravastar in equilibrium given by ( [ sigma - stat ] ) . the temporal component of the 4-momentum conservation corresponds to the energy conservation which can be seen to be equivalent to the dynamical master equation of gravastars ( [ sigma - dyn ] ) . using the parameters of the newly formed gravastar the master equation can be given in terms of the above - introduced dimensionless parameters as @xmath195 the substitution of ( [ v3 ] ) into ( [ sigma3 ] ) then yields the following equation : @xmath196 with @xmath197 equation ( [ eq ] ) is an algebraic relation for @xmath198 possessing the only positive root @xmath199 where @xmath198 denotes the energy density of the boundary of the newly formed gravastar . note that the value of @xmath198 is independent from the eos . whenever @xmath186 the value of @xmath198 can be determined by ( [ sigma3 ] ) where the substitution @xmath200 has to be used . as discussed above once the collision occurs @xmath198 and @xmath201 can be determined . note that to be able to determine what happens after the collision we also need to know the eos of the newly formed surface . without this information we can not determine the new surface energy density as a function of the radius , which is essential in assigning the new effective potential . concerning the new eos , since the pressure of the falling dust shell is zero , it is tempting to assume that the surface tension remains intact during the process of collision . this idea could be supported by a simple - minded summation of partial pressures . however , if this happened the particles of the dust should have stopped at the surface of the gravastar without interacting with the particles forming the surface of the gravastar . this appears to be baseless , and more importantly , to be incompatible with the physical picture of inelastic collisions . to resolve the associated discrepancy , hereafter we shall assume that instead of the surface tension the eos of the boundary remains intact during the inelastic collision . this means that the dust shell ` melts ' into the matter of the surface of the gravastar . in determining the new potential , as previously , ( [ v ] ) may be applied since after the collision a modified three - layer gravastar is formed . the only distinction is that the new dynamical variables indicated by tilde have to be used in ( [ v ] ) . for the new gravastar @xmath202 holds as the mass of the outer schwarzschild region , used as a reference unit , changes from @xmath176 to @xmath183 . accordingly , the dimensionless surface density and the dimensionless radius change as @xmath203 and @xmath204 . combining all these observations the new potential may be given as @xmath205 where @xmath206 is the solution of ( [ general - eos ] ) with the initial condition @xmath207 provided that the eos remains intact as assumed . by inspection of the new effective potential the dynamics of the gravastar can be determined . if the new potential possesses the form of graph 2 of figure [ figure1 ] the motion of the gravastar surface remains bounded and we can determine the minimum and maximum value of the radius . in particular , the maximum mass which may be dropped onto the original gravastar without converting it into a bh can also be determined . first note that after collision the initial velocity points inward ( unless @xmath186 ) ; therefore , the investigation has to start by deriving whether the new effective potential has any root in the interval @xmath208 $ ] . if not then an event horizon will form and the gravastar collapses to a bh . the examination of the potential is made by a numerical maximum finding algorithm . if the maximum is found to be less than zero , the collapse to a bh will occur . using an interval bisection method , changing the initial gravitational mass of the dust shell to be dropped , the numerical value of the maximal gravitational mass , @xmath209 , which may fall onto the gravastar without converting it into a bh , can be determined . whenever @xmath191 and there is a root in the interval @xmath208 $ ] , the motion is bounded and it happens so that @xmath210 , where @xmath65 is the largest root that is smaller than @xmath53 and @xmath211 is the smallest root that is larger than @xmath53 . if @xmath186 then @xmath53 coincides with either the upper or lower bound of the motion . it is an inherent property of the gravastar model that we can not have information about the de sitter energy density inside the gravastar , @xmath17 , nor about the eos of the shell before and after the collision . to have some quantitative results suitable hypotheses have to be applied . in order to avoid the use of a flood of parameters each of which has to possess definite values in numerical investigations , as indicated in subsection [ eossubsec ] , in this paper we analyze three types of eos : homogeneous linear , broken linear and polytrop . if we fix an eos with all of its parameters , then , as a function of @xmath17 , the stable equilibrium radius , @xmath53 , is to be determined . note , that this radius can be determined by the intersection of the specific eos curve and the corresponding @xmath17-level curve in the figure [ figure3 ] . next we have to make an assumption about the eos after the collision . since we have no observational clues , as it was indicated above , we shall assume that the eos remains intact , which means that the dust shell melts into the matter of the gravastar shell yielding an increase of the pressure according to its mass contribution . note that in the case of an initially homogeneous linear eos the possibility of keeping the surface tension to be intact will also be investigated . this latter case might be a good approximation whenever the contribution of the dust shell to the new shell s composition is negligible . as was mentioned in subsection [ eossubsec ] , the simplest possible eos is the homogeneous linear one . if the value of @xmath17 and @xmath212 are given that of @xmath53 is determined by the implicit relation @xmath213 . it is clearly visible in figure [ figure3 ] that for a given value of @xmath214)$ ] the @xmath215 line intersects any @xmath17 constant curve at most at two points . one of these intersections corresponds to a stable configuration , with the smaller @xmath51 value , while the other is unstable . accordingly , once the choice for @xmath17 and @xmath212 is made first we determine the value of @xmath53 by numerically solving the implicit relation @xmath213 . then , for any choice of the pair @xmath216 , we determine the stable gravastar configuration , and as it was outlined above , the maximal mass of the shell that can be dropped onto it without converting it into a bh can also be determined . the results of the corresponding investigations are depicted by figure [ w0surf ] . it is worthy to mentioned that some of the special cases depicted on these panels play distinguished role in various arguments . for instance , the case @xmath217 corresponds to ultrarelativistic gas , which based on certain dimension reduction arguments that might not be straightforward to be applied in the present case is represented in @xcite by a system with @xmath218 . is shown as a function of @xmath17 and @xmath212 for the homogeneous linear eos , where @xmath53 stands for the maximal dimensionless radius allowed by the dec . on the top two panels @xmath185 , while on the bottom two ones @xmath186 . the panels to the left depict the intact eos cases , while in the right ones the surface tension was kept to be intact . , width=377 ] note that to have a stable configuration the value of @xmath219 has to be larger than @xmath220 given by ( [ omegamin ] ) . in particular , @xmath219 may take all the values from the interval @xmath221 $ ] only for @xmath168 , i.e. whenever the inner region is vacuum . also note that @xmath222 has to be smaller than @xmath223 . if @xmath224 occurred , the system composed of the gravastar and the dust shell would get beyond its own schwarzschild radius before the collision occurred . this makes the use of the rescalled quantity @xmath225 to be advantageous . before interpreting figures [ w0surf ] and [ v2p2 ] let us recall that a homogeneous linear eos , @xmath226 , has @xmath212 as the only parameter . therefore , it is also straightforward to investigate the case when both the surface tension and the functional form of the eos remain intact during the collision . it is done by determining the value of the new @xmath227 via the relation @xmath228 . if this happens the particles of the dust shell do not adopt the properties of the particles comprising the surface of the gravastar as only the rest mass of the surface will be increased by the collision . in figures [ w0surf ] and [ v2p2 ] four different subcases are considered . these subcases are yielded , on the one hand , by keeping the eos or the surface tension to be intact and , on the other hand , by choosing the velocity of the particles of the dust shell at the moment of the collision to be such that either @xmath185 or @xmath186 , as discussed in section [ collision - section ] . sections of the four panels of figure [ w0surf ] are shown providing a better comparison of @xmath222 for the considered configuration . the order of the curves from top to bottom coincides with that of the legend at the upper - right corner . , width=377 ] figure [ v2p2 ] depict only the @xmath100referred to as stiff matter in @xcite sections of the four cases indicated in figure [ w0surf ] which offers a better comparison of the corresponding four different dynamical subcases . figure [ bananas ] does also refer to these four basic configurations with @xmath100 . the horizontal sections , corresponding to an @xmath167 value , of the grey regions indicate the dynamical range of the radius of gravastars . for the particular value @xmath229 the pertinent new potentials are also indicated . considered in figure [ v2p2 ] . the @xmath167 sections of the gray regions represent the dynamical range of the radius relevant for the considered four subcases . the @xmath17 dependence of @xmath53 of the gravastars in advance to the collision is indicated by the dashed curves . whenever the initial velocity of the dust shell is nonzero ( e.g. , @xmath185 ) the dashed line is in the interior of the grey region , whereas whenever @xmath186see , e.g. , the right panels the gravastar stars its motion at the edge ( minimal or maximal @xmath29 value ) of the dynamical range . for the particular choice @xmath229 the corresponding potentials , are also plotted , where the zeros of the potentials are indicated by the dotted horizontal lines and the applied scales are the same on each of these panels . the dotted vertical line with the larger @xmath230 value correspond to the buchdahl limit , while the other one with @xmath231 indicates the corresponding lower bound to the size of most compact conventional stars satisfying the dec . , width=377 ] there are two interesting points to be mentioned that are indicated in figure [ bananas ] . first , there are two dotted vertical lines on each of the panels . the one with the larger @xmath29 value corresponds to the buchdahl limit with @xmath230 @xcite , while the other with a smaller @xmath29 value indicates the corresponding lower bound , with @xmath231 , to the size of most compact conventional stars . on these panels of figure [ bananas ] it is also visible that dynamical gravastars can not get below the buchdahl limit so they definitely remains less compact than the most compact conventional stars satisfying the dec . second , on the right two panels with @xmath186 the grey regions on one side are bounded by the dashed lines representing the @xmath17 dependence of @xmath53 of the initial gravastar . this is , on one hand , in accordance with intuition as the particles of the dust shell are simply simultaneously placed on the surface of the gravastar . what might be unexpected , on the other hand , is the following . whenever the eos remains intact ( see the right - top panel of figure [ bananas ] ) there is a subregion where the dashed line bounds from above and another where it bounds from below . from zero to @xmath222 first the radius increases . however , for configurations with @xmath232 there is a turning point in @xmath29 and before reaching @xmath222 the radius is smaller than the initial @xmath53 . ] these subregions are separated by a point @xmath233 where the initial gravastar remains apparently at rest for @xmath234 , while whenever the surface tension remains intact as it is indicated by the bottom - right panel of figure [ bananas] the dynamical range for @xmath234 is always bounded from above by the initial gravastar configurations . since the gravastar radius has always to decrease at the moment of the collision the surface density must increase at this moment . thus , the functional form of the new effective potential depends only on that of the eos relevant for mass density greater than @xmath51 . motivated by this observation we have used the following particular broken linear eos . @xmath194 was chosen to be the breaking point , i.e. the choice @xmath235 was made , and only the slope was varied for @xmath236 . since we have found that the higher the value of @xmath212 the more stable is a gravastar , in the broken linear eos we have chosen @xmath237 and @xmath63 to be varied form zero to 1 , in steps @xmath238 , as it is indicated on the right panel of figure [ w1dec ] . is chosen to be maximal such that the dec is still guaranteed to hold . on the graphs of the left panels the value of @xmath63 decreases form 1 to zero , in steps @xmath238 , while moving downward . , width=377 ] in figure [ w1dec ] the @xmath17 dependence of @xmath222 , i.e the maximal value of the gravitational mass of the shell that may be dropped onto a stable gravastar without converting it into a bh is shown for configurations with @xmath185 and @xmath186 . on both of the left panels the value of @xmath63 decreases from 1 to zero while moving from the top to the bottom . it is also visible that for each particular value of @xmath63 there is a maximal value of @xmath17 that is indicated by the vertical ticks on the horizontal axis of the the top - left panel . these figures do also justify for the pertinent linear approximation of the eos that for given @xmath17 and @xmath53 the larger the value of @xmath239 the more stable the gravastar becomes in the sense that the allowed value of @xmath222 increases together with @xmath240 . a polytrop eos of the form of ( [ polytrop ] ) has two parameters , @xmath74 and @xmath75 . given a point @xmath241with @xmath242it is not obvious to choose at all the parameters @xmath74 and @xmath75 to get an eos fitting to this point except for the choice @xmath243 . to overcome this slight technical difficulty and also because for a given value of @xmath74 the choice @xmath243 always yields the most stable configuration , we have chosen @xmath243 and varied only the value of @xmath74 . it is also informative to compare the behavior of a gravastar with a polytrop eos and with a linear one having the same slope at @xmath51 as it is indicated by the nested panel in figure [ kappa2 ] relevant for the particular choice @xmath243 . for any choice of @xmath17 the value of @xmath53 is chosen to be determined by the intersection of the @xmath167 line and the dashed - dotted curve in figure [ figure2 ] . note that if @xmath53 was chosen to be the maximal value for a given @xmath17 such that the dec holds , the polytrop eos would degenerate into the homogeneous linear one with @xmath100 . and the linear one , having the same slope at @xmath51 are illustrated . left : for any choice of @xmath17 the corresponding value of @xmath53 is determined by the intersection of the @xmath167 line and the dashed - dotted curve in figure [ figure2 ] . the @xmath17 dependence of @xmath225 is shown for two types of eos and for the two types of initial velocities , @xmath185 and @xmath186 . the order of the curves from top to bottom coincides with that of the legend at the upper right corner . , width=264 ] the @xmath17 dependence of the maximum of the normalized gravitational mass of the dust shell , @xmath225 , where @xmath222 is the maximal mass that may be dropped onto a gravastar without converting it into a bh is shown in figure [ kappa2 ] for two types of eos and for the two types of choices concerning the initial velocities , @xmath185 and @xmath186 . as it might have been anticipated , figure [ kappa2 ] justifies that a gravastar characterized by a polytrop eos is more stable than the corresponding one with the linear eos in consequence of the fact that the slope of the polytrop eos is larger than that of the linear one for @xmath236 . in this paper , besides providing a short survey of gravastars in equilibrium , the radial stability of gravastars was investigated . the latter was done by determining the response of a gravastar to the arrival of a dust shell onto its surface . concerning the matter model of the surface of the gravastar three different types of eos were applied . we also assumed that the dec holds and the speed of sound @xmath244 satisfies the relations @xmath245 . our most important findings are as follows . among the investigated eoss the homogeneous linear one with @xmath246 appears to provide the largest possible stability for gravastars in the sense that for a chosen @xmath17 the value of @xmath222 attains its largest possible value for @xmath246 . it would be interesting to find an analytical justification of this observation . it also follows from our investigations that once an eos is fixed , for a given value of @xmath53 , the smaller the value of @xmath17 , or alternatively , for a given value of @xmath17 , the larger the value of @xmath53 the more stable is a gravastar . the maximal mass of the shell that may be dropped onto a gravastar in equilibrium without converting it into a bh is also determined . the normalized value of this gravitational mass @xmath225 was found to be the largest @xmath247 whenever the matter comprising the surface of the gravastar possesses a homogeneous linear eos with @xmath100 and @xmath186 . correspondingly the value of @xmath248 is of the order 1 for a gravastar with radius @xmath249 which indicates that gravastars may be significantly more stable ( although not smaller ) than ordinary stars . in the appendix a generic framework describing the totally inelastic collision of two shells is also shown . the reason behind considering inelastic collisions is that the de sitter region of a gravastar is unstable against hydrodynamical excitations . thereby , it is reasonable to assume that the particles of the dust shell merge to the particles of the surface of the gravastar . in most of the considered cases , for the sake of simplicity , we also assume that the eos of the surface of the gravastar remains intact during the collision . all these assumptions indicate that there might be various generalizations of the simple model we applied . nevertheless , our findings in all the investigated cases clearly manifest that even a dynamical gravastar can not be more compact than the smallest possible ordinary stars of the same mass provided that for both types of models the dec is guaranteed to hold . this means that the main issue is not that whether gravastars can be distinguished from a bh . assuming that they exist in reality it is more appropriate to ask whether they can be distinguished from compact regular stars . let us finally turn the case of a generic bh mimicking gravastar without restricting our considerations by referring to the dec or to any particular eos of the surface layer . note that in this case the size of the gravastar is not bounded from below so , in particular , the assumption @xmath250 corresponding to the requirement of mazur and mottola may also hold . now assume that the mass of this gravastar is @xmath2 and that the gravitational mass of the dust shell falling onto it is @xmath251 . for simplicity we assume that the particles of the dust shell move toward the gravastar with velocity as if they started from rest at infinity , i.e. @xmath185 . then , the mass of the composed system is @xmath252 . in order to avoid the situation in which the composed system gets beyond its own schwarzschild radius @xmath253 , in advance to the collision , for the radius of the initial gravastar , @xmath254 , the inequality @xmath255 must hold . this immediately implies then that @xmath256 , and in turn that the closer the radius of the initial gravastar is to the schwarzschild radius the smaller the mass of the dust shell , which may be dropped onto it without converting it into a bh , must be . in other words , the better is a gravastar in mimicking a bh the easier is to destroy it by the described process . note that since this argument does not refer to the internal structure of the gravastar model it applies not only to the simplified three - layer model of visser and wiltshire or to the original model of mazur and mottola but essentially to any analogous construction . this research was supported in part by otka grant k67942 . this appendix is to introduce the basic equations governing the dynamics of a totally inelastic collision of a pair of concentric spherical shells , the outer of which is moving in a vacuum spacetime region . these equations are derived by making use of the conservation of 4-momentum . in determining them we have applied a suitable adaptation of the method of nakao , ida and sugiura which was originally worked out in @xcite to describe the conservation of 4-momentum in the special case of totally transparent shell crossings . in what follows the abstract index notation will be applied . in particular , the lowercase latin indices will indicate the type of tensors while greek indices will refer to the components of tensors with respect to some specified basis fields . to start off consider two spherically symmetric concentric shells . each of these shells is assumed to be infinitely thin and their history ( before the collision ) is assumed to be represented by separate timelike hypersurfaces . these hypersurfaces will also be referred as shells hereafter . once the collision occurs it is assumed that the initial shells merge such that a single infinitesimally thin shell is formed . as is depicted by the simple schematic picture in figure [ space - time - diagram ] the spacetime may be divided into three characteristically different regions . ( 60,60)(-30,-30 ) ( 0,0)(-1,-1)25 ( 0,0)(1,-1)25 ( 0,0)(0,1)25 ( -25,-27)(0,0)[t]shell 1 ( 25,-27)(0,0)[t]shell 2 ( 0,27)(0,0)[b]shell 3 ( -20,11)(0,0)region 1 ( 0,-18)(0,0)region 2 ( 20,11)(0,0)region 3 in advance to the collision shell 1 and shell 2 indicate the interior and exterior shells , respectively , while the shell yielded by the merging of these two shells is represented by shell 3 . the inside spacetime region which has shells 1 and 3 on its the boundary is called region 1 , the region between shells 1 and 2 is called region 2 , while the outside region bounded by shells 2 and 3 is called region 3 . for notational convenience we introduce region 4 which is identical to region 1 . hereafter we use capital latin index @xmath257 to indicate the label either of a shell or of a region . accordingly , the value of @xmath257 runs from 1 to 3 unless otherwise stated . the metric in the disjoint spherical symmetric regions , labelled by @xmath257 , will be assumed to be given as @xmath258 with @xmath259 note that this class is generic enough to include both the schwarzschild and de sitter vacuum spacetimes , but the electrovacuum reisner - nordstrm solution does also fit to this form . in restricting our attention to a system consisting of a gravastar and a spherical shell falling onto it the same system investigated in this paper in regions 2 and 3 , in virtue of birkhoff s theorem , the metric has to be ( locally ) isometric to that of the schwarzschild spacetime . accordingly , in regions 2 and 3 the metric can be given by ( [ met1 ] ) and ( [ met2 ] ) with @xmath260 , where the constants @xmath261 denote the corresponding gravitational mass parameters of the pertinent schwarzschild spacetimes . in addition , in region 1 the metric is ( locally ) isometric to the de sitter spacetime with @xmath262 , where @xmath3 is the energy density in the de sitter vacuum . in this case , @xmath263 stands for the total gravitational mass of the system while @xmath176 denotes the total gravitational mass of the entire system . note , however , that in carrying out the following calculations there is no need to make any specific choice , i.e. the metric in region 1 will only be required to possess the form ( [ met1 ] ) and ( [ met2 ] ) . this provides us the freedom that whenever the interior is chosen to be de sitter type the corresponding equations will immediately be applicable to a system formed by a shell falling onto a gravastar , while if the interior is chosen to be schwarzschild type the relevant equations describe the inelastic collision of two confocal spherical shells in a schwarzschild vacuum spacetime . in the following denote by @xmath264 the 4-velocity of shell @xmath257 . the components of the 4-velocity , with respect to the coordinates in regions bounding shell @xmath257 , can be given as @xmath265 where @xmath266 denotes the proper time on the shell @xmath257 and the suffixes @xmath267 or @xmath268 indicate whether the pertinent region is outside or inside with respect to the shell @xmath257 . note that as opposed to the area radius , @xmath269 , which is everywhere continuous the time coordinate is not continuous across a shell . once the components of @xmath264 are fixed the spacelike outward pointing unit vector , @xmath270 , normal to the shell @xmath257 can be determined by the orthogonality condition , @xmath271 , and its components can be given as @xmath272 the evaluation of ( [ uup ] ) and ( [ nup ] ) requires the determination of the radial velocity , @xmath273 . there are various ways to get the value of @xmath273 . the most widely known method is based on the thin - shell formalism of israel - sen - lanczos - darmois . for a comprehensive summary of this formalism see @xcite . after an appropriate reformulation of the ` master equation ' ( see equation ( 38 ) of @xcite ) , which is in fact an energy balance equation , @xmath273 , can be given as @xmath274 where @xmath275 denotes the rest mass , while @xmath276 stands for the gravitational mass of the shell @xmath257 . accordingly , @xmath277 for @xmath278 and @xmath279 . note that the rest mass of the shell of a gravastar , in general , is a function of the area radius , nevertheless , the rest mass of a dust shell is constant . recall that the components of the 4-velocity can not be independent as @xmath280 which , in particular , implies that @xmath281 the substitution of ( [ rdot ] ) into ( [ tdot ] ) , and some straightforward algebraic manipulations , yields that @xmath282 where @xmath283 , referred as the self - gravity of the shell @xmath257 . returning to the main line of the argument recall that our aim is to determine the motion of the shell yielded by the merging of the two initial ones . in doing so the only guiding principle we can use is the conservation of 4-momentum @xmath284 which has to be evaluated at the moment of the collision . note that the content of ( [ 4-momentum ] ) is nothing but two scalar equations . the temporal component corresponds to the energy conservation , while the radial component expresses to the radial momentum conservation . the other two components vanish identically due to the spherical symmetry . the two non - trivial equations completely determine the motion of the outgoing shell since , as we shall see below , both the radial velocity and the rest mass @xmath182 get to be fixed by them . note that since the collision is inelastic a part of the kinetic energy is converted into binding energy which , in particular , implies that @xmath285 . in applying ( [ 4-momentum ] ) each of the 4-velocities have to be expressed with respect to a common coordinate basis field . start by choosing the coordinate basis field of region 1 as our reference frame . then , in virtue of ( [ uup ] ) , the formal expressions of the components of @xmath286 and @xmath287 can immediately be determined . similarly , in virtue of ( [ uup ] ) and ( [ nup ] ) , the coordinate basis components of @xmath286 and @xmath288 of region 1 are known , therefore @xmath289 may be expressed there as the linear combination @xmath290 by applying ( [ met1 ] ) , ( [ met2 ] ) , ( [ uup ] ) and ( [ nup ] ) @xmath291 which , along with the notation @xmath292 , implies that @xmath293 by substituting these relations into ( [ u2a ] ) the components of @xmath289 in region 1 are determined . finally , by making use of these components , along with ( [ a7 ] ) , the two non - trivial equations in ( [ 4-momentum ] ) can be given as @xmath294 where @xmath295 can be eliminated from ( [ w3 ] ) by using the relation @xmath296 . we would like to emphasize that in deriving these relations no use of the eos of the shells has been made . let us finally restrict our considerations to the case of a dust shell falling onto a gravastar . since the gravastar is in equilibrium @xmath297 . thus , in virtue of ( [ p3 ] ) , we get @xmath298 whenever @xmath299 then @xmath200 also holds . whenever @xmath300 and the gravitational mass of the shell is equal to its proper mass in this case @xmath185by making use of the notation of section 3 the relations @xmath301 , @xmath302 , @xmath303 , @xmath304 and ( [ v2 ] ) can also hold . by substituting them into ( [ forgravastar ] ) the radial velocity , @xmath305 , of the shell , at the moment of the collision , can be written as @xmath306 99 kormendy j 2003 _ the stellar - dynamical search for supermassive black holes in galactic nuclei _ , carnegie observatories astrophysics series , vol . 1 : _ coevolution of black holes and galaxies _ , ed . l. c. ho ( cambridge : cambridge univ . press )

the formalism of linear response theory for a skyrme functional including spin - orbit and tensor terms is generalized to the case of infinite nuclear matter with arbitrary isospin asymmetry . response functions are obtained by solving an algebraic system of equations , which is explicitly given . spin - isospin strength functions are analyzed varying the conditions of density , momentum transfer , asymmetry and temperature . the presence of instabilities , including the spinodal one , is studied by means of the static susceptibility .

1402.4284

the energy density functional ( edf ) is a tool of choice for a systematic and quantitative description of properties of atomic nuclei from drip - line to drip - line @xcite . among the non - relativistic edf those related to the zero - range non - local skyrme interaction are the most widely used . in its standard form @xcite , skyrme s pseudo - potential contains central , spin - orbit and density - dependent terms . tensor terms , present in skyrme s initial proposal @xcite , have been included only recently @xcite , and show to play a major role in finite nuclei , both for ground state properties @xcite and excited states @xcite . the resulting edf @xcite is written as a linear combination of local densities , whose coupling constants are optimally determined by minimizing a multi - dimensional merit function for a given set of observables or pseudo - observables @xcite . this procedure is not simple and requires an accurate selection of the observables included in the fit to guarantee a proper constraint for each coupling constant . the response of atomic nuclei to different probes is the most efficient way to obtain information about the structure and specific manifestations of the effective nucleon - nucleon interaction in the nuclear medium . physical insight can be obtained from inm which , as a homogeneous medium made of interacting nucleons , is an ideal but very useful system , largely employed because of its relative simplicity . this model is not only connected with the inner part of atomic nuclei , but it is also very useful to describe some phenomena in the interior of compact stars @xcite . most of the investigations of responses have been based on random phase approximation ( rpa ) or linear response ( lr ) theory @xcite , in which excitations result from the residual particle - hole ( ph ) interaction between particles below and above the fermi level . working with a skyrme functional in infinite nuclear matter ( inm ) allows one to derive analytic expressions of the nuclear response function , thus making numerical calculations much more rapid as compared to those for finite nuclei . in this respect , several groups have investigated the properties of infinite systems trying to find additional constraints for the coupling constants of the functional . a typical example is the use of landau inequalities to avoid the appearance of instabilities in the infinite system at low momentum @xcite . the landau - migdal approach is only valid in the so - called _ long - wavelength limit _ , thus it is not able to detect eventual _ finite - size _ instabilities that occur at nonzero values of transferred momentum @xmath0 @xcite . the additional constraint coming from the calculations of the response function in inm could be thus used to avoid such kind of problems and further reduce the parameter space one need to explore during the optimization procedure @xcite . the response of atomic nuclei to different probes is the most efficient way to obtain information about the structure and specific manifestations of the effective nucleon - nucleon interaction in the nuclear medium . physical insight can be obtained from inm which , as a homogeneous medium made of interacting nucleons , is an ideal but very useful system , largely employed because of its relative simplicity . this model is not only connected with the inner part of atomic nuclei , but it is also very useful to describe some phenomena in the interior of compact stars @xcite . most of the investigations of responses have been based on random phase approximation ( rpa ) or linear response ( lr ) theory @xcite , in which excitations result from the residual particle - hole ( ph ) interaction between particles below and above the fermi level . in the present article , the edf described in @xcite will be our starting point and it will be used for the determination of spin - isospin response functions . we have decided to use the skyrme edf instead of the interaction form because in this way the formalism can be easily generalized to recent extensions of the functional @xcite . the rpa response of symmetric nuclear matter ( snm ) has been presented for the case of the standard skyrme pseudo - potential in @xcite , and extended to include spin - orbit @xcite and tensor terms @xcite . the formalism has been revised in @xcite to treat the case of a more general skyrme energy density functional ( edf ) @xcite , both for snm and pure neutron matter ( pnm ) . since neutron excess is a common situation in finite nuclei and also of current interest in astrophysical context , a generalization to asymmetric nuclear matter ( anm ) is clearly needed . the rpa response for such a system has been derived in @xcite considering only the central part of a skyrme interaction . the purpose of this work is to extend the formalism by including spin - orbit and tensor terms . we will show that , analogously to the snm case , the specific momentum dependence of the skyrme edf leads to an algebraic system of coupled equations , with however a double number of equations , from which one can deduce the rpa response function for the required spin and isospin channel . we present results for different isospin asymmetry , density and transferred momentum . thermal effects on response functions are also discussed . however , we restrict ourselves to isospin excitations induced by the operator @xmath1 , the case of charge exchange processes being left for a future study . the article is organized as follows . in sec . [ matrix : el ] we give the matrix elements in the spin - isospin space of the residual interaction between particles and holes ( ph ) obtained from a general skyrme edf . the basic formalism to obtain the rpa response function is presented in sec . [ sec : form ] . in sec . [ sec : res ] we give and discuss the main results , including the presence of instabilities . in sec . [ sec : conclusion ] we draw our conclusions . finally , the algebraic system of equations and some technical details are given in the appendices . in this section , we give explicit expressions for the matrix elements of the residual particle - hole ( ph ) interaction , @xmath2 . within the context of edf and lr theory @xcite , they can be obtained by performing a second functional derivative of the total energy as @xmath3 where @xmath4 refer to the momentum and spin and isospin projections of the corresponding particle or hole , and @xmath5 stands for the occupation number matrix . due to momentum conservation , there are only three independent momenta , which we choose as those of the holes @xmath6 and @xmath7 , and the transferred momentum @xmath8 , using the standard notation @xcite . each p - h pair must be coupled to well - defined values of spin @xmath9 and projection @xmath10 . regarding the isospin , it is more convenient in anm to work in the proton / neutron formalism . dropping momentarily the momenta and spin dependence , we thus need the matrix elements @xmath11 , where the indices @xmath12 refer to protons ( @xmath13 ) or neutrons ( @xmath14 ) . due to charge conservation , there are six matrix elements , namely @xmath15 @xmath16 the latter two matrix elements are only relevant for situations involving charge exchange reactions . they have been already discussed in @xcite , and we will not consider them in the present article . the former four matrix elements will be written as @xmath17 . actually , as @xmath18 , there are only three independent elements . to avoid repetition of indices we shall write them as @xmath19 from now on , we specialize to the skyrme functional defined in @xcite , which includes both spin - orbit and tensor terms . we can write @xmath20 where @xmath21 is the relative hole momentum , and following the notation of @xcite , we have introduced the rank-1 tensor @xmath22\,.\ ] ] the coefficients @xmath23 are combinations of the coupling constants of the functional , and are given in appendix [ app : w ] . some of them depend on the transferred momentum @xmath24 and also on the isoscalar and isovector densities , @xmath25 and @xmath26 respectively . from eq . ( [ mel : nn ] ) we observe that , analogously to the symmetric matter case @xcite , the tensor term ( second line ) acts only in the @xmath27 channel , and couples different projections @xmath10 and @xmath28 , while the spin - orbit term ( last line ) couples both spin channels . the method for calculating rpa response functions has been presented in refs . @xcite for snm . it consists in first solving the bethe - salpeter equation for the rpa p - h propagator , taking advantage of the particular form of the p - h interaction in momentum space , and afterwards averaging the p - h propagator over momenta to get the response function . some details on the method have recently been given in @xcite . here we generalize it for anm , which will be characterized by the asymmetry parameter @xmath29 , defined as the quotient between the isoscalar and isovector densities . the first required ingredient is the hartree - fock ( hf ) retarded propagator of a non - interacting p - h pair . as charge - exchange processes are not considered here , the particle and the hole in the same pair share the same isospin number @xmath30 ( either @xmath13 or @xmath14 ) . the hf p - h propagator can thus be written as @xmath31 where @xmath32 is the fermi - dirac occupation number of @xmath30-particles , given by the step function @xmath33 at zero temperature , and @xmath34 is the single - particle energy @xmath35 where @xmath36 is the mean field , excluding the @xmath37 dependence contributing to the effective mass @xmath38 , written as @xmath39 where @xmath40 are edf coupling constants . notice the use of natural units ( @xmath41 ) , as will be done along this article . from now on , vector @xmath24 will define the @xmath42-axis . the hf response function is obtained as @xmath43 where the factor 2 stands for the spin - degeneracy . in the following we will often deal with momentum averages similar to the previous one , which will be indicated within brackets : @xmath44 . notice that the hf propagator is independent of the spin - channel , and so is the response @xmath45 . the correlated rpa p - h propagator depends however on the p - h spin and isospin quantum numbers . since we do not consider charge - exchange processes we can write the bethe - salpeter equation as @xmath46 the residual interaction links two p - h pairs with quantum numbers @xmath47 and @xmath48 , and hole momenta @xmath49 and @xmath50 , respectively . we refer to ref . @xcite for more details about the adopted notation . the linear response function is obtained as @xmath51 by inspecting the p - h interaction ( [ mel : nn ] ) , one can see that a closed system of algebraic equations is obtained by multiplying eq . ( [ bseq ] ) successively with the functions @xmath52 , @xmath37 , @xmath53 , @xmath54 , and @xmath55 , and integrating over the momentum @xmath49 . as compared to the snm case , the number of equations is doubled because of the isospin indices @xmath56 . the tensor term @xmath57 is only effective on the @xmath27 channel , but it can also influence the @xmath58 channel , due to the mixing between both spin channels induced by the spin - orbit term @xmath59 . similarly to snm @xcite this coupling can be absorbed into an effective coefficient @xmath60 ( cf . appendix [ app - wp ] ) , so that we deal in practice with two separate systems for each spin channel . for fixed values of the spin quantum numbers @xmath61 there are four possible isospin combinations , namely @xmath62 . however , due to isospin properties of the residual interaction , equations can be actually decoupled in two subsystems . one of them is for the couple @xmath63 and it can be written in matrix form as @xmath64 where the column vector @xmath65 contains the unkown momentum averages of the rpa propagator , @xmath66 are square matrices which depend on the edf coupling constants and averages of the hf propagator , and @xmath67 are column vectors depending on hf averages . the explicit expressions of these quantities are given in appendix [ app : eqs ] . the other subsystem is for the couple @xmath68 , which is obtained from the previous one by simply replacing @xmath69 . the number of coupled equations for each subsystem is 6 for the channel @xmath58 , and 8 for @xmath27 . the expressions become cumbersome , preventing us to write the response function in a compact form as in snm @xcite or pnm @xcite . instead , it is preferable to numerically solve these systems in the @xmath70-space . we have nevertheless relied on the analytical systems to derive some interesting quantities , as for instance sum rules . as discussed in @xcite , the relevant spin - isospin responses are given by the combinations @xmath71 actually , instead of the response functions we deal with the corresponding strength functions , defined as @xmath72 since all physical properties are actually embedded into it . to present our results , we have chosen the asymmetry parameter values @xmath73 , and 0.5 , which roughly correspond to the isospin asymmetry of @xmath74pb and the @xmath75-equilibrium condition , respectively . to complete the discussion we will also show the extreme cases of snm ( @xmath76 ) and pnm ( @xmath77 ) . notice that only the spin channels @xmath61 are relevant in the latter case . we have performed calculations at densities 0.16 and 0.08 @xmath78 . the former value corresponds to the saturation density of snm and the bulk density of finite nuclei . the latter one will give information about the surface of nuclei or the crust of neutron stars . all the displayed results have been calculated using the skyrme interaction t44 @xcite , as in previous works on snm and pnm @xcite . in fig . [ response : q01 ] , we show the strength functions @xmath79 calculated at density @xmath80 @xmath78 for a momentum transfer @xmath81 @xmath82 . the spin channels @xmath83 are displayed in panels ( a - d ) and ( e - h ) , respectively . as a reference , the hf strength function which is independent of the spin - isospin channel has also been displayed in panels ( a - d ) . it can be seen that for the intermediate values of the asymmetry parameters ( @xmath73 and 0.50 ) the function @xmath84 is in fact the superposition of two strengths , one for protons and the other for neutrons . each one reflects the different fermi momenta and effective masses of protons and neutrons . the residual ph interaction preserves this two - peaks structure in the rpa strength , except in channel @xmath85 , which displays a single and broad resonance . the collective peak in the @xmath86 channel of snm is shifted to the low - energy region as @xmath87 increases , also reducing its height . regarding the @xmath27 channel , one can see that the collective states existing in both snm and pnm is split in two well - separated peaks . their location is nearly the same for any value of @xmath10 and @xmath88 . the tensor interaction manifests in particular as differences in the @xmath89 and 1 strength functions . these differences are more visible in the isospin @xmath90 channel and practically negligible in the @xmath91 one . strengths calculated at @xmath92 and 1.34 @xmath82 and the same density value are plotted in figs . [ response : q04 ] and [ response : q1 ] , respectively . the two - peaks structure in the @xmath27 channel washes out as the momentum transfer is increased , becoming broader and less intense . tensor effects are also magnified as @xmath0 increases , with more pronounced differences between the @xmath93 strengths , irrespective of the isospin @xmath88 . it is worth noticing the huge peak in channel @xmath94 at low values of @xmath95 for all values of @xmath87 . as will be discussed in subsection [ instab ] , this peak evolves to an instability at zero energy @xcite . consider now the lower density @xmath96 @xmath78 . [ response : q0108 ] and [ response : q108 ] display the strengths calculated at the same transferred momenta than figs . [ response : q01 ] and [ response : q1 ] , respectively . a glance to the employed energy scales suffices to notice that the lowering of the density induces a global shift of the strength towards the low - energy region . for the transferred momentum @xmath81 @xmath82 the strengths in the different @xmath27 channels are nearly superimposed . for the asymmetries @xmath97 and 1 , a collective state is clearly visible in the channel @xmath90 , well separated from the continuum edge . analogously to figs . [ response : q01 ] and [ response : q1 ] , the strength function spreads as @xmath0 increases . however , an increase of the @xmath85 strength at low energies is noticeable . actually , as we shall discuss in subsection [ instab ] , this huge peak is related to the spinodal instability . as already discussed in ref . @xcite a very efficient tool to check the calculated linear responses is provided by their energy weighted sum rules . the sum rule or order @xmath13 is defined as @xmath98 and can be numerically calculated from this expression . alternatively , odd - order sum rules can also be obtained from appropriate expansions of response functions in power series of the transferred energy @xcite as @xmath99 the linear and cubic energy - weighted sum rules @xmath100 and @xmath101 can be easily obtained from the first two terms of ( [ mmp : analytic ] ) , while the inverse energy - weighted sum rule @xmath102 is the first term of ( [ mm1:analytic ] ) . we have derived these analytic expressions using symbolic programs , but the expressions are too long to be presented here . for @xmath103 , we have used the independent check of double - commutator technique @xcite to properly verify our results . our aim in comparing numerical and analytical sum rules is to have a good test about the reliability of our calculations . besides , this comparison provides an independent way to localize collective states . the @xmath102 sum rule is also interesting by itself because , apart from a trivial factor , it is the static susceptibility , or response function at zero energy . it is thus a very efficient tool to detect the instabilities related to a zero - energy mode , as already found in the case of snm and pnm in @xcite . such a mode corresponds to a solution of @xmath104 , and has an infinite strength . equivalently , these modes can be seen as divergencies in the sum rule @xmath102 . as an illustrative example , in fig . [ mm1:t44 ] , we present the results for the asymmetry parameter @xmath97 . numerical ( [ mmp : numeric ] ) and analytical ( [ mm1:analytic ] ) sum rules @xmath102 have been plotted as a function of @xmath0 . for the spin @xmath58 channel the two curves stay on top of each other showing that the calculations are reliable . both curves also coincide for @xmath27 , except near some particular value of @xmath0 , where a singularity appears . this kind of singularity will be discussed in subsection [ instab ] . we now consider the effect of temperature in the rpa response function . we refer the reader to ref . @xcite for a study about the inclusion of temperature in the many - body problem . thermal effects in the rpa strength for snm have been discussed in @xcite in the context of skyrme interactions . in practice one has to modify the occupation probability of levels as @xmath105 where @xmath106 is the chemical potential and @xmath34 is the hf single particle energy . the generalization to finite temperature of the bethe - salpeter equation ( [ bseq ] ) and the related algebraic systems of equations can be easily done . from a technical viewpoint one has simply to include it in the auxiliary functions @xmath107 defined in appendix [ app : eqs ] . at temperature @xmath108 , the system is initially in the ground state , and the sole possible effect of the external probe is to excite the system , _ i.e. _ @xmath109 . however at non - null temperature the ground state of the system at equilibrium corresponds to a statistical mixture of states . using the detailed - balance theorem , the strength function is properly defined as @xmath110 in that case , it is possible to transfer energy from the system to the probe , so that negative values of @xmath95 are admissible . actually , thermal effects become relevant at values of temperatures larger than @xmath111 mev , and the limit of evaporation in finite nuclei appears at @xmath112 - 10 mev @xcite . as a case of study we have performed calculations at @xmath113 mev , a temperature which can be relevant for homogeneous systems of astrophysical interest . in fig . [ response : tt ] are displayed the strength functions calculated at density @xmath114 @xmath78 and asymmetry parameter @xmath97 , for two values of the transferred momentum , namely @xmath81 @xmath82 ( left panel ) and 1.34 @xmath82 ( right panel ) . these responses should be compared to those in panels ( c ) and ( g ) of figs . [ response : q01 ] and [ response : q1 ] , respectively . as a rule , temperature tends to wash out the structure of the response and spread its strength . an important part of the strength is shifted to the negative energy region , and the strength in some channels largely increases at zero energy . notice also that each peak at positive energies has a corresponding `` image '' at negative energies , which corresponds to the deexcitation of the heated system . for this specific value of density , the strength in channels @xmath115 develops a huge zero - energy peak . actually , it is the precursor of an unphysical instability , as we will immediately discuss . in the previous subsections we have encountered some cases were the strength function is hugely peaked at zero energy . this peak will become a divergence for specific values of density and momentum transfer . some of these instabilities are unphysical , as they are simply reflecting drawbacks of the employed interaction @xcite . but at low values of density , the instabilities are related to the physical phenomenon of spinodal transition . as mentioned previously , the inverse - energy - weighted sum rule @xmath102 is the tool of choice for the detection of poles of response functions at zero - transferred energy , and was employed in refs @xcite to analyze instabilities in snm and pnm . let us consider first the instabilities appearing in the channel @xmath116 at low values of density . they are related to the thermodynamic spinodal transition of homogenous matter , where density fluctuations induce a decrease of the total free energy and are thus amplified until a separation in two distinct stable phases , liquid and gas , is reached . it has been shown @xcite that the curvature of the free energy contains two terms : one proportional to @xmath117 coming from density gradient terms of the energy functional , and another proportional to @xmath118 related to the coulomb interaction . for small values of @xmath0 the energy cost due to coulomb interaction dominates and implies a sensitive reduction of the spinodal region for proton - rich systems . note that in the case of a neutron star where a background of electrons is present , the net effect of the coulomb interaction in the residual interaction is to reduce the region of the spinodal as well @xcite . in fig . [ spinodal : t44 ] , we have plotted the spinodal contours in the plane @xmath119 of proton / neutron densities for different values of the transferred momentum @xmath0 at zero temperature ( left panel ) and @xmath120 mev ( right panel ) . the homogeneous system is unstable inside these contours . since coulomb interaction has been ignored in our description , the spinodal region is symmetric with respect to the line @xmath121 , and is reduced as @xmath0 is increased , the largest possible region being that at @xmath122 . in general , the effect of temperature is to suppress fluctuations . this can be seen on fig . [ spinodal : t44 ] by comparing the spinodal contours in both panels for a given value of @xmath0 . finally one may wonder about the influence of the tensor interaction on these instabilities : as we have mentioned in sect . [ sec : form ] , although the tensor acts directly only in the @xmath27 channel , it can influence also the @xmath58 channel , due to the mixing between both spin channels induced by the spin - orbit term . this coupling has been absorbed into effective coefficients @xmath60 defined in appendix [ app - wp ] ) . we have found that spin - orbit effects are very small , except at large values of the transferred momentum , as reflected in the @xmath123 power entering explicitly in @xmath60 . as the spinodal instability concerns the @xmath58 channel and implies relatively small values of the transferred momentum , it is very marginally affected by the tensor interaction , which can be safely neglected to analyze the spinodal region . we turn now to the unphysical instabilities , occurring in the @xmath27 channel . they have been analyzed in snm and pnm in terms of landau parameters associated to a specific skyrme interaction @xcite . as these instabilities are related to a specific interaction , fixing them will be of great help to establish bounds for the edf coupling constants . for a given value of the transferred momentum @xmath0 we define a critical density @xmath124 as the value beyond which the homogeneous system becomes unstable . in fig . [ s1:critical : t44 ] are plotted the values of @xmath124 as a function of @xmath0 for @xmath27 channels and asymmetry parameters varying from @xmath76 ( snm ) to @xmath77 ( pnm ) in steps of 0.2 . actually , for pnm there are only two channels , @xmath125 , and the same curve is plotted in panels @xmath91 and @xmath90 . one can see that for values of @xmath126 less that @xmath127 @xmath82 , the critical density is higher than the saturation density of snm ( in the figure , the horizontal line corresponds to the value 0.16 @xmath78 ) . the opposite happens for higher values of the transferred momentum . as a general trend , we can see that as a function of @xmath87 , the critical densities monotonically varies from a maximum value in snm to a minimum value for pnm , with small variations around @xmath128 @xmath82 . however the decrease is not linear and might be actually very pronounced as soon as a small asymmetry is introduced . we should keep in mind that all the curves are actually obtained from a combination of response functions in the isospin channels @xmath129 , @xmath130 , @xmath131 , and @xmath132 . therefore , when a small asymmetry is introduced some of these response functions may suddenly acquire a non - zero value and influence a lot the whole combination . although the calculations have been done with the single t44 interaction , one can extract a pragmatic criterion : to use the bounds imposed by both snm and pnm critical densities into a fitting procedure in order to get a stable interaction below such densities for all values of symmetries . we have also checked that the critical densities are reduced by only a few per cent at temperature @xmath113 mev . in this work we have generalized the formalism presented in refs . @xcite to calculate rpa response functions in asymmetric nuclear matter for a general skyrme energy density functional including spin - orbit and tensor terms . the responses are obtained by solving closed algebraic systems of equations , which have been explicitly presented . analytical expressions for the energy - weighted sum rules @xmath102 , @xmath100 , and @xmath101 in terms of interaction parameters are easily derived from them by using an algebraic code . however , since the number of equations is doubled as compared to the snm and pnm cases , the analytical expressions are rather cumbersome and it is preferable to compute them numerically . we have investigated the response functions in the different spin - isospin channels for specific conditions . to illustrate the general trend concerning the effect of asymmetry we have chosen four values of the parameter @xmath87 . they correspond to the @xmath74pb nucleus , asymmetric nuclear matter in @xmath75-equilibrium , snm and pnm . we have chosen two values of the density characteristic of homogeneous matter at saturation and the surface of atomic nuclei . all calculations have been done using the t44 functional as a typical one , and so our conclusions are not completely general . however , it seems that the behavior of anm is not just a simple interpolation between snm and pnm even if one has clearly identified how to go from one limiting case to the other in an analytical way . we have found that the tensor plays an important role in the response , leading to the presence of two well - defined collective states . in particular , as compared to the hf response , the tensor significantly amplify the separation between the two fermi surfaces . even small values of asymmetry have a sensible effect on response functions . this is at variance with the results of ref . @xcite where it was concluded that varying the asymmetry parameter produces no spectacular variations in the response . however , these results were obtained using purely central forces , while important tensor effects have already been observed in our previous studies concerning snm and pnm . concerning thermal effects , we have shown that up to temperatures of 16 mev , the response function is not sizably modified . the static susceptibility is a key quantity to get some information about instabilities , and we have computed it through the related inverse energy - weighted sum rule . our results concerning the spinodal instability compare favorably with those of @xcite . the general behavior obtained here for the spinodal region with respect to asymmetry and temperature reproduce nicely the physical features demonstrated in a very different context , as is the study of free energy curvature in fluctuation space @xcite . apart from this physical instability , we have also put into evidence the unphysical instabilities , thus generalizing to anm the findings of @xcite . however it is not so simple to define finite - size instabilities and critical densities . depending on the channel , the asymmetry induces an important modification of the location of the poles or a minor one . moreover , the lowest pole is not obtained for the same asymmetry , even if pnm was generally favored . as a practical rule , the bounds imposed by pnm critical densities are sufficient as a criterion to get stable interactions within a standard fitting procedure . we would like to thank c. ducoin and j. meyer for interesting discussions and remarks about this paper . this work was partly supported by grant fis2011 - 28617-c02 - 02 , mineco ( spain ) . we give here the complete expressions of the coefficients @xmath133 in terms of the coupling constant of the functional @xcite . we distinguish the cases @xmath134 ( i.e. both indices equal to @xmath13 or @xmath14 ) and @xmath135 ( i.e. indices equal to @xmath136 or @xmath137 ) . @xmath138 with @xmath139 and @xmath140 . for the remaining indices ( @xmath141 ) we have @xmath142 the coefficients @xmath143 are @xmath144q^{2}\\ \frac{1}{4}w^{(0,1)}_1&=&2c^{\rho 0}_1 + 2c^{\rho,\gamma_{}}_1\rho_0^{\gamma _ { } } -\left[2 c_1^{\delta \rho}+\frac{1}{2}c_1^{\tau}\right]q^{2}\,,\\ \frac{1}{4}w^{(1,0)}_1&= & 2c_0^{s,0}+2c_0^{s\gamma_{}}\rho_0^{\gamma _ { } } -\left[2c_0^{\delta s}+\frac{1}{2}c_0^{t } \right]q^{2}\,,\\ \frac{1}{4}w^{(1,1)}_1&= & 2c_1^{s,0}+2c_1^{s\gamma_{}}\rho_0^{\gamma_{}}- \left[2c_1^{\delta s}+\frac{1}{2}c_1^{t } \right ] q^{2}\,,\\ \frac{1}{4}w^{(0,0)}_2&=&c_0^{\tau}\ , , \\ \frac{1}{4}w^{(0,1)}_2&=&c_1^{\tau}\,,\\ \frac{1}{4}w^{(1,0)}_2&=&c_0^{t}\,,\\ \frac{1}{4}w^{(1,1)}_2&=&c_1^{t}\,.\end{aligned}\ ] ] for vanishing isovector density , they coincide with those previously given for symmetric nuclear matter @xcite . the remaining coefficients are related to the tensor and spin - orbit components of the interaction @xmath145 we also define for convenience : @xmath146 and @xmath147 . before giving the explicit expressions we have to define some averages of the hf ph propagator @xmath148 where the functions @xmath149 are defined as @xmath150^{2},\frac{(\mathbf{k}\cdot q)k^{2}}{q^{4}},\frac{k^{4}}{q^{4}},\left[\frac{\mathbf{k}\cdot q}{q^{}}\right]^{3},\left[\frac{\mathbf{k}\cdot q}{q}\right]^{4},\frac{(\mathbf{k}\cdot q)^{2}k^{2}}{q^{4}}\,,\nonumber\ ] ] for @xmath151 to 8 , respectively . the explicit expressions of these functions can be found in ref.@xcite up to @xmath152 . the three remaining functions ( for @xmath153 ) are required by the inclusion of tensor terms . its generalization is straightforward and will not be given here . obviously , for the extreme values @xmath76 and @xmath77 one recovers the expressions of the @xmath75 functions given in @xcite . for finite temperature , only the imaginary parts are determined explicitly . the real parts are obtained as usual through a dispersion relation . we give now the explicit form of the column vectors @xmath154 , @xmath155 , @xmath156 and the matrices @xmath157 , @xmath158 . the blocks @xmath159 and @xmath160 are obtained from @xmath157 and @xmath158 by simply exchanging @xmath69 . these quantities define completely the system ( [ pepet ] ) and obtain @xmath161 and @xmath162 . to determine @xmath163 and @xmath164 one must replace @xmath165 in the previous expressions . the coefficients @xmath166 and @xmath167 entering the following expressions are defined in appendix [ app - wp ] . @xmath168 @xmath169 @xmath170 @xmath172 @xmath173 @xmath174 @xmath175 @xmath176 @xmath177 the matrices of the previous appendix contain some short - hand notations which are defined in the following , channel by channel . we remind that coefficients @xmath178 are defined after partly solving the system of equations for the coupling between spin channels induced by the spin - orbit interaction . @xmath179^{2}(\beta_{2}^{\tau}-\beta_{3}^{\tau } ) ( 1+x_{-\tau-\tau})}{(1+x_{\tau\tau})({1+x_{-\tau-\tau}})-{x_{\tau-\tau}x_{-\tau\tau}}}\nonumber\\ & -&4q^4 \frac{c_{+}^{\nabla j}c_{-}^{\nabla j}{(\beta_{2}^{\tau}-\beta_{3}^{\tau})x_{-\tau\tau}}-[c_{-}^{\nabla j}]^{2}(\beta_{2}^{-\tau}-\beta_{3}^{-\tau } ) ( { 1+x_{-\tau-\tau } } ) } { ( 1+x_{\tau\tau})({1+x_{-\tau-\tau}})-{x_{\tau-\tau}x_{-\tau\tau } } } \ , , \\ % { \widetilde{w}}_{1}^{(\tau-\tau;00)}&=&w_{1}^{(\tau-\tau,0)}- 4q^4 \frac{[c_{+}^{\nabla j}]^{2}{(\beta_{2}^{-\tau}-\beta_{3}^{-\tau})x_{\tau-\tau}}-c_{-}^{\nabla j}c_{+}^{\nabla j}(\beta_{2}^{\tau}-\beta_{3}^{\tau } ) ( { 1+x_{-\tau-\tau } } ) } { ( 1+x_{\tau\tau})({1+x_{-\tau-\tau}})-{x_{\tau-\tau}x_{-\tau\tau}}}\nonumber\\ & - & 4q^4 \frac{[c_{-}^{\nabla j}]^{2}{(\beta_{2}^{\tau}-\beta_{3}^{\tau})x_{-\tau\tau}}-c_{-}^{\nabla j}c_{+}^{\nabla j}(\beta_{2}^{-\tau}-\beta_{3}^{-\tau})({1+x_{\tau\tau } } ) } { ( 1+x_{\tau\tau})({1+x_{-\tau-\tau}})-{x_{\tau-\tau}x_{-\tau\tau } } } \ , , \end{aligned}\ ] ] @xmath182\nonumber\\ & -&2q^{4}z^{0}c^{\nabla j}_{-}\left [ z^{-\tau \tau,0}c^{\nabla j}_{+}(\beta_{2}^{\tau}-\beta_{3}^{\tau})-(1+z^{\tau \tau,0})c^{\nabla j}_{-}(\beta_{2}^{-\tau}-\beta_{3}^{-\tau})\right]\nonumber\\ & + & [ c^{f}_{+}]^{2}q^{4}(\beta_{5}^{\tau}-\beta_{7}^{\tau})+[c^{f}_{-}]^{2}q^{4}(\beta_{5}^{-\tau}-\beta_{7}^{-\tau})\nonumber\\ & -&q^{2}\left ( \frac{\nu_{\tau}}{k_{\tau}}-1\right)\left [ c^{f}_{+}(z^{\tau \tau,1}b^{-\tau \tau}_{+}+z^{\tau -\tau,1}b^{\tau -\tau}_{-})\right]-q^{2}\left ( \frac{\nu_{-\tau}}{k_{-\tau}}-1\right)\left[c^{f}_{-}(z^{-\tau -\tau,1}b^{\tau -\tau}_{-}+z^{-\tau \tau,1}b^{-\tau \tau}_{+})\right]\,,\\ % \widetilde{w}_{1}^{(\tau -\tau,11)}&=&w_{1}^{(\tau -\tau,1)}-2q^{4}z^{0}c^{\nabla j}_{+}\left [ z^{\tau -\tau,0}c^{\nabla j}_{+}(\beta_{2}^{-\tau}-\beta_{3}^{-\tau})-(1+z^{-\tau -\tau,0})c^{\nabla j}_{-}(\beta_{2}^{\tau}-\beta_{3}^{\tau})\right]\nonumber\\ & -&2q^{4}z^{0}c^{\nabla j}_{-}\left [ z^{-\tau\tau,0}c^{\nabla j}_{-}(\beta_{2}^{\tau}-\beta_{3}^{\tau})-(1+z^{\tau\tau,0})c^{\nabla j}_{+}(\beta_{2}^{-\tau}-\beta_{3}^{-\tau})\right]\nonumber\\ & + & c^{f}_{+}c^{f}_{-}q^{4}\left [ ( \beta_{5}^{\tau}-\beta_{7}^{\tau})+(\beta_{5}^{-\tau}-\beta_{7}^{-\tau } ) \right]\nonumber\\ & -&q^{2}\left ( \frac{\nu_{\tau}}{k_{\tau}}-1\right)\left [ c^{f}_{+}(z^{\tau\tau,1}b^{-\tau\tau}_{-}+z^{\tau-\tau,1}b^{\tau-\tau}_{+})\right]-q^{2}\left ( \frac{\nu_{-\tau}}{k_{-\tau}}-1\right)\left[c^{f}_{-}(z^{-\tau-\tau,1}b^{\tau-\tau}_{+}+z^{-\tau\tau,1}b^{-\tau\tau}_{-})\right]\,,\end{aligned}\ ] ] @xmath183 is defined as the coefficient in front @xmath184 for the unknown parameter @xmath185 for the first four equations . they read @xmath186\,,\\ \alpha_1^{1p } & = & 2q^{2}\left[c^{f}_{+}b^{np}_{-}+c^{f}_{-}b^{pn}_{+ } \right]\,,\\ % \alpha_3^{1n } & = & 2q(a^{pn}_{+}c^{f}_{+}+c^{f}_{-}a^{np}_{-})-2q w_{2}^{(nn,1)}\ , , \\ \alpha_3^{1p } & = & 2q(a^{pn}_{+}c^{f}_{-}+c^{f}_{+}a^{np}_{-})-2q w_{2}^{(pn,1)}\ , , \\ % \alpha_3^{0n } & = & q \left(\frac{\nu_{n}}{k_{n}}-1\right ) \left[c^{f}_{+}(z^{nn,1}a^{pn}_{+}+z^{np,1}a^{np}_{-})\right]+q \left(\frac{\nu_{p}}{k_{p}}-1\right)\left[c^{f}_{-}(z^{pp,1}a^{np}_{-}+z^{pn,1}a^{pn}_{+ } ) \right]\nonumber\\ & - & 2 q^{3}\left ( [ c^{f}_{+}]^{2}(\beta_{4}^{n}-\beta_{6}^{n})+[c^{f}_{-}]^{2}(\beta_{4}^{p}-\beta_{6}^{p})\right ) \,,\\ \alpha_3^{0p } & = & q \left(\frac{\nu_{p}}{k_{p}}-1\right ) \left[c^{f}_{+}(z^{pp,1}a^{np}_{-}+z^{pn,1}a^{pn}_{+})\right]+q \left(\frac{\nu_{n}}{k_{n}}-1\right)\left[c^{f}_{-}(z^{nn,1}a^{pn}_{+}+z^{np,1}a^{np}_{- } ) \right]\nonumber\\ & - & 2 q^{3 } c^{f}_{+}c_{-}^{f}\left [ ( \beta_{4}^{n}-\beta_{6}^{n})+ ( \beta_{4}^{p}-\beta_{6}^{p})\right ] \,,\\ % \alpha_5^{1n } & = & 2q^{2}\left [ c^{f}_{+}b^{pn}_{-}+c^{f}_{-}b^{np}_{+}\right]\,,\\ \alpha_5^{1p } & = & 2q^{2}\left[c^{f}_{+}b^{np}_{+}+c^{f}_{-}b^{pn}_{- } \right]\,,\\ % \alpha_7^{1n } & = & 2q(a^{np}_{+}c^{f}_{-}+a^{pn}_{-}c^{f}_{+})-2q w_{2}^{(np,1)}\,,\\ \alpha_7^{1p } & = & 2q(a^{np}_{+}c^{f}_{+}+c^{f}_{-}a^{pn}_{-})-2q w_{2}^{(pp,1 ) } \,,\\ % \alpha_7^{0n } & = & q \left(\frac{\nu_{n}}{k_{n } } -1\right)\left [ c^{f}_{+}(z^{nn,1}a^{pn}_{-}+z^{np,1}a^{np}_{+})\right]+q \left(\frac{\nu_{p}}{k_{p } } -1\right)\left[c^{f}_{-}(z^{pp,1}a^{np}_{+}+z^{pn,1}a^{pn}_{-})\right]\nonumber\\ & -&2c^{f}_{+}c^{f}_{-}q^{3}\left [ ( \beta_{4}^{n}-\beta_{6}^{n})+(\beta_{4}^{p}-\beta_{6}^{p})\right]\,,\\ \alpha_7^{0p } & = & q \left(\frac{\nu_{p}}{k_{p}}-1\right ) \left[c^{f}_{+}(z^{pp,1}a^{np}_{+}+z^{pn,1}a^{pn}_{-})\right]+q \left(\frac{\nu_{n}}{k_{n } } -1\right)\left[c^{f}_{-}(z^{nn,1}a^{pn}_{-}+z^{np,1}a^{np}_{+ } ) \right]\nonumber\\ & - & 2 q^{3}\left ( [ c^{f}_{+}]^{2}(\beta_{4}^{p}-\beta_{6}^{p})+[c^{f}_{-}]^{2}(\beta_{4}^{n}-\beta_{6}^{n})\right)\,.\end{aligned}\ ] ] in these expressions @xmath187^{-1}.\end{aligned}\ ] ] @xmath188\,,\\ b^{\tau \tau'}_{\pm}&=&q^{2}z^{1}\left [ ( 1+z^{\tau \tau,1})c^{f}_{\pm}(\beta_{4}^{\tau'}-\beta_{6}^{\tau'})-c^{f}_{\mp } z^{\tau ' \tau,1}(\beta_{4}^{\tau}-\beta_{6}^{\tau})\right]\,.\end{aligned}\ ] ] @xmath191\\ & -&q^{2}c^{f}_{-}\left ( \frac{\nu_{-\tau}}{k_{-\tau}}-1\right)\left[\bar{b}^{-\tau\tau}_{+}(z^{-\tau\tau,1}+3s^{-\tau}_{-})+\bar{b}^{\tau-\tau}_{-}(z^{-\tau-\tau,1}+3s^{-\tau}_{+ } ) \right]\\ & + & 4q^{4}\left[(c^{f}_{+})^{2}(\beta_{8}^{\tau}-\beta_{7}^{\tau})+(c^{f}_{-})^{2}(\beta_{8}^{-\tau}-\beta_{7}^{-\tau } ) \right]\ , , \nonumber\\ % \widetilde{w}^{(\tau-\tau,10)}_{1}&=&w_{1}^{(\tau-\tau,1)}-q^{2}c^{f}_{+}\left ( \frac{\nu_{\tau}}{k_{\tau}}-1\right)\left[\bar{b}^{\tau-\tau}_{+}(z^{\tau-\tau,1}+3s^{\tau}_{-})+\bar{b}^{-\tau\tau}_{-}(z^{\tau\tau,1}+3s^{\tau}_{+ } ) \right]\\ & -&q^{2}c^{f}_{-}\left ( \frac{\nu_{-\tau}}{k_{-\tau}}-1\right)\left[\bar{b}^{\tau-\tau}_{+}(z^{-\tau-\tau,1}+3s^{-\tau}_{+})+\bar{b}^{-\tau\tau}_{-}(z^{-\tau\tau,1}+3s^{-\tau}_{- } ) \right]\\ & + & 4q^{4}c^{f}_{+}c^{f}_{-}\left[(\beta_{8}^{\tau}-\beta_{7}^{\tau})+(\beta_{8}^{-\tau}-\beta_{7}^{-\tau } ) \right]\ , .\nonumber\end{aligned}\ ] ] @xmath193\ , , \\ \alpha^{1p}_{1}&=&2q^{2}\left [ c^{f}_{+}\bar{b}^{np}_{-}+c^{f}_{-}\bar{b}^{pn}_{+}\right]\ , , \\ % \alpha^{1n}_{3}&=&2q\left[-w_{2}^{nn,1}-2c^{f}_{+}+2\bar{a}^{pn}_{+}c^{f}_{+}+2\bar{a}^{np}_{-}c^{f}_{-}\right]\ , , \\ \alpha^{1p}_{3}&=&2q\left [ -w_{2}^{pn,1}-2c^{f}_{-}+2c^{f}_{+}\bar{a}^{np}_{-}+2c^{f}_{-}\bar{a}^{pn}_{+}\right]\ , , \\ % \alpha^{0n}_{3}&=&2q\left(\frac{\nu_{n}}{k_{n}}-1\right)c^{f}_{+}\left [ \bar{a}^{pn}_{+}(z^{nn,1}+3s^{n}_{+})+\bar{a}^{np}_{-}(z^{np,1}+3s^{n}_{-})-s^{n}_{+}\right]\nonumber\\ & + & 2q\left(\frac{\nu_{p}}{k_{p}}-1\right)c^{f}_{-}\left [ \bar{a}^{np}_{-}(z^{pp,1}+3s^{p}_{+})+\bar{a}^{pn}_{+}(z^{pn,1}+3s^{p}_{-})-s^{p}_{-}\right]\ , , \\ \alpha^{0p}_{3}&= & 2q\left ( \frac{\nu_{p}}{k_{p}}-1\right)c^{f}_{+}\left[\bar{a}^{np}_{-}(z^{pp,1}+3s^{p}_{+})+\bar{a}^{pn}_{+}(z^{pn,1}+3s^{p}_{- } ) -s^{p}_{-}\right]\nonumber\\ & + & 2q\left ( \frac{\nu_{n}}{k_{n}}-1\right)c^{f}_{-}\left[\bar{a}^{pn}_{+}(z^{nn,1}+3s^{n}_{+})+\bar{a}^{np}_{-}(z^{np,1}+3s^{n}_{- } ) -s^{n}_{+}\right]\ , , \\ % \alpha^{1n}_{5}&=&2q^{2}\left [ c^{f}_{+}\bar{b}^{pn}_{-}+c^{f}_{-}\bar{b}^{np}_{+}\right]\ , , \\ \alpha^{1p}_{5}&=&2q^{2}\left [ c^{f}_{+}\bar{b}^{np}_{+}+c^{f}_{-}\bar{b}^{pn}_{- } \right]\ , , \\ % \alpha^{1n}_{7}&=&2q\left [ -w_{2}^{np,1}-2c^{f}_{-}+2c^{f}_{+}\bar{a}^{pn}_{-}+2c^{f}_{-}\bar{a}^{np}_{+}\right ] \ , , \\ \alpha^{1p}_{7}&=&2q\left[-w_{2}^{pp,1}-2c^{f}_{+}+2\bar{a}^{np}_{+}c^{f}_{+}+2\bar{a}^{pn}_{-}c^{f}_{-}\right ] \ , , \\ % \alpha^{0n}_{7}&=&2q\left ( \frac{\nu_{n}}{k_{n}}-1\right)c^{f}_{+}\left[\bar{a}^{pn}_{-}(z^{nn,1}+3s^{n}_{+})+\bar{a}^{np}_{+}(z^{np,1}+3s^{n}_{- } ) -s^{n}_{-}\right]\nonumber\\ & + & 2q\left ( \frac{\nu_{p}}{k_{p}}-1\right)c^{f}_{-}\left[\bar{a}^{np}_{+}(z^{pp,1}+3s^{p}_{+})+\bar{a}^{pn}_{-}(z^{pn,1}+3s^{p}_{- } ) -s^{p}_{+}\right]\ , , \\ \alpha^{0p}_{7}&=&2q\left(\frac{\nu_{p}}{k_{p}}-1\right)c^{f}_{+}\left [ \bar{a}^{np}_{+}(z^{pp,1}+3s^{p}_{+})+\bar{a}^{pn}_{-}(z^{pn,1}+3s^{p}_{-})-s^{p}_{+}\right]\nonumber\\ & + & 2q\left(\frac{\nu_{n}}{k_{n}}-1\right)c^{f}_{-}\left [ \bar{a}^{pn}_{-}(z^{nn,1}+3s^{n}_{+})+\bar{a}^{np}_{+}(z^{np,1}+3s^{n}_{-})-s^{n}_{-}\right]\ , .\end{aligned}\ ] ] @xmath195\ , , \\ \bar{b}^{\tau\tau'}_{+}&=&z'\left[\left(\frac{\nu_{\tau'}}{k_{\tau'}}-1\right)(1+z^{\tau\tau,1}+3s^{\tau}_{+})s^{\tau'}_{+}-\left(\frac{\nu_{\tau}}{k_{\tau}}-1\right)(z^{\tau'\tau,1}+3s^{\tau'}_{-})s^{\tau}_{- } \right]\ , .\end{aligned}\ ] ] 99 m. bender , p .- h . heenen , and p .- reinhard , rev . * 75 * , 121 ( 2003 ) . d. vautherin and d.m . brink , phys . c * 5 * , 626 ( 1972 ) . t. h. r. skyrme , phil . mag . * 1 * , 1043 , ( 1956 ) . t. lesinski , m. bender , k. bennaceur , t. duguet , and j. meyer , phys . c * 76 * , 014312 ( 2007 ) . m. kortelainen , j. mcdonnell , w. nazarewicz , e. olsen , p .- g . reinhard , j. sarich , n. schunck , s.m . wild , d. davesne , j. erler , a. pastore , arxiv:1312.1746 ( 2013 ) . h. sagawa and g. col , arxiv 1401.6691 . to be published in progr . li - gang cao , h. sagawa , and g. col , phys . rev . c * 83 * , 034324 ( 2011 ) . e. perliska , s. g. rohoziski , j. dobaczewski , and w. nazarewicz , phys . c * 69 * , 014316 ( 2004 ) . m. kortelainen , t. lesinski , j. mor , w. nazarewicz , j. sarich , n. schunck , m.v . stoitsov , and s. wild , phys . c * 82 * , 024313 ( 2010 ) . l. di gallo , m. oertel , and m. urban , phys . c * 84 * , 045801 ( 2011 ) . n. chamel , d. page , s. reddy , arxiv:1310.4059 ( 2013 ) . f. gulminelli , ad . r. raduta , m. oertel , and j. margueron , phys . c * 87 * , 055809 ( 2013 ) . a.l . fetter and j.d . walecka , _ quantum theory of many - particle systems _ , mcgraw - hill , new york ( 1971 ) . p. ring and p. schuck , _ the nuclear many - body problem _ , springer - verlag ( 1980 ) . j. margueron , j. navarro , and nguyen van giai , phys . c * 66 * , 014303 ( 2002 ) . li - gang cao , g. col , and h. sagawa , phys . c * 81 * , 044302 ( 2010 ) . j. navarro and a. polls , phys . c * 87 * , 044329 ( 2013 ) . t. lesinski , k. bennaceur , t. duguet , and j. meyer , phys . c * 74 * , 044315 ( 2006 ) . m. kortelainen , t. lesinski , journal of physics g * 37 * , 064039 ( 2010 ) . v. hellemans , a. pastore , t. duguet , k. bennaceur , d. davesne , j. meyer , m. bender , and p .- h . heenen , phys . c * 88 * , 064323 , ( 2013 ) . a. pastore , d. davesne , k. bennaceur , j. meyer , and v. hellemans , phys . scripta * t154 * , 014014 ( 2013 ) . n. chamel , s. goriely , and j.m . pearson , phys . c * 80 * , 065804 ( 2009 ) . j. margueron , m. grasso , s. goriely , g. col , and h. sagawa , prog . theor . * 196 * , 172 ( 2012 ) . carlsson , j. dobaczewski , and m. kortelainen , phys . c * 78 * , 044326 ( 2008 ) . j. sadoudi , t. duguet , j. meyer , and m. bender , phys . c * 88 * , 064326 ( 2013 ) . d davesne , a pastore and j navarro , j. phys . g : nucl . part . * 40 * 095104 ( 2013 ) . c. garca - recio , j. navarro , n. van giai , and l. l. salcedo , ann . * 214 * , 293 ( 1992 ) . j. margueron , j. navarro , and nguyen van giai , phys . c * 74 * 015805 ( 2006 ) . d. davesne , m. martini , k. bennaceur , and j. meyer , phys . c * 80 * , 024314 ( 2009 ) ; _ ibid._*84 * 059904 ( 2011 ) . a. pastore , d. davesne , y. lallouet , m. martini , k. bennaceur , and j. meyer , phys . c * 85 * , 054317 ( 2012 ) . a. pastore , m. martini , v. buridon , d. davesne , k. bennaceur , and j. meyer , phys . c * 86 * , 044308 ( 2012 ) . hernndez , j. navarro , and a. polls , phys . * b413 * , 1 ( 1997 ) . hernndez , j. navarro , and a. polls , nucl . phys . * a627 * , 460 ( 1997 ) . hernndez , j. navarro and a. polls , nucl . phys . * a658 * , 327 ( 1999 ) . a. pastore , d. davesne and j. navarro , arxiv:1312.4696 , j. phys . g : nucl . part . ( 2014 ) in press . o. bohigas , a.m. lane and j. martorell , phys . rep . * 51 * , 267 ( 1979 ) . e. lipparini and s. stringari , phys . rep . * 175 * , 103 ( 1989 ) . j. meyer , private communication . d. vautherin , adv . * 22 * , 123 ( 1996 ) . f. l. braghin , d. vautherin , and a. abada , phys . c * 52 * , 2504 ( 1995 ) . hernndez , j. navarro , a. polls , and j. ventura , nucl . phys . * a595 * , 1 ( 1996 ) . p. bonche , s. levit and d. vautherin , nucl . a427 * , 278 ( 1984 ) . p. bonche , s. levit and d. vautherin , nucl . phys . * a436 * , 265 ( 1985 ) . c. ducoin , c. providencia , a. m. santos , l. brito and ph . chomaz , phys . rev c * 78 * , 055801 , ( 2008 ) . m. baldo and c. ducoin , phys . c * 79 * , 035801 ( 2009 ) . a. vidaurre , j. navarro and j. bernabeu , astron . astrophys . * 135 * , 361 ( 1984 ) . a. rios , a. polls and i. vidaa , phys . c * 71 * , 055802 ( 2005 ) n. chamel and s. goriely , phys . c * 82 * , 045804 ( 2010 ) . c. ducoin , j. margueron , ph . chomaz , nucl . phys . * a809*,30 , ( 2008 ) .

we prove a decomposition theorem for the class of triangle - free graphs that do not contain a subdivision of the complete graph on four vertices as an induced subgraph . we prove that every graph of girth at least 5 in this class is 3-colorable . * ams classification : * 05c75

1407.6531

here graphs are simple and finite . we say that @xmath0 _ contains _ @xmath1 when @xmath1 is isomorphic to an induced subgraph of @xmath0 . we say that a graph @xmath0 is _ @xmath2-free _ if @xmath0 does not contain @xmath2 . for a family of graphs @xmath3 , we say that @xmath0 is _ @xmath3-free _ if for every @xmath4 , @xmath0 does not contain @xmath2 . _ subdividing _ an edge @xmath5 of a graph @xmath0 means deleting @xmath6 , and adding a new vertex @xmath7 of degree 2 adjacent to @xmath8 and @xmath9 . a _ subdivision _ of a graph @xmath0 is any graph @xmath1 obtained from @xmath0 by repeatedly subdividing edges . note that @xmath0 is a subdivision of @xmath0 . we say that @xmath1 is an _ isk4 of a graph @xmath0 _ when @xmath1 is an induced subgraph of @xmath0 and @xmath1 is a subdivision of @xmath10 ( where @xmath10 denotes the complete graph on four vertices ) . isk4 stands for `` induced subdivision of @xmath10 '' . in @xcite , a decomposition theorem for isk4-free graphs is given ( see theorem [ th : decisk4 ] ) and it is proved that their chromatic number is bounded by a constant @xmath11 . the proof in @xcite follows from a theorem by khn and osthus @xcite , from which it follows that @xmath11 is at least @xmath12 . it is conjectured in @xcite that every isk4-free is 4-colorable . the goal of this paper is to prove a stronger decomposition theorem for isk4-free graphs , under the additional assumption that they are triangle - free ( see theorems [ decomp ] and [ maindecomp ] ) . we also propose the following conjectures , prove that the first one implies the second one ( see theorem [ th : imply ] ) , and prove both of them for graphs of girth at least 5 ( see theorem [ th : color ] ) . a complete bipartite graph with partitions of size @xmath13 and @xmath14 , is denoted by @xmath15 . [ conj1 ] every \{triangle , isk4 , @xmath16}-free graph contains a vertex of degree at most [ conj2 ] every \{triangle , isk4}-free graph is 3-colorable . in section [ sec : dec ] , we state several known decomposition theorems , and derive easy consequences of them for our class . in particular we prove that conjecture [ conj1 ] implies conjecture [ conj2 ] . in section [ sec : proofdec ] , we prove our main decomposition theorem . in section [ sec : chordless ] , we give some properties needed later for the class of chordless graphs ( graphs where all cycles are chordless ) . in section [ sec : deg2 ] , we prove conjecture [ conj1 ] for graphs of girth at least 5 . in this section , we provide the notation needed to state decomposition theorems for isk4-free graphs , and we state them . a graph @xmath0 is _ series - parallel _ if no _ subgraph _ of @xmath0 is a subdivision of @xmath10 . clearly , every series - parallel graph is isk4-free . when @xmath17 is a graph , the _ line graph _ of @xmath17 is the graph @xmath0 whose vertex - set is @xmath18 and such that two vertices of @xmath0 are adjacent whenever the corresponding edges are adjacent in @xmath17 . for a graph @xmath0 , when @xmath19 is a subset of @xmath20 , we write @xmath21 instead of @xmath22 $ ] . cutset _ in a graph @xmath0 is a set @xmath19 of vertices such that @xmath23 is disconnected . star cutset _ is a cutset @xmath19 that contains a vertex @xmath11 , called a _ center _ of @xmath19 , adjacent to all other vertices of @xmath19 . note that a star cutset may have more than one center , and that a cutset of size 1 is a star cutset . double star cutset _ is a cutset @xmath19 that contains two adjacent vertices @xmath24 and @xmath25 , such that every vertex of @xmath26 is adjacent to @xmath24 or @xmath25 . note that a star cutset of size at least 2 is a double star cutset . a path from a vertex @xmath27 to a vertex @xmath28 is refered to as an _ _ proper 2-cutset _ of a connected graph @xmath30 is a pair of non - adjacent vertices @xmath31 such that @xmath32 can be partitioned into non - empty sets @xmath33 , @xmath34 and @xmath35 so that : there is no edge between @xmath33 and @xmath34 ; and both @xmath36 $ ] and @xmath37 $ ] contain an @xmath29-path and neither of @xmath38 $ ] nor @xmath37 $ ] is a chordless path . we say that @xmath39 is a _ split _ of this proper 2-cutset . the following is the main decomposition theorem for isk4-free graphs . [ th : decisk4 ] if @xmath0 is an isk4-free graph , then @xmath0 is series - parallel , or @xmath0 is the line graph of a graph of maximum degree at most 3 , or @xmath0 has a proper 2-cutset , a star cutset , or a double star cutset . our first goal is to improve this theorem for triangle - free graphs . some improvements are easy to obtain ( they trivially follow from the absence of triangles ) . the non - trivial one is done in the next two sections : we show that double star cutsets and proper 2-cutsets are in fact not needed . we state the result now , but it follows from theorem [ maindecomp ] that needs more terminology and is slightly stronger . [ decomp ] if @xmath0 is a \{triangle , isk4}-free graph , then either @xmath0 is a series - parallel graph or a complete bipartite graph , or @xmath0 has a clique cutset of size at most two , or @xmath0 has a star cutset . follows directly from theorem [ maindecomp ] . we state now several lemmas from @xcite that we need . a _ hole _ in a graph is a chordless cycle of length at least 4 . prism _ is a graph made of three vertex disjoint paths of length at least 1 , @xmath40 , @xmath41 and @xmath42 , with no edges between them except the following : @xmath43 , @xmath44 , @xmath45 , @xmath46 , @xmath47 , @xmath48 . note that the union of any two of the paths of a prism induces a hole . wheel @xmath49 _ is a graph that consists of a hole @xmath1 plus a vertex @xmath50 that has at least three neighbors on @xmath1 . [ l : begin ] if @xmath0 is an isk4-free graph , then either @xmath0 is a series - parallel graph , or @xmath0 contains a prism , or @xmath0 contains a wheel or @xmath0 contains @xmath16 . a _ complete tripartite graph _ is a graph that can be partitioned into three stable sets so that every pair of vertices from two different stable sets is an edge of the graph . [ l : deck33 ] if @xmath0 is an isk4-free graph that contains @xmath16 , then either @xmath0 is a complete bipartite graph , or @xmath0 is a complete tripartite graph , or @xmath0 has a clique - cutset of size at most 3 . we now state the consequences of the lemmas above for triangle - free graphs . [ l : begint ] if @xmath0 is a \{triangle , isk4}-free graph , then either @xmath0 is a series - parallel graph , or @xmath0 contains a wheel or @xmath0 contains @xmath16 . clear from lemma [ l : begin ] and the fact that every prism contains a triangle . [ l : deck33 t ] if @xmath0 is an \{isk4 , triangle}-free graph that contains @xmath51 , then either @xmath0 is a complete bipartite graph , or @xmath0 has a clique - cutset of size at most 2 . clear from lemma [ l : deck33 ] and the fact that complete tripartite graphs and clique - cutsets of size 3 contain triangles . it is now easy to prove the next theorem . [ th : imply ] if conjecture [ conj1 ] is true , then conjecture [ conj2 ] is true . suppose that conjecture [ conj1 ] is true . let @xmath0 be a \{triangle , isk4}-free graph . we prove conjecture [ conj2 ] by induction on @xmath52 . if @xmath53 , the outcome is clearly true . if @xmath0 contains @xmath16 , then by lemma [ l : deck33 t ] , either @xmath0 is bipartite and therefore 3-colorable , or @xmath0 has a clique - cutset @xmath54 . in this last case , we recover a 3-coloring of @xmath0 from 3-colorings of @xmath55 $ ] , , @xmath56 $ ] where @xmath57 are the connected components of @xmath58 . if @xmath0 contains no @xmath16 , then by conjecture [ conj1 ] it has a vertex @xmath8 of degree at most 2 . by the induction hypothesis , @xmath59 has a 3-coloring , and we 3-color @xmath0 by giving to @xmath8 a color not used by its two neighbors . when @xmath24 is a vertex of a graph @xmath0 , @xmath60 denotes the neighborhood of @xmath24 , that is the set of all vertices of @xmath0 adjacent to @xmath24 . we set @xmath61 = n(x ) \cup \{x\}$ ] . when @xmath62 , we set @xmath63 . when @xmath0 is a graph , @xmath54 an induced subgraph of @xmath0 , and @xmath19 a set of vertices disjoint from @xmath64 , the _ attachment _ of @xmath19 to @xmath54 is @xmath65 , that we also denote by @xmath66 . when @xmath67 is a path and @xmath68 , we denote by @xmath69 the @xmath70-subpath of @xmath71 . let @xmath72 and @xmath73 be two disjoint vertex sets such that no vertex of @xmath72 is adjacent to a vertex of @xmath73 . a path @xmath74 _ connects @xmath72 and @xmath73 _ if either @xmath75 and @xmath76 has a neighbor in @xmath72 and a neighbor in @xmath73 , or @xmath77 and one of the two endvertices of @xmath71 is adjacent to at least one vertex in @xmath72 and the other is adjacent to at least one vertex in @xmath73 . @xmath71 is a _ direct connection between @xmath72 and @xmath73 _ if in @xmath78 $ ] no path connecting @xmath72 and @xmath73 is shorter than @xmath71 . the direct connection @xmath71 is said to be _ from @xmath72 to @xmath73 _ if @xmath76 is adjacent to a vertex of @xmath72 and @xmath79 is adjacent to a vertex of @xmath73 . let @xmath1 be a hole . a chordless path @xmath74 in @xmath80 is an _ appendix _ of @xmath1 if no vertex of @xmath81 has a neighbor in @xmath1 , and one of the following holds : * @xmath75 , @xmath82 and @xmath83 is not an edge , or * @xmath77 , @xmath84 , @xmath85 and @xmath86 . so @xmath87 is an attachment of @xmath71 to @xmath1 . the two @xmath83-subpaths of @xmath1 are called the _ sectors _ of @xmath1 w.r.t . @xmath71 . let @xmath88 be another appendix of @xmath1 , with attachment @xmath89 . appendices @xmath71 and @xmath88 are _ crossing _ if one sector of @xmath1 w.r.t . @xmath71 contains @xmath90 , the other contains @xmath91 and @xmath92 . [ ca ] if @xmath0 is an \{isk4 , @xmath93}-free graph , then no two appendices of a hole of @xmath0 can be crossing . let @xmath94 and @xmath95 be appendices of a hole @xmath1 of @xmath0 , and suppose that they are crossing . let @xmath96 be the attachment of @xmath71 to @xmath1 , and let @xmath97 be the attachment of @xmath88 to @xmath1 . so @xmath98 and w.l.o.g . @xmath99 appear in this order when traversing @xmath1 . @xmath100 is adjacent to @xmath76 , and @xmath90 to @xmath101 . a vertex of @xmath71 must be adjacent to or coincident with a vertex of @xmath88 , since otherwise @xmath102 induces an isk4 . note that @xmath103 . let @xmath104 be the vertex of @xmath71 with lowest index that has a neighbor in @xmath88 , and let @xmath105 ( resp . @xmath106 ) be the vertex of @xmath88 with lowest ( resp . highest ) index adjacent to @xmath104 . note that @xmath104 is not coincident with a vertex of @xmath88 . first suppose that @xmath107 . if @xmath108 then @xmath109 induces an isk4 . so @xmath110 . in particular , no vertex of @xmath71 is coincident with a vertex of @xmath88 , and @xmath111 is the only edge between @xmath71 and @xmath88 . if @xmath112 then @xmath113 induces an isk4 . so @xmath75 and by symmetry @xmath114 . since @xmath115 can not induce a @xmath93 , w.l.o.g . @xmath116 is not an edge . let @xmath117 be the @xmath116-subpath of @xmath1 that does not contain @xmath118 and @xmath91 . then @xmath119 induces an isk4 . therefore @xmath120 . if @xmath104 has a unique neighbor in @xmath88 , then @xmath121 induces an isk4 . let @xmath122 be the sector of @xmath1 w.r.t . @xmath88 that contains @xmath100 . if @xmath104 has exactly two neighbors in @xmath88 , then @xmath123 induces an isk4 . so @xmath104 has at least three neighbors in @xmath88 . in particular , @xmath124 . but then @xmath125 induces an isk4 . [ ca2 ] let @xmath0 be an \{triangle , isk4 , @xmath93}-free graph . let @xmath1 be a hole and @xmath74 , @xmath77 , a chordless path in @xmath126 . suppose that @xmath127 or 2 , @xmath128 or 2 , no vertex of @xmath129 has a neighbor in @xmath1 , @xmath130 and @xmath131 . then @xmath71 is an appendix of @xmath1 . assume not . if one of @xmath76 or @xmath79 has one neighbor in @xmath1 and the other one has two neighbors in @xmath1 , then @xmath132 induces an isk4 . so @xmath133 , and hence ( since @xmath0 is triangle - free ) both @xmath76 and @xmath79 are appendices of @xmath1 . by lemma [ ca ] , @xmath76 and @xmath79 can not be crossing . so for a sector @xmath117 of @xmath1 w.r.t . @xmath76 , @xmath134 . but then @xmath135 induces an isk4 . let @xmath49 be a wheel contained in a graph @xmath0 . a _ sector _ is a subpath of @xmath1 whose endvertices are adjacent to @xmath24 and interior vertices are not . two sectors are _ consecutive _ or _ adjacent _ if they have an endvertex in common . throughout this section we use the following notation for a wheel @xmath49 . we denote by @xmath136 the neighbors of @xmath24 in @xmath1 , appearing in this order when traversing @xmath1 . in this case , we also say that @xmath137 is an _ @xmath138-wheel_. for @xmath139 , @xmath140 denotes the sector of @xmath49 whose endvertices are @xmath141 and @xmath142 ( here and throughout this section we assume that indices are taken modulo @xmath138 ) . a path @xmath71 is an _ appendix _ of a wheel @xmath49 if the following hold : * @xmath71 is an appendix of @xmath1 , * each of the sectors of @xmath1 w.r.t . @xmath71 properly contains a sector of @xmath49 , and * @xmath24 has at most one neighbor in @xmath71 . [ wa1 ] let @xmath0 be an isk4-free graph . let @xmath71 be an appendix of a wheel @xmath49 of @xmath0 , and let @xmath143 be a sector of @xmath1 w.r.t . . then @xmath143 contains at least three neighbors of @xmath24 . in particular , @xmath143 contains at least two sectors of @xmath49 . let @xmath87 be the attachment of @xmath71 to @xmath1 . since @xmath71 is an appendix of @xmath49 , @xmath143 contains at least two neighbors of @xmath24 . suppose @xmath143 contains exactly two neighbors of @xmath24 . if @xmath24 has a neighbor in @xmath71 , then @xmath144 induces an isk4 . so @xmath24 does not have a neighbor in @xmath71 . since @xmath71 is an appendix of @xmath49 , @xmath143 properly contains a sector of @xmath49 and so w.l.o.g . @xmath24 is not adjacent to @xmath118 . let @xmath145 be the other sector of @xmath1 w.r.t . @xmath71 , and let @xmath146 be the neighbor of @xmath24 in @xmath145 that is closest to @xmath118 . note that since @xmath147 can not induce an isk4 , @xmath148 , and hence @xmath149 and @xmath150 is not an edge . let @xmath117 be the @xmath151-subpath of @xmath145 . then @xmath152 induces an isk4 . a wheel @xmath137 of @xmath0 is _ proper _ if vertices @xmath153)$ ] are one of the following types : * type 0 : @xmath154 ; * type 1 : @xmath155 ; * type 2 : @xmath156 and for some sector @xmath140 of @xmath49 , @xmath157 . [ w1 ] let @xmath0 be a \{triangle , isk4}-free graph . if @xmath49 is a wheel of @xmath0 with fewest number of vertices , then @xmath49 is a proper wheel . let @xmath153)$ ] . it follows from the following two claims that @xmath7 is of type 0 , 1 or 2 w.r.t . @xmath49 , and hence that @xmath49 is proper . [ w1 - 1 ] for every sector @xmath140 of @xmath49 , @xmath158 . otherwise , @xmath159 induces a wheel with fewer vertices than @xmath49 , a contradiction . [ w1 - 2 ] for some sector @xmath140 of @xmath49 , @xmath160 . assume otherwise , and choose @xmath161 so that @xmath7 has a neighbor in @xmath162 and in @xmath163 , @xmath164 is not contained in a sector of @xmath49 and @xmath165 is minimized . w.l.o.g . @xmath166 and @xmath167 ( since the case when @xmath168 is symmetric to the case when @xmath169 ) . let @xmath170 ( resp . @xmath171 ) be the neighbor of @xmath7 in @xmath172 that is closest to @xmath173 ( resp . @xmath174 ) . let @xmath175 ( resp . @xmath176 ) be the neighbor of @xmath7 in @xmath177 that is closest to @xmath178 ( resp . @xmath179 ) . let @xmath71 be the @xmath180-subpath of @xmath1 that contains @xmath174 . note that by the choice of @xmath181 , vertex @xmath7 has no neighbor in the interior of @xmath71 and @xmath182 . since @xmath0 is triangle - free , @xmath183 is not an edge , and if @xmath184 then @xmath185 is not an edge . first suppose that @xmath7 has at least two neighbors in @xmath172 . then by ( [ w1 - 1 ] ) , @xmath7 has exactly two neighbors in @xmath172 . if @xmath186 then @xmath184 ( by the choice of @xmath181 ) , and hence @xmath187 induces an isk4 . so @xmath188 . if @xmath189 is an edge , then @xmath190 is not an edge ( since @xmath0 is triangle - free ) and hence @xmath191 induces an isk4 . so @xmath189 is not an edge . but then @xmath192 together with @xmath193 induces an isk4 . therefore @xmath7 has exactly one neighbor in @xmath172 , and by symmetry it has exactly one neighbor in @xmath177 . if @xmath169 then @xmath194 induces an isk4 . so @xmath195 . if @xmath71 contains at least three neighbors of @xmath24 , then ( since @xmath196 and @xmath182 ) @xmath197 induces a wheel with center @xmath24 that has fewer vertices than @xmath49 , a contradiction . therefore @xmath71 contains exactly two neighbors of @xmath24 . but then @xmath198 and hence @xmath194 together with @xmath199 induces an isk4 . [ w2 ] let @xmath0 be a \{triangle , isk4 , @xmath16}-free graph . let @xmath49 be a proper wheel of @xmath0 with fewest number of spokes . if @xmath49 has an appendix , then @xmath49 is a 4-wheel . assume @xmath49 has an appendix . note that by lemma [ wa1 ] , if @xmath71 is an appendix of @xmath49 then each of the sectors of @xmath1 w.r.t . @xmath71 contains at least two sectors of @xmath49 . if @xmath49 has an appendix such that one of the sectors of @xmath1 w.r.t . this appendix is @xmath200 for some @xmath201 , then let @xmath71 be such an appendix and @xmath202 . otherwise , let @xmath71 be an appendix of @xmath49 such that for a sector @xmath143 of @xmath1 w.r.t . @xmath71 , there is no appendix @xmath88 of @xmath49 such that @xmath143 properly contains a sector of @xmath1 w.r.t . @xmath88 . assume further that such a @xmath71 is chosen so that @xmath203 is minimized . let @xmath145 be the other sector of @xmath1 w.r.t . we may assume that @xmath143 contains @xmath204 , @xmath205 , and does not contain @xmath206 . let @xmath207 be the attachment of @xmath71 to @xmath1 such that @xmath208 and @xmath209 . let @xmath74 and w.l.o.g . assume that @xmath76 is adjacent to @xmath210 , and @xmath79 to @xmath211 . let @xmath212 be the @xmath213-subpath of @xmath214 , and @xmath215 the @xmath216-subpath of @xmath217 . let @xmath117 be the hole induced by @xmath218 . note that since @xmath205 , @xmath219 is a wheel . [ w2 - 1 ] @xmath219 is a proper wheel . let @xmath220)$ ] and assume that @xmath7 is not of type 0 , 1 or 2 w.r.t . @xmath219 . note that @xmath221 . since @xmath49 is a proper wheel , @xmath7 must have a neighbor in @xmath71 . let @xmath222 . let @xmath100 ( resp . @xmath118 ) be the neighbor of @xmath7 in @xmath71 that is closest to @xmath210 ( resp . @xmath211 ) . note that sectors @xmath223 of @xmath49 are also sectors of @xmath219 . first suppose that @xmath224 . since @xmath49 is proper , @xmath7 can not have a neighbor in both @xmath212 and @xmath215 . if @xmath7 has a neighbor in @xmath225 , then by lemma [ ca2 ] and since @xmath7 has a neighbor in @xmath71 , @xmath226 contains an appendix @xmath88 of @xmath49 such that a sector of @xmath1 w.r.t . @xmath88 is properly contained in @xmath143 , contradicting our choice of @xmath71 . so @xmath7 does not have a neighbor in @xmath225 , and by symmetry it does not have a neighbor in @xmath227 . since @xmath7 is not of type 0 , 1 or 2 w.r.t . @xmath219 , @xmath228 , and since @xmath0 is triangle - free , @xmath83 is not an edge . let @xmath229 be the chordless path from @xmath210 to @xmath211 in @xmath226 that contains @xmath7 . since the length of @xmath229 can not be less than the length of @xmath71 , by the choice of @xmath71 , there is a vertex @xmath230 such that @xmath231 and @xmath232 are edges . since @xmath0 is triangle - free , @xmath7 is not adjacent to @xmath233 , and hence @xmath7 has exactly two neighbors in @xmath234 . since @xmath7 is not of type 2 w.r.t . @xmath219 , @xmath24 must be adjacent to @xmath233 . but then @xmath235 induces an isk4 . therefore @xmath7 must have a neighbor in @xmath236 . since @xmath49 is proper , for some @xmath237 , @xmath157 . suppose that @xmath7 is of type 2 w.r.t . @xmath49 . w.l.o.g . @xmath7 is not adjacent to @xmath173 . let @xmath238 be the vertex of @xmath71 with lowest index adjacent to @xmath7 . if @xmath239 then , since @xmath0 is triangle - free , not both @xmath79 and @xmath7 can be adjacent to @xmath240 , and hence @xmath241 induces an isk4 . so @xmath242 , and hence @xmath243 induces an isk4 . therefore @xmath7 is of type 1 w.r.t . @xmath49 . recall that by our assumption that @xmath7 has a neighbor in @xmath236 , @xmath7 is not adjacent to @xmath173 nor @xmath240 . but then @xmath244 contains an isk4 . so @xmath219 is a proper wheel . since it can not have fewer sectors than @xmath49 and by lemma [ wa1 ] , it follows that @xmath245 , @xmath246 and @xmath247 . but then by the choice of @xmath71 , @xmath248 , and hence @xmath49 is a 4-wheel . a _ short connection _ between sectors @xmath140 and @xmath249 of @xmath49 is a chordless path @xmath74 , @xmath77 , in @xmath250)$ ] such that the following hold : * @xmath251 , @xmath252 , * @xmath253 , @xmath254 , and * the only vertex of @xmath1 that may have a neighbor in @xmath255 is @xmath142 . [ w3 ] let @xmath0 be a \{triangle , isk4 , @xmath93}-free graph . let @xmath49 be a proper wheel of @xmath0 with fewest number of spokes . then @xmath49 has no short connection . suppose @xmath49 has a short connection @xmath74 . assume that @xmath49 and @xmath71 are chosen so that @xmath203 is minimized . w.l.o.g . @xmath76 is adjacent to @xmath256 and @xmath79 to @xmath257 . let @xmath258 be the @xmath259-subpath of @xmath172 , and let @xmath260 be the @xmath261-subpath of @xmath262 . let @xmath263 be the @xmath83-subpath of @xmath1 that contains @xmath174 . let @xmath117 be the hole induced by @xmath264 . [ w3 - 1 ] @xmath265 if @xmath266 then @xmath147 induces an isk4 . if @xmath267 then @xmath268 induces an isk4 . therefore , @xmath265 . [ w3 - 2 ] @xmath49 has no appendix . follows from ( [ w3 - 1 ] ) and lemma [ w2 ] . [ w3 - 3 ] vertex @xmath174 has a neighbor in @xmath129 . assume not . if @xmath174 has no neighbor in @xmath71 , then @xmath269 induces an isk4 . so w.l.o.g . @xmath174 is adjacent to @xmath76 . then @xmath270 is not an edge , since @xmath0 is triangle - free . if @xmath174 is not adjacent to @xmath79 , then @xmath271 induces an isk4 . so @xmath174 is adjacent to @xmath79 , and hence @xmath272 is not an edge . but then @xmath273 induces an isk4 . [ w3 - 4 ] @xmath219 is a proper wheel . by ( [ w3 - 1 ] ) @xmath219 is a wheel . assume it is not proper and let @xmath220)$ ] be such that it is not of type 0 , 1 or 2 w.r.t . @xmath219 . note that @xmath221 , and hence @xmath7 is of type 0 , 1 or 2 w.r.t . it follows that @xmath7 must have a neighbor in @xmath71 . let @xmath104 ( resp . @xmath238 ) be the vertex of @xmath71 with lowest ( resp . highest ) index adjacent to @xmath7 . first suppose that @xmath274 . then @xmath275 . if @xmath7 does not have a neighbor in @xmath276 , then @xmath7 has at least three neighbors in @xmath71 and hence @xmath277 is a short connection of @xmath49 that contradicts our choice of @xmath71 . so @xmath7 has a neighbor in @xmath278 . @xmath279 . since @xmath49 is proper , @xmath280 . if @xmath7 has a unique neighbor in @xmath172 , then @xmath195 ( since @xmath0 is triangle - free and @xmath7 has at least two neighbors in @xmath71 ) and hence @xmath281 is a short connection of @xmath49 that contradicts our choice of @xmath71 . so @xmath7 is of type 2 w.r.t . @xmath49 . if @xmath282 then @xmath283 and hence @xmath284 induces an isk4 . so @xmath285 . if @xmath190 is an edge then @xmath195 ( since @xmath0 is triangle - free and @xmath7 has at least three neighbors in @xmath117 ) and hence @xmath281 is a short connection of @xmath49 that contradicts our choice of @xmath71 . so @xmath190 is not an edge . if @xmath174 has no neighbor in @xmath286 and @xmath287 , then let @xmath288 . otherwise let @xmath88 be the chordless path from @xmath238 to @xmath174 in @xmath289 . then @xmath290 induces an isk4 . therefore , for some @xmath291 , @xmath7 has a neighbor in @xmath292 . since @xmath49 is proper , @xmath293 . if @xmath282 let @xmath294 , if @xmath107 let @xmath295 , and otherwise let @xmath88 be a chordless path in @xmath296 from @xmath7 to a vertex of @xmath129 that is adjacent to @xmath174 ( note that such a vertex exists by ( [ w3 - 3 ] ) ) . by lemma [ ca2 ] @xmath88 is an appendix of @xmath1 . in particular , @xmath7 is of type 1 w.r.t . @xmath49 . let @xmath170 be the neighbor of @xmath7 in @xmath1 . since by ( [ w3 - 2 ] ) @xmath88 can not be an appendix of @xmath49 , w.l.o.g . @xmath282 and @xmath297 . note that @xmath298 and @xmath299 . let @xmath300 be the vertex of @xmath71 with lowest index adjacent to @xmath174 ( such a vertex exists by ( [ w3 - 2 ] ) ) . if @xmath301 then @xmath302 induces an isk4 . so @xmath303 . if if @xmath304 then @xmath305 together with the @xmath306-subpath of @xmath172 induces an isk4 . so @xmath307 . but then @xmath308 induces an isk4 . by ( [ w3 - 4 ] ) @xmath219 is a proper wheel that has fewer sectors than @xmath49 , a contradiction . [ w4 ] let @xmath0 be a \{triangle , isk4 , @xmath93}-free graph . let @xmath49 be a proper wheel of @xmath0 with fewest number of spokes . then for every @xmath201 , @xmath309 { \setminus}h ) \cup \ { x_i , x_{i+1}\}$ ] is a star cutset separating @xmath140 from @xmath310 . assume not . then w.l.o.g . there is a direct connection @xmath74 from @xmath311 to @xmath312 in @xmath313 { \setminus}h ) \cup \ { x_1,x_2 \})$ ] . note that the only vertices of @xmath1 that may have a neighbor in the interior of @xmath71 are @xmath173 and @xmath174 . since @xmath49 is proper , @xmath77 and @xmath76 and @xmath79 are of type 1 or 2 w.r.t . let @xmath314 be such that @xmath315 . first suppose that no vertex of @xmath316 has a neighbor in @xmath129 . by lemma [ ca2 ] , @xmath71 is an appendix of @xmath1 . in particular , @xmath76 and @xmath79 are both of type 1 w.r.t . @xmath49 . if @xmath317 then @xmath71 is an appendix of @xmath49 . it follows from lemma [ w2 ] that @xmath267 and @xmath318 . but then , since @xmath76 has a neighbor in @xmath311 , @xmath49 and @xmath71 contradict lemma [ wa1 ] . so @xmath319 , and hence @xmath71 is a short connection , contradicting lemma [ w3 ] . therefore , a vertex of @xmath316 has a neighbor in @xmath129 . let @xmath238 be the vertex of @xmath320 with highest index adjacent to a vertex of @xmath316 . @xmath321 is an edge . we now show that if @xmath76 has two neighbors in @xmath322 , then @xmath173 has a neighbor in @xmath320 . assume not . then @xmath76 has two neighbors in @xmath322 and @xmath173 does not have a neighbor in @xmath320 . since @xmath76 is of type 2 w.r.t . @xmath49 , @xmath323 is not an edge . but then @xmath1 and @xmath324 ( where @xmath325 is the vertex of @xmath71 with lowest index adjacent to @xmath174 ) contradict lemma [ ca2 ] . suppose @xmath326 . if @xmath79 has two neighbors in @xmath327 , then @xmath328 induces an isk4 . so , @xmath79 has a unique neighbor in @xmath327 . if @xmath173 does not have a neighbor in @xmath320 , then @xmath76 has a unique neighbor in @xmath322 and hence @xmath71 is a short connection of @xmath49 , contradicting lemma [ w3 ] . so @xmath173 has a neighbor in @xmath320 . let @xmath300 be such a neighbor with highest index . then @xmath329 is a short connection of @xmath49 , contradicting lemma [ w3 ] . so @xmath330 . if @xmath331 then either @xmath286 is a short connection of @xmath49 contradicting lemma [ w3 ] , or @xmath332 contains an isk4 . therefore , @xmath333 and @xmath79 has a neighbor in @xmath334 . by lemma [ ca2 ] applied to @xmath1 and @xmath286 , vertex @xmath79 is of type 1 w.r.t . @xmath49 and @xmath335 is not an edge . but then @xmath286 is an appendix of @xmath49 . by lemma [ w2 ] it follows that @xmath267 and @xmath318 . but then , since @xmath79 has a neighbor in @xmath336 , @xmath49 and @xmath286 contradict lemma [ wa1 ] . we say that a graph @xmath0 has a _ wheel decomposition _ if for some wheel @xmath49 , for every @xmath337 , @xmath309 { \setminus}h ) \cup \ { x_i , x_{i+1 } \}$ ] is a cutset separating @xmath140 from @xmath338 . we say that such a wheel decomposition is _ wheel @xmath49_. note that if a graph has a wheel decomposition , then it has a star cutset . [ maindecomp ] if @xmath0 is a \{triangle , isk4}-free graph , then either @xmath0 is a series - parallel graph or a complete bipartite graph , or @xmath0 has a clique cutset of size at most two , or @xmath0 has a wheel decomposition . assume @xmath0 is not series - parallel nor a complete bipartite graph . by lemma [ l : begint ] @xmath0 contains a wheel or @xmath93 . by lemma [ l : deck33 t ] if @xmath0 contains a @xmath93 then it has a clique cutset of size at most two . so we may assume that @xmath0 does not contain a @xmath93 . so @xmath0 contains a wheel . by lemma [ w1 ] @xmath0 contains a proper wheel , and hence by lemma [ w4 ] @xmath0 has a wheel decomposition . [ maindecomp2 ] if @xmath0 is a \{triangle , isk4 , @xmath93}-free graph , then either @xmath0 is series - parallel or @xmath0 has a wheel decomposition . assume @xmath0 is not series - parallel . by lemma [ l : begint ] @xmath0 contains a wheel . by lemma [ w1 ] @xmath0 contains a proper wheel , and hence by lemma [ w4 ] @xmath0 has a wheel decomposition . the following corollary is needed in the next section . [ cmain ] if @xmath0 is an isk4-free graph of girth at least 5 , then either @xmath0 is series - parallel or @xmath0 has a star cutset . follows directly from theorem [ maindecomp2 ] because @xmath16 contains a cycle of length 4 . a graph @xmath0 is _ chordless _ if no cycle in @xmath0 has a chord . chordless graphs were introduced in @xcite as roots of wheel - free line graphs , and it is a surprise to us that we need them here for a completely different reason in a very similar class . a graph is _ sparse _ if every edge is incident to at least one vertex of degree at most 2 . a sparse graph is chordless because any chord of a cycle is an edge between two vertices of degree at least three . recall that proper 2-cutsets are defined in section [ sec : dec ] . [ t : chordless ] if @xmath0 is a 2-connected chordless graph , then either @xmath0 is sparse or @xmath0 admits a proper 2-cutset . the following theorem is mentioned in @xcite without a proof , and we need it in the next section . so , we prove it for the sake of completeness . [ th : cchordless ] in every cycle of a 2-connected chordless graph that is not a cycle , there exist four vertices @xmath339 that appear in this order and such that @xmath340 have degree 2 and b , d have degree at least 3 . we prove the result by induction on @xmath52 . if @xmath0 is sparse ( in particular , if @xmath341 ) , then it is enough to check that every cycle of @xmath0 contains at least two vertices of degree at least 3 , because these vertices can not be adjacent in a sparse graph . but this true , because a cycle with all vertices of degree 2 must be the whole graph ( since @xmath0 is connected ) , and a cycle with a unique vertex of degree 3 can not exists in a 2-connected graph ( the vertex of degree at least 3 would be a cut - vertex ) . so , by theorem [ t : chordless ] we may assume that @xmath0 has a proper 2-cutset with split @xmath342 . we now build two blocks of decompositions of @xmath0 as follows . the block @xmath343 is obtained from @xmath344 $ ] by adding a marker vertex @xmath345 adjacent to @xmath7 and @xmath8 , and the block @xmath343 is obtained from @xmath346 $ ] by adding a marker vertex @xmath347 adjacent to @xmath7 and @xmath8 . by the definition of proper 2-cutsets , @xmath348 . also , @xmath343 and @xmath349 are chordless and 2-connected . so we may apply the induction hypothesis to the blocks of decomposition . let @xmath19 be a cycle of @xmath0 . if @xmath350 , then @xmath19 is a cycle of @xmath343 , so by the induction hypothesis we get four vertices @xmath351 in @xmath19 . we now check that a vertex @xmath352 has degree 2 in @xmath0 if and only if it has degree 2 in @xmath343 . this is obvious , except if @xmath353 . but in this case , because of @xmath345 and because @xmath9 lies in a cycle of @xmath343 that does not contain @xmath345 , @xmath9 has degree at least 3 in both @xmath0 and @xmath343 . this proves our claim . it follows that we obtain by the induction hypothesis the condition that we need for the degrees of @xmath27 , @xmath28 , @xmath11 and @xmath354 . the proof is similar when @xmath355 . we may now assume that @xmath19 has vertices in @xmath33 and @xmath34 . it follows that @xmath19 edge wise partitions into a path @xmath356 whose interior is in @xmath33 and a path @xmath357 whose interior is in @xmath34 . we apply the induction hypothesis to @xmath343 and @xmath358 . so , we get four vertices @xmath27 , @xmath28 , @xmath11 and @xmath354 in @xmath359 and they have degree 2 , @xmath360 , 2 , @xmath360 respectively ( in @xmath343 ) . these four vertices are in @xmath19 and have the degrees we need ( in @xmath0 ) , except possibly when @xmath361 . in this case , we may assume w.l.o.g . that @xmath362 or @xmath363 , and we find in place of @xmath27 a vertex of degree 2 in @xmath364 , where @xmath365 . this vertex exists by the induction hypothesis applied to @xmath349 and @xmath366 . we need the following application of menger s theorem . [ l : tree ] let @xmath367 be a tree , and suppose that the vertices of @xmath367 are labelled with labels @xmath24 and @xmath25 ( each vertex may receive one label , both labels , or no label ) . one and only one of the following situations occures . * in @xmath367 , there exist two vertex - disjoint paths @xmath71 and @xmath88 , and each of them is from a vertex with label @xmath24 to a vertex with label @xmath25 ( possibly , @xmath71 and/or @xmath88 have length 0 ) . * there exists @xmath368 and two subtrees of @xmath367 , @xmath369 and @xmath370 such that : 1 . @xmath371 ; 2 . @xmath372 ; 3 . @xmath369 contains all vertices of @xmath367 with label @xmath24 and @xmath370 contains all vertices of @xmath367 with label @xmath25 . it is clear that not both outcomes hold , because if the second holds , then @xmath8 must be on every path from a vertex with label @xmath24 to a vertex with label @xmath25 , so no two such paths can be vertex - disjoint . if at most one vertex of @xmath367 has label @xmath24 , then we nominate this vertex as @xmath8 and we set @xmath373 $ ] , @xmath374 . it is easy to see that @xmath375 @xmath369 and @xmath370 satisfy all the requirements of the second outcome . hence , we may assume that at least two vertices of @xmath367 have label @xmath24 , and similarly , at least two vertices of @xmath367 have label @xmath25 . by the classical menger s theorem , if the first outcome does not hold , then there exists a vertex @xmath8 in @xmath367 such that every path from a vertex with label @xmath24 to a vertex with label @xmath25 contains @xmath8 . in particular , every component of @xmath376 contains at most one label . so , the second outcome holds : we define @xmath369 as the subtree of @xmath367 formed by @xmath8 and all components containing vertices with label @xmath24 , and we define @xmath370 as the subtree of @xmath367 formed by @xmath8 and all the other components . a graph @xmath0 together with two of its vertices @xmath24 and @xmath25 such that @xmath377 or @xmath378 , have the _ @xmath379-property _ if @xmath380 \cup n[y])$ ] contains a vertex of degree 2 in @xmath0 . instead of @xmath381-property , we simply write _ @xmath24-property_. the @xmath382-property is very convenient for us , because it ensures the existence of vertices of degree 2 , and also because it is well preserved in proofs by induction . unfortunately , not all graphs in our class have the @xmath379-property , for intance the graphs represented in fig . [ fig ] do not have the @xmath24-property when @xmath24 is a vertex with maximum degree in the graph . * @xmath0 is a graph and @xmath24 and @xmath25 are vertices of @xmath0 . * @xmath0 does not have the @xmath379-property . * if @xmath384 , then @xmath0 has the @xmath24-property and @xmath0 has the @xmath25-property . indeed , if @xmath0 does not have the @xmath24-property , then @xmath386 is a bad triple . similarly , if @xmath0 does not have the @xmath25-property , then @xmath387 is a bad triple . so , we may assume that @xmath0 has the @xmath24-property and the @xmath25-property . if @xmath0 has the @xmath379-property , then we are done , and otherwise , all the requirements in the definition of bad triples are fulfilled . it follows that a structural description of bad triples really discribes all triples @xmath383 such that @xmath0 does not have the @xmath379-property . such a description is given in the next lemma for triangle - free 2-connected series - parallel graphs with no clique cutset . note that in the next lemma , we do not require the girth of the graph to be at least 5 . [ l : desc ] let @xmath383 be a bad triple , and suppose that @xmath0 is triangle - free , 2-connected , series - parallel , has least 5 vertices , and has no clique cutset . then , @xmath0 can be constructed as follows ( and conversely , all graphs constructed as follows are triangle - free , 2-connected , series - parallel , have least 5 vertices , and have no clique cutset ) . * if @xmath384 , then build two non - empty trees @xmath369 and @xmath370 , not containing @xmath388 , and consider the tree @xmath367 obtained by gluing @xmath369 and @xmath370 along some vertex @xmath8 ( so @xmath389 ) . * if @xmath378 build a non - empty tree @xmath367 , and set @xmath390 . * add vertices of degree 2 , each of them either adjacent to @xmath24 and to some vertex in @xmath369 , or to @xmath25 and some vertex in @xmath370 , in such a way that the following conditions are satisfied : 1 . @xmath391 ; 2 . every vertex of @xmath367 that has degree 2 in @xmath367 has at least one neighbor in @xmath309 \cup n[y ] ) { \setminus}\{x , y\}$ ] ; 3 . every vertex of @xmath367 that has degree 1 in @xmath367 ( so , a leaf of @xmath367 ) has at least two neighbors in @xmath309 \cup n[y ] ) { \setminus}\{x , y\}$ ] ; 4 . every vertex of @xmath367 that has degree 0 in @xmath367 ( this happens only when @xmath392 ) has at least three neighbors in @xmath309 \cup n[y ] ) { \setminus}\{x , y\}$ ] . suppose that some cycle @xmath19 of @xmath0 has a chord @xmath29 . thereofore , @xmath19 is formed of two @xmath29-paths @xmath17 and @xmath393 , both of length at least 2 . since @xmath394 is not a clique cutset of @xmath0 , some path @xmath395 of @xmath0 is disjoint from @xmath394 , has one end in @xmath17 , the other one in @xmath393 , and is internally disjoint from @xmath19 . therefore , @xmath396 and @xmath395 form a subdivision of @xmath10 ( that is in @xmath0 as a subgraph ) , and this contradicts @xmath0 being series - parallel . suppose w.l.o.g . that a neighbor of @xmath24 , @xmath398 has degree at least 3 . let @xmath7 and @xmath8 be two neighbors of @xmath146 distinct from @xmath24 . because @xmath0 is triangle - free , @xmath7 and @xmath8 are non - adjacent to @xmath24 . since @xmath0 is 2-connected , there is a path @xmath356 in @xmath399 . so , @xmath400 is a cycle of @xmath0 . note that @xmath19 does not go through @xmath24 , for otherwise @xmath401 would be a chord a @xmath19 , a contradiction to ( [ c : chordless ] ) . by ( [ c : cdeg2 ] ) , @xmath19 contains two vertices @xmath27 and @xmath28 of degree 2 . vertices @xmath27 and @xmath28 must be adjacent to @xmath24 or @xmath25 because they have degree 2 and @xmath0 does not have the @xmath402-property . if @xmath27 and @xmath28 are both adjacent to @xmath24 ( in particular , when @xmath378 ) , then @xmath19 must go through @xmath24 ( because @xmath27 and @xmath28 have degree 2 ) , a contradiction . hence , at least one of @xmath27 or @xmath28 is adjacent to @xmath25 , not to @xmath24 , and in particular , @xmath384 . if @xmath19 does not go through @xmath25 , then @xmath19 , @xmath24 and @xmath25 form a subdivision of @xmath10 , contradiciting @xmath0 being series - parallel , so @xmath19 goes through @xmath25 . if @xmath24 is adjacent to @xmath27 or @xmath28 , then again @xmath19 , @xmath24 and @xmath25 form a subdivision of @xmath10 . so , @xmath27 and @xmath28 are both adjacent to @xmath25 , not to @xmath24 . since @xmath0 has the @xmath25-property ( from the definition of bad triples ) , there is a non - neighbor @xmath403 of @xmath25 that has degree 2 , and since @xmath0 does not have the @xmath379-property , @xmath403 is a neighbor of @xmath24 , distinct from @xmath146 because @xmath146 has degree 3 . since @xmath0 is 2-connected , there exists a path @xmath404 from @xmath403 to @xmath19 in @xmath405 . if @xmath406 , then @xmath24 , @xmath88 and @xmath19 form a subdivision of @xmath10 , a contradiction to @xmath0 being series - parallel . otherwise , @xmath407 and @xmath408 is a chord of some cycle of @xmath0 , a contradiction to ( [ c : chordless ] ) . otherwise , @xmath409\cup n[y]$ ] and all vertices of @xmath0 are adjacent to @xmath24 or @xmath25 . if @xmath378 , then @xmath409 $ ] , a contradiction since @xmath0 is triangle - free and 2-connected . so , @xmath384 . let @xmath398 be a neighbor of @xmath24 . by ( [ c : deg2 ] ) , @xmath146 has degree 2 . its other neighbor @xmath410 is non - adjacent to @xmath24 ( because @xmath0 is triangle - free ) , so it must be adjacent to @xmath25 , and by ( [ c : deg2 ] ) , @xmath410 has degree 2 . since @xmath0 contains at least five vertices , there must be other vertices , so w.l.o.g . another neighor @xmath403 of @xmath24 . again , @xmath403 has degree 2 , a neighbor @xmath411 ( distinct from @xmath410 because @xmath410 has degree 2 ) , and @xmath411 is adjacent yo @xmath25 . now , @xmath412 is a cycle of @xmath0 that has a chord ( namely @xmath413 ) , a contradiction to ( [ c : chordless ] ) . otherwise , let @xmath7 and @xmath8 be two adjacent vertices in @xmath414 . by ( [ c : deg2 ] ) they have degree 2 , so @xmath415 is a clique cutset that separates them from @xmath416\cup n[y])$ ] that is non - empty by ( [ c : nonempty ] ) . since @xmath383 is a bad triple , @xmath416 \cup n[y])$ ] contains only vertices of degree at least 3 ( in @xmath0 ) , so by ( [ c : cdeg2 ] ) , it can not contain a cycle . also @xmath417 \cup n[y])$ ] is connected , because otherwise let @xmath72 and @xmath73 be two components of @xmath416 \cup n[y])$ ] . these two components must attach to disjoint sets of neighbors of @xmath24 and @xmath25 , because all neighbors of @xmath24 and @xmath25 ( except @xmath24 and @xmath25 ) have degree 2 by ( [ c : deg2 ] ) . it follows that @xmath418 is a clique cutset , a contradiction . if the first outcome holds ( so in @xmath367 , there exist two vertex - disjoint paths @xmath71 and @xmath88 , and each of them is from a vertex with label @xmath24 to a vertex with label @xmath25 ) , then we reach a contradiction as follows . let @xmath419 and @xmath420 where @xmath421 have label @xmath24 and @xmath422 have label @xmath25 . so , let @xmath423 be a neighbor of @xmath424 that is adjacent to @xmath24 , and let @xmath425 , @xmath426 and @xmath427 be defined similarly . observe that @xmath423 , @xmath425 , @xmath426 and @xmath427 are distinct , because all neighbors of @xmath24 and @xmath25 distinct from @xmath24 and @xmath25 have degree 2 by ( [ c : deg2 ] ) . now the cycle @xmath428 has a chord ( namely @xmath413 ) , a contradiction to ( [ c : chordless ] ) . hence , the second outcome holds , so we keep the notation @xmath8 , @xmath369 and @xmath370 from lemma [ l : tree ] . note that @xmath367 can be obtained by gluing @xmath369 and @xmath370 along @xmath8 . it follows that @xmath0 can be constructed as we claim it should be . by ( [ c : tree ] ) and ( [ c : nonempty ] ) , we really need to consider a non - empty tree . by ( [ c : deg2 ] ) , we have to add vertices of degree 2 , and by ( [ c : stable ] ) , they all have one neighbor in @xmath418 and the other one in @xmath367 . the last three conditions are here to ensure that the vertices of @xmath367 really all have degree at least 3 . we do not prove the converse statement ( every graph constructed as above is a bad triple , is triangle - free , 2-connected , series parallel , has at least 5 vertices and has no clique cutset ) . it is easy to check and we do not need it in what follows . [ xy ] let @xmath0 be a 2-connected series - parallel graph , of girth at least 5 that has no clique cutset . if @xmath24 and @xmath25 are vertices of @xmath0 such that @xmath378 or @xmath377 , then @xmath0 has the @xmath379-property . if @xmath430 , then the unique vertex of @xmath367 has at least three neighbors in @xmath309\cup n[y ] ) { \setminus}\{x , y\}$ ] , and at least two of them are neighbors of @xmath24 , or are neighbors of @xmath25 . therefore , @xmath0 contains a 4-cycle , a contradiction to our assumption on the girth . if @xmath431 , then we consider a leaf of @xmath367 that is distinct from the vertex @xmath8 defined in the outcome of lemma [ l : desc ] . this leaf has at least two neighbors that are both neighbors of @xmath24 , or that are both neighbors of @xmath25 . again , there exists a 4-cycle in @xmath0 . we prove the statement by induction on @xmath52 . if @xmath0 has no star cutset , then it has no clique cutset , and it is series parallel by corollary [ cmain ] ( all this happens in particular when @xmath433 which is the base case of our induction ) . so , we have the result directly by lemma [ xy ] . hence , we may assume that @xmath0 has a star cutset @xmath19 . we suppose that @xmath19 is inclusionwise minimal among all possible star cutsets and is centered at @xmath11 . w.l.o.g . we suppose that @xmath24 and @xmath25 are both in @xmath434 $ ] where @xmath33 is a component of @xmath21 , and we consider another component @xmath34 . we claim that @xmath435 $ ] is 2-connected . indeed , suppose for a contradiction that @xmath435 $ ] has a cutvertex @xmath8 . since @xmath0 is 2-connected , @xmath436 . by the minimality of @xmath19 , every vertex of @xmath437 has a neighbor in @xmath34 and if @xmath438 , then @xmath11 also has neighbors in @xmath34 . it follows that @xmath439 . so , @xmath8 is in fact a cutvertex of @xmath0 , a contradiction . we proved that @xmath435 $ ] is 2-connected . it is enough to prove that every isk4-free graph of girth at least 5 contains a vertex of degree at most 2 . for the sake of induction , we prove by induction on @xmath52 a slightly stronger statement : every isk4-free graph of girth at least 5 on at least two vertices contains at least two vertices of degree at most 2 . if @xmath441 , this is clearly true . if @xmath0 is 2-connected , it follows from lemma [ l:2-conn ] applied twice ( once to find a vertex @xmath24 of degree 2 , and another time to find the second one in @xmath442 $ ] ) . so , we may assume that @xmath0 is not 2-connected and has at least 3 vertices , so it has a cutvertex @xmath8 . the result follows from the induction hypothesis applied to @xmath443 $ ] and to @xmath444 $ ] where @xmath33 and @xmath34 are connected components of @xmath445 .

a formalism and its numerical implementation is presented which allows to calculate quantities determining the spacetime structure in the large directly . this is achieved by conformal techniques by which future null infinity ( ) and future timelike infinity ( ) are mapped to grid points on the numerical grid . the determination of the causal structure of singularities , the localization of event horizons , the extraction of radiation , and the avoidance of unphysical reflections at the outer boundary of the grid , are demonstrated with calculations of spherically symmetric models with a scalar field as matter and radiation model . [ satz]theorem [ satz]corollar [ satz]corollary [ satz]lemma

gr-qc9409029

from the singularity theorems by s. hawking and r. penrose it is known that the appearance of singularities in general relativity is an unavoidable feature for `` strong '' initial data @xcite . conversely , it has been proven recently that for small initial data the future of the initial value hypersurface looks in the large in a certain sense like the future of flat space data @xcite . unfortunately , it has turned out to be extremely difficult to obtain reasonable conjectures not to mention the proofs about the properties of a spacetime given by an initial value problem , as illustrated , for example , by the history of the cosmic censorship hypothesis . + getting an overview for the phenomena appearing and making reasonable conjectures is an area numerical relativity can contribute a lot to and actually already has ; most impressive illustrated by the discovery of the echoing property by m. choptuik @xcite . but most of the codes used so far are designed to analyze local behaviour . statements about the behaviour in the large , global issues , are obtained by assuming that the extension of the grid used approximates an infinite grid well enough . this way it is very difficult to get a reliable error estimate and to decide , what is sufficiently far , if not impossible at all . + but what global issues are of interest and why are they interesting ? + firstly , there are questions related to cosmic censorship , especially the location and penetration of an event horizon , which is the boundary of the region of spacetime from where no null geodesic reaches null infinity . in a recent review article @xcite about cosmic censorship c. clarke writes about the location of event horizons : `` in terms of numerical simulations , this means that it is essential to perform the simulations in the compactified picture in which @xmath2 is represented explicitly '' . + secondly , there is the whole topic of gravitational radiation on asymptotically flat spacetimes . due to the gauge freedom and the non - linearity of the theory the classification of radiation as ingoing and outgoing is a difficult issue . in- and outgoing gravitational waves are defined with respect to null infinity only . the related difficulties in extracting gravitational waves from the grid at finite distances and the problem with the avoidance of an unphysical reflection on the outer boundary of the grid are well known and have been a topic for research for a long time . in most methods developed so far the error made consists of the error from reading off at finite distances and the discretization error . + in this paper i will present the numerical implementation of a formalism allowing to calculate a `` compactified '' spacetime including @xmath2 and @xmath1 . the solution of the problems concerning radiation is trivial by construction as will be seen . the only errors which may appear are caused by the discretization error of the numerical partial differential equation solver . these errors can be estimated and controlled by grid extrapolation techniques ( e. g. richardson extrapolation ) . although the calculations have been done under the assumption of spherical symmetry with a scalar field as model for radiation the simplicity and exactness of radiation extraction hold for arbitrary symmetry assumptions . + furthermore , due to the spherical symmetry a special coordinate gauge could be found allowing to cover the complete domain of dependence of the initial hypersurface and to calculate the causal structure of a singularity . the location of the event horizon is straightforward . + the formalism is based on conformal techniques developed by r. penrose to describe asymptotically flat spacetimes . by a rescaling @xmath3 the physical spacetime @xmath4 is mapped to an unphysical spacetime @xmath5 with boundary . the boundary represents null infinity . @xmath6 is a `` compactified '' is not really compact in minkowski spacetime there are three points , future and past timelike and spacelike infinity , missing . in general @xmath6 can not be smoothly extended to contain those points . ] version of @xmath7 . gravitational radiation is the value of certain components of the weyl spinor on that boundary . in the originally suggested form the formalism is not suited to describe initial value problems . h. friedrich has modified it to describe initial value problems one has to solve a set of evolution equations for the unphysical spacetime @xmath5 . the rescaling factor @xmath8 acts like an artificial matter field for the einstein equations . subsection [ mathform ] reviews the equations for the unphysical spacetime . + j. winicour and r. gomez have developed an approach which uses bondi s ideas for describing gravitational radiation : outgoing null cones , with the area distance and the direction angle as coordinates on it , are compactified to represent future null infinity by grid points . their method gives simpler equations but has certain disadvantages : the existence of smooth outgoing null cones is essential . null caustics , which are caused e.g. by gravitational lensing effects , must not appear . furthermore the evolution scheme can not penetrate event horizons . the extraction of radiation is complicated slightly as the weyl spinor is not a variable of the system the determination of gravitational radiation requires to calculate numerically derivatives on . their formalism solves a characteristic initial value , the formalism presented here solves a `` normal '' , spacelike initial value problems in unphysical spacetime . shortly after h. bondi et . @xcite proved that gravitational radiation is not a gauge effect r. penrose suggested a coordinate independent way to characterize asymptotically flat spaces and gravitational radiation . a thorough discussion of the ideas and the interpretation can be found at various places in the literature , e.g. @xcite . the definitions of asymptotical flatness given in the literature differ slightly . the following will be used here : [ asymflat ] a spacetime @xmath9 is called * asymptotically flat * if there is another `` unphysical '' spacetime @xmath10 with boundary and a smooth embedding by which @xmath7 can be identified with @xmath11 such that : 1 . there is a smooth function @xmath12 on @xmath6 with @xmath13 2 . on @xmath14 3 . [ nullcompleteness ] each null geodesic in @xmath15 acquires a past and a future endpoint on . because of item [ nullcompleteness ] null geodesically incomplete spacetimes like schwarzschild are not asymptotically flat . the next definition includes those spacetimes which have only an asymptotically flat part : [ weakasymflat ] a spacetime is called * weakly asymptotically flat * if definition [ asymflat ] with the exception of item [ nullcompleteness ] is fulfilled . definition [ asymflat ] and [ weakasymflat ] classify spacetimes , they do not require that the einstein equation is fulfilled . for a physical problem one would like to give `` asymptotically flat data '' and evolve them according to the einstein equation . + nevertheless the geometrically description was extremely helpful in analyzing asymptotically flat spacetimes and it has been successfully used as guideline to construct a formalism for analyzing initial value problems . this method has been developed and applied to various matter sources by h. friedrich @xcite and myself @xcite . + the idea is to choose a spacelike initial value surface in the unphysical spacetime @xmath10 and to evolve it . + for minkowski space the unphysical spacetime @xmath10 can be smoothly extended with three points , future @xmath16 and past @xmath17 timelike infinity , the end and the starting point of all timelike geodesics of @xmath18 respectively , and spacelike infinity @xmath19 , the end point of all spacelike geodesics of @xmath18 . the point @xmath20 divides into two disjunct parts , future @xmath21 and past @xmath22 null infinity . it is well known and has been discussed elsewhere that there are unsolved problems in smoothly extending a `` normal '' cauchy hypersurface of @xmath7 to @xmath20 if the spacetime has non - vanishing adm mass . certain curvature quantities blow up at @xmath20 , reflecting the non - invariance of the mass under rescalings . + by choosing a spacelike ( with respect to @xmath23 ) hypersurface @xmath24 not intersecting @xmath20 but @xmath2 ( @xmath25 ) we avoid the problems with @xmath20 . @xmath24 is called a hyperboloidal hypersurface the corresponding initial value problem is called a hyperboloidal initial value problem . + the well - posedness of the hyperboloidal initial value problem for general relativistic scalar fields has been discussed in @xcite . there , a precise definition of a hyperboloidal initial value problem can be found , too . + figure [ regunrz ] shows a diagram of the unphysical spacetime with an example of a hyperboloidal surface in it . in figure [ regrz ] the corresponding physical spacetime is shown . in both figures only one space coordinate is drawn . all points , except those on the axis and @xmath20 , represent spheres . = = the domain of dependence @xmath26 of @xmath24 will not contain the complete spacetime . the interior of @xmath24 corresponds to an everywhere spacelike hypersurface in the physical spacetime which approaches a null hypersurface @xmath27 asymptotically . if @xmath27 is a light cone @xmath28 then the domain of dependence of @xmath24 is @xmath28 . therefore the hyperboloidal initial value problem is well suited to describe the future ( past ) of data on the spacelike hypersurface @xmath24 , e. g. a stellar object and the gravitational radiation caused by its future time evolution . it is not well suited to investigate the structure near @xmath20 . + why is it advantageous to solve the initial value problem in the unphysical spacetime ? the future of the data in physical spacetime @xmath18 is completely determined by the data on the interior of @xmath24 . since the rescaling factor @xmath8 is known , the calculation of quantities of physical spacetime from the unphysical quantities is merely algebra . in unphysical spacetime @xmath2 is on the grid . @xmath0 is an ingoing null cone starting at the 2-surface in @xmath24 where @xmath8 vanishes . by extending @xmath24 a little bit beyond the intersection with in unphysical spacetime we do not change the physical spacetime @xmath18 . but with a finite number of gridpoints and thus a finite computation time the whole future of the initial hypersurface is covered . and since @xmath0 is an ingoing null line the values at @xmath7 and @xmath0 describing the physics do not depend on the values at the outer boundary of the extended @xmath24 . the numerical dependence on the outer boundary must converge to @xmath29 . + future null infinity , @xmath0 , which can be found by searching for the @xmath30 contour , is on the grid . by an appropriate choice of gauge it can even be arranged that the position of on the grid is known by analytical considerations . in the worst case , one has to interpolate between neighbouring grid points to find the radiation emitted . in the case of gravitational radiation , it is given by certain components of the weyl tensor , which are variables in the system of equations for the unphysical spacetime . the choice of the model is always a compromise between generality and manageability . for this work it was necessary to have some model for radiation to demonstrate the simplicity of the radiation extraction in the formalism . on the other side the spacetime symmetry should be as high as possible to reduce the demand for computational resources and to simplify coordinate choice related questions . + the spacetime is assumed to be spherically symmetric , the radiational degree of freedom is modeled by a conformally invariant scalar field . the equations are [ model ] @xmath31 where the second is equivalent to @xmath32 with @xmath33 the notation used is explained in appendix [ konventionen ] . + the form invariance of ( [ wllngl ] ) under the rescalings @xmath34 and @xmath35 was the reason for not choosing the massless klein - gordon scalar field with its simpler equations in physical spacetime . in @xcite it has been shown that the initial value problems for these matter models are equivalent for practical purposes . in @xcite it has been argued that the einstein equation , if simply translated to the unphysical spacetime , becomes `` singular '' on and thus is not suitable to calculate the unphysical spacetime @xmath10 . for a derivation of regular equations for the unphysical spacetime the reader is referred to @xcite , here only the final first order set of equations is given . + in this first order system , the variables are the components of the frame @xmath36 with respect to the coordinates @xmath37 , @xmath38 , the frame components @xmath39 of the ricci rotation coefficients @xmath40 , certain combinations of the components of the tracefree part @xmath41 of the ricci tensor and the regularized weyl tensor @xmath42 , the conformal factor @xmath8 , the frame components @xmath43 of its gradient @xmath44 , the trace of its second derivative @xmath45 , the conformal scalar field @xmath46 , the frame components @xmath47 of its gradient @xmath48 and combinations of its second derivatives @xmath49 . the first order system , written in abstract index notation is : [ quasys ] @xmath50c } + \frac{1}{12 } ( \nabla_{[a } r ) \ , g_{b]c } - \omega_d \ , d_{abc}{}^d + \omega \ , m_{abc } - \frac{2}{3 } \ , \omega \ , m_{[a|d|}{}^d g_{b]c } \nonumber \\ & = & 0 \\ \label{nd } { { \cal n}{{}_{\mathrm\mathbf d \ , } { } } } _ { abc } & : = & \nabla_d d_{abc}{}^d - m_{abc } + \frac{2}{3 } \ , m_{[a|d|}{}^d \ , g_{b]c } = 0 \\ \label{no } { { \cal n}{{}_{\mathrm\mathbf \omega \ , } { } } } _ a & : = & \nabla_a \omega - \omega_a = 0 \\ \label{ndo } { { \cal n}{{}_{\mathrm\mathbf d\omega \ , } { } } } _ { ab } & : = & \nabla_a \omega_b + \frac{1}{2 } \ , \omega \ , { \mbox{$\hat{r}$}}_{ab } - \omega \ , g_{ab } - \frac{1}{2 } \ , \omega^3 \ , t_{ab } = 0 \\ \label{no } { { \cal n}{{}_{\mathrm\mathbf \omega \ , } { } } } _ { ab } & : = & \nabla_a \omega + \frac{1}{2 } \ , { \mbox{$\hat{r}$}}_{ab } \ , \omega^b + \frac{1}{12 } \ , r \ , \omega_a + \frac{1}{24 } \ , \omega \ , \nabla_a r - \frac{1}{2 } \ , \omega^2 \ , t_{ab } \ , \omega^b \nonumber \\ & = & 0 \\ \label{np } { { \cal n}{{}_{\mathrm\mathbf \phi \ , } { } } } _ a & : = & \nabla_a \phi - \phi_a = 0 \\ \label{ndp } { { \cal n}{{}_{\mathrm\mathbf d\phi \ , } { } } } _ { ab } & : = & \nabla_a \phi_b - \hat{\phi}_{ab } - \frac{1}{4 } \phi_c{}^c \ , g_{ab } = 0 \\ \label{nbp } { { \cal n}{{}_{\mathrm\mathbf \box\phi \ , } { } } } & : = & \phi_a{}^a - \frac{r}{6 } \ , \phi = 0 \\ \label{nddp } { { \cal n}{{}_{\mathrm\mathbf dd\phi \ , } { } } } _ { abc } & : = & \nabla _ { [ a } \hat{\phi}_{b]c } + \frac{1}{6 } \ , ( \phi \ , \nabla_{[a } r + r \ , \phi_{[a } ) \ , g_{b]c } - \frac{1}{2 } \ , r{{}_{\mathrm\mathbf alg \,}{}}_{abc}{}^d \ , \phi_d = 0 \\ \label{ndbp } { { \cal n}{{}_{\mathrm\mathbf d\box\phi \ , } { } } } _ a & : = & \nabla _ { a } \phi_b{}^b - \frac{1}{6 } \ , ( \phi\ , \nabla_a r + r \ , \phi_{a } ) = 0,\end{aligned}\ ] ] where @xmath51 is the torsion whose components are given by @xmath52 the tensor @xmath53 is an abbreviation for @xmath54d } - g_{d[a } { \mbox{$\hat{r}$}}_{b]c } + \frac{1}{6 } \ , g_{c[a } \ , g_{b]d } \ , r,\ ] ] @xmath55 has the components @xmath56 the `` energy momentum tensor '' in unphysical spacetime is @xmath57 and @xmath58c } - \frac{1}{2 } \ , g_{c[a } \phi_{b]d } \phi^d + \frac{1}{4 } \ , \phi \ , \omega d_{abc}{}^d \phi_d + \frac{1}{4 } \ , \phi \ , g_{c[a } { \mbox{$\hat{r}$}}_{b]}{}^d \phi_d - \frac{3}{4 } \ , \phi \ , \phi_{[a } { \mbox{$\hat{r}$}}_{b]c } \\ & & \qquad \quad - \frac{1}{12 } \ , \phi \ , \phi_{[a } g_{b]c } r + \frac{1}{4 } \ , \omega \ , \phi^2 d_{abc}{}^d \omega_d \big ] \\ & & \quad - 3 \ , \omega_{[a } \big [ \phi_{b ] } \phi_c - \frac{1}{2 } \ , \phi \ , \phi_{b]c } + \frac{1}{4 } \ , \phi^2 { \mbox{$\hat{r}$}}_{b]c } + \frac{1}{36 } \ , \phi^2 g_{b]c } r - \frac{1}{3 } \ , g_{b]c } \ , \phi^d \phi_d \big ] \\ & & \quad + \omega^d g_{c[a } \big [ \phi_{b ] } \ , \phi_d - \frac{1}{2 } \ , \phi \ , \phi_{bd } + \frac{1}{4 } \ , \phi^2 { \mbox{$\hat{r}$}}_{b]d } \big ] \quad \big).\end{aligned}\ ] ] unfortunately the terms introduced into the system by the scalar field , @xmath59 and @xmath60 , complicate the equations a lot due to the form of the energy momentum tensor for the conformally invariant scalar field . + furthermore there is an additional equation , @xmath61 which must be satisfied at one point to be automatically fulfilled everywhere . + if we fulfill equation ( [ spgl ] ) at one point the function @xmath62 can be given freely . it determines the gauge freedom in the conformal factor @xmath8 to a certain extent . if @xmath63 is a solution of the unphysical equations so is @xmath64 for @xmath65 . in all the calculations , @xmath66 has been set to @xmath67 , the value obtained if the compactification for minkowski space given in @xcite is used . the program code does allow to specify @xmath62 . the conformal gauge freedom has been discussed in more detail in @xcite . the first order system ( [ quasys ] ) is an underdetermined system . to make it complete the coordinate and frame gauge freedom must be fix . this will be done in this section . + furthermore the number of variables will be reduced significantly by making use of the spherical symmetry . + a spacetime is said to be spherically symmetric if it possesses a isometry group @xmath68 which contains a subgroup @xmath66 which is isomorphic to the 3-dimensional group @xmath69 and whose orbits are orientable submanifolds of dimension @xmath70 . the orbits of @xmath66 are either ( fix ) points or spheres @xmath71 . the fix points form at most two timelike geodesics as can be seen by symmetry arguments . it is assumed that the initial hypersurface intersects with at least one line of fix points , i.e. there is a ( regular ) center on the initial hypersurface . + but no attempt was made to use every simplification spherical symmetry provides for the following reason : it is well known that einstein s equations can be reduced to two ordinary differential equations ( constraints ) @xcite plus matter equations in spherical symmetry . the circumstance that no equation with a time derivative for the geometry variables must be solved can be viewed as expressing the lack of gravitational radiation in spherical symmetry . i have not made any attempt to incorporate this kind of simplification into the system for the unphysical spacetime , since the calculations are supposed to be a playground for testing and learning about the advantages and disadvantages of the formalism for a later application to models with less symmetry . + in the gauge used the system is symmetric hyperbolic and the coordinates cover the whole domain of dependence of the initial hypersurface . it is not known to the author how to construct such a coordinate system in less symmetric spacetimes . the very nice feature of semilinearity is certainly an artifact of spherical symmetry . in the analytical analysis of the initial value problem with scalar fields as source terms , a gaussian gauge was used @xcite . this gauge gives a symmetric hyperbolic subsidiary system of equations . if the energy momentum tensor fulfills certain conditions , the coordinate system will break down because of the formation of caustics @xcite . whether the region of the unphysical spacetime representing the physical spacetime can be covered by the use of a gaussian gauge in unphysical spacetime is not obvious since the conformal rescaling factor @xmath8 acts as a kind of artificial energy - momentum source in unphysical spacetime . by numerical calculations it was straightforward to show that even in unphysical spacetime a gaussian gauge will lead to caustics @xcite , and is thus inappropriate to analyze the strong field regime . + the system of equations is very similar to the one obtained in `` double null '' coordinates except that it is not semilinear . it will not be discussed any further . the construction of the coordinate system and the frame are described starting from an initial hypersurface . the remarks about the differentiability class of the objects assume that the construction is done with respect to a given @xmath72 manifold ( for a definition see @xcite ) . the investigation of the differentiability is necessary to ensure that a discontinuity of a value or even a singular value of a variable at the center is not caused by lacking smoothness of the coordinates . + the gauge realized in this way turns out to be an obvious adherent of double null coordinates . + assume we are given a spacelike @xmath72 hypersurface @xmath24 with @xmath73 . this hypersurface is factored by the orbits of the group . by the geodesics running from the center in the different directions the angle coordinates @xmath74 with line element @xmath75 on the unit sphere are defined on all orbits . by isotropy the geodesics are perpendicular to the orbits . the orbits are labeled by a monotonically increasing coordinate @xmath76 , defined to be @xmath29 at the center . auxiliary coordinates @xmath77 are defined by @xmath78 and @xmath79 . set @xmath80 along the future directed , outgoing null lines , @xmath81 along the future directed , ingoing null line . on passing through the center set @xmath82 . due to the spherical symmetry there can not be any null caustic , a break down of the coordinates is aligned with a spacetime singularity . by @xmath77 a timelike coordinate @xmath83 and a spacelike coordinate @xmath84 are defined everywhere in the domain of dependence of @xmath24 . an orthonormal frame field @xmath85 is defined except in the center ( polar coordinate singularity ) by normalizing the orthogonal vector fields @xmath86 . + as shown in @xcite the @xmath87 hypersurfaces are at least @xmath88 and hence sufficiently smooth . the proof proceeds in two steps , firstly it is shown that there is a @xmath72 transformation to radar coordinates , secondly a theorem by u. proff ( * ? ? ? * theorem 1.2.4 ) about radar coordinates completes the proof . + in this coordinate system the following relations hold : the frame coordinate matrix is diagonal , @xmath89 and @xmath90 all ricci rotation coefficients except @xmath91 vanish . + scalars invariant under rotations are functions of @xmath92 only , rotationally invariant vectors @xmath93 are of the form @xmath94 and symmetric covariant 2-tensors @xmath95 , e.g. the ricci tensor @xmath96 , invariant under rotations , look like @xmath97 this follows from the assumptions about the symmetry . + all components of @xmath98 are either zero or proportional to @xmath99 . due to the complicated form of the matter terms the set of equations is very lengthy . the equations are given in appendix [ appsphsys ] , they are derived from system ( [ quasys ] ) and @xmath100 the system ( [ quasys ] ) has some remarkable features which should be mentioned . firstly it is semi linear . secondly every equation has one of the following forms : @xmath101 where @xmath102 , @xmath103 and @xmath104 are variables propagating along @xmath105 , @xmath106 and @xmath107 . @xmath102 stands for @xmath108 , @xmath109 and @xmath110 , @xmath103 for @xmath108 , @xmath111 and @xmath112 , @xmath104 for @xmath113 , @xmath114 , @xmath115 , @xmath116 , @xmath117 , @xmath8 , @xmath118 , @xmath119 , @xmath46 , @xmath120 , @xmath121 and @xmath122 . for every @xmath104 there also exists a constraint , @xmath123 there are no constraints for the @xmath102 and @xmath103 . polar coordinates cause regularity conditions in the center , in physical spacetime as well as in unphysical spacetime . for the variables in the system ( [ sphsys ] ) those conditions are given in appendix [ regularityconditions ] . they express that locally the center behaves like minkowski space , and indicate how a wave hitting the @xmath124 ( inner ) boundary is reflected there ( passes through the center ) . the regularity conditions at the inner boundary are the part of the code that would change if a spacetime with a throat were calculated . + but in unphysical spacetime there are also necessary conditions for the regularity at expressing an appropriate fall - off in physical spacetime and therefore reflecting asymptotical flatness . + it is not known whether those conditions are also sufficient for regularity on in the general case . the answer to this requires the investigation of the constraints on a given hypersurface . for the case that a neighbourhood of is free of matter and for a special choice of a hypersurface ( for technical reasons ) it has been shown that the necessary conditions are also sufficient @xcite . in the calculations presented here those conditions are fulfilled . + there are as many regularity conditions ( at the center and at ) as there are variables @xmath104 . in this section the numerical methods used will be described shortly and the reasons given for choosing those methods . there are 12 constraints and 12 regularity conditions to be solved . the free functions in the constraints are the six variables @xmath102 and @xmath103 . but giving those makes the interpretation of a parameter study difficult . @xmath102 and @xmath103 represent higher derivatives of the primary quantities , the metric @xmath23 , the rescaling factor @xmath8 fixing the relation between physical and unphysical spacetime and the scalar field @xmath46 . furthermore the regularity conditions at are easier to handle if it is known where @xmath8 vanishes . a straightforward way to realize this is to give @xmath8 on the initial slice . + in the code @xmath125 , @xmath126 , @xmath127 , @xmath8 , @xmath46 , and @xmath128 are given as free functions ( because of regularity they all must be even at the center ) . @xmath129 determines the @xmath130 respectively the @xmath131 component of the metric , @xmath126 and @xmath127 the extrinsic curvature of the initial slice in unphysical spacetime . the constraints which contain derivatives of these quantities are interpreted as algebraic conditions for the quantities on the right hand side . to ensure the invertibility of the resulting system , including the points where @xmath30 , instead of equation ( [ nd1om ] ) its derivative is used . a constraint for @xmath132 results which becomes an algebraic condition on @xmath30 if @xmath132 is at least @xmath133 ( condition [ smoothscri ] ) . this is a necessary condition for a regular . + the differential algebraic system has boundary conditions in the center and at @xmath30 . as there are as many boundary conditions as variables to solve for all the initial value freedom is coded in the free functions @xmath125 , @xmath126 , @xmath127 , @xmath8 , @xmath46 , and @xmath128 . the system is solved with a relaxation scheme combined with a newton raphson solver derived with minor modifications from the code given in @xcite . firstly the system is solved between @xmath124 and @xmath30 . if desired the same scheme can be used to extend the initial hypersurface beyond the intersection with . for this integration boundary conditions are given at @xmath30 , namely the values obtained at @xmath30 when solving the constraints `` inside '' . to avoid a dependency of the values at on the treatment of the outer grid boundary the initial hypersurface has always be slightly extended beyond . + to improve accuracy the system is solved with different grid sizes and the results are richardson ( or bulirsch stoer ) @xcite extrapolated to vanishing grid size . in the test cases , where an exact solution is known , the accuracy was limited by the rounding error of the inversion of the matrix in the newton raphson part of the code . typically the numerical and the exact solution differed in the last two digits for calculations with 8 byte floating point numbers . constrainted evolution schemes are known to give in general more accurate results than free evolution schemes . but there are disadvantages of constrainted evolution schemes . in a constrainted evolution scheme the values at a certain grid point depend on the values at many other grid points on the same time slice . thus if one grid point becomes singular the property of being singular is spread over the time slice into regions which are not causally connected with the singular point . + in most approaches used in numerical relativity the singularity is avoided by an appropriate choice of the coordinate system , actually the necessity of singularity avoidance has become a dogma of numerical relativity @xcite . a quite often used example for a singularity avoiding coordinate system is obtained by the radial gauge in spherically symmetric spacetimes @xcite . this coordinate system can not penetrate apparent horizons which are supposed to wrap around singularities by the cosmic censorship hypothesis . the very interesting region of spacetime , where the gravitational field has become strong enough to prevent light rays from expanding , is excluded from being calculated . + the double nulllike coordinate system used does not avoid singularities , the coordinate lines can only end at a singularity . in this subsection it will be described how a program crash is avoided if singular points occur and how the calculation is continued in the region of spacetime which are outside the future domain of dependence of singular points . in the chosen approach the singular boundary of spacetime is represented by grid points . to avoid the spreading of the singular property out of the domain of influence of the continuous equations it is necessary to run the scheme , at least in the very neighbourhood of a singularity , with a courant factor of @xmath134 . different schemes have been tested @xcite . the only second order schemes which could be run at a courant factor of 1 were the second order schemes of the class @xmath135 @xcite with @xmath136 . a second order lax wendroff scheme as follows has been chosen : for the equation @xmath137 , with @xmath138 and a vector of functions @xmath139 , the predictor step is @xmath140 then the corrector step is @xmath141 if @xmath142 were a function of @xmath139 , as it is the case for other coordinate choices , it would be discretized as @xmath143 . + the second order lax wendroff scheme is used to evolve gridpoints not depending on the values at the boundary . the treatment at the outer boundary is irrelevant as long as the scheme remains stable for the following reason : the initial value surface is extended beyond the intersection with . running with a courant factor of @xmath134 the points inside and on , representing the physical spacetime , do not even depend on the values at the outer boundary which has no intersection with . even if the treatment of the outer boundary causes an instability it will not influence the physics , i.e. the values on @xmath7 and . in most runs with coraunt factor @xmath134 the grid points depending on the outer boundary have not even been calculated . for a run with a courant factor @xmath144 the influence of the outer boundary treatment on the physical spacetime must converge to @xmath29 with the same order the scheme converges . thus as long as the treatment on the outer boundary is numerically stable the values at @xmath7 and are in the limit of vanishing grid size independent of the outer boundary treatment at any physical time . + finding a stable treatment of the center was a very difficult problem . so far i could not find a treatment which remains stable if a fourth order scheme is built from the second order lax wendroff by richardson extrapolation . + especially it was not possible to extend the grid from the gridpoints at @xmath124 and @xmath145 to @xmath146 and run the lax wendroff scheme on the extended grid . + the solution of the stability problems at the inner boundary was to replace the values near the center after every time step by the values obtained by other methods . + in the method used for my ph.d . thesis the constraints were integrated from grid point number @xmath147 towards the center ( the constraints together with the regularity conditions determine all the variables at the center ) . the values at grid point @xmath134 have been obtained by interpolation . the solution did not look smooth on the innermost gridpoints for coarse grids in the very strong field regime . for very large fields ( values of @xmath148 beyond @xmath149 ) this method even became unstable . therefore another method has been developed . + in the calculations for this paper a kind of polynomial extrapolation with dissipation has been used to replace the innermost gridpoints . the idea is the following : use the values at gridpoints @xmath147 and higher and the regularity conditions to extrapolate towards the center by polynomial fitting and get solution i. do the polynomial fitting again starting at grid point @xmath150 to get solution ii . solution i and ii can now be added in such a way that the simplest grid mode with values @xmath151 is eliminated . this adding of dissipation is necessary to ensure stability . i call this method polynomial extrapolation with dissipation . since the coordinate system does not avoid singularities some variables may become singular . depending on the default setting of the compiler this causes a crash of the program by a floating point exception ( on unix systems the signal sigfpe is sent ) . the programming language c allows to specify what to do in case of a certain signal . in the program used the action on sigfpe is to flag the corresponding grid point as singular and to continue the calculation on the rest of the grid . in addition to the reception of a sigfpe signal a grid point is flagged as singular if either the principal part , i.e. @xmath113 or @xmath152 , of the equation changes sign , i.e. the system of equations becomes singular in the sense of @xcite , or if the evaluation of the values according to the scheme would involve points already flagged as singular ( hence called influence singularity ) . according to the latter reason every point whose values depend on values at singular points must be flagged singular . + a necessary condition for the stability of a scheme solving symmetric hyperbolic systems is the courant friedrichs lewy condition . the domain of dependency of the discretized equations must be a superset of the domain of dependency of the continuous equations , the courant factor must be @xmath153 . a courant factor of @xmath134 means that both domains of dependency coincide . on the other hand , every grid point depending on singular gridpoints is singular . thus all points in the future null cone @xmath28 of a singular point must be flagged singular . if the courant factor is @xmath144 points outside @xmath28 also . for a scheme with a courant factor of exactly @xmath134 at least in the very vicinity of the singularity it is possible to distinguish timelike / nulllike singularities from spacelike singularities . if a line of gridpoints is numerically singular only because of a influence singularity it is a strong hint for a timelike singularity behind . pure influence singularities appeared only due to instabilities during the tests of different treatments of the center . this is in agreement with cosmic censorship . + the dependence of the position of a singular line on the grid size has been tested intensely . in all calculated cases presented the position changed only by a negligible amount corresponding to a few grid points for calculations with about 10000 spatial points . the dependence of the first appearance of a singularity on the courant factor has also been tested . it only changed by a few grid points when using courant factors significantly smaller than @xmath134 . on the exact solutions given below a number of tests have been performed with grids varying from @xmath154 to @xmath155 spatial grid points and different courant factors . for most plots a coarser grid has been used , typically with @xmath156 grid points . + on the plotting grid the scheme shows the expected convergence behaviour . for finer grids ( @xmath157 spatial points and more ) the error is dominated by second order terms . for coarser grids fourth order terms ( oscillations corresponding to high fourier modes ) dominate the error near regions with steep gradients . + the error predictions by richardson s extrapolation are in good agreement with the actual error . + the violation of the constraints converges at least with second order . convergence is partly dominated by the error in the discretization of the derivative for an evaluation of the constraints . the violation of the constraints turned out to be an inappropriate measure for the quality of the solution . a scalar field on a flat background is obtained by setting @xmath158 in the evolution equations . since the physical energy - momentum tensor @xmath159 is tracefree the ricci scalar @xmath160 in physical spacetime vanishes . @xmath161 is determined by @xmath162 . the corresponding solution in unphysical spacetime is @xmath163 . on @xmath164 the following rescaling has been used : @xmath165 with an obvious choice of the sign , depending on the position relative to and @xmath166 . @xmath167 with constants @xmath168 and @xmath169 . + this solution is useful for testing the matter part and the geometry part of the code separately since @xmath158 decouples their evolution equations . but as most of the geometry variables are just constants this is only a first step as test for the geometry part . for any positive function @xmath170 the spacetimes @xmath171 and @xmath172 correspond to the same physical spacetime if @xmath172 is a solution of the unphysical equations . this gauge freedom can be used to obtain a unphysical representation of minkowski spacetime where the geometry variables evolve in time and are not constant ( except the weyl tensor component @xmath122 , which is zero for every representation of minkowski space ) . for @xmath173 an ingoing wave pulse with the initial form ( [ epeak ] ) plus a constant @xmath174 , such that @xmath175 everywhere , has been chosen . @xmath46 has been set to @xmath29 . + this solution has been used to make the geometry variables vary in time and space and is thus a better test for the geometry part of the code than the test in the previous subsection . in physical spacetime @xmath177 is a solution for minkowski spacetime . if we use @xmath178 @xmath179 is still a solution on @xmath7 in unphysical spacetime . but this solution is no longer regular on thus the calculation can be done in the interior of @xmath7 only . for the tests that poses no restriction , since in the unphysical spacetime there is nothing special in the evolution equations on . + since in this solution @xmath180 and @xmath181 the coupling between the geometry and the matter part can be tested . for the nonlinear regime two ways have been used to check the accuracy of the solution . + the first is the richardson extrapolation to vanishing grid - size with the resulting error estimates . the second uses the fact that the position of is known and thus it is possible to compare the calculated value @xmath8 there with the exact value @xmath29 . the error estimate for @xmath8 by the richardson extrapolation was in all cases in agreement with that deviation . + on an approach to the critical parameter , the smallest parameter where a singularity appears , there is a dramatic loss of accuracy although on the coarse plotting grid the scheme still converges quadratically . viewing the fine grid used for the calculations one sees that the loss of accuracy is largest around gridpoints 510 , and this maxima of the error decreases linear with the grid size , meaning that there is only quadratic convergence in a @xmath182 norm , but not pointwise . + for models not to close to the critical value the accuracy problem described can be cured by using more gridpoints . but with about 50000 spatial gridpoints that brute force method reaches its limitation . since the scheme converges quadratically in a @xmath182 norm only , building a higher order scheme by richardson extrapolation did not work . + a detailed inspection of the reasons hints , that this is partly a problem of numerical regularization as described in @xcite(pages 1315 ) . when the calculations for this paper were done there was no way known to the author to numerically regularize the lax wendroff scheme . in the calculations presented the free functions defining the geometry of the initial slice are given as follows : @xmath183 the value of the time coordinate on the initial slice @xmath184 is @xmath185 . the form of @xmath8 is an often used choice for the rescaling of minkowski space . + for the scalar field the initial value is given by @xmath186 with @xmath187 and @xmath188 . + this function @xmath189 is the uniquely defined @xmath88-function with compact support @xmath190 $ ] , being a polynomial of degree 8 on the support and with maximal value @xmath148 at @xmath169 . @xmath148 is the parameter to be varied . @xmath189 should be at least @xmath88 in order to have all variables in the system at least @xmath133 avoiding the known problems of a lax wendroff scheme at discontinuities . from the numerical viewpoint this form of the pulse is much better than a @xmath72 partition of the one like ( [ epeak ] ) , since the variation of @xmath189 and its spatial derivatives is better distributed over the support . + @xmath128 is chosen in such a way that the pulse would be purely ingoing on a flat background , which is the einstein cylinder in unphysical space , i. e. @xmath191 for @xmath192 - 1/\sigma,1/\sigma[$ ] and @xmath29 otherwise . + a pulse with compact support provides several advantages when checking the accuracy and interpreting the numerical solution since part of the qualitative behaviour of the structure in the large is known from analytic considerations . figure [ globalstruct ] shows the qualitative behaviour in the large for a regular solution . = region i is the compact support of the ingoing pulse in the linear model ( @xmath158 ) , passing through the center and then crossing , the thick line . every deviation from a schwarzschild solution in region iii is caused by a pure back scattering effect . the deviations from flat space in region ii are also an back scattering effect the field gets caught near the center for some time . as the characteristics have slope @xmath134 , region iva is a portion of flat space , region ivb is a portion of schwarzschild . for stronger data the solution will become singular before reaching the tip a singularity will appear . to investigate the structure of this singularity is one of the goals of this work . in a regular spacetime timelike infinity , @xmath1 , lies at the point @xmath193 . with increasing parameter @xmath148 more and more matter is put onto the initial slice . at some critical parameter @xmath194 the spacetime is expected to become singular . in this subsection the change of the global structure will be discussed and a conjecture for the critical case presented . + when interpreting the behaviour with varying parameter two things have to be kept in mind . + firstly the characteristics , null geodesics , are a common structure of all models . as the coordinates are adapted to this structure it is straightforward and well defined to compare the models . + secondly the area distance @xmath195 to the center in physical spacetime , expressed in unphysical variables , is @xmath196 even on the initial slice @xmath114 is determined as a solution of differential equations , the constraints . depending on @xmath148 the area distance of the inner and outer edge of the shell of compact support of @xmath46 vary ( in this parameter study they monotonically grow with @xmath148 ) . in principle it is possible and for a better comparison with m. choptuik s work desirable to give @xmath114 as a free function , but parts of the initial value solver code would have to be rewritten . but in doing so one can no longer give either @xmath8 or components of the extrinsic curvature as free functions . the first case causes problems since the location of @xmath197 on the grid is no longer known , in the second case it is unknown whether the necessary conditions for regularity on are also sufficient . figure [ omega025 ] = = shows a contour plot of the rescaling factor @xmath8 for modestly strong initial data corresponding to a parameter @xmath198 . that is well below the critical value @xmath194 , which is somewhere in the interval @xmath1990.48,0.49[$ ] . apart from some minor deformations in the contour lines of @xmath8 , the figure looks like minkowski , although nonlinear effects like backscattering are already significant . the @xmath30 contour coincides with the analytic location of @xmath200 represented by the thick dashed line . the values depending on the outer boundary of the grid have not even been calculated . that is why there are no contours in the upper triangle . + in figure [ omega049 ] the spacetime has become singular . the thick line is certainly a singular boundary of the future domain of dependence of the initial data since here and in all the following cases scalar invariants become singular . near the center the singularity is spacelike . further outside its slope is numerically indistinguishable from a null line . the edgy look is an effect of the plotting program and the coarse plotting grid , which will be seen later when the singularities will be examined more thorough a finer grid near the singularity showing more details will be used there . + with increasing amplitude of the initial scalar field ( figure [ omega055 ] ) = = the spacelike part of the singularity has grown . the change of the shape of the singularity on approaching the critical parameter from above suggest the following conjecture : in the critical case the singularity is a null line . + for further increasing @xmath148 the outer edge of the singularity becomes spacelike . this part grows inwards and finally we get a picture like figure [ omega150 ] . note the contour lines of @xmath8 are focused at the intersection of with the singularity . outside the scalar @xmath8 goes to @xmath201 , inside to @xmath202 on approach to the singularity . in subsection [ physicalreality ] the structure of it will be examined . + the critical case can also be approached from the subcritical side . in figure [ omegachange ] = the time dependence of @xmath8 in the center is plotted for the three subcritical values @xmath203 and the supercritical model @xmath204 . + approaching the critical case @xmath205 seems to develop a zero point before @xmath1 . for spacetimes above the critical case the value of @xmath205 at the singularity is greater than @xmath29 but again approaching @xmath29 on approach to the critical case . outside the center and away from , the scalar @xmath8 does not seem to go to @xmath29 even in the critical case . although the values of @xmath8 are influenced by gauge , the statement that @xmath8 vanishes is not . the author can not interpret this is exceptional conformal structure for the critical case yet . nevertheless the calculations show the critical model might be recognized by this behaviour of @xmath8 . the figures of the preceding subsection show that there are singularities for large values of @xmath148 . in subsection [ singcatcher ] it has been described when and how the program flags a point as singular . that does not necessarily mean that those points represent a singularity in physical spacetime . in this subsection arguments will be given to decide that issue . furthermore the quantities used classify the singularity . + in their singularity theorems r. penrose and s. hawking have shown that a singularity is unavoidable if there are trapped surfaces and the energy momentum tensor fulfills certain energy conditions . the conformally invariant scalar field does not fulfill those energy condition . nevertheless i regard the appearance of trapped surfaces as a strong hint for singularities . with the transformation to the solution of a massless klein gordon field @xcite it is easy to check for the existence of an apparent horizon in a massless klein gordon field model which satisfies the relevant energy conditions . + the null expansion of in- and outgoing null directions in spherical symmetry can be written as @xmath206 where is the outgoing null vector and @xmath207 is the ingoing null vector . written in unphysical quantities : @xmath208 there is the freedom of null boost , thus @xmath209 and @xmath210 are not gauge invariant , but their product is . on the first view one might expect that @xmath211 vanish for @xmath212 . the gauge for @xmath213 chosen makes @xmath214 not vanishing on @xmath0 , @xmath215 not vanishing on @xmath216 . that is a pure gauge effect , by an appropriate null boost @xmath214 and @xmath215 can be made vanishing on . + another criteria for or against a `` real '' singularity is the behaviour of curvature invariants of physical spacetime , e.g. @xmath217 figure [ trapped055 ] = = shows the spacetime region near the singularity of a @xmath218 model . the thick , dashed line is a part of @xmath0 , the thick line the singularity @xmath219 . the singularity is covered by a region of trapped surfaces , the thin line shows the apparent horizon . in the corresponding initial value problem for a massless klein gordon field the same qualitative picture arises . + the proper time of a bondi observer ( see equation ( [ bondipropertime ] ) ) goes to infinity on approach to the intersection point @xmath220 of the @xmath219 with @xmath2 . this is a strong hint that @xmath220 is a singular timelike infinity , a singular @xmath1 . thus the last null line from the center which reaches @xmath2 at @xmath1 is the event horizon ( dotted line ) . + these are already strong hints that @xmath219 is a singularity of the physical spacetime . it is further strengthened by the fact that the curvature scalar @xmath221 blows up at least on approach to the inner ( spacelike part ) of @xmath219 . there are also hints that the area radius goes to @xmath29 on @xmath219 , but to decide that issue more numerical accuracy near the singularity would be needed . + all these considerations `` prove '' that @xmath219 is a real singularity of physical spacetime . + the situation is quite different in figure [ trapped150 ] . along @xmath219 the scalar @xmath8 goes to @xmath222 inside @xmath0 , and to @xmath201 outside @xmath0 . @xmath219 is spacelike everywhere . @xmath223 does not show any sign of blow up near @xmath219 , there is no region of trapped surfaces wrapped around @xmath219 . due to the discontinuity of @xmath8 near @xmath220 the formula for the proper time of an observer at @xmath0 is not applicable . + since only quantities of the unphysical spacetime become singular on @xmath219 i conclude that @xmath219 is a conformal singularity caused by the gauge chosen by specifying @xmath224 . it is very similar to a coordinate singularity . only minor , unsuccessful effort has been spent to try to avoid the conformal singularity by using different choices of @xmath62 . + there is another very interesting point in the models with @xmath225 . on the initial slice there are regions where both @xmath210 and @xmath209 are positive ( `` antitrapped surfaces '' ) . a calculation back in time shows a singularity in the past , covered by an antitrapped region , i.e. a spacetime with a white hole . although the model @xmath226 shows a conformal singularity in the future the existence of an apparent horizon suggests that the physical singularity is hidden behind the conformal singularity . it is natural to assume that beyond @xmath194 there is always a physical singularity which is sometimes hidden by a conformal singularity . in his study of scalar fields m. choptuik found the critical mass scaling behaviour @xmath227 with @xmath228 independent of @xmath148 and an exponent @xmath116 of approximately @xmath229 . + the hawking mass @xmath230 can be written as @xmath231 \nonumber \\ & & + \frac{1}{8 } \ , \frac{r}{e } \ , \kappa \ , \phi^2 \omega.\end{aligned}\ ] ] the use of equation ( [ singglsph ] ) on is essential to write the hawking mass in a form not containing terms with @xmath232 factors . on the hawking mass is identical with the bondi mass . for scalar field data with compact support on the initial slice in @xmath164 the bondi mass on the initial slice coincides with the adm mass . + figure [ scaling ] = = shows an logarithmic plot of the black hole mass obtained for the parameter study performed here ( the stars @xmath233 mark calculated models ) . the black hole mass is determined as the bondi mass `` at @xmath1 '' . + the value of @xmath234 for the critical parameter @xmath194 has been found by optimizing the straight line look in figure [ scaling ] . + the critical exponent is @xmath229 , but due to the uncertainty in @xmath194 the error is certainly @xmath235 . hence the results are compatible with the results of others , but i would not claim more . + although the mass caught by the black hole increases the percentage of the caught mass has a significant maximum at @xmath236 . for large @xmath148 almost all the mass escapes to @xmath0 . actually @xmath237 the given value is only an upper bound since @xmath1 is hidden behind the conformal singularity and thus the bondi mass at the intersection of @xmath0 with the conformal singularity may further decrease . the figures for the black hole mass of the corresponding klein gordon models show the same behaviour . + for models with large initial scalar field amplitudes @xmath148 there are number of significant changes in the structure of spacetime . although most of the scalar field is still contained in region i the shell with most of the hawking mass in it expands and crosses in region iii . mass and scalar field `` decouple '' , propagation of mass is almost completely determined by nonlinear terms . d. christodoulou investigated the general relativistic massless klein gordon field in spherically symmetric spacetimes in a number of papers , for a list of references see @xcite . he found sufficient conditions for the appearance of a spacetime singularity . the theorems proven there contain the following statement : [ chrissatz]@xcite consider on the initial future null geodesic cone @xmath238 an annular region bounded by the two spheres @xmath239 and @xmath240 with @xmath240 in the exterior of @xmath239 and areal radii @xmath241 and @xmath242 . let @xmath243 @xmath244 and @xmath245.\ ] ] a sufficient condition for a non timelike singular boundary in the future of @xmath246 is @xmath247 the model with a conformal scalar field calculated here is ( almost ) equivalent to the massless klein gordon field @xcite . if @xmath248 is a solution for the general relativistic conformal scalar field model @xmath249 is a solution for the general relativistic massless klein gordon field with @xmath250 the area radius @xmath251 and the hawking mass @xmath252 are given by @xmath253 @xmath254 . @xmath255 is the area radius , @xmath256 the hawking mass , @xmath257 the null expansion of the out- and ingoing null directions with null vectors @xmath258 and @xmath258 in the conformal scalar field model . @xmath252 can be written in a form which does not , even implicitly , contain terms proportional to @xmath232 . for the purpose of this subsection it is not necessary to write @xmath252 in a form obviously regular on . + in figure [ chrisinitial ] = @xmath259 is plotted for three outgoing light cones versus the area radius @xmath251 . for a parameter of @xmath260 , a model far beyond the critical value with a distinct singularity in physical spacetime , the solution has been printed out for 100 time slices with 100 grid points . on outgoing null cones @xmath261 has been evaluated . the sufficient condition is satisfied if @xmath262 . the gridpoints are marked with squares , diamonds , and triangles . the first light cone ( squares ) lies outside the event horizon , the criteria is nowhere fulfilled . the second lightcone ( diamonds ) approximately coincides with the event horizon . the criteria is not fulfilled either . only if the apparent horizon is crossed ( triangles ) @xmath263 becomes larger than @xmath29 . although many outgoing lightcones in various singular models have been checked the criteria was only fulfilled when the lightcone crossed the apparent horizon . thus i conclude that the criteria is not very sharp . this is in agreement with the results in @xcite . in addition to the already presented determination of the bondi mass i am going to demonstrate the simplicity of radiation extraction on two more examples in this subsection . in the first case the effects of the nonlinearities in the equations are analyzed the purely backscattered radiation in region iii and the decay of the radiation in region iv of a regular spacetime model . in the second case values of the bondi mass in a singular model are compared with values of the mass read off at finite radii . figure [ radall ] = compares @xmath46 on , i.e. the coefficient of the @xmath264 term of the scalar field @xmath161 in physical spacetime , for a model with an initial amplitude of @xmath265 for the linear model ( @xmath158 , dotted line ) and the nonlinear model ( @xmath266 , continuous line ) . in the model with gravitation a significant amount of radiation , reflected outward by backscattering , has already crossed before the linear signal reaches ( region iii ) . the main signal is stronger in the nonlinear case and there is some scalar field left in region ii . + in the figure [ purebackscatter ] = = the bondi mass is shown in the pure backscatter region iii . later times correspond to the backscattering of a matter shell which has decreased in size during the infall . + the scalar field @xmath46 is displayed in a double logarithmic plot of the scalar field versus the proper time @xmath267 of an observer at . this observer is obtained by taking a world line at fixed area radius @xmath255 ( and angle ) and taking the limit @xmath268 . in the unphysical variables one gets @xmath269 where @xmath270 is the coordinate @xmath76 in unphysical spacetime of at unphysical time @xmath107 and @xmath271.\end{aligned}\ ] ] the dots represent grid points of the calculation with 5000 grid points . since on approach to @xmath1 the center is approached and the values become inaccurate , only those points have been included in the plot where the results of a richardson extrapolation with 5000 and 10000 grid points are visually indistinguishable from those of an extrapolation with 10000 and 20000 grid points . in this plot the last 10 ( of 5000 ) points on before @xmath1 are missing . + the slope in the asymptotical region is approximately @xmath272 . thus for very late times @xmath273 . + unfortunately the bondi mass falls off so fast that rounding errors of the graphic program , with which the calculation of the mass is done and which has only single precision , do not allow to enter the asymptotical region with the constant slope for @xmath46 . + but in spite of the accuracy problem near the center , figure [ purerest ] illustrates an advantage of the conformal method : the time scale of the collapse of the initial matter shell is of order @xmath134 , the fall off of the decay could be investigated for almost @xmath274 times the dynamical time scale . a rough estimate of the errors which are to be expected by reading of radiative quantities at finite radius has been made in the following way : given a lower bound for the area radius , the grid on a time slice has been searched for the first grid point @xmath275 with larger area radius . at that point @xmath275 the value for the hawking mass @xmath276 and the `` finite '' bondi mass , @xmath277 have been calculated . those values have been compared with the bondi mass on the intersection point of the future directed light cone of @xmath275 and @xmath0 . the model used corresponds to the parameter @xmath218 and develops strong gravitational fields , the spacetime has a singularity which catches approximately @xmath278 of the adm mass of @xmath279 . the outer boundary of the compact support of the scalar field lies at an area radius @xmath280 , @xmath275 has been chosen two , five and ten times that value , expressed in units of the adm mass , @xmath275 is at @xmath281 , @xmath282 or @xmath283 . + the maximal relative error , i. e. error over adm mass , is approximately @xmath284 , @xmath285 and @xmath286 . but the relative error in the backscattering region iii ( error / mass loss in region iii ) is by a factor of 10 larger . those errors are the errors made by approximating null infinity by grid which ends at a finite physical distance . there is almost no difference between @xmath276 and @xmath287 . in the work reported in this paper a mathematical formalism , designed to investigate the structure of a spacetime in the large by conformal techniques , has been introduced into numerics . this paper demonstrates that global properties of a spacetime are numerically calculable and must not necessarily be estimated from a numerical simulation by heuristic arguments . + the set of equations can be solved either by a spacelike or a characteristic initial value problem . the form of the equations differs from einstein s equations . the technical tricks developed for the numerical solution of einstein s equations can not be used unchanged . the accuracy problem as described in section [ accuracyproblem ] shows that there is still the potential for further progress . recently , too late to be included in this paper , a new way of discretizing was found , which propagates the regularity condition at the center without switching schemes near the center , as done here . the accuracy has improved by a factor of @xmath281 for models near the critical parameter . + in this formalism important properties of spacetimes are automatically fulfilled , the adm mass is always conserved , the energy radiated away equals the change in the bondi mass , and there is no spurious ingoing wave caused by an unphysical reflection at the outer boundary . furthermore the location of event horizons is straightforward and the calculation yields `` penrose '' diagrams allowing to read off the causal structure of singularities . + the obvious advantages might not be worth the effort for the investigation of astrophysical questions where most of the uncertainty originates in an approximate equation of state . but if one is interested in the use of numerical relativity as a substitute for the lacking experimental approach to the open problems of mathematical relativity , the conformal techniques provide a very promising approach . in this circumstance it should be mentioned that h. friedrich has developed the conformal techniques further @xcite . there the rescaling factor @xmath8 is no longer a variable of the system fixed in a complicated way by the gauge source function @xmath62 . the gauge freedom in the rescaling is fixed directly by specifying @xmath288 , and the problem with spacelike infinity @xmath20 may be solved @xcite . + it is a pleasure for me to thank the relativity group in garching for the cooperation during my ph . d. thesis , j. winicour for his advice on my first steps in numerical relativity during his stay in garching , and w. kley for helpful comments on the manuscript . the signature of the lorentzian metric @xmath23 is @xmath289 . + whenever possible i use abstract indices as described in @xcite . small latin letters denote abstract indices , underlined small latin letters are frame indices . for the components of a tensor with respect to coordinates small greek letters are used . the frame @xmath290 is constructed from the coordinates @xmath291 , @xmath292 denotes an arbitrary frame . in this notation @xmath293 is a covector , @xmath294 a scalar , namely @xmath295 . + @xmath296 is defined to be the action of the vector @xmath297 on the function @xmath298 , i.e. for every covariant derivative @xmath299 : @xmath300 . + the transformation between abstract , coordinate , and frame indices is done by contracting with @xmath292 and @xmath301 . all indices may be raised and lowered with the metric @xmath302 and the inverse @xmath303 . @xmath304 , @xmath148 and @xmath305 are arbitrary indices , e.g. @xmath306 and @xmath307 . + for a frame @xmath292 and a covariant derivative @xmath299 the ricci rotation coefficients are defined as @xmath308 from this definition follows @xmath309 with respect to a coordinate frame @xmath310 the components @xmath311 are the christoffel symbols @xmath312 . + the torsion @xmath51 is defined by @xmath313 the riemann tensor @xmath314 by @xmath315 contraction gives the ricci tensor , @xmath316 and the ricci scalar @xmath317 the einstein tensor is given by @xmath318 the speed of light @xmath174 is set to @xmath134 . the gravitational constant @xmath319 in @xmath320 has been set to @xmath134 or @xmath29 in the calculations , corresponding to the full nonlinear theory or a scalar field on a flat background respectively . from the system ( [ quasys ] ) the following set of equations can be derived , where the abbreviations @xmath321 and the substitutions @xmath322 @xmath323 and @xmath324 which follow from spherical symmetry , have been used . @xmath325 where @xmath256 stands for @xmath326 from ( [ quasys ] ) the following set of constraint equations can be derived : @xmath327 equation ( [ spgl ] ) , which must be fulfilled at one point , becomes @xmath328 because of spherical symmetry there is also the identity @xmath329 c. bona and j. mass , in _ approaches to numerical relativity _ , proceedings of the international workshop on numerical relativity , edited by r. dinverno ( cambridge university press , 1992 ) , pp . 258264 .

we present an analysis of the chiral property of 1621 field galaxies having radial velocity 3000 km s to 5000 km s . a correlation between the chiral symmetry breaking and the preferred alignment of galaxies in the leading and trailing structural modes is studied using chi - square , auto - correlation and the fourier tests . we noticed a good agreement between the random alignment of the position angle ( pa ) distribution and the existence of chirality in both the leading and trailing arm galaxies . chirality is found stronger for the late - type spirals ( sc , scd , sd and sm ) than that of the early - types ( sa , sab , sb and sbc ) . a significant dominance ( 17% 8.5% ) of trailing modes is noticed in the barred spirals . in addition , chirality of field galaxies is found to remain invariant under the global expansion . the pa - distribution of the total trailing arm galaxies is found to be random , whereas preferred alignment is noticed for the total leading arm galaxies . it is found that the rotation axes of leading arm galaxies tend to be oriented perpendicular the equatorial plane . a random alignment is noticed in the pa - distribution of leading and trailing spirals . example.eps gsave newpath 20 20 moveto 20 220 lineto 220 220 lineto 220 20 lineto closepath 2 setlinewidth gsave .4 setgray fill grestore stroke grestore

1307.1830

an object that is not superimposable on its mirror image is termed as chiral object . chiral objects do not show reflection symmetry , but may exhibit rotational symmetry . they are not necessarily asymmetric , because they can have other types of symmetry . an object is achiral ( non - chiral ) if and only if it has an axis of improper rotation ( rotation by 360@xmath2/@xmath3 ) followed by a reflection in the plane perpendicular to this axis . thus , an object is chiral if and only if it lacks such an axis . chiral symmetry breaking or restoration is a key ingredient in different problems of theoretical physics : from nonperturbative quantum cromodynamics to highly doped semiconductors ( garcia - garcia & cuevas 2006 ) . for a pseudoscalar meson , it is found that the chiral symmetry breaking decreases with increasing current - quark mass ( chang et al . bagchi et al . ( 2006 ) studied the large color approximation of the compact strange stars and discussed the chiral symmetry restoration . they claimed that the chiral symmetry restoration can be understood by exploring the possible existence of strange stars in the universe . in principle , microscopic process have triggered huge astrophysical large scale structure ( liddle & lyth 2000 ) . it is therefore interesting to study the breaking or restoration of chiral symmetry not only in microscopic phenomena ( l - neutrino , mesons , l - aminoacids , d - sugars , etc ) , but also in macroscopic ones ( stars , molecular clouds , galaxies ) . according to the basic concepts of cosmology , initial quantum fluctuations have been hugely enhanced during the inflationary epoch , leading the formation of the large scale structure in the universe ( peacock 1999 , liddle & lyth 2000 ) . thus , the macroscopic chirality should be related to some primordial microscopic process which led to the today s observed large scale structures ( fall 1982 ) . differential rotation in a galaxy s disc generate density waves in the disc , leading to spiral arms . according to gravitational theory , the spiral arms born as leading and subsequently transform to trailing modes . with the passage of time , the spiral pattern deteriorates gradually by the differential rotation of the equatorial plane of the galaxy , but the bar structure persists for a long time ( oort 1970a ) . this structure can again regenerate spiral pattern in the outer region . thus , a close relation between the origin of the arms in the spirals and barred spirals can not be denied ( oort 1970b ) . by considering the group of transformations acting on the configuration space , capozziello & lattanzi ( 2006 ) revealed that the spiral galaxies possesses chiral symmetry in the large scale structure . in addition , they predicted that the progressive loss of chirality might have some connection with the rotationally - supported ( spirals , barred spirals ) and randomized stellar systems ( lenticulars , ellipticals ) . aryal , acharya & saurer ( 2007 ) carried out a study to test the capozziello and lattanzi s ( 2006 ) prediction regarding the progressive loss of chirality in the large scale structure . they analysed the distribution of leading and trailing arm galaxies in the local supercluster ( lsc ) and concluded the existence of chiral symmetry for both the spirals and the barred spirals . however , the virgo cluster galaxies show a preferred alignment : the galactic rotation axes of leading and trailing structures are found to lie in the equatorial plane . aryal & saurer ( 2005a ) noticed a preferred alignment for the late - type spirals and barred spirals in the local supercluster . in addition , they found that the spin vector projections of early- and late - type galaxies show opposite alignment . their results hint the existence of the chiral characterization . in this work , we present an analysis of chiral property of field galaxies having radial velocity ( rv ) 3000 km s@xmath0 to 5000 km s@xmath0 . we intend to study the importance of chiral symmetry in order to understand the structural modes of the galaxy . in addition , we expect to study the following : ( 1 ) does chirality exist for the spirals and barred spirals in the field ? ( 2 ) is there any correlation between the chirality and preferred alignment of galaxies ? ( 3 ) does the group of galaxies exhibit chiral characterization ? and finally , ( 4 ) what can we say about the chiral and achiral ( non - chiral ) properties of the large scale structure ? this paper is organized as follows : in sect . 2 we describe the method of data reduction . in sect . 3 we give a brief account of the methods and the statistics used . finally , a discussion of the statistical results and the conclusions are presented in sects . 4 and 5 . ) and trailing ( @xmath4 ) arm field galaxies having rvs in the range 3000 km s@xmath0 to 5000 km s@xmath0 . the morphology ( c ) , radial velocity ( d ) , axial ratio ( e ) and the magnitude ( f ) distribution of leading and trailing arm galaxies in our databse . the statistical @xmath11@xmath5 error bars are shown for the trailing ( @xmath6 ) subsample . the dashed line ( e ) represents the expected distribution . , title="fig:",height=158 ] ) and trailing ( @xmath4 ) arm field galaxies having rvs in the range 3000 km s@xmath0 to 5000 km s@xmath0 . the morphology ( c ) , radial velocity ( d ) , axial ratio ( e ) and the magnitude ( f ) distribution of leading and trailing arm galaxies in our databse . the statistical @xmath11@xmath5 error bars are shown for the trailing ( @xmath6 ) subsample . the dashed line ( e ) represents the expected distribution . , title="fig:",height=170 ] ) and trailing ( @xmath4 ) arm field galaxies having rvs in the range 3000 km s@xmath0 to 5000 km s@xmath0 . the morphology ( c ) , radial velocity ( d ) , axial ratio ( e ) and the magnitude ( f ) distribution of leading and trailing arm galaxies in our databse . the statistical @xmath11@xmath5 error bars are shown for the trailing ( @xmath6 ) subsample . the dashed line ( e ) represents the expected distribution . , title="fig:",height=154 ] ) and trailing ( @xmath4 ) arm field galaxies having rvs in the range 3000 km s@xmath0 to 5000 km s@xmath0 . the morphology ( c ) , radial velocity ( d ) , axial ratio ( e ) and the magnitude ( f ) distribution of leading and trailing arm galaxies in our databse . the statistical @xmath11@xmath5 error bars are shown for the trailing ( @xmath6 ) subsample . the dashed line ( e ) represents the expected distribution . , title="fig:",height=154 ] ) and trailing ( @xmath4 ) arm field galaxies having rvs in the range 3000 km s@xmath0 to 5000 km s@xmath0 . the morphology ( c ) , radial velocity ( d ) , axial ratio ( e ) and the magnitude ( f ) distribution of leading and trailing arm galaxies in our databse . the statistical @xmath11@xmath5 error bars are shown for the trailing ( @xmath6 ) subsample . the dashed line ( e ) represents the expected distribution . , title="fig:",height=154 ] ) and trailing ( @xmath4 ) arm field galaxies having rvs in the range 3000 km s@xmath0 to 5000 km s@xmath0 . the morphology ( c ) , radial velocity ( d ) , axial ratio ( e ) and the magnitude ( f ) distribution of leading and trailing arm galaxies in our databse . the statistical @xmath11@xmath5 error bars are shown for the trailing ( @xmath6 ) subsample . the dashed line ( e ) represents the expected distribution . , title="fig:",height=154 ] the method of data compilation and the identification of the structural modes of the galaxies were the same as in aryal , acharya & saurer ( 2007 ) . we describe the selection criteria and the data reduction process in brief here . a galaxy had to fulfill the following selection criteria in order to be selected : ( 1 ) the rv should lie in the range 3000 km s@xmath0 to 5000 km s@xmath0 , ( 2 ) the morphology should be known , ( 3 ) should not be the cluster galaxy , ( 4 ) the diameters , magnitude and the position angle should be given and , ( 5 ) the arms should be visible . there were two clusters abell 0426 ( @xmath7(j2000 ) = 03@xmath818@xmath9 36.4@xmath10 , @xmath11(j2000 ) = + 41@xmath23054 ) and abell 3627 ( @xmath7(j2000 ) = 16@xmath815@xmath932.8@xmath10 , @xmath11(j2000 ) = 60@xmath25430 ) in our region . these clusters have mean rvs 5366 km s@xmath0 ( 75 @xmath1 5 mpc ) and 4881 km s@xmath0 ( 63 @xmath1 4 mpc ) , respectively ( abell , corwin & olowin 1989 , struble & rodd 1999 ) . we removed the galaxies belong to the cluster abell 0426 using the catalog established by brunzendorf & meusinger ( 1999 ) . for the cluster abell 3627 galaxies , we used photometric atlas of northern bright galaxies ( kodaira , okamura , & ichikawa 1990 ) and uppsala galaxy catalogue ( nilson 1973 ) . there were 174 galaxies belongs to these clusters in our database . the rvs were compiled from las campanas redshift survey ( shectman 1996 ) . the pas and the diameters of galaxies were added from the uppsala galaxy catalogue ( nilson 1973 ) , uppsala obs . general catalogue , addendum ( nilson 1974 ) , photometric atlas of northern bright galaxies ( kodaira , okamura , & ichikawa 1990 ) , eso / uppsala survey of the european southern observatory ( lauberts 1982 ) , southern galaxy catalogue ( corwin et al . 1985 ) and third reference catalogue of bright galaxies ( de vaucouleurs et al . 1991 ) . in the ned ( nasa / ipac extragalactic database , http : //nedwww.ipac.caltech.edu/ ) , 6493 galaxies having rvs 3000 km s@xmath0 to 5000 km s@xmath0 were listed until the cutoff date ( october 2006 ) . the diameters were given for 5324 ( 82% ) galaxies . both the diameters and pas were listed for 3971 ( 61% ) galaxies . morphological information was given in the catalogues for 3276 ( 50% ) galaxies . we inspected the dss images of 3276 galaxies using ned and aladin2.5 softwares . there were 2107 ( 32% ) galaxies having distinct spiral arms in our database . in order to understand true structural modes ( leading or trailing ) of spiral galaxies , we need to know the direction of the spiral pattern ( s- or z - shaped ) , the approaching and receding sides and the near and far parts , since galaxies are commonly inclined in space to the line of the sight . the s and z - patterns can be determined from the image of the galaxy . similarly , the approaching and receding sides can be defined if spectroscopy data on rotation is available . the third one is fairly hard to established . for this , pasha ( 1985 ) used ` tilt ' criteria and studied the sense of winding of the arms in 132 spirals . he found 107 spirals to have trailing arm . it should be remembered that the classical ` tilt ' criteria is based on the visible asymmetry of a dust matter distribution . now it is well known that the dark matter halo dominates the dynamics of the galaxies throughout . in our own galaxy , the observed rotation of the stars and gas clouds indicates that the visible matter is surrounded by a halo of this dark matter containing the major portion of the total galaxy mass and extending very far beyond the visible matter . the nature of dark matter in the galactic halo of spiral galaxies is still undetermined . thomasson et al . ( 1989 ) studied theoretically and performed @xmath12-body simulations in order to understand the formation of spiral structures in retrograde galaxy encounters . interestingly , they noticed the importance of halo mass . they concluded that the spirals having halos with masses larger than the disk mass exhibit leading pattern . thus , the makeup of galactic haloes is important to cosmology in order to understand the evolution of galaxies . sugai & iye ( 1995 ) used statistics and studied the winding sense of galaxies ( s- and z - shaped ) in 9825 spirals . no significant dominance from a random distribution is noticed . aryal & saurer ( 2005a ) studied the spatial orientations of spin vectors of 4073 galaxies in the local supercluster . no preferred alignment is noticed for the total sample . these results hint that the distribution of angular momentum of galaxies is entirely random in two- ( s- and z - shaped ) and three - dimensional ( spin vector ) analysis provided the database is rich . thus , choice of rotation ( approaching or receding , near or far parts ) of individual galaxy might be random for a observer . we used computer and gave random direction to our galaxies . each galaxy got a virtual direction by the computer . after then , the arm patterns ( s- or z - type ) the galaxies were studied visually by one of the authors ( pl ) in order to maintain homogeneity . the digitized sky survey ( dss ) image and the contour maps of the galaxies were studied in order to identify their structural modes . for this , we use midas software . the leading mode is one whose outer tip points towards the direction of galactic rotation ( see fig similarly , the outer tip of trailing mode directs in the direction opposite to the galactic rotation . the re - examination of the arms using iraf software resulted the rejection of more than 23% of the objects . these rejected galaxies were nearly edge - on galaxies . as expected , it was relatively easier to identify the structural modes of nearly face - on than that of nearly edge - on galaxies . in this way , we compiled a database of 1621 galaxies showing either leading or trailing structural mode . * for individual galaxies , these structural modes might be incorrect . as a gross , it might reveal the results in the subjective sense . * there were 807 leading and 821 trailing patterns in our database . all sky distribution of leading and trailing arm galaxies is shown in fig . the symbols @xmath4 " and @xmath13 " represent the positions of the trailing and the leading arm galaxies , respectively . several groups and aggregations of the galaxies can be seen in the figure . the inhomogeneous distribution of the positions of the galaxies might be the selection effects for the galaxy orientation study ( aryal & saurer 2000 ) . the morphology , radial velocity , axial ratio and the magnitude distributions of leading and the trailing arm field galaxies are shown in figs . 1c , d , e , f . the spirals ( 47% ) dominate our database in the morphological distribution ( fig . however , a significant dominance of trailing modes are noticed in the barred spirals whereas a weak dominance of leading modes are found in the spirals . the galaxies in the rv distribution ( @xmath14rv = 500 km s@xmath0 ) were nearly equal ( fig . the axial ratio distribution shows a good agreement with the expected _ cosine _ curve in the limit 0.2 @xmath15 @xmath16/@xmath17 @xmath18 0.9 ( fig . the absolute magnitude lie between 13 and 16 for 82% galaxies in our database ( fig . we classified the database into 34 subsamples for both the leading and trailing modes on the basis of the morphology , radial velocity , area and the group of the galaxies . a statistical study of these subsamples are given in table 1 and discussed in sect . basic statistics is used to study the dominance of leading and trailing modes . at first , morphology and rv dependence is studied concerning the chiral property of galaxies . secondly , sky is divided into 16 equal parts in order to observe the violation of chirality locally . several galaxy groups are identified in the all - sky map where the structural dominance are noticed . finally , we study the dominance of leading or trailing arm galaxies in these groups . we assume isotropic distribution as a theoretical reference and studied the equatorial pa - distribution in the total sample and subsamples . in order to measure the deviation from isotropic distribution we have carried out three statistical tests : chi - square , auto - correlation and the fourier . we set the chi - square probability p(@xmath19 ) = 0.050 as the critical value to discriminate isotropy from anisotropy , this corresponds to a deviation from isotropy at the 2@xmath5 level ( godlowski 1993 ) . auto correlation test takes account the correlation between the number of galaxies in adjoining angular bins . we expect , auto correlation coefficient c@xmath200 for an isotropic distribution . the critical limit is the standard deviation of the correlation coefficient c. if the deviation from isotropy is only slowly varying with angles ( in our case : pa ) the fourier test can be applied . a method of expanding a function by expressing it as an infinite series of periodic functions ( _ sine _ and _ cosine _ ) is called fourier series . let @xmath12 denote the total number of solutions for galaxies in the sample , @xmath12@xmath21 the number of solutions in the k@xmath22 bin , @xmath12@xmath23 the mean number of solutions per bin , and @xmath12@xmath24 the expected number of solutions in the k@xmath22 bin . then the fourier series is given by ( taking first order fourier mode ) , @xmath25 here the angle * @xmath26@xmath21 * represents the polar angle in the k@xmath22 bin . the fourier coefficients @xmath27 and @xmath28 are the parameters of the distributions . we obtain the following expressions for the fourier coefficients @xmath27 and @xmath28 , @xmath29 @xmath30 the standard deviations ( @xmath5(@xmath27 ) ) and ( @xmath5(@xmath28 ) ) can be estimated using the expressions , @xmath31 @xmath32 the probability that the amplitude @xmath33 greater than a certain chosen value is given by the formula @xmath34 with standard deviation @xmath35 the fourier coefficient @xmath27 gives the direction of departure from isotropy . the first order fourier probability function @xmath36(@xmath37@xmath38 ) estimates whether ( smaller value of @xmath36(@xmath37@xmath38 ) or not ( higher value of @xmath36(@xmath37@xmath38 ) a pronounced preferred orientation occurs in the sample . first we present the statistical result concerning the distribution of leading and trailing modes of galaxies in the total sample and subsamples . then , we study the distribution of the leading and trailing arm galaxies in the unit area of the sky and the groups . the equatorial pa - distribution of galaxies in the total sample and subsamples is presented . at the end , a general discussion and a comparison with the previous results will be presented . a statistical comparison between the total sample and subsamples of the leading and trailing arm field galaxies is given in table 1 . 2 shows this comparison in the histogram . the @xmath14(% ) in table 1 and fig . 2 represent the percentage difference between the number of trailing and leading arm galaxies . we studied the standard deviation ( @xmath39 ) of the major diameters ( @xmath17 ) of galaxies in the total sample and subsamples for both the leading and trailing modes . in table 1 , @xmath40 represents the difference between the standard deviation of the major diameters of leading and the trailing arm galaxies . an insignificant difference ( 0.4% @xmath1 0.2% ) between the total number of trailing and the leading arm galaxies are found ( table 1 ) . the difference between the standard deviation of the major diameters ( @xmath40 ) of the trailing and leading arm galaxies is found less than 0.019 ( seventh column , table 1 ) . interestingly , the sum of the major diameters of total trailing and leading arm galaxies coincide . this result strongly suggests the existence of chirality of field galaxies having rv 3000 km s@xmath0 to 4000 km s@xmath0 . in fig . 2 , the slanting - line ( grey - shaded ) region corrosponds to the region showing @xmath18 10% ( 5% ) @xmath14 value . almost all subsamples lie within this region , suggesting the existence of chirality within 10% error limit . we present the distribution of the subsamples of leading and trailing arm galaxies classified according as their morphology , rvs , area and the groups below . in the spirals , leading structural modes are found 3.7% @xmath1 1.8% more than that of trailing modes . the chirality is found stronger for the late - type spirals ( sc , scd , sd and sm ) than that of early - type ( sa , sab , sb and sbc ) : @xmath14 value turned out to be 9.5% ( @xmath1 4.8% ) and 1.8% ( @xmath1 1.0% ) for early- and late - types ( table 1 ) . thus , the late - type spirals are the best candidate of the chiral object in our database . ( % ) = @xmath41-@xmath42 , where @xmath41 and @xmath42 represent the number of trailing and leading arm galaxies , respectively . the statistical error bars @xmath5(% ) shown in the figure are calculated as : @xmath5(% ) = @xmath5/(@xmath43+@xmath44)@xmath45100 , where @xmath5 = ( @xmath43-@xmath44 ) . the grey - shaded and the slanting - line region represent the @xmath18@xmath15% and @xmath18@xmath110% @xmath14 value , respectively . , title="fig:",height=166 ] ( % ) = @xmath41-@xmath42 , where @xmath41 and @xmath42 represent the number of trailing and leading arm galaxies , respectively . the statistical error bars @xmath5(% ) shown in the figure are calculated as : @xmath5(% ) = @xmath5/(@xmath43+@xmath44)@xmath45100 , where @xmath5 = ( @xmath43-@xmath44 ) . the grey - shaded and the slanting - line region represent the @xmath18@xmath15% and @xmath18@xmath110% @xmath14 value , respectively . , title="fig:",height=166 ] @xmath46 p0.23rccrrrrrr sample / subsample & @xmath47 & @xmath42 & @xmath41 & @xmath48 & @xmath46@xmath49 & @xmath46@xmath49 & @xmath46@xmath50a@xmath51 & @xmath46@xmath50a@xmath52 + total & @xmath41 & 814 & 807 & 7 & 0.4 & 0.2 & 0.019 & 0.0 + spiral & @xmath53 & 395 & 367 & -28 & -3.7 & -1.8 & 0.031 & 3.0 + spiral ( early - type ) & @xmath54 & 150 & 124 & -26 & -9.5 & -4.8 & 0.058 & 8.1 + spiral ( late - type ) & @xmath55 & 131 & 126 & -5 & -1.9 & -1.0 & 0.031 & 0.3 + barred spiral & @xmath56 & 191 & 269 & 78 & 17.0 & 8.5 & 0.062 & 15.2 + barred spiral ( early - type ) & @xmath57 & 97 & 140 & 43 & 18.1 & 9.1 & 0.091 & 14.6 + barred spiral ( late - type ) & @xmath58 & 68 & 83 & 15 & 9.9 & 5.0 & 0.066 & 9.2 + unknownmorphology & @xmath59 & 31 & 37 & 6 & 8.8 & 4.4 & 0.095 & 3.9 + irregular & @xmath60 & 20 & 14 & -6 & -17.6 & -8.9 & 0.105 & 5.2 + 3000@xmath15rv ( km s@xmath0)@xmath183500 & @xmath61 & 194 & 208 & -14 & -3.5 & -1.7 & 0.046 & 2.6 + 3500@xmath15rv ( km s@xmath0)@xmath184000 & @xmath62 & 201 & 201 & 0 & 0.0 & 0.0 & 0.031 & 3.1 + 4000@xmath15rv ( km s@xmath0)@xmath184500 & @xmath63 & 182 & 172 & 10 & 2.8 & 1.4 & 0.034 & 0.6 + 4500@xmath15rv ( km s@xmath0)@xmath185000 & @xmath64 & 237 & 226 & 11 & 2.4 & 1.2 & 0.041 & 1.0 + grid 1 & @xmath65 & 21 & 20 & 1 & 2.4 & 1.2 & 0.068 & 6.8 + grid 2 & @xmath66 & 116 & 121 & -5 & -2.1 & -1.1 & 0.014 & 1.6 + grid 3 & @xmath67 & 112 & 88 & 24 & 12.0 & 6.0 & 0.076 & 9.3 + grid 4 & @xmath68 & 14 & 12 & 2 & 7.7 & 3.9 & 0.647 & 9.7 + grid 5 & @xmath69 & 8 & 11 & -3 & -15.8 & -7.9 & 0.042 & 22.3 + grid 6 & @xmath70 & 80 & 75 & 5 & 3.2 & 1.6 & 0.081 & 4.5 + grid 7 & @xmath71 & 56 & 62 & -6 & -5.1 & -2.5 & 0.095 & 8.9 + grid 8 & @xmath72 & 33 & 22 & 11 & 20.0 & 10.1 & 0.073 & 12.9 + grid 9 & @xmath73 & 31 & 20 & 11 & 21.6 & 10.9 & 0.028 & 18.1 + grid 10 & @xmath74 & 108 & 124 & -16 & -6.9 & -3.5 & 0.004 & 5.4 + grid 11 & @xmath75 & 52 & 61 & -9 & -8.0 & -4.0 & 0.025 & 6.2 + grid 12 & @xmath76 & 20 & 20 & 0 & 0.0 & 0.0 & 0.409 & 7.4 + grid 13 & @xmath77 & 66 & 78 & -12 & -8.3 & -4.2 & 0.039 & 2.9 + grid 14 & @xmath78 & 44 & 37 & 7 & 8.6 & 4.3 & 0.356 & 10.3 + grid 15 & @xmath79 & 44 & 47 & -3 & -3.3 & -1.6 & 0.050 & 5.2 + grid 16 & @xmath80 & 9 & 9 & 0 & 0.0 & 0.0 & 0.191 & 8.6 + group 1 & @xmath81 & 30 & 37 & -7 & -10.4 & -5.2 & 0.032 & 7.1 + group 2 & @xmath82 & 70 & 48 & 22 & 18.6 & 9.4 & 0.097 & 12.6 + group 3 & @xmath83 & 37 & 31 & 6 & 8.8 & 4.4 & 0.027 & 4.3 + group 4 & @xmath84 & 40 & 34 & 6 & 8.1 & 4.1 & 0.031 & 3.9 + group 5 & @xmath85 & 85 & 107 & -22 & -11.5 & -5.7 & 0.089 & 11.2 + group 6 & @xmath86 & 45 & 42 & 3 & 3.4 & 1.7 & 0.024 & 1.6 + @xmath46 the dominance of trailing structural modes are significant ( 17% @xmath1 8.5% ) in spiral barred galaxies . the @xmath14 value is found @xmath37 9% for both early- ( sba , sbab , sbb and sbbc ) and late - type ( sbc , sbcd , sbd , sbm ) barred spirals . we suspect that the field sb galaxies are not the best candidates of chiral objects . similar result ( i.e. , @xmath14 @xmath37 8% ) is found for the irregulars and the morphologically unidentified galaxies . one interesting similarity is noticed between the late - type spirals and barred spirals . the @xmath14 value for both the late - types are found less than that of early - types ( see table 1 ) . thus , the chirality is favourable for the late - types rather than the early - types . the difference between the standard deviation of the major diameters ( @xmath40 ) for trailing and leading arm galaxies is found less than 0.050 arc minute for the total sample , spirals and the late - type spirals ( seventh column , table 1 ) . these samples showed @xmath14 value @xmath15 5% ( grey - shaded region , fig . thus , we found a good correlation between the @xmath14(% ) and @xmath40 value . probably , this result hints the fact that the size of the non - superimposable mirror images should lie within a limit . in our database , this limit should not exceed 0.050 arc minute for @xmath40 . the difference between the sum of the major diameters ( in percentage ) are found greater than 10% for the barred spirals and early - type barred spirals . interestingly , these two subsamples showed @xmath14 value greater than 15% ( fig . we suspect that the sb galaxies possess chiral symmetry breaking . thus , we found that the chirality between the total leading and trailing arm galaxies exist in our database . this behavior is found prominent for the spirals , mainly for early - type spirals . a very good correlation between the number of leading and trailing arm galaxies can be seen in the rv classifications ( fig . all 4 subsamples show the @xmath14 and @xmath40 value less than 5% and 0.050 , respectively ( table 1 ) . in addition , @xmath14(@xmath17 ) is found to be @xmath15 5% . this result is important in the sense that the statistics in these subsamples is rich ( number of galaxies @xmath37 170 ) enough . thus , we could not observe the violation of chirality in the low and high rv galaxies in our database . a difference is noticed : dominance of leading and trailing modes in low ( rv1 ) and high ( rv3 , rv4 ) rv subsamples , respectively . however , this dominance is not significant ( i.e. , @xmath14 @xmath15 5% ) . an equal number of leading and trailing arm galaxies are found in the subsample rv2 ( 3500 @xmath15 rv ( km s@xmath0 ) @xmath18 4000 ) ( table 1 ) . this might be a coincidence . in order to check the binning effect , we further classify the total galaxies in 6 ( @xmath14rv = 333 km s@xmath0 ) and 8 bins ( @xmath14rv = 250 km s@xmath0 ) and study the statistics . no significant dominance of either trailing or leading structural modes are noticed . thus , it is found that the chirality of field galaxies remain invariant with the global expansion ( i.e , expansion of the universe ) . this is an important result . we further discuss this result below . we study the distribution of leading and trailing arm galaxies by dividing the sky into 16 equal parts ( fig . the area of the grid ( g ) is 90@xmath2 @xmath45 45@xmath2 ( ra @xmath45 dec ) . the area distribution of leading and trailing arm galaxies are plotted , that can be seen in fig . 3a. the statistical parameters are given in table 1 . 1@xmath5 is shown . the positions of the clusters abell 0426 and abell 3627 are shown by the symbol @xmath45 " ( a , b).,title="fig:",height=173 ] 1@xmath5 is shown . the positions of the clusters abell 0426 and abell 3627 are shown by the symbol @xmath45 " ( a , b).,title="fig:",height=113 ] 1@xmath5 is shown . the positions of the clusters abell 0426 and abell 3627 are shown by the symbol @xmath45 " ( a , b).,title="fig:",height=173 ] 1@xmath5 is shown . the positions of the clusters abell 0426 and abell 3627 are shown by the symbol @xmath45 " ( a , b).,title="fig:",height=113 ] a significant dominance ( @xmath372@xmath5 ) of trailing structures is noticed in grid 3 ( ra : 0@xmath2 to 90@xmath2 , dec : 0@xmath2 to 45@xmath2 ( j2000 ) ) ( fig . 3a , a ) . an elongated subcluster like structure can be seen in this grid . in this grid , @xmath14 , @xmath40 and @xmath87 are found to be 12% @xmath1 6% , 0.076 and 9.3% , respectively . these figures suggest that the galaxies in g3 lost their chiral property . probably , this might be due to the apparent subgroupings or subclusterings of the galaxies . the trailing arm galaxies dominate in the grids g8 and g9 ( fig . 3a ) . however , the statistics is poor ( @xmath15 40 ) in these grids ( table 1 ) . in addition , no groupings or subclustering are noticed . thus , we conclude nothing for these area grids . a dominance ( @xmath881.5@xmath5 ) of leading structures is noticed in g10 ( ra : 180@xmath2 to 270@xmath2 , dec : 45@xmath2 to 0@xmath2 ( j2000 ) ) and g13 ( ra : 270@xmath2 to 360@xmath2 , dec : 90@xmath2 to 45@xmath2 ( j2000 ) ) ( fig . 3a , a ) . in both the grids , a large aggregation of the galaxies can be seen . a subcluster - like aggregation can be seen in g10 . an elongated structure can be seen in g13 . in both the grids , @xmath14 value is found to be greater than 5% ( table 1 ) . thus , we suspect that the galaxies in these grids ( g10 , g13 ) are loosing chiral symmetry . this result might reveal the effects of the cluster evolution on chiral symmetry of galaxies . no dominance of either leading or trailing structures is noticed in the groups g1 , g2 , g4 , g5 , g6 , g7 , g11 , g12 , g14 , g15 and g16 . thus , the chirality is found intact in @xmath88 80% area of the sky . we suspect that the groupings or subclusterings of the galaxies lead the violation of chirality in the grids g3 , g10 and g13 . we study the existence of chirality in the groups below . in all - sky map , several groups of galaxies can be seen ( fig . it is interesting to study the existence of chirality in these groups . for this , we systematically searched for the groups fulfilling following selection criteria : ( a ) major diameter @xmath37 30@xmath2 , ( b ) cutoff diameter @xmath15 2 times the background galaxies , ( c ) number of galaxies @xmath37 50 . we found 6 groups fulfilling these criteria ( fig . all 6 groups ( gr ) are inspected carefully . in 3 groups ( gr2 , gr5 and gr6 ) , subgroups can be seen . the number of galaxies in the groups gr2 and gr5 are found more than 100 . the clusters abell 0426 and abell 3627 are located close to the groups gr2 and gr6 . the symbol @xmath45 " represents the cluster center in fig . the mean radial velocities of these clusters are 5366 km s@xmath0 and 4881 km s@xmath0 , respectively . however , we have removed the member galaxies of these clusters from our database . a significant dominance ( @xmath372@xmath5 ) of trailing structures is noticed in the group gr2 ( fig . 3b , b ) . the @xmath14 , @xmath40 and @xmath87 values are found to be 18.6% @xmath1 9.4% , 0.097 and 12.6% , respectively ( table 1 ) . these values indicate that the galaxies in this group might lost their chiral symmetry . we suspect that the galaxies in this group is under the influence of the cluster abell 0426 , due to which apparent subclustering of the galaxies can be seen in this group . this subclustering lead the violation of chiral symmetry . the galaxies in gr5 shows an opposite preference : a significant dominance of the leading arm galaxies ( @xmath372@xmath5 ) ( fig . 3b , b ) . in this group , @xmath14 , @xmath40 and @xmath87 are found to be 11.5% @xmath1 5.7% , 0.089 and 11.2% , suggesting the violation of chirality ( table 1 ) . no humps or dips can be seen in the groups gr1 , gr3 , gr4 and gr6 ( fig . 3.2b , b ) . thus , the galaxies in these groups show chiral property . it is interesting that the number of galaxies in these groups are less than 100 . the groups gr2 and gr5 , which showed a significant difference between the number of leading and trailing arm galaxies , have a very good statistics ( i.e. , @xmath37 100 ) . thus , we conclude that the large aggregation of the galaxies lead the violation of chirality . in the group 6 , we could not notice the influence of the cluster abell 3627 . this might be due to the off location of the cluster center from the group center . 1@xmath5 error bars are shown . pa = 90@xmath2@xmath145@xmath2 ( grey - shaded region ) corresponds to the galactic rotation axes tend to be oriented perpendicular with respect to the equatorial plane.,title="fig:",height=151 ] 1@xmath5 error bars are shown . pa = 90@xmath2@xmath145@xmath2 ( grey - shaded region ) corresponds to the galactic rotation axes tend to be oriented perpendicular with respect to the equatorial plane.,title="fig:",height=151 ] we study the equatorial position angle ( pa ) distribution of trailing and leading arm galaxies in the total sample and the subsamples . a spatially isotropic distribution is assumed in order to examine non - random effects in the pa - distribution . in order to discriminate the deviation from the randomness , we use three statistical tests : chi - square , auto correlation and the fourier . the bin size was chosen to be 20@xmath2 ( 9 bins ) in all these tests . the statistically poor bins ( number of solution @xmath15 5 ) are omitted in the analysis . the conditions for anisotropy are the following : the chi - square probability p(@xmath19 ) @xmath15 0.050 , correlation coefficient @xmath89/@xmath90 @xmath37 1 , first order fourier coefficient @xmath27/@xmath91 ) @xmath37 1 and the first order fourier probability p(@xmath92)@xmath150.150 as used by godlowski ( 1993 , 1994 ) . table 2 lists the statistical parameters for the total samples and subsamples . in the fourier test , @xmath27 @xmath15 0 ( i.e. , negative ) indicates an excess of galaxies with the galactic plane parallel to the equatorial plane . in other words , a negative @xmath27 suggests that the rotation axis of galaxies tend to be oriented perpendicular with respect to the equatorial plane . because the galactic plane is perpendicular to the rotation axis of the galaxy . similarly , @xmath27 @xmath37 0 ( i.e. , positive ) indicates that the rotation axis of galaxies tend to lie in the equatorial plane . in the histograms ( see figs . 4 - 7 ) , a hump at 90@xmath2@xmath145@xmath2 ( grey - shaded region ) suggests that the galactic planes of galaxies tend to lie in the equatorial plane . in other words , the rotation axes of galaxies tend to be oriented perpendicular with respect to the equatorial plane when there is excess number of solutions in the grey - shaded region in the histograms . all three statistical tests show isotropy in the total trailing arm galaxies . thus , no preferred alignment is noticed for the total trailing arm field galaxies ( solid circles in fig . interestingly , all three statistical tests show anisotropy in the total leading arm galaxies . the chi - square and fourier probabilities ( p@xmath93 , p(@xmath92 ) ) are found 1.5% ( @xmath15 5% limit ) and 8.5% ( @xmath15 15% limit ) , respectively ( table 2 ) . the auto correlation coefficient ( c / c(@xmath5 ) ) turned 3.2 ( @xmath941 ) . the @xmath27/@xmath91 ) value is found to be negative at @xmath88 2@xmath5 level , suggesting that the rotation axes of leading arm galaxies tend to be oriented the equatorial plane . three humps at 50@xmath2 ( @xmath371.5@xmath5 ) , 90@xmath2 ( @xmath372@xmath5 ) and 130@xmath2 ( 1.5@xmath5 ) can be seen ( fig . 4a ) . all these humps lie in the grey - shaded region . we checked the binning biasness in the statistics by increasing the number of bins to 12 and 16 . a similar statistical result is found for both structural modes . 4b shows the pa - distribution histogram for the total sample in 18 bins . the leading arm galaxies show three significant humps in the grey - shaded region , supporting the above mentioned result . thus , we conclude isotropy for trailing whereas anisotropy for leading arm galaxies in the total sample . @xmath46 p0.1ccccccccc sample & & @xmath95 & & & & @xmath96 & & + & @xmath36(>^2)@xmath46 & @xmath97@xmath98 & @xmath46_11@xmath99@xmath50_11@xmath98 & @xmath100>_1@xmath98 & @xmath36(>^2)@xmath46 & @xmath97@xmath98 & @xmath46_11@xmath99@xmath50_11@xmath98 & @xmath100>_1@xmath98 + total & 0.666 & + 0.0 & -0.9 & 0.381 & 0.015 & -3.2 & -1.9 & 0.085 + s & 0.511 & -0.7 & -1.2 & 0.434 & 0.225 & + 0.4 & + 0.8 & 0.383 + se & 0.973 & + 0.1 & -0.9 & 0.569 & 0.031 & + 2.0 & + 2.8 & 0.015 + sl & 0.234 & + 0.5 & + 0.8 & 0.209 & 0.460 & -0.1 & -0.5 & 0.345 + sb & 0.729 & + 0.3 & + 1.0 & 0.454 & 0.285 & -1.0 & -0.2 & 0.497 + sbe & 0.739 & + 0.1 & -0.5 & 0.566 & 0.230 & -0.7 & + 0.1 & 0.521 + sbl & 0.043 & + 1.8 & + 1.7 & 0.046 & 0.620 & -0.9 & -0.2 & 0.872 + rv1 & 0.910 & + 0.3 & + 0.8 & 0.362 & 0.369 & -0.9 & -0.6 & 0.285 + rv2 & 0.790 & + 0.3 & -1.0 & 0.496 & 0.925 & -0.4 & -0.2 & 0.887 + rv3 & 0.050 & + 1.6 & -1.5 & 0.083 & 0.033 & -1.8 & -2.3 & 0.046 + rv4 & 0.043 & -2.3 & -1.5 & 0.116 & 0.636 & + 0.2 & -0.7 & 0.692 + gr2 & 0.455 & + 0.6 & + 0.8 & 0.861 & 0.033 & -1.8 & + 1.7 & 0.116 + gr5 & 0.033 & -1.4 & -2.0 & 0.085 & 0.516 & + 0.4 & -0.4 & 0.548 + @xmath46 in the spirals , the chi - square and auto correlation tests show isotropy for both the trailing and leading modes . the first order fourier probability is found greater than 35% , suggesting no preferred alignment . however , the @xmath27 value exceeds 1@xmath5 limit ( 1.2@xmath5 ) in the trailing spirals . a hump at 90@xmath2 is not enough to turn the @xmath27/@xmath91 ) @xmath37 1.5 ( fig . 5a ) . similarly , a hump at 150@xmath2 is not enough to make the @xmath27/@xmath91 ) @xmath37 1.5 in the leading spirals . hence , the preferred alignment is not profounded in both the leading and trailing spirals . thus , we conclude a random orientation of trailing and leading arm spirals . early- and late - type trailing arm spirals show isotropy in all three statistical tests ( table 2 ) . no humps and the dips are seen in the histograms ( solid circles in fig . thus , the trailing arm spirals show a random alignment in the pa - distribution . in the subsample se , all three statistical tests show anisotropy ( table 2 ) . two significant humps at @xmath37 150@xmath2 cause the first order fourier coefficient ( @xmath27 ) @xmath37 + 2.5@xmath5 ( hollow circle in fig . thus , a preferred alignment is noticed in the early - type leading arm spirals : the galactic rotation axes tend to lie in the equatorial plane . the late - type leading spirals show a random alignment . the spiral barred galaxies show a random alignment in both the trailing and leading modes . in fig . 5d , no deviation from the expected distribution can be seen . all three statistical tests support this result ( table 2 ) . a similar result is found for the early - type sb galaxies in both structural modes ( table 2 , fig . the p@xmath93 and p(@xmath92 ) are found less than 5% , suggesting a preferred alignment for the late - type sb galaxies having trailing arm ( table 2 ) . the auto correlation coefficient ( c / c(@xmath5 ) ) and the hump at @xmath37 150@xmath2 support this result ( fig . 5f ) . the @xmath27/@xmath91 ) is found to be positive at 1.7@xmath5 level , suggesting that the trailing arm sbl galaxies tend to lie in the equatorial plane . thus , the late - type trailing and the leading arm sb galaxies show preferred and random alignments , respectively . the subsamples rv1 and rv2 show isotropy in all three statistical tests ( table 2 ) . no humps or dips can be seen in figs . thus , the galaxies having radial velocity in the range 3000 km s@xmath0 to 4000 km s@xmath0 show a random alignment for both the leading and the trailing structural modes . the humps at 90@xmath2 ( @xmath372@xmath5 ) and 110@xmath2 ( @xmath372@xmath5 ) are found in the leading and trailing arm rv3 galaxies , respectively ( fig . these two significant humps lead the subsample show anisotropy in the statistical tests ( table 2 ) . the @xmath27 values are found negative at @xmath1011.5 level , suggesting a similar preferred alignment for both modes : the galaxy rotation axes tend to be directed perpendicular to the equatorial plane . a hump at 70@xmath2 ( @xmath371.5@xmath5 ) and a dip at 150@xmath2 ( @xmath882@xmath5 ) cause the trailing arm rv4 galaxies to show anisotropy in all three statistical tests ( fig . thus , the trailing arm galaxies having radial velocity in the range 4500 km s@xmath0 to 5000 km s@xmath0 show a similar alignment as shown by the subsample rv3 : galactic planes of galaxies tend to lie in the equatorial plane . the leading arm galaxies in the subsample rv4 show a random alignment ( table 2 , fig . 6d ) . we do not study pa - distribution of leading and trailing arm galaxies in the groups gr1 , gr3 , gr4 and gr6 because of poor statistics ( number @xmath15 50 ) . fortunately , a very good correlation between the leading and trailing arm galaxies are noticed in these groups . in other words , we noticed that the chirality is not violated in these groups . we study the pa - distribution of leading and the trailing arm galaxies in the groups gr2 and gr5 , where the chiral symmetry seem to be violated . in addition , the statistics is relatively better in these two groups . in the group gr2 , leading arm galaxies dominate the trailing arm galaxies . in this group , the leading arm galaxies show a preferred alignment whereas trailing arm galaxies show a random alignment in the pa - distribution . all three statistical tests suggest anisotropy in the leading arm galaxies ( table 2 ) . the humps at @xmath37 150@xmath2 cause the @xmath27 value to be positive at @xmath37 1.5@xmath5 level ( fig . 7a ) , suggesting that the rotation axes of leading arm galaxies in gr2 tend to be oriented parallel the equatorial plane . the trailing arm galaxies dominate in the group gr5 . interestingly , a preferred alignment of trailing arm galaxies is noticed in the pa - distribution . in fig . 7b , two significant humps at 90@xmath2 ( @xmath882@xmath5 ) and 110@xmath2 ( @xmath372@xmath5 ) can be seen . these humps lead the subsample ( trailing gr5 ) to show anisotropy in the statistical tests ( table 2 ) . no preferred alignment is noticed in the leading arm galaxies in this group . thus , the structural modes ( leading or trailing ) whose population dominates in the groups show a preferred alignment in the pa - distribution . this is an interesting result . 8a shows a comparison between the number ( @xmath14 ) and position angle ( @xmath27/@xmath91 ) ) distribution of leading and trailing arm galaxies in the total sample and subsamples . this plot deals the possible correlation between the chirality ( non - chirality ) and the random ( preferred ) alignment in the subsamples . the grey - shaded region represents the region of isotropy and chirality for the @xmath27/@xmath91 ) and @xmath14(% ) , respectively . twenty five ( out of 39 , 64% ) subsamples lie in the grey - shaded region ( fig . 8a ) , suggesting a good agreement between the chiral property and the random alignment of the rotation axes of galaxies . in four subsamples ( se , sbl , gr2 and gr5 ) , a good correlation between the preferred alignment and the achiral ( i.e. , non - chiral ) property is noticed ( fig . thus , it is found that the random alignment of the pas of galaxies lead the existence of chiral property of galaxies . % ) and the position angle ( @xmath27/@xmath91 ) ) distribution of leading and trailing arm galaxies in the total sample and subsamples . ( b ) a comparison with the previous work ( aryal , acharya & saurer 2007 ) . error bars as in fig . 2.,title="fig:",height=181 ] % ) and the position angle ( @xmath27/@xmath91 ) ) distribution of leading and trailing arm galaxies in the total sample and subsamples . ( b ) a comparison with the previous work ( aryal , acharya & saurer 2007 ) . error bars as in fig . 2.,title="fig:",height=188 ] here we discuss our results with the results obtained by aryal et al . aryal et al . ( 2007 , hereafter paper 1 ) compiled a database of 667 leading and trailing arm galaxies in the lsc ( rv @xmath15 3000 km s@xmath0 ) and studied the chiral property in the total sample and 17 subsamples . in the present study , the database is compiled from the field galaxies having rv in the range 3000 to 5000 km s@xmath0 . thus , we are moving deep inside the sky and studying the existence of chiral property in the present work . the distribution of the total trailing and leading structures in the lsc ( paper 1 ) as well as in the field galaxies ( present work ) are found homogeneous ( fig . the difference between the leading and trailing arm galaxies are turned well within 5% error limit . this result indicate the fact that the chiral property is a global phenomenon rather than a local phenomenon . for the spirals , we noticed no deviation from the chirality . however , paper 1 found a slight deviation because of the presence of the virgo cluster . interestingly , we noticed the violation of chirality in the barred spirals ( sb ) whereas paper 1 concluded the sb galaxies as a chiral object . for other morphological types ( irregulars , morphologically unidentified galaxies ) , our result is similar to that of the previous result . in the pa - distribution , we noticed a random alignment for the trailing arm galaxies whereas paper 1 concludes no preferred alignments for the leading arm galaxies . in this work , a preferred alignment is noticed for the leading arm galaxies . in contrast to this , a preferred alignment for the trailing arm galaxies is concluded in paper 1 . this contradiction is interesting in the sense that the nature of the databases are different in the past ( lsc galaxies having rv @xmath15 3000 km s@xmath0 ) and the present ( field galaxies having 3000 @xmath15 rv ( km s@xmath0 ) @xmath18 5000 ) study . the preferred alignment in both the studies are obviously different . for leading arm lsc galaxies , they found that the galactic rotation axes tend to lie in the equatorial plane . in our case , we noticed that the rotation axes of the trailing arm galaxies tend to be oriented perpendicular the equatorial plane . paper 1 concludes a similar preferred alignment in the leading and trailing structures in the virgo cluster galaxies : the galactic rotation axes tend to lie in the equatorial plane . we noticed a similar result for the early - type spirals ( leading mode ) , late - type barred spirals ( trailing mode ) and the group gr2 ( leading mode ) . an important similarity is found in the past and present study : late - type galaxies show chiral property , whereas this property is found to be violated in the early - type galaxies . in this way , we noticed few inconsistencies in the present and the previous work . it should be remembered that these inconsistencies are profounded either due to the poor statistics or because of the bias in the sample classification . we combined the database of the paper 1 and the present work , and studied the basis statistics . a strong chirality ( @xmath14 @xmath15 1% , @xmath40 @xmath15 0.010 and @xmath87 @xmath15 5% ) is found in the total and the spiral galaxies . however , the @xmath14 value is found to be greater than 10% for the barred spirals . thus , the total and the spiral galaxies having rvs less than 5000 km s@xmath0 strongly exhibit chiral symmetry . it seems that the chirality loss sequence ( spiral @xmath102 barred spiral @xmath102 elliptical ) as predicted by capozziello and lattanzi ( 2006 ) might be true . aryal & saurer ( 2004 , 2005b , 2006 ) and aryal , paudel & saurer ( 2007 ) studied the spatial orientation of galaxies in 32 abell clusters of bm type i ( 2004 ) , bm type iii ( 2005b ) , bm type ii - iii ( 2006 ) and bm type ii ( 2007 ) and found a significant preferred alignment in the late - type cluster ( bm type ii - iii , bm type iii ) . they concluded that the randomness decreases systematically in galaxy alignments from early - type ( bm type i , ii ) to late - type ( bm type ii - iii , iii ) clusters . thus , the existance of chirality in bm type i cluster , as predicted by capozziello and lattanzi ( 2006 ) might be true . because we noticed a very good correlation between the randomness and the chirality . probably , this result reveals the fact that the progressive loss of chirality might have some connection with the rotationally supported ( spirals , barred spirals ) to the randomized ( lenticulars , ellipticals ) system . thus , we suspect that the dynamical processes in the cluster evolution ( such as late - type clusters ) give rise to a dynamical loss of chirality . in other word , a good correlation between the achirality and anisotropy can be suspected for the late - type clusters . it would be interesting to test this prediction by analysing the chiral property of spirals in the late - type clusters in the future . we studied the chiral property of 1621 field galaxies having radial velocity ( rv ) in the range 3000 km s@xmath0 to 5000 km s@xmath0 . the distribution of leading and trailing structural modes is studied in the total sample and 34 subsamples . to examine non - random effects , the equatorial position angle ( pa ) distribution of galaxies in the total sample and subsamples are studied . in order to discriminate anisotropy from the isotropy we have performed three statistical tests : chi - square , auto - correlation and the fourier . our results are as follows : 1 . the homogeneous distribution of the total trailing and the total leading arm galaxies is found , suggesting the existence of chiral symmetry in the field galaxies having rvs 3000 km s@xmath0 to 5000 km s@xmath0 . the pa - distribution of trailing arm galaxies is found to be random , whereas preferred alignment is noticed for leading arm galaxies . it is found that the galactic rotation axes of leading structural modes tend to be oriented perpendicular the equatorial plane . leading structural modes are found 3.7% @xmath1 1.8% more than that of the trailing modes in the spirals whereas a significant dominance ( 17% @xmath1 8.5% ) of trailing modes are noticed in the barred spirals . this difference is found @xmath37 8% for the irregulars and the morphologically unidentified galaxies . a random alignment is noticed in the pa - distribution of leading and trailing spirals . thus , it is noticed that the random alignment of the pas of galaxies lead the existence of chiral property of galaxies . in other words , a good correlation between the preferred alignment and the chiral symmetry breaking is found . this result verifies the previous result ( paper 1 ) . the chirality is found stronger for the late - type spirals ( sc , scd , sd and sm ) than that of early - type ( sa , sab , sb and sbc ) . similar result is found for the late - type barred spirals . thus , the late - type galaxies are the best candidates for the chiral objects than that of the early - types . 4 . a very good correlation between the number of leading and trailing arm galaxies are found in the rv subsamples . all 4 subsamples show the @xmath14 value less than 5% . thus , we conclude that the chirality of field galaxies remain invariant with the global expansion the galaxies having rvs 3000 km s@xmath0 to 4000 km s@xmath0 show a random alignment for both the leading and the trailing structural modes . the rotation axes of leading and trailing arm galaxies having 4000 @xmath15 rv ( km s@xmath0 ) @xmath18 4500 tend to be oriented perpendicular the equatorial plane . the galaxies possesses the chiral symmetry in @xmath88 80% area of the sky . this property is found to be violated in few groups of galaxies . two such groups ( gr2 and gr8 ) are identified . in these groups , the structural dominance and the preferred alignments of galaxies are found to oppose each other . aryal et al . ( 2007 ) studied the chiral property of the local supercluster galaxies ( rv @xmath15 3000 km s@xmath0 ) . if we include their database , we found a strong ( @xmath14 @xmath15 1% ) chiral behavior by the total and the spiral galaxies . however , the @xmath14 value is found to be greater than 10% for the barred spirals . hence , the total and the spiral galaxies having rv @xmath15 5000 km s@xmath0 show chiral property . thus , the chirality - loss sequence ( spiral @xmath102 barred spiral @xmath102 elliptical ) as proposed by capozziello and lattanzi ( 2006 ) might be true . the true structural mode of a galaxy must involve a determination of which side of the galaxy is closer to the observer ( binney and tremaine 1987 ) . three - dimensional determination of the leading and the trailing arm patterns in the galaxies is a very important problem . we intend to address this problem in the future . this research has made use of the nasa / ipac extragalactic database ( ned ) which is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration . we acknowledge prof . udayraj khanal and prof . mukunda mani aryal for insightful discussions . one author ( rp ) acknowledge central department of physics , tribhuvan university , kirtipur , for providing various forms of support for their masters thesis . 3 abell , g.o . , corwin , h.g . , olowin , r.p . : _ astrophys . _ * 70 * , 1 ( 1989 ) aryal , b. , paudel , s. , saurer , w. : _ astronom . astrophys . _ * 479 * , 397 ( 2008 ) aryal , b. , acharya , s. , saurer , w. : _ astrophys . space sci . co. tmp.4a , in press ( 2007)(paper 1 ) aryal , b. , saurer , w. : _ monthly notices royal astron . soc . _ * 366 * , 438 ( 2006 ) aryal , b. , saurer , w. : _ astronom . astrophys . _ * 432 * , 841 ( 2005a ) aryal , b. , saurer , w. : _ astronom . astrophys . _ * 432 * , 431 ( 2005b ) aryal , b. , saurer , w. : _ astronom . astrophys . _ * 425 * , 871 ( 2004 ) aryal , b. , saurer , w. : _ astronom . lett . _ * 364 * , l97 ( 2000 ) bagchi , m. et al . : _ astronom . _ * 450 * , 431 ( 2006 ) binney j. , tremaine , s. : _ galactic dynamics _ , princeton univ . press , princeton , new jersey ( 1987 ) brunzendorf , j. , meusinger h. : _ astronom . ser . _ * 139 * , 141 ( 1999 ) capozziello , s. , lattanzi , a. : _ astrophys . space sci . _ * 301 * , 1 - 4 , 189 ( 2006 ) chang , l. et al . rev . c. _ , * 75 * , id015201 ( 2007 ) corwin , h.g . , de vaucouleurs , a. , de vaucouleurs , g. : _ univ . texas monogr . astron . _ * 4 * , 1 ( 1985 ) de vaucouleurs , g. , de vaucouleurs , a. , corwin , et al . : _ third reference catalogue of bright galaxies _ , springer - verlag , new york ( 1991 ) fall , s.m . : in : _ progress in cosmology _ , proc . of the oxford intl . oxford , dordrecht , d. reidel publishing co. , 347 - 356 ( 1982 ) garcia - garcia , a.m. , cuevas , e. : _ phy . b. _ , * 74 * , id113101 ( 2006 ) godlowski , w. : _ monthly notices royal astron . soc . _ * 265 * , 874 ( 1993 ) godlowski , w. : _ monthly notices royal astron . soc . _ * 271 * , 19 ( 1994 ) kodaira , k. , okamura , s. , ichikawa , s. : _ photometric atlas of northern bright galaxies _ , univ . of tokyo press , tokyo ( 1990 ) lauberts , a. : _ eso / uppsala survey of the eso b atlas _ , garching bei muenchen ( 1982 ) liddle , a.r . , lyth , d.h . : _ cosmological inflation and large - scale structure _ , cambridge univ . press , cambridge ( 2000 ) nilson , p. : _ uppsala general catalogue of galaxies _ , nova acta uppsala university , ser . v : a , vol.1 ( 1973 ) nilson , p. : _ upps . _ , 5 ( 1974 ) oort , j.h . : _ science _ , * 170 * , 1363 ( 1970a ) oort , j.h . : _ astronom . _ , * 7 * , 405 ( 1970b ) peacock , j.a . : _ cosmological physics _ , cambridge univ . press , cambridge ( 1999 ) shectman , s.a . , landy , s.d . , oemler , a. et al . : 1996 , _ astrophys . j. _ * 470 * , 172 ( 1996 ) struble , m.f . , rodd , h.j . : _ astrophys . j. supp . _ * 125 * , 355 ( 1999 )

time - resolved spectral analysis , though a very promising method to understand the emission mechanism of gamma - ray bursts ( grbs ) , is difficult to implement in practice because of poor statistics . we present a new method for pulse - wise time - resolved spectral study of the individual pulses of grbs , using the fact that many spectral parameters are either constants or smooth functions of time . we use this method for the two pulses of grb 081221 , the brightest grb with separable pulses . we choose , from the literature , a set of possible models which includes the band model , blackbody with a power - law ( bbpl ) , a collection of black bodies with a smoothly varying temperature profile , along with a power - law ( mbbpl ) , and two blackbodies with a power - law ( 2bbpl ) . first , we perform time - resolved study to confirm the spectral parameter variations , and then construct the new model to perform a joint spectral fit . we find that any photospheric emission in terms of black bodies is required mainly in the rising parts of the pulses and the falling part can be adequately explained in terms of the band model , with the low energy photon index within the regime of synchrotron model . interestingly , we find that 2bbpl is comparable or sometimes even better , though marginally , than the band model , in all episodes . consistent results are also obtained for the brightest grb of fermi era grb 090618 . we point out that the method is generic enough to test any spectral model with well defined parameter variations .

1302.6091

the spectrum of the prompt emission of a gamma - ray burst ( grb ) is generally fitted with the celebrated band spectral model ( band et al . this model represents a non - thermal spectrum , and it can be described in terms of two smoothly joined power laws . generally speaking , band model adequately fits most of the time - integrated prompt emission spectra of grbs , though additional spectral components show up for some grbs ( e.g. , preece et al . 1996 ; gonzalez et al . 2003 ; shirasaki et al . but , to emphasise , the exceptions are very few in number in comparison with the large set of grbs , which can be fitted with a simple band only function ( e.g. , kaneko et al . 2006 ; nava et al . zhang et al . ( 2011 ) have found that 15 out of 17 lat detected grbs could be fitted with band only model ( category i , in their notation ) . hence , the band function , till date , is the simplest , standalone model for grb spectra , whether time - integrated or time - resolved . although the band model is statistically the most appropriate model of grb data , its physical origin is yet to be identified . over the years , many authors have investigated the underlying mechanism of the prompt emission . in the fireball model of grb , the observed radiation during the prompt emission is attributed to a highly relativistic optically thick outflow , which thermalizes photons due to random collisions . this thermal energy is expected to be seen from the photosphere ( goodman 1986 ; paczynski 1986 ; meszaros & rees 2000 ; peer 2008 ) , where the fireball becomes optically thin and radiation decouples from the matter . in this scenario , the light curve ( lc ) is expected to be a single , smooth pulse and the spectrum should be a lorentz boosted bb , while the temperature of the bb adiabatically cools . hence , the time - integrated spectrum should be a superposition of many bbs . though some grbs do have a single pulse , most grb light curves are either superposition of many pulses or they are highly variable . moreover , a bb in the rayleigh - jeans region has a photon index 1.0 , which is much harder than that actually observed for grbs ( @xmath0 ) . many models have been proposed to overcome these difficulties . for example , the internal shock model ( rees & meszaros 1994 ; woods & loeb 1995 ; sari & piran 1997 ; kobayashi et al . 1997 ) assumes that the major radiation is not due to photospheric emission , but optically thin synchrotron radiation ( sr ) from internal shocks . one major problem with the synchrotron radiation is that the low energy photon index is limited to @xmath1 ( preece et al . crider , liang & preece ( 1998 ) , using time - resolved spectra of 99 grbs , have shown that the instantaneous spectra and their evolution can not be explained by sr @xmath2 often crosses the line of death , set by the synchrotron model . another possibility is that the radiation is due to inverse compton ( ic ) of the thermal photons near the photosphere ( thompson 1994 ; peer et al . 2005 , 2006 ; beloborodov 2010 ; lazzati & begelman 2009 ) . this process can indeed produce a band like spectrum , but with a rather hard value of @xmath3 at the best , assuming slow heating . a unique prompt emission model of a grb is yet to be settled . from the phenomenological point of view , the correct model can be identified by segregating the details from the average properties . for example , grbs are superpositions of pulses ( nemiroff 2000 ; norris 2005 ) . hence , one should use the individual pulses for spectral study , instead of the full grb . the next step is to study the spectral evolution within the individual pulses . hence , one should do time - resolved spectroscopy in order to extract greater information than merely an average spectral property of a pulse , e.g. , average peak energy , isotropic energy etc . but , performing such a detailed study is difficult , as one loses photon counts . for example , ghirlanda et al . ( 2010 ) have done time - resolved spectral study of 9 selected grbs , detected by gamma - ray burst monitor ( gbm ) onboard the fermi satellite . as the photon count is low in each time bin , they could model the spectra only with a cut - off powerlaw . parameters of a more complicated model , such as band , can not be well constrained from the time - resolved data . the solution to this problem lies in the realisation that spectral evolution is not totally unpredictable , and one can suitably parametrize this evolution in order to reduce the number of free parameters of the description . for example , the spectral evolution of a grb pulse is generally described as a hard - to - soft evolution ( e.g. , liang & kargatis 1996 ; kocevski & liang 2003 ; nemiroff 2012 ) . recently , basak and rao ( 2012a ; b ) have assumed this hard - to - soft evolution of the individual pulses of the set of 9 grbs of ghirlanda et al . ( 2010 ) to generate simultaneous spectral and timing model of the pulses , with essentially two parameters , namely , the peak energy at the start of the pulse ( @xmath4 ) and the characteristic evolution parameter ( @xmath5 ) . the basic assumptions in this approach , however , are not well established . for example , it was assumed that the spectral softening happens throughout the pulse , though there are evidences that some grb pulses show a different behaviour like the intensity tracking spectral evolution ( see e.g. , lu et al . 2012 ) . further , it was also assumed that the applicable model is the band function throughout . hence , it is essential to critically examine all the applicable spectral models and their evolution to arrive at a correct pulse - wise description of a grb . in this paper , we discuss a new method for pulse - wise spectral analysis where we parametrize the spectral evolution in order to arrive at the correct spectral description with a minimum set of free parameters . we apply this method to study grb 081221 , the brightest grb with clean , separable pulses . we compare the results obtained for this grb with those for grb 090618 the brightest grb in the fermi era . the organisation of this paper is as follows in section 2 , we describe the analysis techniques and the basic assumptions of our model . results are given in section 3 , and major conclusions are drawn in section 4 . the basic necessity for a good spectral analysis of grb pulses is wide band coverage to identify additional spectral components . the gamma - ray burst monitor ( gbm ) on board the fermi satellite , with its wide band width and excellent sensitivity , provides a good data base for such studies . it has two scintillation detectors : the sodium iodide ( nai ) detector is sensitive in the @xmath6 8 kev to @xmath7 900 kev range while the bgo energy range is @xmath7 200 kev to @xmath7 40 mev ( meegan et al . 2009 ) . we examined the nava catalog ( nava et al . 2011 ) of fermi/ gbm grbs and found that there are 112 bright ( fluence @xmath8 erg ) , long ( @xmath9 s ) grbs and 11 of these grbs have single/ separable pulses . grb 081221 is the brightest among them . in figure [ lc ] , we have plotted the light curve ( lc ) of this grb with norris model ( norris et al . 2005 ) fitted for the two pulses . we have also made a systematic analysis of the other 10 grbs and the time - integrated spectral analysis for all of them is given later . we use the cspec data for time - integrated study and the time tag event ( tte ) data for the time - resolved spectral analysis . we choose 2 or more nai detectors having high count rate and one/ both bgo detector ( s ) . for source selection and background subtraction , we use the _ rmfit v3.3pr7 _ tool , developed by user contributions of fermi science support center ( fssc ) . the background exposure time is chosen before and after the burst . this background is modelled by a polynomial of different degrees , according to the need . the pha files are binned in energy channels so as to get a minimum count in each spectral bin . typically , the nai detectors are binned by minimum count @xmath10 , while the bgo detectors are binned by requiring a minimum count of @xmath11 . first , we perform time - integrated analysis for all the 11 bright , long grbs having single/ separable pulses . we implement both c - stat and @xmath12 minimization methods in _ rmfit_. we fit either band or cut - off powerlaw ( cpl ) model . in table 1 , we report the best fit parameter values along with the corresponding 3@xmath13 errors . the reduced c - stat and @xmath14 with degrees of freedom ( dof ) are also reported . for comparison , we quote the results of nava et al . ( 2011 ) for these grbs . it is clearly seen that all these values are matching quite well . the sources of very minor deviations between the values of this work , done by c - stat minimisation and nava et al . ( 2011 ) are : ( i ) mismatch between actual start and stop time , ( ii ) exact background selection and modelling and , ( iii ) the exact number of detectors and the channels used . comparing the deviation of the parameter values , it is clear that the deviation due to using different statistics other than c - stat , i.e. @xmath12 minimisation , is much less than that due to these other reasons . hence , we conclude that the statistics plays a minimal role in actual parameter estimation . in fact , the grbs taken in our analysis are all bright grbs ( fluence @xmath8 erg ) . hence , by default , the @xmath12 minimisation is a correct technique for parameter estimation of grbs with high count rates . we select four models for the spectral study : a ) band model , b ) blackbody with a power - law , c ) a modified blackbody with a power - law and d ) two blackbodies with a power - law . the band function can be written in terms of the spectral indices ( @xmath2 and @xmath15 ) and the peak energy ( @xmath16 ) as : here , @xmath18 is the normalisation constant . a model consisting of thermal ( blackbody with temperature kt ) and non - thermal ( power law with index @xmath19 ) components has been used earlier ( see , e.g. , ryde 2004 , ryde et al . 2006 , ryde and peer 2009 , peer and ryde 2011 ) . we name this function as bbpl . this can be written as : @xmath20}+k_{2}e^{-\gamma } \label{bbpl}\ ] ] where k1 and k2 are normalisation constants . there are suggestions ( e.g. , ryde et al . 2010 ) in the literature of a modified blackbody ( mbb ) , which may exist due to angular dependence of the optical depth and the observed temperature ( peer 2008 ) . hence , we also investigate this model with a powerlaw ( mbbpl ) . the mbb model is a multi - colour blackbody disk model ; the local disk temperature kt(r ) is proportional to @xmath21 . in several grbs there are distinct additional thermal components ( see eg . shirasaki et al . 2008 ) and further , if the grb spectrum is due to thermal inverse compton ( ic ) of seed photons , then there may be , in principle , multiple photon baths . for brevity , we take one more bb component and call it 2bbpl model . this has essentially two bbs with two temperatures @xmath22 and @xmath23 and two normalisations @xmath24 and @xmath25 . we use all the four models for our subsequent analysis . we note here that it is also possible to have combinations of the above models ( like band+pl , band + bb + pl ) , but , at this juncture , we consider only these four generic models . this is essentially because of the fact that the time resolved spectra of grbs generally consists of a broad peak with wings , which can be adequately captured by any of the above four models . time - resolved spectral studies require a large number of parameters . if ` n ' is the number of time bins , then a four - parameter model , such as band or bbpl , requires 4n parameters for a full description . our motivation , in this study , is to reduce the number of parameters with some reasonable assumptions . the basic assumption we make is that the temperature ( kt ) and the peak energy ( @xmath26 ) follow smooth time evolution . this evolution has a break at the peak of the pulse ( see also ryde and peer 2009 ) . in the following discussion we shall refer to them as two episodes rising and falling part . we assume that the time evolution law of kt and @xmath26 is simple powerlaw of time i.e. , @xmath27 , with different @xmath28 in different episodes . if ` m ' and ` n ' are the number of time bins in these two episodes , respectively , then this parametrization reduces the number of free parameters by ` m+n-2 ' . note that , ideally one should use one more parameter to account for the start time , but we have chosen it to be zero for all the cases , except for the rising part of grb 090618 , where the pulse starts from -1 sec . hence , only for this case , we have chosen this start time to be -10 sec . as the parametrization is not corrected for the actual start time , the actual values of @xmath28 are not unique and hence , these values should not be used to compare different pulses . we also assume that for the band model the photon indices ( @xmath2 and @xmath15 ) are constants in each episode . their values can be determined by simultaneously fitting all the time - resolved spectra in an episode , with the photon indices tied . for bbpl model , we assume that the index ( @xmath19 ) of the powerlaw ( pl ) is constant in each episode . this index can be determined in the similar way as the photon indices of the band . the parametrization of the norm of bbpl model is more complicated , as , unlike band , we have two norms . in the band model , norm is a free parameter . in order to have equal number of parameters as band , either we can use an overall free norm with suitable parametrization of bbpl norms , or , we can parametrize one of the norms and treat the other as a free parameter . ryde and peer ( 2009 ) have shown that the parameter @xmath29 either increases with time or remains constant . if we assume that the observed flux of the bb varies as a simple function of time as @xmath30 , then the variation of @xmath31 can be parametrized as @xmath32 . for @xmath33 , @xmath31 will show the expected behaviors . the bb flux variation can be translated to the norm variation of equation [ bbpl ] as @xmath34 . now , if we parametrize the bb norm and treat the powerlaw ( pl ) norm as a free parameter of the model , we have the same number of free parameters as band model . but , this makes the bbpl more constrained , in the sense that the overall norm is not a free parameter , as in the case of band model . to overcome this difficulty , let us assume ( which will be justified later ) that the powerlaw ( pl ) component has norm variation as smooth as that of bb , i.e. , @xmath35 . then the overall norm ( k ) can be made a free parameter by parametrizing the ratio of the bb norm and pl norm as @xmath36 . thus , both band and bbpl models have equal number of free parameters ` m+n+8 ' . for band model , ` m+n ' are norms of band function in ` m+n ' time bins . the other 8 parameters are @xmath2 , @xmath15 , @xmath28 and the @xmath26 at the starting bin , @xmath37 , in the two episodes . peak energy at any time ` t ' is determined by @xmath38 . for the bbpl model , ` m+n ' are overall norms ( k ) and the other four parameters are powerlaw index ( @xmath19 ) , @xmath28 , @xmath39 and kt at the initial bin , kt(@xmath40 ) , in the two episodes . for mbbpl model , we have an added parameter , namely , p , in each episode . hence , the number of parameters is ` m+n+10 ' . for 2bbpl model , the photons are boosted by the same material , hence , the bb parameters can not be arbitrary . we assume that the ratio of temperatures and norms of these bbs are fixed , which should be determined by tying the ratios in all bins . hence , the number of parameters in this model is ` m+n+12 ' . as an example , if there are 5 time bins preceding to the peak of a pulse and 10 bins afterwards , then the parametrization for band and bbpl reduces the number of free parameters from 60 to 23 ; for a total of 25 bins this factor is 100 to 33 and so on . to compare the significance of a model with respect to another , we have performed the f - test . the f value is defined in most general case as , @xmath41 , where index 1 is used for the primary model , while 2 is used for the alternative model . we compute the probably ( p ) of a given f value and thereby find the @xmath13 significance ( and the % confidence level ) of the alternative model preferred over the primary model . if the primary model is a subset of the alternative model , then the f value is defined as , @xmath42 and the difference between the dofs ( m ) is deemed as the dof of the primary model . we start with a time - resolved spectral analysis so that some of the assumptions sketched above can be examined and validated . the major challenge in time - resolved spectroscopy is to define the time bin size , which crucially depends on two factors a ) timescale of spectral evolution and b ) the minimum bin size allowed by the data , which in turn depends on the subjective decision of signal - to - noise ratio ( snr ) . one can not violate the latter condition for a given snr . the peak count rate of this grb is @xmath7 4000 , while we have demanded that each pha bin should have at least 40 counts ( snr @xmath7 6.3 , i.e. , total @xmath7 5000 counts , for 128 channels ) . hence , we can not choose smaller than @xmath7 1 sec time bin . first , we choose uniform time bins of 3 second to extract time - resolved spectra . later , we reduce the time bin to 1 second to check any improvement due to finer time bins . in table [ 081221_a ] , we report the results of band and bbpl fit to the time - resolved data of 3 second bin size . the time bin starts from -1 sec , with 14 bins ( numbered 0 to 13 ) . approximately , first 4 bins belong to pulse 1 , last 8 bins belong to pulse 2 and the 2 intermediate bins belong to the overlapping region . for the bbpl fit , we first fit the spectra with the powerlaw index ( @xmath19 ) free . we note that @xmath19 is more or less constant for the major portion of the burst ( bin # 0 - 2 for pulse 1 and # 6 - 11 for pulse 2 ) . we take the average of @xmath19 over these bins , separately for the two pulses , and found in both cases , @xmath19=1.83 with standard deviation ( @xmath13 ) 0.14 and 0.10 for pulse 1 and 2 , respectively . @xmath19 has large error bars in the last bin of pulse 1 and last 2 bins of the second pulse . the values in the overlapping region ( # 4 and # 5 ) , apart from having large errors , may be ambiguous , and hence neglected . we freeze @xmath19 to 1.83 for all the bins and redo the analysis . the corresponding values are also reported in table [ 081221_a ] . note that , by doing this we are gaining one degree of freedom in each time bin . it is clear from table [ 081221_a ] that for the second pulse , the time - resolved spectral fit with the band model are much better than those of the bbpl fit in terms of the @xmath43 . on the contrary , these values are comparable in the first pulse . hence , the first pulse may be dominated by thermal emission . note , however , that @xmath43 of bbpl is poor in bin # 1 of the first pulse , which has , in fact , the highest flux . hence , this points to a different radiation mechanism than a simple bb , which may appear in the high flux regions . in figure [ chi ] , we have plotted the @xmath43 of bbpl with @xmath19 thawed , bbpl with @xmath19 frozen and band fit with filled circles , open circles and stars , respectively . the bbpl model is clearly inferior to the band model for a major portion of the second pulse , specially in the regions where photon counts are high . in the first pulse ( -1 to 11 sec ) , the bbpl model , except for the one bin is comparable to the band model . we also fit mbbpl and 2bbpl , which are shown by filled boxes and pluses . the mbbpl and 2bbpl are as good as the band model , in terms of @xmath43 . hence , the correct model for the major portion of the burst is either of these three . in the time - resolved analysis of section 3.1.1 , the band and other models show superior fits compared to the bbpl model in terms of @xmath43 . but , as zhang ( 2011 ) pointed out for grb 090902b , this can be an effect of evolution of the bbpl within a time bin . to investigate this , we make finer time bins of 1 second and redo the time - resolved spectroscopy . note that this is the finest bin possible for a snr @xmath7 6.3 . in table [ av_chi ] , we report the average values of @xmath43 for different model fittings , obtained for different bin sizes . apparently , there are improvements in the @xmath43 for finer bins , but we also see that these improvements are of the same orders for different models . hence , it seems that the spectrum is not due to the evolution of bbpl , but one of these other models . but , all of these models are comparable . hence , we can not hope to find the correct model by taking finer bins . presence of a simple blackbody ( bb ) is apparent in the first pulse , with the exception of the second bin , where one of the other models is correct . the second pulse is dominated either by a modified blackbody or a fully non - thermal radiation ( band ) or inverse compton ( 2bbpl ) . in order to find the right answer , one should carefully parametrize the spectral evolution in each episode of the pulses separately and reduce the set of free parameters in the description . the spectral evolution in the pulses are not arbitrary . for example , the temperature of the blackbody evolves with time in a smooth way . this can be seen from figure [ t_kt_epeak ] , where we have plotted the kt of bbpl ( both @xmath19 free and frozen ) by circles . the peak energy variation is shown by pluses . the parameters evolve smoothly as a function of time . for a single pulse , ryde and peer ( 2009 ) have shown that the temperature remains constant or slowly declines with an average powerlaw index , @xmath44 with @xmath45 , during the rise of the pulse and decays faster with an average index @xmath46 with @xmath47 . the break time of this evolution has a strong positive correlation with the pulse peak time . we find a similar behaviour for both kt and @xmath26 evolution . hence , the spectral evolution can be described by a simple time evolution of temperature ( @xmath48 ) or the peak energy ( @xmath49 ) . the index , @xmath28 , in principle may have two values in the two episodes , namely , the rising and the falling part . liang and kargatis ( 1996 ; hereafter lk96 ) showed for fred pulses that the peak energy of the ef(e ) ( or @xmath50 ) spectrum follows a more complicated evolution . @xmath26 decreases exponentially with the running fluence as : @xmath51 where , @xmath4 and @xmath52 are the constants of the @xmath16 evolution law , @xmath53=@xmath54 , is the fluence at time t , @xmath55 being the flux . in figure [ kt_epeak_fluence ] ( upper panels ) , we have plotted the ln ( @xmath26 ) with the fluence . note that the fluence of each pulse is calculated from their respective start time . clearly , the @xmath26 evolution strictly follows the lk96 law in the first pulse . in the second pulse , however , the variation is not smooth throughout . in the falling part , the variation is clearly lk96 type , but in the rising part , the variation is rather `` soft - to - hard '' . this effect may come due to overlap between two pulses . in fact , the first two bins of the second pulse belong to the overlapping region . hence , they might be contaminated with the preceding pulse . however , the third bin , where there should not be any effect of the first pulse , also deviates from the lk96 law . this might indicate that the second pulse is genuinely intensity tracking . kocevski and liang ( 2003 ) argued that the `` intensity tracking '' pulses for which this evolution does not appear very prominent , are rather made of more than one short hard - to - soft pulses . ghirlanda et al . ( 2011 ) , on the other hand , have analysed time - resolved spectra of 11 long and 12 short fermi grbs and found that the long grbs appear to follow a `` soft - hard - soft '' trend , tracking the flux of the grb , rather than a strict `` hard - to - soft '' evolution . lu et al . ( 2012 ) have categorised grb 081221 as one having a strict `` hard - to - soft '' pulse followed by `` intensity - tracking '' pulse . they have simulated overlapping pulses to show that in the overlapping region , the spectral evolution may appear `` intensity - tracking '' . however , they also found some single pulses to have `` intensity - tracking '' spectral evolution . hence , the second pulse may be genuinely `` intensity tracking '' . in figure [ kt_epeak_fluence ] bottom panels , the temperature evolution is plotted against blackbody fluence . here the same behaviour is noticed . hence , kt evolution of the first pulse and the falling part of the second pulse can as well be described by a similar exponential decay : @xmath56 where @xmath57=@xmath54 is the running fluence of the blackbody component at time t , @xmath55 being the flux at @xmath58 . @xmath59 and @xmath60 are the constants of the evolution law . lk96 law is empirical and a simpler version , in principle , can be used instead , such as a simple powerlaw of time . as fluence is a monotonically increasing function of time , either of them can be used for evolution study . the bbpl model has two components . hence , in order to parametrize the norms of this model one has to see the flux evolution of the individual components . in figure [ flux_evolution ] , we have plotted both photon and energy flux of the individual components calculated for 8 - 900 kev energy range . the flux evolutions look similar for both @xmath19 free and frozen cases . interestingly , the pl flux is as smooth as the bb flux , in each pulse . hence , as argued in section 2.3 , we can safely assume that the ratio of their evolutions is a smooth function of time . in figure [ par_evolution ] , we have shown the evolution of @xmath2 , @xmath15 of band and @xmath19 of bbpl . the parameter @xmath15 , in many cases , have either large error bars or only an upper limit could be derived . in some cases , they peg to the value -10 . it is clear from figure [ par_evolution ] that the parameters remain reasonably constant at all episodes ( rising and falling part ) of a pulse . hence , we tie them over all the time bins in a given episode to determine them with greater accuracy . this reduces the number of free parameters of the description of spectral evolution to a great extent , as described in section 2.3 . the fact that the parameters are well behaved functions of time , makes the time - resolved spectroscopy more tractable , as we can reduce the number of free parameters in the pulse - wise description ( see section 2.3 ) . following the parametrization and tying scheme of setion 2.3 , we do the spectral analysis for the individual pulses of the grb . in the following analysis , we use the tte data of nai n0 , n1 , n2 and bgo b0 for our analysis . the constant , which takes care of the relative normalisation of the detectors should not vary throughout the burst . hence , we freeze them to the values obtained in the time - integrated analysis , i.e. , 2.25 , 2.32 , 2.34 and 3.24 respectively . additionally , we make the following changes compared to the time resolved spectral analysis discussed earlier . we divide the data into spectra of equal total counts rather than equal time bins so that equal importance is given to all individual spectra . further , the spectral data in each bin is regrouped into spectral channels to provide an uniform snr . we also note that the 30 to 40 kev region of the spectrum of this grb has the known calibration issues due to k - edge of nai ( see e.g. , guiriec et al . this does not matter much for parameter estimations , but , if one wants to compare different models in terms of @xmath12 then it is wise to neglect these bins . in the following , we have done the spectroscopy by neglecting the 30 - 40 kev band . this pulse constitutes the major portion of the burst . the count rate is @xmath61 times higher than pulse 1 . hence , we can analyse this pulse with greater accuracy and later use our experience to analyse the other one . we perform the analysis for two cases as follows : _ case i analysis for count per time bin @xmath6 3000 _ : this analysis is done by dividing the second pulse from 17.0 s onwards , requiring @xmath6 3000 counts per time bin . we divide the pulse into two parts . 17 to 21.45 s is the rising part and the rest up to 40.45 s is the falling part . in the rising part , we get 3 time bins and the falling part has 9 of them . the spectral bins in the energies > 100 kev sometimes show less than 2@xmath13 count , while < 15 kev show less than 3@xmath13 count . hence , we merge the 8 kev to 15 kev bins to form 1 bin ; 100 to 900 kev bins are merged into 7 bins , with progressively higher binning at higher energies . similarly , spectral bins of the bgo ( 200 kev to 30 mev ) are merged into 5 large bins . all the spectral fit parameters are listed in table [ p2 ] . \(a ) _ rising part _ : in the rising part ( first four rows of table [ p2 ] ) , the bbpl , compared to the other models is inferior with @xmath12 ( dof ) 455.92 ( 354 ) . if we parametrize only the bb norm , and treat the pl norm as free parameter , the corresponding @xmath12 ( dof ) is 487.68 ( 354 ) . a bbpl fitting with no constraint gives @xmath12 ( dof ) = 451.09 ( 348 ) . while we have gained dof , the @xmath14 remains of the same order , which confirms that the parametrization works . in comparison to bbpl , mbbpl is a better fit with @xmath12 ( dof ) = 355.68 ( 353 ) , and significance of 2.55 @xmath13 ( 98.93% confidence ) . band is better than bbpl model , with @xmath12 ( dof ) = 364.41 ( 354 ) , and significance of 2.37 @xmath13 ( 98.23% confidence ) . this suggests that the radiation mechanism in the rising part of the second pulse may be a photospheric emission , but the thermal part is a modified blackbody ( mbb ) rather than a simple bb . however , if we compare these values with a 2bbpl model , then we immediately see that this model is the best with @xmath12 ( dof ) = 351.78 ( 352 ) . as the set of parameters of bbpl model is a subset of 2bbpl , the significance of 2bbpl compared to bbpl is much higher 9.29 @xmath13 ( 100% confidence ) . compared to the band model , 2bbpl model has a significance of 0.86 @xmath13 ( 60.94% confidence ) , which shows that 2bbpl is only marginally better than band model . \(b ) _ falling part _ : in the falling part ( see table [ p2 ] ) , the band model is better compared to the bbpl as well as the mbbpl model . compared to the mbbpl model , the band model has 35.22 less @xmath12 with one more dof . a comparison with 2bbpl model , on the other hand , shows that 2bbpl is better than band model at 1.03 @xmath13 significance ( 69.71% confidence ) . compared to band model , 2bbpl has 40.64 less @xmath12 with two less dof . the band model does not show much difference in terms of residuals of individual spectral fit . but , it is only when we perform a parametrized joint fit , then we realise that the 2bbpl model is marginally better than band model ( table 4 ) . hence , in this region either band or 2bbpl is the best model , with 2bbpl marginally better than band model . in figure [ spectrum ] , we have shown the significance of 2bbpl fitting over the bbpl fitting as a case study . the residual of bbpl model shows excess at various channels . no such structure is visible in the residual of 2bbpl fit . note that the nai k - edge is present in both the residuals between 30 - 40 kev . we have done the fitting both by including and excluding this band . when all the channels are used the @xmath12 ( dof ) of bbpl and 2bbpl are 340.85 ( 217 ) and 239.82 ( 215 ) respectively . 2bbpl is preferred over bbpl at a significance of 8.42 @xmath13 ( 100% confidence , p=@xmath62 ) . if we exclude the 30 - 40 kev bins , the corresponding @xmath12 ( dof ) are 300.26 ( 197 ) and 193.17 ( 195 ) , while 2bbpl is preferred at a significance of 9.01 @xmath13 ( 100% confidence , p=@xmath63 ) _ case ii analysis for count per time bin @xmath6 1000 _ : to check whether lowering the size of the time bins improves the bbpl fitting , we perform the same analysis for count per time bin @xmath6 1000 . as the count rate is lower , we merge the 100 to 900 kev of nai detectors into 5 channels rather than 7 . the rest of the binning remains the same . we obtain 10 bins in the rising and 29 bins in the falling part . \(a ) _ rising part _ : the values of @xmath12 ( dof ) for bbpl , band , mbbpl and 2bbpl are 1247.91 ( 1187 ) , 1328.50 ( 1187 ) , 1239.20 ( 1186 ) and 1222.76 ( 1185 ) , respectively . compared to bbpl model , the band model is preferred at 1.36 @xmath13 ( 82.76% confidence ) , mbbpl is preferred at 1.56 @xmath13 ( 88.17% confidence ) , while 2bbpl is preferred at 9.66 @xmath13 ( 100% confidence ) . 2bbpl is preferred over band model at 0.89 @xmath13 ( 62.60% confidence ) . hence , the conclusions of case i remains unaltered . \(b ) _ falling part _ : similarly , in the falling part , the finer bin makes equal impact on all the models and hence , the conclusion remains the same . note that compared to mbbpl , 2bbpl model has 113.84 less @xmath12 , with the cost of one more dof . hence , 2bbpl model is preferred over mbbpl in the falling part ( 1.31 @xmath13 with 68.99% confidence ) . compared to bbpl model , band , mbbpl and 2bbpl are preferred at 3.12 @xmath13 ( 99.82% confidence ) , 2.67 @xmath13 ( 99.24% confidence ) and 19.61 @xmath13 ( 100% confidence ) , respectively . 2bbpl is marginally better than band at 0.95 @xmath13 ( 65.64% ) . the time - resolved spectra of this pulse are extracted by requiring @xmath64 counts per bin , as the photon count is @xmath65rd of pulse 2 . as before the spectral bins of nai are binned in 8 - 15 kev and 100 - 900 kev , while bgo spectral channels are merged to form 5 broad channels . the 30 - 40 kev band is neglected . the results of different fits are reported in table [ p1 ] . \(a ) _ rising part _ : the rising part has only one bin from -1.0 to 2.15 s. interestingly the bbpl model is marginally better than the band model in the rising part ( see table [ p1 ] ) at 0.87 @xmath13 ( 61.46% confidence ) . the mbbpl model has comparable @xmath12 as bbpl , but with one more parameter . 2bbpl model has a significance of 1.04 @xmath13 ( 70.17% confidence ) compared to band model , while the same model has a significance of 2.19 @xmath13 ( 97.12% confidence ) compared to bbpl model . \(b ) _ falling part _ : in the falling part of this pulse , the bbpl model is no longer the best model . the band is the best model with @xmath12 ( dof ) = 544.65 ( 473 ) . mbbpl and 2bbpl models are comparable to band with @xmath12 ( dof ) 548.22 ( 472 ) and 544.79 ( 471 ) . hence , the spectrum may be still thermal , though the thermal part is no longer a simple bb , but either a multi - color bb ( mbb ) or has multiple spectral component ( one more bb ) or simply synchrotron dominated ( band ) . note also that the low energy photon index ( @xmath66 ) , in the falling part is within the regime of synchrotron model , which is clearly in contrast with the rising part ( @xmath67 ) . this phenomenon of softening of photon index at the falling part of a pulse can be seen for all the pulses ( see table [ p2 ] and [ p1 ] ) . in table [ sigma_level ] , we have listed all the significance levels ( in terms of p - value , sigma level and confidence level c.l . ) of a model over another . model@xmath68 is the primary model , while model@xmath69 is the alternative model . it is clear from the table that 2bbpl model is preferred over band model , though marginally , in some cases . the p values denote the probably that the alternative hypothesis is incorrect . hence , lower the value of p , better is the alternative model over the primary model . only in one case , namely the falling part of pulse 1 , band is preferred over 2bbpl . but the p - value of this case is 0.48 , which signifies that they are only comparable . interestingly , if we use finer bin size , the significance of band and mbbpl over bbpl model decreases in the second pulse . the significance of 2bbpl , however , increases . the smooth variations of the parameters demand that the temperature , kt ( of bbpl or mbbpl or 2bbpl ) or peak energy , @xmath26 ( of band model ) should be continuous functions of time , even during the pulse peak time . hence , these parameters should match at the peak within errors to the one predicted by the empirical law . we follow the evolution law of the rising part to predict these parameters in the first bin of the falling part . we compare these values with the corresponding observed values . in figure [ connection ] , we have plotted the observed values with respect to the predicted values . note that the error bars of the observed values are much less compared to those of the predicted values . the sources of errors in the predicted values are errors in the evolution parameter , @xmath28 and the errors in the actual parameter at the starting bin of the rising part . generally , the parameter @xmath28 has large errors , which affects the errors of the predicted values considerably . the data points are essentially the same for both wider bin ( open circles ) and finer bin ( filled circles ) . the dot - dashed line , which shows the equality of the observation and prediction , goes through all the points . grb 090618 is an interesting grb in many aspects ( for details see ghirlanda et al . 2010 ; rao et al . 2011 ; basak & rao 2012a ) . this is a grb with very high fluence . the redshift ( z ) of this object is 0.54 and the total fluence is @xmath70 erg @xmath71 , when integrated over its duration ( 182.27 s ) . in terms of fluence , this is the brightest among all grbs detected by fermi . this is a very long grb with multiple peaks . it has four broad pulses as follows . pulse 1 is rather a clean precursor from -1 s to 40 s. the second pulse is well separated from this precursor , and occurs from 50 s to 75 s , with two structures in 50 to 61 s. the third pulse , occurring from 75 s to 100 s , is contaminated with the falling part of the second pulse , and the rising part of the fourth pulse . the fourth pulse occurs from 100 s to 124 s. though the secondary pulses ( i.e. , other than the precursor ) are sometimes overlapping , we can still examine the spectral variation in the first pulse and in the major portions of the other pulses . grb 081221 has only one secondary pulse , which makes it more convenient . however , in contrast with grb 090618 , the precursor of grb 081221 has overlaps with the secondary pulse . in order to compare the results of grb 081221 , we shall take the precursor and the second pulse of grb 090618 . the time - resolved spectra of this pulse are obtained by requiring minimum of 1000 counts per bin . we obtain 10 spectra in the rising part ( -1.0 to 14.15 s ) and 11 spectra in the falling part ( 14.15 to 40.85 s ) . for the rising part , the parameters , @xmath28 and @xmath39 are obtained by assuming the start time at -10.0 s ( see section 2.3 ) . the results of spectral fitting by different models are reported in table [ 090618_a ] . the advantage of this pulse 1 over 081221 is it is longer and brighter , enabling us to parametrize the rising part . also , this pulse is fully separated from the secondary events . it is clear from table [ 090618_a ] that bbpl model is inferior to the mbbpl model for this pulse . in case of grb 081221 , we found that the bbpl and mbbpl models are comparable to each other , and marginally better than band model . in this case , we definitely need a mbbpl rather than a bbpl for comparable fit as band . note that 2bbpl model is only comparable , but not better than mbbpl . hence , the spectrum in the rising part may be modified blackbody dominated . in the falling part , mbbpl is again comparable to band model . the 2bbpl model is superior than all the models in the falling part . the same conclusion was drawn for grb 081221 . hence , there is hardly any difference of spectral evolution in the precursor pulse between these two grbs . this pulse is more difficult to analyse , because , it has two small structures in the rising part and two pulses at the peak . hence , we ignore the 50 to 61 s of data and analyse only 61 to 75 s. the falling part covers 64.35 to 74.95 s , where we obtain 24 time - resolved spectra requiring 2000 counts per bin . this , in principle , can be lowered as we have higher count rate , but , following the analysis of grb 081221 , we restrict ourselves to a moderate count per bin . the results of joint spectral fit with various models are reported in table [ 090618_b ] . it is clear from this table that bbpl and mbbpl are not the correct models in the falling part . in comparison band model is much better . however , the 2bbpl model , which shows comparable or better fit than band in all cases , is only comparable to mbbpl in this particular case . this may arise due to the fact that this pulse is actually a combination of two highly overlapping pulses ( see rao et al . hence , we redo the analysis on 11 spectra from 69.25 s to 74.95 s , which covers only the falling part of the second pulse . we obtain the following @xmath12 ( dof ) : 1137.67 ( 991 ) , 1179.13 ( 990 ) and 1156.25 ( 989 ) for band , mbbpl and 2bbpl , respectively . hence , the band and 2bbpl models are comparable , though we can not rule out the possibility of contamination even in this falling part . in the rising part ( see table [ 090618_b ] ) , though mbbpl is better than bbpl , it is inferior to 2bbpl and band . hence , the rising part of this pulse is probably synchrotron dominated . 2bbpl is comparable to band . in summary of this pulse , the whole episode can be described by band model . this is in contrast with the second pulse of grb 081221 , where mbbpl clearly dominates the rising part , and then it is taken over by band . from purely data analysis point of view , there are essentially two parts in a grb spectrum the peak of the spectrum and the wings , which extend to very low and very high energies . the fundamental difference between a band only model and the models with thermal and non - thermal parts is : while band accounts for the peak position , with exponential fall in the wings , the other models have thermal component to account for the peak position , and a powerlaw , falling slower than an exponent , holding the spectra at the wings . in principle , all these spectra should be equally good at the peak position , except for the fact that band and mbb have broader peak than simple bb , while 2bbpl has double hump . hence , the difference between these models arises mainly in the wings . the bb is inferior to the others if the peak is not narrow . as photon count at the peak is larger , the residual should show up immediately . but , this is not easy to see if the difference occurs in the wings . consequently , the three very different models , namely , band , mbbpl and 2bbpl show comparable @xmath12 , while fitting time - resolved data . hence , re - binning at these wings plays a very important role to pin down the correct model . however , we can not expect an order of magnitude improvement in the @xmath12 , because , binning in these wings gives 6 - 8 broad bins ( see section 3.3.1 ) with large errors , while the major statistics comes from the peak position . we have found , in our analysis , that the spectrum changes from one pulse to the other , and even within a pulse . the fact that one of these four models is superior than the others , in a particular episode , points to the fundamental radiation mechanism . we see that this change of superiority is not random . for example , the first pulse of both the grbs have shown that a mbbpl model is better , though marginally , than band in the rising part . for grb 081221 , this could be described even by bbpl model in the rising part , which is really pointing towards the thermal origin of radiation in the first pulses . similar observations are reported in the literature e.g. , ryde et al . ( 2009 ) showed , in the 1 - 3 s time bin of batse detected grb 981021 , that a bbpl is better fit than a band model . note that the low energy photon index ( @xmath2 ) of the band model crosses the synchrotron limit in the rising part , where the thermal models are adequate . however , in the falling part of all the pulses , where band is better than mbbpl , @xmath2 is consistent with the synchrotron limit . hence , we can safely conclude that the radiation mechanism starts with a thermal origin , but is rapidly overtaken by synchrotron mechanism . the first pulse may be dominated by the photospheric emission in all episode , but the second pulse is mostly synchrotron dominated . the second pulse may or may not have a thermal origin . for example , the second pulse of grb 081221 shows a mbbpl model in the rising part , which then becomes synchrotron dominated in the falling part . on the other hand , the second pulse of 090618 is always synchrotron dominated . hence , the transition between these different radiation paradigms is smooth and repeatable . in comparison to the mbbpl model , we note that the 2bbpl is particularly better in all episodes . this model sometimes shows superiority to the band model , even at the falling part of a pulse , except for the second pulse of grb 090618 , though we can not rule out the possibility of two highly overlapping pulses in this particular case . softer component than band were reported for a few batse grbs by preece et al . shirasaki et al . ( 2008 ) , using the time - resolved spectral data of grb 041006 , detected by hete-2 ( 2 kev to 400 kev ) , found multiple spectral components , each having characteristic evolution . after the launch of fermi satellite , these earlier claims were reconfirmed in some cases . for example , guiriec et al . ( 2011 ) , fitting the time - integrated spectrum of grb 100724b , have shown the presence of an additional blackbody ( bb ) component along with the traditional band spectrum ( also see burgess et al . in our analysis , we have used two blackbodies to account for the softer components . the origin of these two components is speculative . they might be different locations of the boosted front of the fireball having same temperature , but different boosting factors . alternatively , they can be different seed photon baths , up - scattered by the bulk material . irrespective of its origin , this model shows superiority to all other models in all episodes . note that , though 2bbpl model has double hump in the peak , one of the peaks may occur in the lower wing ( i.e. , < 15 kev ) . hence , it is easy to identify this model , only if the difference occurs at the peak . figure [ spectrum ] clearly shows the double hump in the residuals of the bbpl fit . hence , it is easy to visualise the 2bbpl model from this figure . the band model , however , has similar residuals as the 2bbpl model . hence , the data are not sufficient to distinguish between these two models , except when we perform a parametrized joint fit . in figure [ spectrum_compare ] , we have shown the marginal superiority of the 2bbpl fit over the band model as a case study of the rising part of the first pulse ( -1.0 to 2.15 s ) of grb 081221 . in the right panels , we have plotted the fitted data neglecting the 30 - 40 kev channels . the upper panels are 2bbpl fits , while the lower panels are band model fits . residuals of the band model shows structures with excesses in 15 kev , 50 - 60 kev , and 150 kev regions . these are not present in the residuals of the 2bbpl model . of course , the difference is not as prominent as the case of figure [ spectrum ] . the 2bbpl model is preferred over the band model at 1.04 @xmath13 ( p=0.298 , 70.17% confidence ) . hence , 2bbpl is only marginally better than the band model . to visualise the evolution of the lower black body component , we have plotted in figure [ lower_bb ] , the residuals of 2bbpl fit , with the lower bb omitted . this technique is well known for finding iron line profile in the inner accretion disks of black holes ( see e.g. , miller 2007 ) . we fit the spectrum with the 2bbpl model and then omit the lower bb . the residual ( expressed as normalized counts kev@xmath72 s@xmath72 ) of the fit clearly shows this bb component . residuals of different detectors are shown by different markers . we have overplotted the lower bb model ( in terms of normalized counts kev@xmath72 s@xmath72 ) on the residual to guide the eye . we have plotted these residuals for second , sixth and ninth time bins from top to bottom panels to show that the bb peaks at lower energies at later times . in summary , we have rigorously used the evolution of parameters in the pulses of a grb to construct various spectral models with a minimum number of parameters . we have constructed band model with parametrized peak evolution and tied photon indices , bbpl with parametrized norm ratio of the bb and pl , parametrized temperature , and tied pl indices . apart from these , we have used mbbpl and 2bbpl , which , other than the parametrizations of bbpl , have tied p indices , and tied ratios of temperatures and norms , respectively . this new method is quite general in the sense that any such model can be incorporated with suitable parametrization . the fact that the parametrization works demands a close look into the theoretical predictions of various radiation models . these models , irrespective of their complexities , should produce such smooth variations of parameters within a pulse of a grb . also , if there is really a transition from one radiation mechanism to another , one should correctly model the mechanism of such transition . the fact that the synchrotron model is applicable at the falling part of the pulses , without invoking any other component , is really intriguing and demands a close look at the predictions of the internal shock model . one of the surprising result obtained in this work is that the 2bbpl model is statistically superior to the other models in most of the episodes in these two grbs . basak & rao ( 2012c ) have used this model for grb 090902b . the residual of bbpl fit clearly shows double humps ( see figure 3 of basak & rao 2012c ) which are taken care by the two peaks of the 2bbpl model . we selected this model purely in a phenomenological and data analysis perspective : distinct blackbody components ( apart from the main peak in the spectrum ) are seen in a few grbs and while looking at the residuals , two humps are clearly discernible in a few time bins . since these two are the brightest grbs for such analysis ( grb 081221 is the brightest grb in the category of grbs with single/ separable pulses and grb 090618 is the brightest grb in the fermi era ) , it is unlikely that we can reinforce this result by analysing data from other grbs . one method could be to get the pulse - wise spectral parameters of a sample of grbs and relate them to other properties of grbs like redshift , afterglow properties etc . this will not only help us to identify the most appropriate spectral description but also to identify the emission mechanism operating during the prompt emission . this research has made use of data obtained through the heasarc online service , provided by the nasa / gsfc , in support of nasa high energy astrophysics programs . 080904 & -4.096 , 21.504 & @xmath2=-1.22@xmath75 & @xmath2=-1.21@xmath76 & @xmath2=-1.14@xmath77 + ( cpl ) & & @xmath78 & @xmath79 & @xmath80 + & & @xmath81=1.08 ( 597 ) & @xmath14=1.23 ( 597 ) & c=1.14(587 ) + 080925 & -3.840 , 32.0 & @xmath2=-1.06@xmath82 & @xmath2=-1.06@xmath82 & @xmath2=-1.03@xmath83 + ( band ) & & @xmath84=-2.34@xmath85 & @xmath84=-2.24@xmath86 & @xmath84=-2.29@xmath87 + & & @xmath88 & @xmath89 & @xmath90 + & & c=1.17 ( 712 ) & @xmath14=1.14 ( 712 ) & c=1.13(716 ) + 081118 & 0.003 , 19.968 & @xmath2=-0.42@xmath91 & @xmath2=-0.37@xmath92 & @xmath2=-0.46@xmath93 + ( band ) & & @xmath84=-2.18@xmath94 & @xmath84=-2.14@xmath95 & @xmath84=-2.29@xmath96 + & & @xmath97 & @xmath98 & @xmath99 + & & c=1.17 ( 716 ) & @xmath14=1.02 ( 716 ) & c=1.16(601 ) + 081207 & 0.003 , 103.426 & @xmath2=-0.58@xmath100 & @xmath2=-0.58@xmath101 & @xmath2=-0.58@xmath102 + ( band ) & & @xmath84=-2.15@xmath103 & @xmath84=-2.13@xmath104 & @xmath84=-2.22@xmath105 + & & @xmath106 & @xmath107 & @xmath108 + & & c=1.43 ( 713 ) & @xmath14=1.02 ( 713 ) & c=1.74(596 ) + 081217 & -28.672 , 29.696 & @xmath2=-1.09@xmath109 & @xmath2=-1.10@xmath110 & @xmath2=-1.05@xmath111 + ( cpl ) & & @xmath112 & @xmath113 & @xmath114 + & & c=1.19 ( 715 ) & @xmath14=1.06 ( 715 ) & c=1.46(599 ) + 081221 & 0.003 , 39.425 & @xmath2=-0.84@xmath115 & @xmath2=-0.84@xmath116 & @xmath2=-0.82@xmath117 + ( band ) & & @xmath84=-4.24@xmath118 & @xmath84=-3.89@xmath119 & @xmath84=-3.73@xmath120 + & & @xmath121 & @xmath122 & @xmath123 + & & c=1.64 ( 595 ) & @xmath14=1.49 ( 595 ) & c=1.67(600 ) + 081222 & -0.768 , 20.736 & @xmath2=-0.89@xmath124 & @xmath2=-0.89@xmath124 & @xmath2=-0.90@xmath83 + ( band ) & & @xmath84=-2.46@xmath125 & @xmath84=-2.32@xmath126 & @xmath84=-2.33@xmath93 + & & @xmath127 & @xmath128 & @xmath129 + & & c=1.12 ( 595 ) & @xmath14=1.07 ( 595 ) & c=1.23(604 ) + 090129 & -0.256 , 16.128 & @xmath2=-1.43@xmath130 & @xmath2=-1.46@xmath131 & @xmath2=-1.46@xmath111 + ( cpl ) & & @xmath132 & @xmath133 & @xmath134 + & & c=1.09 ( 596 ) & @xmath14=1.03 ( 596 ) & c=1.12(602 ) + 091020 & -3.584 , 25.088 & @xmath2=-1.31@xmath140 & @xmath2=-1.32@xmath141 & @xmath2=-1.20@xmath142 + ( cpl@xmath143 & & @xmath144 & @xmath145 & @xmath84=-2.29@xmath146 + & & c=1.03 ( 354 ) & @xmath14=0.95 ( 354 ) & @xmath147 + & & & & c=1.18(354 ) + 091221 & -2.048 , 37.889 & @xmath2=-0.62@xmath148 & @xmath2=-0.62@xmath149 & @xmath2=-0.57@xmath96 + ( band ) & & @xmath84=-2.40@xmath150 & @xmath84=-2.26@xmath151 & @xmath84=-2.22@xmath93 + & & @xmath152 & @xmath153 & @xmath154 + & & c=1.42 ( 474 ) & @xmath14=1.12 ( 474 ) & c=1.44(466 ) + @xmath155 the errors quoted from nava et al . ( 2011 ) are symmetric errors . errors for this work are 3@xmath13 errors . @xmath156 c is the reduced c - stat value , the number in the parentheses are dof . @xmath157 the band spectrum showed unbound 3@xmath13 errors , we found better fit with cpl for this grb band & 623.94 ( 547 ) & @xmath162 & & @xmath163 & @xmath164 & @xmath165 & & & & + bbpl & 661.77 ( 547 ) & @xmath166 & @xmath167 & & & & & @xmath168 & @xmath169 & + mbbpl & 621.93 ( 546 ) & @xmath170 & @xmath171 & & & & @xmath172 & @xmath173 & @xmath174 & + 2bbpl & 624.55 ( 545 ) & @xmath170 & @xmath175 & & & & & @xmath176 & @xmath177 & @xmath178 + + band & 602.53 ( 571 ) & @xmath179 & & @xmath180 & @xmath181 & @xmath182 & & & & + bbpl & 642.37 ( 571 ) & @xmath183 & @xmath184 & & & & & @xmath185 & @xmath186 & + mbbpl & 602.30 ( 570 ) & @xmath187 & @xmath188 & & & & @xmath189 & @xmath190 & @xmath191 & + 2bbpl & 593.74 ( 569 ) & @xmath192 & @xmath193 & & & & & @xmath194 & @xmath195 & @xmath196 + band & 2012.93 ( 1707 ) & @xmath197 & & @xmath198 & @xmath199 & @xmath200 & & & & + bbpl & 2173.47 ( 1707 ) & @xmath201 & @xmath202 & & & & & @xmath203 & @xmath204 & + mbbpl & 2099.21 ( 1706 ) & @xmath205 & @xmath206 & & & & @xmath207 & @xmath208 & @xmath209 & + 2bbpl & 2030.83 ( 1705 ) & @xmath210 & @xmath211 & & & & & @xmath212 & @xmath213 & @xmath214 + + band & 1574.79 ( 1317 ) & @xmath215 & & @xmath216 & @xmath217 & @xmath218 & & & & + bbpl & 2402.54 ( 1317 ) & @xmath219 & @xmath220 & & & & & @xmath221 & @xmath222 & + mbbpl & 1794.05 ( 1316 ) & @xmath223 & @xmath224 & & & & @xmath225 & @xmath226 & @xmath227 & + 2bbpl & 1794.06 ( 1315 ) & @xmath228 & @xmath229 & & & & & @xmath230 & @xmath231 & @xmath232 +

it is investigated whether a massive abelian vector field , whose gauge kinetic function is growing during inflation , can be responsible for the generation of the curvature perturbation in the universe . particle production is studied and it is shown that the vector field can obtain a scale invariant superhorizon spectrum of perturbations with a reasonable choice of kinetic function . after inflation the vector field begins coherent oscillations , during which it corresponds to pressureless isotropic matter . when the vector field dominates the universe its perturbations give rise to the observed curvature perturbation following the curvaton scenario . it is found that this is possible if , after the end of inflation , the mass of the vector field increases at a phase transition at temperature of order 1 tev or lower . inhomogeneous reheating , whereby the vector field modulates the decay rate of the inflaton , is also studied .

0705.3334

observations suggest that the formation of structure in the universe is due to the existence of a primordial spectrum of superhorizon curvature perturbations . the fact that they are superhorizon strongly supports the hypothesis that these perturbations were generated through an acausal process . the most compelling mechanism for this to date is cosmic inflation . during inflation , all light , non - conformally invariant fields undergo particle production and obtain a superhorizon spectrum of perturbations . these perturbations can be responsible for the generation of the curvature perturbation in the universe , if their spectrum is compatible with the observations . traditionally , it has been considered that it is the perturbations of the inflaton field itself , which give rise to the curvature perturbation . however , this inflaton hypothesis typically results in overconstraining inflation model - building , which leads to fine - tuning . for this reason , alternative suggestions have been recently put forward . according to such proposals , the field responsible for the curvature perturbation may have little or nothing to do with the dynamics of inflation . one possibility is to consider a field whose contribution to the density is negligible during inflation but , after the end of inflation , it manages to dominate ( or nearly dominate ) the universe before its decay , thereby imposing its own curvature perturbation spectrum . this is the so - called curvaton hypothesis @xcite . under this hypothesis the fine - tunning problems of inflation are alleviated @xcite , while one can attain inflation at low energy scales @xcite . many suggestions in the literature offer realistic candidates in theories beyond the standard model , which can play the role of the curvaton field . another suggestion along similar lines is that the field responsible for the curvature perturbation is not related to the dynamics of inflation but it affects the reheating process by modulating the decay rate of the inflaton . this is the inhomogeneous reheating mechanism @xcite , which can also allow for low - scale inflation @xcite . until now the literature considers that the curvature perturbation in the universe is due to particle production of a suitable _ scalar field _ , typically through one of the above mechanisms . however , even though theories beyond the standard model ( in particular supersymmetric theories ) contain a plethora of scalar fields , the fact that no scalar field has been observed as yet undermines somewhat the predictability and falsifiability of these models . in contrast , in this paper , we consider the possibility that the curvature perturbation is due to particle production of a _ vector field _ during inflation . a massive vector field is non - conformally invariant and can indeed undergo particle production during inflation . in ref . @xcite this scenario has been investigated for a massive abelian vector field . it was shown that a scale - invariant spectrum of perturbations can be generated provided the mass of the vector field satisfies the condition during inflation , where @xmath0 is the inflationary hubble scale . however , this condition is hard to realise in a theoretically well motivated way . this problem is overcome in this paper by considering a non - trivial evolution of the kinetic term for the vector field , during inflation . in supergravity the kinetic term of vector fields is determined , in general , by the gauge kinetic function which is a holomorphic function of the fields of the theory . we consider a similar setup here and assume that , the kinetic function is dominated by a degree of freedom which varies substantially during inflation , while the cosmological scales exit the horizon . we find that a scale - invariant spectrum of vector field perturbations can be attained , without the need for a negative mass - squared for the vector field , if the kinetic function is growing with time during inflation . we then investigate how such a spectrum of vector field perturbations can give rise to the observed curvature perturbation in the universe . in general , a homogeneous vector field generates an anisotropic pressure , which , if dominant , results in a large - scale anisotropy that contradicts the observations ( isotropy of the cmb ) . this is why , the vector field can not play the role of the inflaton ( see , however , @xcite ) . on the other hand , as shown in ref . @xcite , a massive oscillating vector field has zero average pressure and behaves as pressureless , isotropic matter . thus , it can safely dominate the universe without generating a long - range anisotropy . hence , one can employ the curvaton mechanism to generate the curvature perturbation in the universe , using as curvaton a massive vector field , which has assumed a scale - invariant spectrum of perturbations during inflation . in this paper we study in detail the use of such a vector field as curvaton . one other way to attempt to generate the curvature perturbation from the vector field without the latter ever dominating the universe , is by considering that the vector field controls the decay rate of the inflaton , resulting in inhomogeneous reheating . we briefly investigate this scenario as well . the paper is structured as follows . in sec . ii we derive the equations of motion for the perturbations of a massive vector field with a varying kinetic function and mass . in sec . iii we study particle production during inflation of this vector field and obtain the necessary conditions to attain the desired scale - invariant spectrum . in sec . iv we study the dynamics of the scalar field which controls the kinetic function for our vector field . in sec . v we obtain the spectrum of the produced perturbations in the case when the vector field has constant mass and also when its mass is controlled by the scalar field which also controls the kinetic function . in sec . vi we study analytically the curvaton scenario . by obtaining the energy - momentum tensor for the vector field we find the scaling of its density during and after inflation and reheating . we then implement this to find the parameter space in which the vector field can generate the curvature perturbation . we find that the lower bound on the inflationary scale is too stringent to allow the scenario to work . in sec . vii we employ the mass increment mechanism to lower further the inflationary scale . the mechanism assumes that the mass of the vector field grows at a phase transition after the end of inflation . in sec . viii we study possible complications to the scenario due to the dynamics and to particle production of the scalar field that controls the kinetic function . in sec . ix we present a concrete example of our vector curvaton model , taking all the constraints into account . in sec . x the inhomogeneous reheating mechanism is tried out , using as the inflaton the field that controls the kinetic function . the mechanism is shown to be ineffective . finally , in sec . xi we discuss our results and present our conclusions . throughout the paper we use natural units , where and newton s gravitational constant is , with being the reduced planck mass . the signature of the metric is . consider the following lagrangian density for a massive vector field with mass @xmath1 @xmath2 where @xmath3 is a function of cosmic time @xmath4 reminiscent of the gauge kinetic function in supergravity . being the dilaton . this setup has been used to break the conformality of electromagnetism and generate a primordial magnetic field during inflation @xcite ( see also @xcite ) . ] in general , the mass of the vector field can also depend on time , i.e. . for an abelian field , the field strength tensor is @xmath5 employing the above one obtains the field equations for the vector field : @xmath6 [ f(\partial^\mu a^\nu-\partial^\nu a^\mu)]+m^2a^\nu=0\ , , \hspace{-.5 cm } \label{fe}\ ] ] where is the determinant of the metric tensor . since we are interested in particle production during inflation we assume that , to a good approximation , the spacetime is spatially flat , homogeneous and isotropic . hence , we use the flat - frw metric : @xmath7 where @xmath8 is the scale factor of the universe , @xmath9 are cartesian spatial coordinates with @xmath10 and einstein summation is assumed . employing the above metric into eq . ( [ fe ] ) and following the process detailed in ref . @xcite we obtain the following temporal and spatial components of the field equations respectively : @xmath11 and @xmath12 where the dot denotes derivative with respect to the cosmic time and stands for the divergence or the gradient while is the laplacian . we expect inflation to homogenise the vector field and , therefore , @xmath13\ , . \label{hom0}\ ] ] enforcing this condition into eq . ( [ v=0 ] ) we obtain @xmath14 using eqs . ( [ hom0 ] ) and ( [ at=0 ] ) into eq . ( [ eom ] ) we find @xmath15 the above is reminiscent of the klein - gordon equation of a homogeneous scalar field in an expanding universe , with the crucial difference that the coefficient in the `` friction '' term does not feature a factor of @xmath16 . we are interested in the generation of superhorizon perturbations of the vector field , which might be responsible for the curvature perturbations in the universe . therefore , we perturb the vector field around the homogeneous value @xmath17 as follows : @xmath18 where we took into account eq . ( [ at=0 ] ) . in the above @xmath19 satisfies eq . ( [ eomhom ] ) . in view of eqs . ( [ eomhom ] ) and ( [ pert ] ) , eqs . ( [ v=0 ] ) and ( [ eom ] ) become @xmath20 now , let us switch to momentum space by fourier expanding the perturbations : @xmath21 using the above , eq . ( [ v=0pert ] ) becomes @xmath22 where . using this and eq . ( [ fourier ] ) we can write eq . ( [ eompert ] ) as @xmath23 we can rewrite the above in terms of the components parallel and perpendicular to , defined as : @xmath24 thus , we obtain the following equations of motion for the vector field perturbations in momentum space : @xmath25 \delta\mbox{\boldmath $ \cal a$}^\perp = 0 \label{eomperp}\\ & & \nonumber\\ & & \hspace{-1 cm } \left\{\partial_t^2 + \left[h+\frac{\dot f}{f}+ \frac{\left(2h+2\frac{\dot m}{m}-\frac{\dot f}{f}\right)k^2}{k^2 + \frac{(am)^2}{f}}\right ] \partial_t\,+\right . \nonumber\\ & & \hspace{-1 cm } \left.+\;\frac{m^2}{f}+\left(\frac{k}{a}\right)^2\right\ } \delta\mbox{\boldmath $ \cal a$}^\parallel=0\ , . \label{eomparal}\end{aligned}\ ] ] to investigate particle production during inflation for the vector field we need to solve the equation of motion for the perturbations of the field . the integration constants are then evaluated by matching the solution to the vacuum at early times ( when ) , i.e. by demanding @xmath26 where and we note that at early times the perturbation in question is well within the horizon , which means that and . afterwards we evaluate the solution at late times , when the perturbation is superhorizon in size ( i.e. when ) . the power spectrum is obtained by @xmath27 we assume that , during inflation , @xmath28 is constant . we also assume that @xmath29 is proportional to some power of the scale factor , such that @xmath30 with @xmath31 being a constant . we will concern ourselves only with the transverse component of the vector field perturbations eq . ( [ eomperp ] ) , whose equation of motion we write as @xmath32 \delta{\cal a}_k = 0\ , , \label{eomfin}\ ] ] where @xmath33 is a constant associated with the mass @xmath1 of the vector field ( see below ) and we have dropped the ` @xmath34 ' for simplicity . solving eq . ( [ eomfin ] ) and matching to the vacuum in eq . ( [ match ] ) , we obtain the solution @xmath35 where with @xmath36 we denote the hankel function of the first kind and @xmath37 the above solution at late times approaches @xmath38\!. \label{late1}\end{aligned}\ ] ] hence , using eq . ( [ pa ] ) we find that the dominant contribution to the power spectrum is @xmath39 therefore , we may obtain a scale - invariant spectrum if @xmath40 in this case we find that a scale invariant spectrum of perturbations is recovered with @xmath41 as in the case of a massless scalar field . parameterising the scale dependence of the perturbations in the usual manner @xmath42 and comparing with eq . ( [ paperp ] ) we obtain , for the spectral index , the result @xmath43 where we also used eq . ( [ v ] ) . in the case when we find @xmath44 which , when @xmath45 , is the usual finding in the case of a light scalar field . from eq . ( [ fa ] ) we see that , if @xmath46 constant then @xmath47 and we can have a scale invariant spectrum of perturbations only if ( c.f eq . ( [ v=3/2 ] ) ) , i.e. only if the effective mass - squared of the vector field is negative @xcite . to avoid this , we need to consider that @xmath48 is controlled by a degree of freedom which undergoes non - trivial evolution during inflation , at least during the period when the cosmological scales exit the horizon . this is natural to expect in supergravity . indeed , in supergravity , the scalar fields of the theory receive a contribution to their mass of order the hubble scale @xmath28 during inflation , due to corrections to the scalar potential generated by a non - minimal khler potential @xcite . hence , these scalar fields are expected to evolve substantially during inflation as they fast - roll down the potential slopes . hence , dependence of @xmath29 on these scalar fields is expected to yield naturally during inflation . to parametrise this behaviour we assume that @xmath29 is a function of some scalar field @xmath49 , whose value varies during inflation . the gauge kinetic function in supergravity is a holomorphic function of the fields of the theory . hence , we consider that @xmath50 can be expanded around the origin as , where @xmath51 is some cutoff scale and @xmath52 are constant coefficients . we assume that this sum is dominated by a term of @xmath53-th order , so that we can write @xmath54 where we have absorbed @xmath52 into @xmath51 . inserting the above into eq . ( [ fa ] ) we find @xmath55 let us introduce the following lagrangian density for the scalar field @xmath56 : @xmath57 where , for the scalar potential , we consider @xmath58 where the dots denote higher order terms which stabilise the potential at , such that @xmath59 hence , the kinetic term of the vector field becomes canonical ( @xmath60 ) after @xmath56 settles at its vacuum expectation value ( vev ) . assuming that the field has been homogenised by inflation , its equation of motion , when , is @xmath61 the solution of the above during inflation has a growing mode of the form @xmath62\right\ } , \label{phidt}\ ] ] where @xmath63 is the initial value at the onset of inflation and @xmath64 is the elapsed time . comparing the above with eq . ( [ phia ] ) we find @xmath65 where we considered that @xmath66 . the above means that , if then during inflation . this is naturally expected for scalar fields in supergravity due to corrections introduced to the scalar potential when considering a generic form of the khler potential @xcite . this is the source of the so - called @xmath67-problem , which is endemic to inflation when a scalar field is used to produce the curvature perturbation in the universe . from eq . ( [ phia ] ) it is easy to obtain the number of e - folds it takes for @xmath56 to reach the minimum of @xmath68 : @xmath69 after inflation , which means that @xmath56 rushes toward its vev , , in less than a hubble time . in this section we concentrate on two particular possibilities , which may be realised in this model . other possibilities exist but the following appear to be the most straightforward for investigation . suppose at first that the mass of the vector field is constant . in this case , the mass term in eq . ( [ eomperp ] ) is @xmath70 let us choose @xmath45 . ) , such a choice implies that which means that ( c.f . ( [ fn ] ) ) . this case , therefore , is already explored in ref . @xcite and requires a negative mass - squared for the vector field . ] then the above suggests that , which means that the mass term in eq . ( [ eomperp ] ) scales as the @xmath71 term . thus , the resulting equation of motion is of the form of eq . ( [ eomfin ] ) with @xmath45 and @xmath72 under the substitution : @xmath73 where @xmath74 with @xmath75 being the value of the scale factor at the onset of inflation and we have used eq . ( [ nphi ] ) . the solution of eq . ( [ eomperp ] ) is , therefore , the one described in eq . ( [ solu ] ) , which , at late times , approaches the result in eq . ( [ late1 ] ) , with and @xmath76 . hence , in view of eq . ( [ pa ] ) we obtain the dominant contribution to the power spectrum : @xmath77 thus , when , the power spectrum is approximately scale invariant . in the opposite case , . if these perturbations are to give rise to the curvature perturbations in the universe the cosmological scales should correspond to scales with . hence , we require : @xmath78 where @xmath79 is the scale factor at the time when the scale @xmath80 exits the horizon during inflation , @xmath81 is the scale factor at the end of inflation , the subscript ` * ' denotes the time when the cosmological scales exit the horizon , denotes the total number of e - folds of inflation and we have used eq . ( [ kc ] ) . the condition in eq . ( [ mcons ] ) can be better understood when considering the `` effective '' mass of the vector field during inflation as featured in the equation of motion ( [ eomhom ] ) : @xmath82 where we have used eqs . ( [ kc ] ) and ( [ nphi ] ) with . in view of the above we see that the constraint in eq . ( [ mcons ] ) corresponds to the requirement : @xmath83 if @xmath56 were also responsible for inflation we would have and the above constraint would reed . according to ref . @xcite , satisfying this bound allows the generation of a scale - invariant perturbation spectrum for the longitudinal component of the vector field , which may also be used to generate the curvature perturbation in the universe . however , as discussed in ref . @xcite , such a bound on the mass of the vector field is very hard to satisfy . hence , most probably , @xmath56 needs to be some scalar field other than the inflaton ( see also footnote [ mmp ] ) . in this case too , though , we need inflation not to last too long because the cosmological scales have to exit the horizon while @xmath56 is still rolling . otherwise , the roll of @xmath56 is irrelevant and we are back to the case studied in ref . @xcite . the spectrum is , while when the spectrum becomes approximately scale invariant . hence , the cosmological scales should correspond to momentum @xmath84.,width=302 ] suppose , now that the mass of the vector field is due to an interaction between the former and the scalar field @xmath56 . in this case , the lagrangian of the model is @xmath85 where @xmath68 is given by eq . ( [ vphi ] ) , is the covariant derivative in field space and @xmath86 is the ( gauge ) coupling ( @xmath56 is taken to be real for simplicity ) . then , the mass term of the vector field is @xmath87 i.e. the mass of the vector field , in this case , is . consequently , this time , the mass term in eq . ( [ eomperp ] ) is @xmath88 where we used eqs . ( [ fn ] ) and ( [ phia ] ) . now , one might be interested to obtain a scale invariant spectrum in the same manner as the previous subsection , i.e. by taking . as we have seen , this case corresponds to in eq . ( [ eomfin ] ) . then , according to eq . ( [ v=3/2 ] ) , scale invariance requires . however , in view of eq . ( [ mf1 ] ) , this is only possible when , which is not realistic . thus , it seems that is not realisable in this case . another option is to consider that . from eq . ( [ mf1 ] ) we see that this is possible if either @xmath47 or @xmath89 . the former case implies that ( cf . ( [ phia ] ) ) , which means that ( cf . ( [ fn ] ) ) . this case has been explored in ref . @xcite and requires a negative mass - squared for the vector field . let us then concentrate in the latter case , when . assuming @xmath89 means that eq . ( [ eomhom ] ) becomes @xmath90 where we have used eq . ( [ fa ] ) and that @xmath91 with @xmath92 therefore , the mass term in eq . ( [ eomperp ] ) becomes @xmath93 which suggests that in this case . hence , according to eq . ( [ v=3/2 ] ) , scale invariance requires @xmath94 which means that . in view of the above , the solution to eq . ( [ eom ] ) , during inflation , is @xmath95 where @xmath96 and @xmath97 are constants of integration and , in the proportionality relation , we have considered only the `` growing mode '' . hence we see that , for , the magnitude of the vector field is decreasing . this is undesirable , as will be made clear later ( see footnote [ foot ] ) . hence , we choose , in which case @xmath98 const . to satisfy , therefore , eq . ( [ scinv ] ) , we need to enforce the constraint @xmath99 the above constraint is necessary in order to obtain an approximately scale invariant spectrum of perturbations . however , if these perturbations are to account for the curvature perturbations in the universe then the above constraint is tightened further by spectral index considerations . indeed , from eq . ( [ ns ] ) we readily find @xmath100 hence , the spectrum obtained is blue in contrast to the observational preferences . since @xmath101 is still marginally acceptable and the precision of the observational data is at the level of a few percent , we obtain the following bound @xmath102 one mechanism for generating the curvature perturbation in the universe starting from a superhorizon spectrum of vector field perturbations , follows the curvaton scenario . in this case , the vector field , while subdominant during inflation , may come to dominate ( or nearly dominate ) some time afterwards . when it does so , it imposes its own curvature perturbation onto the universe @xcite . to compute if and when the vector field dominates the universe , in order to imprint its superhorizon perturbation spectrum , we follow the evolution of the energy - momentum tensor of the vector field . using eq . ( [ l ] ) , the energy momentum tensor for @xmath103 is @xmath104 assume that the homogenised vector field lies along the @xmath105-direction @xmath106 then the energy - momentum tensor can be written in the form @xmath107 where @xmath108 with @xmath109 from eq . ( [ tdiag ] ) we see that the energy momentum tensor for our vector field resembles the one of a perfect fluid , with the crucial difference that the pressure along the longitudinal direction is of opposite sign to the pressure along the transverse directions . thus , if the pressure is non - zero and the vector field dominates the universe , then large scale anisotropy will be generated . this is the reason we did not consider that @xmath103 can play the role of the inflaton field in the first place . however , in ref . @xcite it was shown that , once , the vector field undergoes quasi - harmonic oscillations , during which , where the overline denotes average over a large number of oscillations . and @xmath60 , the treatment and the results of ref . @xcite are directly applicable here . ] this result suggests that and the oscillating vector field behaves as isotropic pressureless matter . therefore , it can indeed dominate the radiation background , _ without introducing a large scale anisotropy_. in ref . @xcite it was indeed confirmed that , during the oscillations , the density of the vector field scales as @xmath110 how does the density of the vector field scale before the onset of the oscillations ? by virtue of eqs . ( [ mlight ] ) , ( [ meff1 ] ) and ( [ massless ] ) we have that , @xmath111 during inflation . then , it can be easily shown that eq . ( [ eomhom ] ) suggests that remains frozen . hence , , while @xmath112 where we considered eq . ( [ va ] ) . as shown in ref . @xcite , the scaling of the vector density remains as such after inflation ( when ) as well , provided . thus , we see that , despite the fact that @xmath113 is frozen before the onset of the oscillations , the density of the vector field decreases . using the results in the previous section we can trace the evolution of the density of the vector field during and after inflation . as noted above , to avoid a large scale anisotropy , we need that the vector field begins oscillating before its decay and before it dominates the universe . thus , we require : @xmath114 where @xmath115 and @xmath116 are the decay rates of the inflaton field and the vector curvaton field respectively and the subscript ` dom ' denotes the time when the curvaton dominates the universe ( if it does not decay earlier ) . let us define the density parameter of the vector field as @xmath117 where @xmath118 is the background density typically corresponding to either the oscillating inflaton field or the thermal bath of its decay products . in the standard picture , after the end of inflation the inflaton field undergoes quasi - harmonic oscillations until it decays at reheating . during these coherent oscillations the inflaton corresponds to a collection of massive particles ( inflatons ) which behave like pressureless matter hence , for the background density in this period we have . after the decay of the inflaton ( when ) the universe becomes dominated by the relativistic decay products , in which case . in view of the above and eqs . ( [ raosc ] ) and ( [ rfrz ] ) it is easy to obtain the density parameter of the vector field at the onset of its oscillations ( denoted by ` osc ' ) : @xmath119 similarly , if the curvaton decays before domination , we obtain @xmath120 where ` dec ' denotes the time of the vector field decay ( ) . finally , if the curvaton dominates the universe before its decay ( i.e. ) we find @xmath121 let us now estimate @xmath122 . since during inflation @xmath123 const . hence , using that , we obtain @xmath124 where is the magnitude of the _ physical _ vector field at the onset of inflation . in ref . @xcite it is was explained that @xmath103 is the _ comoving _ vector field , which has the expansion of the universe factored out . in a homogeneous and isotropic universe the associated _ physical _ vector field is @xmath125 this can be understood easily by considering the mass term in the lagrangian in eq . ( [ l ] ) . using the flat frw metric in eq . ( [ frw ] ) one has @xmath126 since the lagrangian corresponds to a physical ( observable ) quantity we readily see that the spatial components of the physical vector field are , as in eq . ( [ w ] ) . note also , that this is the explanation of the explicit appearance of the scale factor in the results shown in eqs . ( [ paflat ] ) , ( [ rkin ] ) and ( [ va ] ) . for example , in view of eq . ( [ pa ] ) , the value of the scale invariant power spectrum of the physical vector field @xmath127 is , i.e. identical to the case of a massless scalar field @xcite . from the above we see that , even though @xmath113 is frozen during inflation , is gradually decreasing , which explains the exponential suppression of @xmath122 in eq . ( [ oend ] ) . the curvature perturbation associated with the vector field is @xmath128 where we considered that , before its decay , the vector field is undergoing coherent oscillations , for which as suggested by eq . ( [ raosc ] ) . since during oscillations we have , we find @xmath129 where we took into account that , during the oscillations , both @xmath130 and @xmath113 obey the same equation of motion , since eq . ( [ eomhom ] ) is linear . we also considered that , where by @xmath131 we denote the amplitude of the oscillations , which is equal to @xmath113 at the onset of the oscillations . before the onset of the oscillations we have , which means that @xmath113 is frozen . however , as evident from eq . ( [ paflat ] ) , @xmath130 grows as . that is , _ although the spectrum of the perturbations of the vector field is scale invariant , its amplitude grows with the scale factor of the universe_. this implies that @xmath132 where we have used that and we have assumed that at all times . the above shows that the growth of the amplitude of the perturbations before the onset of the oscillations is due to the decrease of the physical vector field , according to eq . ( [ w ] ) . using eq . ( [ w ] ) , it is easy to find @xmath133 where we assumed that @xmath113 is frozen throughout inflation . putting together eqs . ( [ zaosc ] ) , ( [ da / a ] ) and ( [ wosc ] ) we obtain @xmath134 substituting from the above @xmath135 into eq . ( [ oend ] ) we get @xmath136 which , remarkably , is independent of @xmath137 . plugging eq . ( [ oend1 ] ) into eqs . ( [ odec ] ) and ( [ hdom ] ) we find that , if the vector curvaton decays before domination @xmath138 while if the vector curvaton dominates the universe before its decay @xmath139 solving eqs . ( [ odec1 ] ) and ( [ hdom1 ] ) for @xmath0 we obtain @xmath140 where we used the fact that , in the curvaton mechanism , where is the observed curvature perturbation . now , considering that , and , it can be easily verified that @xmath141 the lower bound in the above is attained when , , and . this case corresponds to almost prompt reheating and curvaton decay as soon as the latter dominates the universe . demanding that the decay of the curvaton occurs before big bang nucleosynthesis ( bbn ) imposes the bound , which suggests @xmath142 where is the temperature at bbn . hence , under this mechanism , the inflationary energy scale can not be lower than , which agrees with the generic bound for the curvaton mechanism @xcite . however , there is an important subtlety that needs to be considered here . even though @xmath137 drops our from the calculations , one still must take into account the evolution of @xmath143 during inflation . this is because , in the above , we have assumed that @xmath144 decreases as @xmath145 since @xmath113 is frozen . however , because , the decrease of @xmath144 will be halted if . thus , we need to postulate that . note that this bound can be much more stringent if @xmath113 is not frozen but diminishes with time . this is why we have chosen in sec . [ hvf ] . ] for this we need to obtain an estimate of @xmath146 . as discussed above , the parameter space for @xmath0 is maximised if the vector curvaton decays when it is about to dominate the density of the universe . this means that , after the decay of the inflaton field , the universe remains radiation dominated , in which case , the number of e - folds corresponding to the horizon at present is @xmath147 the parameter space for @xmath0 is maximised when , which also results in minimising @xmath148 . now , postulating and considering ( so that inflation solves the horizon and flatness problems ) we obtain the bound : ! such huge values of @xmath137 are unacceptably unrealistic . if , on the other hand , we demand then we find that the bound requires @xmath149 in view of eqs . ( [ h*bound ] ) and ( [ nh ] ) , the above bound can not satisfy , i.e. inflation is not enough to solve the horizon and flatness problems . it seems , therefore , that some modification is required , which will allow low scale inflation , for the vector curvaton scenario to work . in ref . @xcite the possibility of low scale inflation in the context of the curvaton mechanism was investigated . it was shown that this is indeed possible in two ways . one possibility is to consider as curvaton a pseudo - nambu goldstone boson , whose order parameter increases after the cosmological scales exit the horizon during inflation . this mechanism was implemented in ref . @xcite and it was shown that inflation with @xmath0 at least as low as 1 tev was possible to attain . the other technique involves a phase transition after the end of inflation , which gives rise to a sudden increment of the curvaton s mass ( see also ref . it is this mechanism that we attempt to implement in this paper to the case of a vector curvaton . we assume , therefore , that a phase transition takes place at some time after the end of inflation but before the onset of the vector field oscillations . the mass of the vector field is increased from @xmath1 to @xmath150 at this phase transition to become larger than the hubble scale at the time , so that oscillations begin immediately . hence , the phase transition corresponds to hubble scale . the sudden increment of the mass of the vector field results in a corresponding growth of the density of the vector field . since , before the oscillations ( c.f . ( [ va ] ) ) , we find that @xmath151 grows by a factor of @xmath152 . hence , we have @xmath153 the above is directly obtainable by eq . ( [ oosc ] ) with the substitution and taking also the growth factor @xmath152 into account . an important constraint here is that the increment of the density of the vector field does not surpass the overall density available at the phase transition , i.e. @xmath154 using the above , in the case when the curvaton decays before domination , we obtain @xmath155 while in the case when the curvaton dominates the universe before its decay @xmath156 note that , in the above , @xmath122 is still given by eq . ( [ oend ] ) . it is easy to see that eqs . ( [ wosc ] ) and ( [ z ] ) remain unaffected by the mass increment , apart from the substitution . thus , we have @xmath157 and @xmath158 then , working as in the previous section , we obtain @xmath159 which is the equivalent of eq . ( [ oend1 ] ) . using this we find that , if the vector curvaton decays before domination @xmath160 while if the vector curvaton dominates the universe before its decay @xmath161 solving eqs . ( [ odec1 + ] ) and ( [ hdom1 + ] ) for @xmath0 we obtain @xmath162 where we used the fact that , in the curvaton mechanism . comparing the above with eq . ( [ h * ] ) we see that , apart from the substitution , there is an extra factor of @xmath163 in the right - hand - side . this means that , if , the lower bound on @xmath0 can be substantially relaxed to the desired level . in contrast to the previous section , due to the extra factor of @xmath163 , the lower bound on @xmath0 can be more relaxed the _ latter _ the oscillations begin . hence , the lowest bound is found when . indeed , in this case it is easy to find @xmath164 the above shown bound corresponds to and , i.e. to the case when the phase transition ( which results in the oscillations of the vector curvaton ) takes place just before the latter decays and as soon as it dominates the universe . since we need some oscillations before the curvaton domination and decay , in order to avoid a long - range anisotropy , the above lower bound is unattainable . the decay rate of the vector curvaton is @xmath165 where the lower bound to the decay coupling @xmath166 corresponds to gravitational decay . using this we have @xmath167 inserting the above into eq . ( [ h*+ ] ) and after a little algebra we obtain @xmath168 where the lower bound is attained when the vector curvaton decays gravitationally . now , from eqs . ( [ osccons ] ) and ( [ oend1 + ] ) we get @xmath169 where we used . the upper bound corresponds the the case when the density of the oscillating vector field dominates the universe immediately after the phase transition . using the above , eq . ( [ h*bound+1 ] ) results in the bound @xmath170 the above bound suggests that the mass increment mechanism can relax the lower bound on @xmath0 _ only if the phase transition occurs much later than the end of inflation_. to show this , consider the opposite case , when . in this case , and considering also that , we find @xmath171 which is not too different from the bound in eq . ( [ h*bound ] ) . let us investigate now whether , under the mass increment mechanism , it is possible to achieve enough e - folds of inflation to solve the horizon and flatness problems while generating the observed amplitude for the curvature perturbation . to maximise the parameter space , we assume , i.e. . also , we consider , which means that the vector curvaton decays as soon as it dominates the universe . finally , since the bounds on the inflationary scale are relaxed for small values of @xmath172 , we assume , that is the phase transition occurs after the decay of the inflaton field . under the above assumptions eq . ( [ z+ ] ) gives @xmath173 while the bound in eq . ( [ h*bound+ ] ) can be written as @xmath174 now , writing eq . ( [ nh ] ) as @xmath175 and using eqs . ( [ ntot1 ] ) and ( [ h*bound+2 ] ) , the requirement results in the bound : @xmath176 taking and also ( prompt reheating ) we find that the phase transition , which results to the growth of the mass of the vector field , can take place at temperature @xmath177 the above upper bound can be saturated when the bound in eq . ( [ h*bound+2 ] ) is saturated , i.e. when the vector curvaton decays gravitationally . thus , we see that it is indeed possible to attain enough inflation to solve the horizon and flatness problems and explain the curvature perturbations in the universe , when the phase transition , which results in mass increment for the vector field , occurs around the time of the breakdown of electroweak unification . apart from the above considerations there are a couple of issues regarding the scalar field @xmath56 , whose evolution is crucial during inflation , since it controls @xmath50 . one issue that needs to be examined is whether @xmath56 also manages to obtain a superhorizon spectrum of perturbations and , if so , whether they may give rise to an acceptable or not contribution to the curvature perturbation . being tachyonic , @xmath56 is guaranteed to undergo particle production during inflation . one can understand this as follows . from eq . ( [ eomphi ] ) one obtains the following equation of motion for the fourier modes of the perturbation @xmath178 of the field @xmath179 \delta\varphi_k=0\ , , & \label{eomphi}\end{aligned}\ ] ] where @xmath180 solving eq . ( [ eomphi ] ) with vacuum boundary conditions in the same manner as in sec . [ pp ] one obtains the following power spectrum @xmath181 where @xmath182 where we used eq . ( [ mphih ] ) , taking as discussed in sec . [ sp ] . from eq . ( [ pphi ] ) it is evident that a scale invariant spectrum is attainable only if . however , as suggested by eq . ( [ vphi ] ) , such a spectrum is attainable only if @xmath53 is very large . for example , if , as discussed in sec . [ hvf ] , then and . if such a spectrum of perturbations contributed significantly to the curvature perturbation then it would be incompatible with the observations . the contribution of the perturbations of @xmath56 to the curvature perturbation is @xmath183 where we considered that and also defined the density parameter of @xmath56 as @xmath184 where we used that , during inflation [ c.f . ( [ v0 ] ) ] and , according to eq . ( [ mphih ] ) . we need to make sure that @xmath56 does not produce an excessive curvature perturbation compared to the observations , which suggest . thus , avoiding conflict with observations is guaranteed if , i.e. @xmath185 note that , as the roll of @xmath56 towards @xmath51 progresses , the associated curvature perturbation is diminished , not only because @xmath56 grows but also because the spectrum of @xmath178 is red . hence , the above bound on @xmath51 can be relaxed if the cosmological scales exit the horizon after the initial outburst of tachyonic perturbations has subsided somewhat . note that , if then @xmath56 can not play the role of the inflaton because , where we used eq . ( [ v0 ] ) and also that , according to eq . ( [ mphih ] ) . ] the dependence on @xmath56 of the vector field kinetic term gives rise to source terms in the field equation of the scalar field . to study their influence let us consider the following lagrangian density : @xmath186 where @xmath68 is given by eq . ( [ vphi ] ) and . the case of constant mass corresponds to , while the higgsed vector field case corresponds to . from the above we find @xmath187\right\}\phi\ , , \label{eomphia}\end{aligned}\ ] ] where the prime denotes derivative with respect to @xmath56 and , in the last line of the above we have used eq . ( [ fn ] ) as well as that and also . assuming as discussed in sec . [ sp ] , eq . ( [ mphih ] ) suggests . on the other hand , due to eqs . ( [ mlight ] ) , ( [ meff1 ] ) and ( [ massless ] ) , we have , which means that is frozen , since the mass term in eq . ( [ eomhom ] ) is negligible compared to the `` friction '' term @xmath188 . this implies that , which means that eq . ( [ eomphia ] ) can be recast as @xmath189 where we used also eq . ( [ w ] ) . now , in the constant mass case and , therefore , the above equation reduces to eq . ( [ eomphi ] ) . in the higgsed vector field case , though , this is not necessarily so . indeed , due to eq . ( [ massless ] ) , we have , which might still dominate , if . since , in principle , this is not impossible , given eq . ( [ mbound ] ) . what happens if ? note , at first , that , since @xmath113 is frozen during inflation , , i.e. @xmath144 is decreasing exponentially , which means that eventually @xmath190 becomes dominant , whatever the initial value of @xmath144 . still , just after the onset of inflation we may well have . according to eq . ( [ eomphib ] ) , a positive mass - squared larger than @xmath28 would rapidly send @xmath56 to the origin . consequently , since in eq . ( [ l2 ] ) in the higgsed vector field case , the vector field is rendered exactly massless . this means that conformal invariance is restored and no perturbations of the vector field are generated @xcite . as @xmath144 decreases , however , the effective mass - squared of @xmath56 : becomes smaller than @xmath191 , in which case particle production of @xmath56 generates a condensate for @xmath56 of order . indeed , eq . ( [ vphi ] ) becomes . hence , particle production begins when , where we used eq . ( [ mphih ] ) with . after , a phase transition sends @xmath56 rolling off the origin and down the potential hill in eq . ( [ vphi ] ) as described in sec . [ frsf ] , while eq . ( [ eomphib ] ) reduces to eq . ( [ eomphi ] ) . thus , when , there is an initial period of inflation , where there is no vector particle production , while @xmath56 is sent to the origin . this period lasts for @xmath192 e - foldings . afterwards , a phase transition occurs which releases @xmath56 from the origin , the conformal invariance of the vector field is broken and particle production takes place as discussed in sec . [ hvf ] . from the above we see that the higgs vector field case has the advantage of explaining the initial condition of @xmath56 on top of the potential hill , if @xmath137 is large enough . to visualise the above findings we briefly study a particular example , considering the case of a higgsed vector field . thus , the lagrangian density is given in eq . ( [ l1 ] ) . we take and , that is we assume that @xmath50 is given by eq . ( [ f2 ] ) . according to eq . ( [ phia ] ) , while eq . ( [ nphi ] ) suggests @xmath193 in order to obtain an approximately scale invariant spectrum of perturbations we have to take the constraint in eq . ( [ massless ] ) into account . we assume that this constraint well satisfied , so that @xmath194 for the scalar field , which controls the mass of the vector field , we consider a higgs - type potential @xmath195 with @xmath196 being a constant . in view of the above potential and also of eq . ( [ mphih ] ) , we have @xmath197 from eqs . ( [ gmh ] ) and ( [ mh ] ) we readily obtain @xmath198 we assume that reheating is prompt and also that the vector curvaton decays as soon as it dominates the universe . this means @xmath199 also , we assume that the vector curvaton decays through gravitational interactions , i.e. @xmath200 furthermore , we assume that . this means that the bound in eq . ( [ hoscbound ] ) becomes . we choose the following value for our example @xmath201 i.e. the phase transition which results in the increment of the mass of the vector filed occurs at the breaking of electroweak unification . similarly , we choose the decay rate of the vector curvaton to be @xmath202 from eqs . ( [ gaex ] ) and ( [ gaex1 ] ) we find @xmath203 now , in view of eq . ( [ assum ] ) , eqs . ( [ oosc+ ] ) and ( [ oend1 + ] ) give @xmath204 similarly , eq . ( [ h*+ ] ) becomes @xmath205 combining the above we find @xmath206 which satisfies the bound in eq . ( [ osccons ] ) . the above estimate for @xmath151 is quite reasonable , assuming equipartition of energy at the phase transition over a large number [ @xmath207 of degrees of freedom . this argument substantiates our choice of @xmath116 in eq . ( [ gaex1 ] ) . using eqs . ( [ hoscex ] ) , ( [ gaex1 ] ) and ( [ m0ex ] ) , eq . ( [ h*consex ] ) gives @xmath208 hence , we have low scale inflation . this means that reheating , even though prompt , will not result in gravitino overproduction . also , one typically expects that the contribution of the inflaton to the curvature perturbation is negligible . inserting eq . ( [ h*ex ] ) into eq . ( [ h*bound+ ] ) and considering also eqs . ( [ assum ] ) and ( [ hoscex ] ) it can be easily shown that the bound in eq . ( [ h*bound+ ] ) is saturated . this is expected since we assumed that the vector curvaton decays gravitationally . similarly , it can be checked that the bound in eq . ( [ m0bound ] ) is also satisfied . employing eq . ( [ h*ex ] ) into eq . ( [ nh1 ] ) we find @xmath209 where we used also eq . ( [ assum ] ) . similarly , using eqs . ( [ assum ] ) , ( [ hoscex ] ) and ( [ h*ex ] ) , eq . ( [ ntot1 ] ) gives @xmath210 where we have used . thus we see that as required for the solution of the horizon and flatness problems . it is easy also to confirm that the bound in eq . ( [ ntotbound ] ) is well satisfied . in order to attain a scale - invariant spectrum of perturbations over cosmological scales up to the horizon at present we need to satisfy the constraint : @xmath211 which ensures that the backreaction of the vector field onto @xmath56 becomes negligible before the current horizon scale exits the horizon during inflation . in view of eqs . ( [ nw ] ) , ( [ nhex ] ) and ( [ ntotex ] ) , the above results in the bound @xmath212 where we considered . hence , we see that the interaction between the vector field and @xmath56 must be quite suppressed . furthermore , the curvature perturbation spectrum must extend down to scales at least as small as the horizon at the time of matter - radiation equality . thus , the e - fold range must be at least @xmath213 where @xmath214 gyrs is the age of the universe . as discussed in sec . [ pp ] , the generation of vector field perturbations ceases when @xmath56 assumes its vev : and . after this moment and , which , according to eq . ( [ paperp ] ) , gives . this is the vacuum spectrum as can be readily confirmed by eqs . ( [ match ] ) and ( [ pa ] ) . hence , after @xmath215 e - folds of inflation , particle production stops . therefore , in order to ascertain that the produced spectrum of perturbations extends over the entire range of the cosmological scales , we need to impose @xmath216 since there is a period of @xmath217 e - folds after the onset of inflation , during which the backreaction of the vector field sends @xmath56 to the origin , we can safely assume that , after the end of this period , the @xmath56 field begins to roll down its potential ( c.f . ( [ higgs ] ) ) with initial value , as determined by its quantum fluctuations . using this , eqs . ( [ nphi1 ] ) and ( [ nphibound ] ) result in the constraint @xmath218 where the upper bound is due to eq . ( [ mbound ] ) . from eqs . ( [ mbound ] ) and ( [ gbound ] ) we also find @xmath219 where we also considered eq . ( [ h*ex ] ) . hence , the vector field is indeed light during inflation . the upper bound on @xmath1 is much more stringent though , due to the requirement . indeed , using eqs . ( [ hoscex ] ) and ( [ h*ex ] ) we find @xmath220 in view of eq . ( [ nshiggs ] ) and considering that , we find that to a high accuracy , provided the contribution from is negligible . this value is marginally acceptable in terms of the observations . since the data prefer a lower value , however , one may assume a large - field inflation model , with non - negligible @xmath221 . in this case , in accordance to the curvaton scenario @xcite , we have @xmath222 for example , with quadratic chaotic inflation , one finds @xmath223 which gives the spectral index and the tensor fraction , which is more acceptable by the latest wmap data @xcite . let us choose , for illustrative purposes , , which lies comfortably within the allowed range in eq . ( [ mrange ] ) . in this case , ( [ mh ] ) and ( [ h*ex ] ) suggest , which is in agreement with eqs . ( [ gl ] ) and ( [ gbound ] ) . also , eq . ( [ nphi1 ] ) gives , as required . in this section we briefly discuss an altogether different possibility from the curvaton mechanism for the use of a vector field to generate the curvature perturbation in the universe . this is the inhomogeneous reheating mechanism , first introduced in ref . @xcite . according to this mechanism the curvature perturbations are due to the modulation of the decay rate of the inflaton field , because of its interaction with another field , which carries a superhorizon spectrum of perturbations . in our setup , one might employ this idea using the @xmath56 field in the higgsed vector case as an inflaton , whose decay rate is modulated by the perturbations of the vector field . according to the modulated reheating mechanism , the resulting curvature perturbation is related with the modulation of the decay rate of the inflaton as follows @xcite : @xmath224 where . we must , therefore , estimate the modulation of @xmath115 at reheating . the decay rate of the inflaton field @xmath56 is of the order @xmath225 where @xmath226 is the coupling of the inflaton field to its decay products . now , for the inflaton mass we have @xmath227 where we used that according to eqs . ( [ a ] ) and ( [ w ] ) . from eqs . ( [ zg ] ) , ( [ ginf ] ) and ( [ minf ] ) we find @xmath228^{-1 } \left.\frac{\delta w}{w}\right|_{\rm reh } , \label{zw}\ ] ] where ` reh ' denotes the time of reheating . as discussed in sec . [ emt ] , before the oscillations , while during the oscillations . since , in both cases ( c.f . ( [ va ] ) ) we find @xmath229 using this , it is easy to obtain @xmath230 where we used that , after the onset of the oscillations remains constant and also that as suggested by eq . ( [ paflat ] ) . combining eqs . ( [ zw ] ) and ( [ dw / w ] ) one gets @xmath231^{-1 } \frac{h_*}{2\pi w_{\rm reh}}\ , . \label{zz}\ ] ] * case 1 : * suppose , at first , that @xmath232 then , eq . ( [ zz ] ) becomes @xmath233 * case 2 : * suppose , now , that @xmath234 then , eq . ( [ zz ] ) becomes @xmath235 combining eqs . ( [ gwm1 ] ) and ( [ gwm2 ] ) with eqs . ( [ gw1 ] ) and ( [ gw2 ] ) respectively , we find that , in all cases @xmath236 since in the higgsed vector case ( sec . [ hvf ] ) a scale invariant perturbation spectrum requires and , eq . ( [ mphih ] ) suggests that . using this and also that , the above results in @xmath237 however , combining this with eq . ( [ massless ] ) suggests @xmath238 where is the vev of the inflaton field @xmath56 , as implied by the fact that ( c.f . ( [ v0 ] ) ) . since and , using eq . ( [ nphi ] ) we obtain @xmath239 where we have considered that the initial value for the inflaton can not be smaller than its quantum fluctuation . the above number of e - folds is far too small to compare with the requirements for the solution of the horizon problem . indeed , from eqs . ( [ nh ] ) and ( [ h*g ] ) we find . therefore , the inhomogeneous reheating mechanism can not be used to account for the curvature perturbation in the universe in this model . we have investigated whether a massive abelian vector field whose kinetic term is evolving during inflation can be responsible for the curvature perturbation in the universe , without the need of a negative mass - squared . in particular , we have studied particle production of this vector field when its kinetic term is determined by a function , similar to the gauge kinetic function in supergravity . we assumed that the dynamics of this kinetic function is dominated by a degree of freedom which is varying during inflation ; at least when the cosmological scales are exiting the horizon . in supergravity the gauge kinetic function is a holomorphic function of the fields of the theory . since scalar fields typically obtain masses comparable to the hubble parameter @xmath0 due to supergravity corrections during ( and after ) inflation @xcite , one typically expects that the value of the gauge kinetic function is indeed varying during inflation , as these fields roll down the potential slopes . we parametrised these fields using a single degree of freedom @xmath56 , which rolls down towards its vev : @xmath51 . with respect to this degree of freedom we expressed the kinetic function as so that the vector field becomes canonically normalised when @xmath56 reaches its vev . we have obtained the condition for the generation of a scale invariant spectrum of perturbations and showed that it can be naturally achieved when , where @xmath190 is the tachyonic mass of @xmath56 . we , then studied two particular cases : i ) the case of a vector field with constant mass @xmath1 and ii ) the case with a vector field higgsed with @xmath56 , whose mass is . then , we argued that the most promising results are obtained when during inflation and also when ( i.e. ) , which can be achieved if . a mass of this order is naturally expected in supergravity , due to khler corrections to the scalar potential @xcite . after obtaining a scale invariant spectrum we attempted to employ the curvaton mechanism in order to generate the observed curvature perturbation . under this mechanism the vector field remains subdominant during inflation when it obtains a scale - invariant superhorizon spectrum of perturbations over the cosmological scales . after inflation , when the hubble parameter decreases below its mass , the vector field begins oscillating . as shown in ref . @xcite a coherently oscillating , homogeneous , massive abelian vector field corresponds to pressureless , _ isotropic _ matter , and can dominate ( or nearly dominate ) the universe without introducing a long - range anisotropy . when it does so , it imprints its own curvature perturbation spectrum , as in the curvaton scenario . we followed the evolution of the vector field and obtained the corresponding bounds on the inflationary scale for the scenario to work . we found , however , that the parameter space does not allow enough inflation for the solution of the flatness and horizon problems . to overcome this problem the lower bound on the inflationary scale must be relaxed , i.e. low - scale inflation is required . to attain low - scale inflation we employed the mass increment mechanism , first introduced in ref . @xcite . in this scenario a phase transition after the end of inflation enlarges the mass of the vector curvaton field . we have explored the parameter space under this mechanism and showed that it is possible to solve the horizon and flatness problems and also produce the required amplitude for the scale - invariant spectrum of curvature perturbations provided the phase transition does not occur much earlier than the breakdown of electroweak unification . we also found that the best results are obtained if the curvaton decays as soon as it comes to dominate the universe . we demonstrated our findings in a concrete example , which serves as an existence proof that the mechanism works with natural values of the parameters . in our example we considered the case of a higgsed vector curvaton , which can also explain the initial conditions of the rolling @xmath56 . we have assumed that the phase transition which enlarges the mass of the field occurs at temperature . the mass of the vector field is roughly comparable to the inflationary scale , which turns - out to be . reheating is assumed prompt , but the reheating temperature is low enough not to result in gravitino overproduction . at the phase transition the vector curvaton assumes roughly 1 % of the density of the universe , which is reasonable on energy equipartition grounds . rapid oscillations of the vector field allow it to dominate the universe at temperature . the vector field is taken to decay at domination so as not to disturb bbn . the scenario works for , which is a comfortably large range of parameter space , including both the grand unified and the electroweak scales . unless one considers a large - field model of inflation the spectral index is indistinguishable from unity . however , for a large - field model ( e.g. chaotic inflation ) , one attains a lower value for the spectral index , which agrees better with the observations . finally , we have also studied the possibility that the vector field generates the observed curvature perturbation spectrum through the so - called inhomogeneous reheating mechanism . in this case , the rolling @xmath56 field is taken to be the inflaton , whose decay rate is modulated by the perturbations in the vector field . even though the idea sounds promising , our results show that the scenario is inviable . in summary , we have investigated the use of a massive abelian vector field for the generation of the observed curvature perturbation spectrum in the universe . we have shown that , it is possible to attain a scale - invariant spectrum with a positive mass - squared for the vector field , provided the kinetic function is growing during inflation . in this case the vector field can act as a curvaton . the mechanism works with low - scale inflation , when the mass of the vector field increases at a phase transition near the breakdown of electroweak unification . the form of the kinetic function as well as other aspects of the mechanism ( such as masses of order the hubble scale ) can be naturally accommodated in the theoretical framework of supergravity . d. h. lyth and d. wands , phys . b * 524 * ( 2002 ) 5 ; k. enqvist and m. s. sloth , nucl . b * 626 * ( 2002 ) 395 ; t. moroi and t. takahashi , phys . b * 522 * ( 2001 ) 215 [ erratum - ibid . b * 539 * ( 2002 ) 303 ] ; s. mollerach , phys . d * 42 * ( 1990 ) 313 . k. dimopoulos and d. h. lyth , phys . d * 69 * ( 2004 ) 123509 . t. moroi , t. takahashi and y. toyoda , phys . d * 72 * ( 2005 ) 023502 ; t. moroi and t. takahashi , phys . d * 72 * ( 2005 ) 023505 . k. dimopoulos , d. h. lyth and y. rodriguez , jhep * 0502 * ( 2005 ) 055 . m. postma , jcap * 0405 * ( 2004 ) 002 . k. dimopoulos and g. lazarides , phys . d * 73 * ( 2006 ) 023525 ; k. dimopoulos , phys . b * 634 * ( 2006 ) 331 g. dvali , a. gruzinov and m. zaldarriaga , phys . d * 69 * ( 2004 ) 023505 ; phys . d * 69 * ( 2004 ) 083505 ; l. kofman , astro - ph/0303614 ; k. enqvist , a. mazumdar and m. postma , phys . d * 67 * ( 2003 ) 121303 . s. matarrese and a. riotto , jcap * 0308 * ( 2003 ) 007 ; f. vernizzi , phys . d * 69 * ( 2004 ) 083526 ; a. gruzinov , astro - ph/0401407 . m. postma , jcap * 0403 * ( 2004 ) 006 . k. dimopoulos , phys . d * 74 * ( 2006 ) 083502 . l. h. ford , phys . d * 40 * ( 1989 ) 967 ; c. m. lewis , phys . d * 44 * ( 1991 ) 1661 . k. bamba and j. yokoyama , phys . d * 69 * ( 2004 ) 043507 ; phys . d * 70 * ( 2004 ) 083508 . k. bamba and m. sasaki , jcap * 0702 * ( 2007 ) 030 . b. ratra , astrophys . j. * 391 * ( 1992 ) l1 . m. dine , l. randall and s. thomas , nucl . b * 458 * ( 1996 ) 291 ; phys . rev . lett . * 75 * ( 1995 ) 398 . d. h. lyth , phys . b * 579 * ( 2004 ) 239 . d. n. spergel _ et al . _ [ wmap collaboration ] , astro - ph/0603449 .

we present a variety of methods to derive the casimir interaction in planar systems containing two - dimensional layers . examples where this can be of use is graphene , graphene - like layers and two - dimensional electron gases . we present results for two free standing layers and for one layer above a substrate . the results can easily be extended to systems with a larger number of layers .

1409.2660

the dispersion interaction is a correlation effect . the interaction is a result of the correlated motion of the charged carriers in different parts of the system . the dispersion energy between two objects is the inter - object correlation energy . in the field of condensed matter one often uses the coulomb gauge . in coulomb gauge the scalar potential is instantaneous and the vector potential is purely transverse . the scalar potential represents coulomb interactions between the carriers while the vector potential represents photon interactions . if the two objects are not too far away from each other one may neglect retardation effects ( effects from the finite speed of light in vacuum ) . then only the coulomb interactions contribute and the properties of the materials enter in terms of the density - density correlation functions . if retardation has to be taken into account the photon interactions have to be included and , in this additional contribution , the properties of the materials enter in terms of the current - current correlation functions . the interaction can be found using many - body theory@xcite with diagrammatic perturbation theory based on feynman diagrams . alternatively the interaction can be viewed as caused by fluctuations , fluctuations in the charge and current densities or fluctuations in the electromagnetic fields . then one way to find the interaction is in terms of the electromagnetic normal modes of the system.@xcite these normal modes are massless bosons and at zero temperature the interaction energy is the sum of the zero - point energy of all these modes . in a translational invariant and homogeneous system there are three types of normal mode ; two with transverse fields and one with longitudinal fields . in the present system there are only two types of normal mode ; one with @xmath0-polarized fields , i.e. , with electric vector in the plane formed by the in - plane momentum @xmath1 and the normal to the layers , experienced as longitudinal in the two - dimensional ( 2d ) sheets ; one with @xmath2-polarized fields , i.e. , with electric vector perpendicular to this plane , experienced as transverse in the 2d sheets . the first type of mode is called tm ( transverse magnetic ) and second te ( transverse electric ) . the dispersion curves for each mode type can have one or several branches . for planar structures the quantum number that characterizes the normal modes is * k * , the 2d wave vector in the plane of the interfaces or planar structures and the interaction energy per unit area for each mode type can be written as @xmath3 where @xmath4 is the area of a planar interface and the summation over @xmath5 is a summation over the branches . to get the expression on the right - hand side we have let @xmath4 be large so that the summation over the discrete @xmath1 variable can be replaced by an integral over a continuous @xmath1 variable . after using an extension of the so - called argument principle and deforming the integration path in the complex frequency plane one arrives at the expression@xcite @xmath6 , \label{equ2}\ ] ] where @xmath7 is the condition for electromagnetic normal modes . equation([equ2 ] ) is valid for zero temperature and the interaction energy is the internal energy . at finite temperature also the thermal population of the modes affects the interaction . the interaction energy is now helmholtz free energy and can be written as @xmath8 } } , \\ { \xi _ n } = \frac{{2\pi n}}{{\hbar \beta } } ; \;n = 0,\,1,\,2,\ , \ldots , \end{array } \label{equ3}\ ] ] where @xmath9 . the integral over frequency has been replaced by a summation over discrete frequencies , the so - called matsubara frequencies.@xcite the prime on the summation sign indicates that the @xmath10 term should be divided by two . the normal modes can be found in different ways . one way is to solve maxwell s equations ( in absence of any external sources ) in all regions of the system and use the proper boundary conditions at the interfaces . another way is to find self - sustained solutions using induced charge- and current - densities in the objects . we here derive the dispersion interaction in a variety of ways . since the non - retarded treatment is much easier to perform we present both the non - retarded and fully retarded derivations in separate sections . in the derivations involving normal modes we derive the mode condition function , @xmath11 . the interaction energy is then found by inserting this function into ( [ equ2 ] ) or ( [ equ3 ] ) . in the derivations involving diagrammatic perturbation theory the interaction energy is obtained directly . we do not present any numerical results , only the final expressions and different ways to arrive at these . however , we give the analytical expressions for the dielectric function of graphene and of the 2d electron gas so it is straight forward for the readers to perform numerical derivations of their own . numerical results for graphene by us and others can be found in the literature.@xcite we begin in section [ not ] with introducing our notation . the following two sections are devoted to background materials , electromagnetic in section [ embg ] and 2d in section [ bg2d ] . the main results when retardation effects are neglected are derived in section [ nonret ] and when retardation effects are taken into account in section [ ret ] . we end in section [ sum ] with a brief summary and conclusions . in this work we discuss planar structures . we let all surfaces , interfaces and 2d films be parallel to the @xmath12-plane . we let @xmath13 be a three - dimensional ( 3d ) wave vector and @xmath14 its 2d in - plane component . the corresponding notation for the spatial coordinates are @xmath15 and @xmath16 , respectively . thus we have @xmath17 we will end up with a formalism only depending on @xmath14 and @xmath16 apart from the distances between the various interfaces . thus the formalism is 2d . we will make extensive use of the fourier transform , both the 3d and 2d versions , and since they can be defined in different ways we list our definitions here . the 3d fourier transform and inverse fourier transform of a function @xmath18 with only spatial dependence are @xmath19 if the function also has a temporal dependence the corresponding relations are @xmath20 in 2d the fourier transform and inverse fourier transform of a function are @xmath21 if the function has only spatial dependence . if the function also has a temporal dependence the relations are @xmath22 we will need yet another fourier transform . we let the function @xmath18 be a function of the 3d spatial variable @xmath15 but we only transform with respect to @xmath16 . then we have for a function with only spatial dependence @xmath23 if the function also has a temporal dependence the relations are @xmath24 note that we have refrained from introducing separate function names for the different fourier transformed versions of a function . by looking at the argument of the function we can tell if it is a 3d or 2d fourier transform , a partial transform , or not a transform at all . this section has two subsections . in the first , section [ me ] , we present the proper version of maxwell s equations and also the boundary conditions at interfaces in the system . in the second , section [ pot ] , we give the scalar and vector potentials in coulomb and lorentz gauges and show how the fields are obtained from these . there are different formulations of electromagnetism in the literature . the difference lies in how the conduction carriers are treated . in one formulation these carriers are lumped together with the external charges to form the group of free charges . then only the bound charges contribute to the screening . we want to be able to treat geometries with metallic regions . then this formulation is not suitable . in the formulation that we use the conduction carriers are treated on the same footing as the bound charges . thus , both bound and conduction charges contribute to the dielectric function . in the two formalisms the * e * and * b * fields , the true fields , of course are the same . however , the auxiliary fields , the * d * and * h * fields , are different . to indicate that we use this alternative formulation we put a tilde above the * d * and * h * fields and also above the dielectric functions , polarizabilities and conductivities . with this formulation maxwell s equations ( mes ) have the form @xmath25 and the boundary conditions at an interface between two media @xmath26 and @xmath27 are @xmath28 where @xmath29 and @xmath30 are volume charge and current densities , respectively , while @xmath31 and @xmath32 are surface charge and current densities , respectively . we note that the sources to the fields in mes are the external charge and current densities and also in the boundary conditions discontinuities in the normal component of the @xmath33 fields and tangential component of the @xmath34 fields are caused by external surface charge densities and external surface current densities , respectively . the unit vector * n * is the surface normal pointing into region @xmath27 . often it is easier to deal with potentials instead of the fields . the potentials are auxiliary functions , help functions . the potentials are not unique , they depend on which gauge we have chosen . two very common gauges used in condensed matter theory are the coulomb gauge ( mentioned in the introduction ) and the lorentz gauge . in coulomb gauge the potentials are @xmath35 here the scalar potential , @xmath36 , is instantaneous ; it describes the coulomb interaction between the charged carriers ; the coulomb interaction is purely longitudinal . the vector potential , @xmath37 , is retarded and transverse ; it describes the photon interactions between the charged particles ; the photon interactions are purely transverse . the potentials in coulomb gauge are not lorentz invariant . however , they are in lorentz gauge , @xmath38 here both potentials are retarded . in all gauges the relations between the true fields and the potentials are the same , viz . when retardation effects are neglected ( @xmath40 ) both gauges coincide and the only potential that remains is the instantaneous scalar potential . the different versions of the fourier transform of the function @xmath41 appear many times in the derivations so we compile the results in this subsection . this function is the coulomb potential between two electrons , @xmath42 , stripped of the charges . the 3d fourier transform is @xmath43 the 2d fourier transform performed over a plane the distance @xmath44 from the @xmath12-plane is @xmath45 the 2d fourier transform performed over the @xmath12-plane is @xmath46 these relations appear in expressions involving the scalar potential in coulomb gauge and when retardation effects are negligible . when retardation is important the function has a temporal dependence , @xmath47 . then the following fourier transforms are important : the 3d fourier transform , @xmath48 the 2d fourier transform performed over a plane the distance @xmath44 from the @xmath12-plane , @xmath49 the 2d fourier transform performed over the @xmath12-plane , @xmath50 where in the last two equations @xmath51 we will need one further relation , viz . when dealing with a 2d layer above a substrate it is fruitful to be able to find the image charge- and current - densities produced in the substrate . we start from a given time dependent charge- and current - density in the 2d layer , calculate the potentials and from these the @xmath53- and @xmath54-fields . we make an ansatz for the image densities and determine these by using the standard boundary conditions . here , we use the lorentz gauge in which the scalar and vector potentials are given in ( [ equ14 ] ) . both are convolution integrals of the type @xmath55 which has the nice property that @xmath56 . in the case of the scalar potential these functions are @xmath57 and @xmath58 , respectively . for the vector potential @xmath59 and @xmath58 , respectively . thus using ( [ equ19 ] ) we find the fourier transforms of the potentials are @xmath60 the electric field in terms of the potentials was given in ( [ equ15 ] ) . this leads to the following relation between the fourier transforms @xmath61 thus , @xmath62 } } + \frac{{i\omega } } { { { c^2}}}\frac{{4\pi { \bf{j}}\left ( { { \bf{q}},\omega } \right)}}{{\left [ { { q^2 } - { { \left ( { \omega /c } \right)}^2 } } \right]}}. \label{equ27}\ ] ] now , in our system the charge- and current - densities are surface densities , @xmath63 and @xmath64 which means that @xmath65 thus , we have for the electric field from surface charge- and current - densities , confined to the @xmath12-plane @xmath66 } } + \frac{{i\omega } } { { { c^2}}}\frac{{4\pi { \bf{k}}\left ( { { \bf{k}},\omega } \right)}}{{\left [ { { q^2 } - { { \left ( { \omega /c } \right)}^2 } } \right ] } } , \label{equ31}\ ] ] in vacuum . if the source densities are embedded in a medium with dielectric function @xmath67 the electric field is @xmath68}}\\ \quad \quad \quad \quad\quad \quad \quad \quad + \frac{{i\omega } } { { { c^2}}}\frac{{4\pi { \bf{k}}\left ( { { \bf{k}},\omega } \right)}}{{\left [ { { q^2 } - { { \tilde \varepsilon } _ m}\left ( \omega \right){{\left ( { \omega /c } \right)}^2 } } \right]}}. \end{array } \label{equ32}\ ] ] the magnetic induction is @xmath69 } } , \label{equ33}\ ] ] in vacuum and @xmath70 } } , \label{equ34}\ ] ] in a medium . we may use the equation of continuity to eliminate @xmath71 in favor of @xmath72 . the equation of continuity reads @xmath73 the relation for the fourier transforms becomes @xmath74 in the present system with surface densities we have from ( [ equ30 ] ) @xmath75 and since * k * is in the plane @xmath76 the scalar product picks out the longitudinal part of the surface current density . thus we have @xmath77 we use this relation to eliminate the charge density from the field relations . from now on we assume that the 2d layer is embedded in a medium . let us now choose the @xmath78-axis to point along @xmath1 . a general vector quantity has a component normal to the planar interfaces . we attach a subscript @xmath79 to this component . the in - plane part of the vector has a longitudinal and a transverse part . the longitudinal is parallel and the transverse perpendicular to @xmath1 . we attach the subscripts @xmath80 and @xmath81 , respectively , to these components . the current density has no normal component . to summarize we have @xmath82 we then have in the source plane @xmath83 where @xmath84 . in a plane parallel to the source plane ( @xmath12-plane ) we have @xmath85 where we have used ( [ equ39 ] ) in the third relation . now , we have all we need to determine the image densities . in the spirit of image theory we assume that the fields outside the substrate can be reproduced by the fields from the actual source densities in the 2d layer ( the distance @xmath86 from the interface ) plus the fields from the mirror densities ( the distance @xmath86 from the interface on the opposite side ) , both sources embedded in the medium with dielectric function @xmath87 . the field inside the substrate we assume can be reproduced by another mirror density at the position of the 2d layer embedded in the medium with dielectric function @xmath88 . we mark the mirror densities of the first kind by a prime and those of the second kind by a double prime . we start with the longitudinal current densities . we first use the boundary condition that the in plane electric field is continuous across the medium - substrate interface . this gives @xmath89\\ = \frac{{{\gamma _ s}\left ( { k,\omega } \right)}}{{{{\tilde \varepsilon } _ s}\left ( \omega \right)}}{e^ { - { \gamma _ s}\left ( { k,\omega } \right)kd}}{k_\parallel } ' ' \left ( { { \bf{k}},\omega } \right ) . \end{array } \label{equ43}\ ] ] then we use the condition that the normal component of the displacement field @xmath90 is continuous across the interface . this gives @xmath91\\ = { e^ { - { \gamma _ s}\left ( { k,\omega } \right)kd}}{k_\parallel } ' ' \left ( { { \bf{k}},\omega } \right ) . \end{array } \label{equ44}\ ] ] combining these two equations results in the mirror densities @xmath92 now , we continue with the transverse current densities . we first use the boundary condition that the in plane transverse electric field is continuous across the interface . this gives @xmath93}\\ { = \frac{1}{{{\gamma _ s}\left ( { k,\omega } \right)}}{e^ { - { \gamma _ s}\left ( { k,\omega } \right)kd}}{k _ \bot } ^{\prime \prime } \left ( { { \bf{k}},\omega } \right ) . } \end{array } \label{equ46}\ ] ] then we use the condition that the in plane longitudinal component of the @xmath94-field is continuous across the interface . this gives @xmath95}\\ { = { e^ { - { \gamma _ s}\left ( { k,\omega } \right)kd}}{k _ \bot } ^{\prime \prime } \left ( { { \bf{k}},\omega } \right ) . } \end{array } \label{equ47}\ ] ] these two equations result in the mirror densities @xmath96 in later sections we will need the single primed mirror densities in ( [ equ45 ] ) and ( [ equ48 ] ) . the relation between the dielectric function , @xmath97 , and dynamical conductivity , @xmath98 , is different in 2d and 3d . the relations are @xmath99 where @xmath100 is the polarizability . the relations between the induced current densities and the electric field are @xmath101 note that both bound and conduction carriers contribute to the polarizabilities and conductivities . it is sometimes convenient to introduce another correlation function , @xmath102 , the polarization bubble ( in the language of feynman diagrams ) , or lowest order contribution to the density - density correlation function , @xmath103 we now give the analytical expression for the 2d dielectric function of graphene and of a 2d electron gas . let us first begin with an undoped graphene sheet . in a general point , @xmath44 , in the complex frequency plane , away from the real axis the density - density correlation function is@xcite @xmath104 where @xmath105 is the carrier velocity which is a constant in graphene ( @xmath106 ) , and @xmath107 represents the degeneracy parameter with the value of 4 ( a factor of 2 for spin and a factor of 2 for the cone degeneracy . ) in our numerical calculations@xcite we use the value@xcite @xmath108 m / s for @xmath105 . in the present work we do not give any numerical results . when the graphene sheet is doped the dielectric function becomes much more complicated . however it has been derived by several groups@xcite . in the two next equations we use dimension - less variables : @xmath109 ; @xmath110 ; @xmath111 . the density - density correlation function in a general point in the complex frequency plane , @xmath44 , away from the real axis is@xcite @xmath112 } \right\};}\\ { f\left ( { x,\tilde z } \right ) = { \rm{asin}}\left ( { \frac{{1 - \tilde z}}{x } } \right ) + { \rm{asin}}\left ( { \frac{{1 + \tilde z}}{x } } \right)}\\ { \qquad \qquad - \frac{{\tilde z - 1}}{x}\sqrt { 1 - { { \left ( { \frac{{\tilde z - 1}}{x } } \right)}^2 } } + \frac{{\tilde z + 1}}{x}\sqrt { 1 - { { \left ( { \frac{{\tilde z + 1}}{x } } \right)}^2 } } , } \end{array } \label{equ53}\ ] ] where @xmath113 is the density of states at the fermi level and @xmath114 is the doping concentration . the same result holds for excess of electrons and excess of holes . in the calculations one needs the function at the imaginary frequency axis . we have derived a very useful analytical expression valid along the imaginary axis ; an expression in terms of real valued functions of real valued variables:@xcite @xmath115 } \right\};}\\ { \quad g(x , y ) = { \rm{atan}}\left [ { h(x , y)k\left ( { x , y } \right ) } \right ] + l\left ( { x , y } \right);}\\ { \quad h\left ( { x , y } \right ) = \frac{{2{{\left\ { { { { \left [ { { x^2}\left ( { { y^2 } - 1 } \right ) + { { \left ( { { y^2 } + 1 } \right)}^2 } } \right]}^2 } + { { \left ( { 2y{x^2 } } \right)}^2 } } \right\}}^{1/4}}}}{{\sqrt { { { \left ( { { x^2 } + { y^2 } - 1 } \right)}^2 } + { { \left ( { 2y } \right)}^2 } } - \left ( { { y^2 } + 1 } \right)}},}\\ { \quad k\left ( { x , y } \right ) = \sin \left\ { { \frac{1}{2}{\rm{atan}}\left [ { \frac{{2y{x^2}}}{{{x^2}\left ( { { y^2 } - 1 } \right ) + { { \left ( { { y^2 } + 1 } \right)}^2 } } } } \right ] } \right\},}\\ { \quad \;l\left ( { x , y } \right ) = } \\ { \frac{{\sqrt { - 2{x^2}\left ( { { y^2 } - 1 } \right ) - 2\left ( { { y^4 } - 6{y^2 } + 1 } \right ) + 2\left ( { { y^2 } + 1 } \right)\sqrt { { x^4 } + 2{x^2}\left ( { { y^2 } - 1 } \right ) + { { \left ( { { y^2 } + 1 } \right)}^2 } } } } } { { { x^2 } } } , } \end{array } \label{equ54}\ ] ] where the arcus tangens function is taken from the branch where @xmath116 . the density - density correlation function on the imaginary frequency axis has been derived before in a compact and inexplicit form ( see ref.@xcite and references therein . ) here we have chosen to express it in an explicit form in terms of real valued functions of real valued variables . the 2d polarizability of an electron gas in rpa ( random phase approximation ) is given by@xcite @xmath117}^{1/2}}/{q^2 } } \right\ } , \end{array } \label{equ55}\ ] ] where @xmath118 now we have presented all background material needed for the actual derivations . in the non - retarded treatment many terms in mes drop out and the derivations become much simpler . there are no photons , only coulomb interactions . the dispersion interactions treated here are called van der waals interactions or sometimes non - retarded casimir interactions . we begin in section [ 22dnr ] with two parallel 2d sheets and use different methods to derive the results . in section [ 2dsnr ] we treat a 2d sheet above a substrate . for a system of parallel 2d sheets the derivation in the language of many - body theory becomes very simple , especially when retardation effects can be neglected . in the present two - sheet system the interaction energy is nothing but the inter - sheet correlation energy@xcite . using diagrammatic perturbation theory the feynman diagrams representing the correlation energy are given in fig.[figu1 ] . to get the inter - sheet contribution we can either subtract the intra - sheet part or subtract the total result when the separation between the sheets goes to infinity ( at that limit only the intra - sheet contribution remains . ) to save space we assume that the two sheets are identical . it is straight forward to extend the treatment to different sheets . each ellipse represents a polarization bubble , @xmath119 , and the number indicates in which sheet the process occurs . each of the four diagrams represents an infinite series of diagrams . since the sheets are identical the diagrams in the second row give contributions identical to the diagrams in the first row . we can use the first row diagrams and multiply the result with a factor of two . the interaction energy per unit area can be written as@xcite @xmath120 , \end{array } \label{equ57}\ ] ] where @xmath121 is the coupling constant and the factor of two has been inserted . now @xmath122 and @xmath123 , respectively . note that with our assumptions @xmath124 and @xmath125 . the distance between the sheets is @xmath86 and @xmath126 is the 2d coulomb interaction , @xmath127 . the exponential factor in the second potential is the result from taking the 2d fourier transform of the coulomb potential in a plane the distance @xmath86 from the center of the potential , see ( [ equ17 ] ) . the interaction lines with double bars represent a series of terms , with zero , one , two @xmath128 number of polarization bubbles . this can be expressed as @xmath129 where we have closed the two infinite series . this system of equations can be solved and the result is @xmath130}}{{{{\left [ { 1 + \tilde { \alpha } \left ( { { \bf{k}},\omega } \right ) } \right]}^2 } - \exp \left ( { - 2kd } \right){\tilde { \alpha } ^2}\left ( { { \bf{k}},\omega } \right)}};\\ { { \tilde v}_{12}}\left ( { { \bf{k}},\omega } \right ) = \frac{{{v^{2d}}\left ( k \right)\exp \left ( { - kd } \right)}}{{{{\left [ { 1 + \tilde { \alpha } \left ( { { \bf{k}},\omega } \right ) } \right]}^2 } - \exp \left ( { - 2kd } \right){\tilde { \alpha } ^2}\left ( { { \bf{k}},\omega } \right ) } } , \label{equ59}\end{aligned}\ ] ] where we have used the relation @xmath131 in accordance with ( [ equ51 ] ) . interpreting the feynman diagrams in figure [ figu1 ] one finds that the square brackets in ( [ equ57 ] ) are @xmath132 $ ] , where the last term comes from the subtraction of the intra - layer correlation energy . each factor of @xmath133 appearing implicitly in the expression should be multiplied by the coupling constant . performing the integration over coupling constant gives @xmath134}^2 } } \right\ } } } . \label{equ60}\ ] ] here for simplicity we assumed that the two 2d layers were identical . if they are not the result is @xmath135\left [ { \frac{{{{\tilde \alpha } _ 2}\left ( { k , i\omega } \right)}}{{1 + { { \tilde \alpha } _ 2}\left ( { k , i\omega } \right ) } } } \right ] } \right\}. \end{array } \label{equ61}\ ] ] this is the result from diagrammatic perturbation theory within the random phase approximation ( rpa ) . in next section we derive the same thing using normal modes . we will now go through one possible way to find the normal modes and their zero - point energies . let us assume that we have an induced charge distribution , @xmath136 , in sheet 1 . this gives rise to the potential @xmath137 and @xmath138 in sheets 1 and 2 , respectively . the resulting potential in sheet 2 after screening by the carriers is @xmath139 $ ] , which gives rise to an induced charge distribution in sheet 2 , @xmath140}}. \label{equ62}\ ] ] in complete analogy , this charge distribution in sheet 2 gives rise to a charge distribution in sheet 1 , @xmath141}}. \label{equ63}\ ] ] to find the condition for self - sustained fields , normal modes , we let this induced charge density in sheet 1 be the charge density we started from . this leads to @xmath142\left [ { \frac{{{{\tilde \alpha } _ 2}\left ( { { \bf{k}},\omega } \right)}}{{1 + { { \tilde \alpha } _ 2}\left ( { { \bf{k}},\omega } \right ) } } } \right ] = 0 . \label{equ64}\ ] ] the left hand side of this equation ( @xmath11 ) is exactly the argument of the logarithm in ( [ equ61 ] ) . inserting this function into ( [ equ2 ] ) we obtain exactly the same energy as in ( [ equ61 ] ) . that result was derived using many - body theory and in that derivation there is no talk about zero - point energies at all . here we solve mes in the three regions separated by the two 2d layers . we treat the induced charge and current densities in the 2d layers as external to our system . since we have vacuum in between the layers our system is vacuum and the @xmath143 and @xmath144 fields are equal to the @xmath53 and @xmath54 fields , respectively . it is enough to study the @xmath53-field in the non - retarded treatment . we follow the procedure in @xcite and make the following ansatz @xmath145 the two mes involving the @xmath53-field , @xmath146 , and @xmath147 , both give @xmath148 the boundary conditions are that the @xmath78-components of the fields are continuous across the two interfaces and there is a jump in the @xmath44-component equal to @xmath149 . thus we have @xmath150 or on matrix form @xmath151 and @xmath152 now , we eliminate the @xmath44-components in favor of the @xmath78-components using ( [ equ66 ] ) and use @xmath153 and assume for simplicity that the two layers are identical . we find @xmath154 or on matrix form @xmath155 combining ( [ equ68 ] ) and ( [ equ71 ] ) we find that the condition for having normal modes is that the determinant of the matrix @xmath156 vanishes . the mode condition function becomes @xmath157 ^ 2}. \label{equ73}\ ] ] we have taken as reference system the system when the two sheets are at infinite separation . this means that we have divided the mode condition function with the corresponding result when @xmath86 is infinite . if the two layers are different we instead find @xmath158\left [ { \frac{{{{\tilde \alpha } _ 2}\left ( { { \bf{k}},\omega } \right)}}{{1 + { { \tilde \alpha } _ 2}\left ( { { \bf{k}},\omega } \right ) } } } \right ] , \label{equ74}\ ] ] which agrees with the results of the previous sections . another way to find the normal modes is to find the frequency where the potential between carriers in the same layer or in different layers diverges . the condition for modes is that the denominators in ( [ equ59 ] ) are equal to zero . this leads to the same mode condition function as in ( [ equ64 ] ) and ( [ equ74 ] ) when we use as reference system the system when the 2d layers are at infinite separation . if we have a 2d layer ( like a graphene sheet ) the distance @xmath86 above a substrate the procedure is very similar to in section [ cisnr ] . we start with an induced mirror charge density , @xmath136 , in the substrate . the induced charge density in the 2d layer is given by the expression in ( [ equ62 ] ) except that now the distance between the mirror charge and the 2d layer is @xmath159 instead of @xmath86 . the mirror charges can be obtained from ( [ equ39 ] ) and ( [ equ45 ] ) and agree with the traditional results for the mirror charges at a planar vacuum dielectric interface.@xcite equation([equ63 ] ) is then replaced by @xmath160 and the condition for normal modes becomes @xmath161 and hence the mode condition function is @xmath162 in section [ menr22d ] we had a geometry with two interfaces , one at @xmath163 and one at @xmath164 . there there were a 2d layer at both interfaces . we use the same geometry here but there is a 2d layer only at @xmath163 . at @xmath164 we place the vacuum - substrate interface . the derivation and equations are equal up till ( [ equ69 ] ) . when the media are no longer vacuum there is a jump in @xmath165 at the 2d layer instead of in @xmath166 . note that here there is only a 2d layer at the left interface so the normal component of the @xmath167-field is continuous at the substrate surface . the new equation is @xmath168 now we eliminate the @xmath44-components in favor of the @xmath78-components and use @xmath169 so that @xmath170 or on matrix form @xmath171 combining ( [ equ68 ] ) and ( [ equ80 ] ) we find that the condition for having normal modes is that the determinant of the matrix @xmath172 is zero . thus , the mode condition function becomes @xmath173\left [ { \frac{{{{\tilde \varepsilon } _ s}\left ( \omega \right ) - 1}}{{{{\tilde \varepsilon } _ s}\left ( \omega \right ) + 1 } } } \right ] , \label{equ82}\ ] ] which agrees with the results of the previous sections . here we take retardation effects fully into account . the formalism becomes much more involved . the dispersion interactions treated here are called casimir interactions or sometimes retarded van der waals interactions . we begin in section [ 22dret ] with two parallel 2d sheets and use different methods to derive the results . in section [ 2dsret ] we treat a 2d sheet above a substrate . we do not perform a detailed derivation here , we just describe how the derivation is performed . we follow the derivation in @xcite and refer to that work for more details . we have chosen to work in coulomb gauge . in this gauge one part of the interaction between the electrons is in the form of the instantaneous , longitudinal coulomb interaction ( scalar potential ) and the other part is via transverse photons ( vector potential ) . the hamiltonian for the system is @xmath174 ^ 2 } + \frac{1}{2}\sum\limits_{ij } { \frac{{{e_i}{e_j}}}{{{r_{ij } } } } } \\ \quad + \sum\limits_{{\bf{q}},\lambda } { \hbar { \omega _ { \bf{q}}}\left ( { a_{{\bf{q}}\lambda } ^\dag { a_{{\bf{q}}\lambda } } + 1/2 } \right ) } , \end{array } \label{equ83}\ ] ] where the first term is the kinetic energy that contains the interactions via the vector potential @xmath175 . the second term represents the scalar potential interaction and the last term is the free photon hamiltonian . the non - retarded results are obtained by letting the speed of light tend to infinity . from the hamiltonian we see that this corresponds to neglecting the vector - potential interaction completely . now , in the specific system we consider here , the electrons are only free to move in a plane . in a plane only one longitudinal and one transverse electric field can exist . it turns out that the longitudinal electric field and the @xmath0-polarized photons combine into one field that is longitudinal in the plane . the induced charge and current densities from such a field will also produce a field of that kind . the @xmath2-polarized photons produce an electric field that is transverse in the plane . the current induced by such a field produces a field of the same type . the two types of interaction involving longitudinal and transverse electric fields in the planes defined by the two sheets will not mix and the energy is given by a sum of two sets of diagrams of the type given in figure[figu1 ] , one set for the longitudinal interaction and one containing only transverse interactions . let us now discuss the longitudinal interaction . the coupling to the longitudinal field and @xmath0-polarized photons is given by the scalar potential and the @xmath176 terms in the hamiltonian . the resulting interaction energy per unit area is @xmath177}^2 } } \right\}. \end{array } \label{equ84}\ ] ] compared with the non - retarded result for the correlation energy , we gave earlier in ( [ equ60 ] ) , we find that the polarizability has attained a factor @xmath178 in front , which is in agreement with the polarizability obtained by stern@xcite . furthermore , the factor appearing in the interlayer interactions is modified with the same factor in the exponent . the response to the @xmath2-polarized fields is a transverse current . it is described by the transverse conductivity , which is dominated by the contribution originating from the @xmath179-term in the hamiltonian . also the @xmath176 terms in the hamiltonian give a contribution to the transverse conductivity , in the form of the current - current correlation function . the energy per unit area from the @xmath2-photon interaction is @xmath180}^2 } } \right\}. \end{array } \label{equ85}\ ] ] we could here proceed in a way analogous to how we derived the modes in section [ cisnr ] by starting with a current density ( longitudinal or transverse ) in layer 1 . this current density gives rise to an electric field in layer 2 ; this field induces a current density in layer 2 ; this current density in layer 2 gives rise to an electric field in layer 1 which causes an induced current density in layer 1 . if we let this current density be the one we started with we have closed the loop and produced self - sustained fields , normal modes . when the current densities and electric fields are longitudinal the modes are of tm type and the te modes are the results when the current densities and fields are transverse . to vary the derivations we instead proceed in a slightly different way . we have two layers , 1 and 2 . the electric field in layer 1 has contributions from both layers and the induced current density in layer 1 is the conductivity of layer 1 times the electric field in layer 1 . the corresponding is true for layer 2 . we begin with the tm modes . the longitudinal field in layer 1 is from ( [ equ41 ] ) and ( [ equ42 ] ) @xmath181 \end{array } , \label{equ86}\ ] ] and the induced current density is @xmath182 \end{array}\\ \begin{array}{l } \quad \quad \quad = - \tilde \alpha _ \parallel ^1\left ( { { \bf{k}},\omega } \right){\gamma ^{\left ( 0 \right)}}\left ( { k,\omega } \right)\\ \quad \quad \quad \quad \times \left [ { k_\parallel ^1\left ( { { \bf{k}},\omega } \right ) + { e^ { - { \gamma ^{\left ( 0 \right)}}\left ( { k,\omega } \right)kd}}k_\parallel ^2\left ( { { \bf{k}},\omega } \right ) } \right ] . \end{array } \end{array } \label{equ87}\ ] ] in analogy we get the result for the induced current density in layer 2 by interchanging the superscripts . thus we have @xmath183 ; \end{array}\\ \begin{array}{l } k_\parallel ^2\left ( { { \bf{k}},\omega } \right ) = - \tilde \alpha _ \parallel ^2\left ( { { \bf{k}},\omega } \right){\gamma ^{\left ( 0 \right)}}\left ( { k,\omega } \right)\\ \quad \quad \quad \quad \times \left [ { { e^ { - { \gamma ^{\left ( 0 \right)}}\left ( { k,\omega } \right)kd}}k_\parallel ^1\left ( { { \bf{k}},\omega } \right ) + k_\parallel ^2\left ( { { \bf{k}},\omega } \right ) } \right ] , \end{array } \end{array } \label{equ88}\ ] ] and on matrix form @xmath184 where we have omitted all function arguments @xmath185 . this system of equations has a trivial solution , in which both current densities are zero . for a non - trivial solution the determinant of the matrix is equal to zero . thus we get the condition for modes by letting the determinant of the matrix be equal to zero . the mode condition function becomes @xmath186\left [ { 1 + { \gamma ^{\left ( 0 \right)}}\left ( { k,\omega } \right){{\tilde \alpha } ^2}_\parallel \left ( { { \bf{k}},\omega } \right ) } \right ] } } \end{array } , \label{equ90}\ ] ] where we have let the reference system be when the layers are at infinite separation . for the te modes we need the transverse fields . they are from ( [ equ41 ] ) and ( [ equ42 ] ) @xmath187 ; \end{array}\\ \begin{array}{l } e _ \bot ^2\left ( { { \bf{k}},\omega } \right ) = \frac{{i\omega } } { { { c^2}}}\frac{{2\pi } } { { k{\gamma ^{\left ( 0 \right)}}\left ( { k,\omega } \right)}}\\ \quad \quad \quad \quad \times \left [ { { e^ { - { \gamma ^{\left ( 0 \right)}}\left ( { k,\omega } \right)kd}}k _ \bot ^1\left ( { { \bf{k}},\omega } \right ) + k _ \bot ^2\left ( { { \bf{k}},\omega } \right ) } \right ] , \end{array } \end{array } \label{equ91}\ ] ] and the induced current densities @xmath188 ; \end{array}\\ \begin{array}{l } k _ \bot ^2\left ( { { \bf{k}},\omega } \right ) = \frac{{{{\left ( { \omega /ck } \right)}^2}\tilde \alpha _ \bot ^2}}{{{\gamma ^{\left ( 0 \right)}}\left ( { k,\omega } \right)}}\\ \quad \quad \quad \quad \times \left [ { { e^ { - { \gamma ^{\left ( 0 \right)}}\left ( { k,\omega } \right)kd}}k _ \bot ^1\left ( { { \bf{k}},\omega } \right ) + k _ \bot ^2\left ( { { \bf{k}},\omega } \right ) } \right ] , \end{array } \end{array } \label{equ92}\ ] ] or on matrix form @xmath189 the mode condition function for te modes is @xmath190\left [ { { { \left ( { \omega /ck } \right)}^2}\tilde \alpha _ \bot ^2 } \right]}}{{\left [ { { \gamma ^{\left ( 0 \right)}}\left ( { k,\omega } \right ) - { { \left ( { \omega /ck } \right)}^2}\tilde \alpha _ \bot ^1 } \right]\left [ { { \gamma ^{\left ( 0 \right)}}\left ( { k,\omega } \right ) - { { \left ( { \omega /ck } \right)}^2}\tilde \alpha _ \bot ^2 } \right ] } } , } \end{array } \label{equ94}\ ] ] where we have chosen as reference system the geometry when the layers are infinitely apart . the normal modes can be found from self - sustained charge and current densities , from self - sustained potentials or from self - sustained fields . here , we will find self - sustained fields . to do that one solves the mes in all regions of the geometry and makes use of the standard boundary conditions at all interfaces between the regions . for the present task we need a geometry consisting of three regions and two interfaces , @xmath191 . this geometry gives the mode condition function @xmath192 where @xmath193 is the amplitude reflection coefficient for a wave impinging on the interface between medium @xmath5 and @xmath194 from the @xmath5 side , @xmath86 is the thickness of region @xmath27 , and @xmath195 . the function @xmath196 is the dielectric function of medium @xmath5 . at an interface where there is no 2d sheet the tm and te amplitude reflection coefficients are@xcite @xmath197 and @xmath198 respectively . note that @xmath199 holds for both mode types . the amplitude reflection coefficient gets modified if there is a 2d layer at the interface . we treat the 2d layer at the interface as external to our system . the fields will then induce external surface charge and current densities at the interface . two of the boundary conditions are enough to get the modified fresnel coefficients . the other two are redundant . we choose the following two , @xmath200 where @xmath201 is the conductivity of the 2d sheet . the modified amplitude reflection coefficient for a tm mode is@xcite @xmath202 where the polarizability of the 2d sheet is obtained from the dynamical conductivity according to ( [ equ49 ] ) . for tm modes the tangential component of the electric field , which will induce the external current , is parallel to * k * , so the longitudinal 2d dielectric function of the layer enters . the bound charges in the 2d sheet also contribute to the dynamical conductivity and the polarizability . the modified amplitude reflection coefficient for a te mode is@xcite @xmath203 where the polarizability of the 2d sheet is obtained from the dynamical conductivity according to ( [ equ49 ] ) . for a te wave the electric field is perpendicular to * k * , so the transverse 2d dielectric function of the layer enters . the bound charges in the 2d sheet also contribute to the dynamical conductivity and the polarizability . now , when we have geometrical structures with 2d layers at some of the interfaces we just substitute the new reflection coefficients at the proper interfaces . note that @xmath204 no longer applies so one has to be careful when making the substitutions . to arrive at the starting expression one might have used this relation . for the three regions system @xmath191 we find the most general mode condition functions as @xmath205\left [ { \frac{{{{\tilde \varepsilon } _ 3}{\gamma _ 2 } - { { \tilde \varepsilon } _ 2}{\gamma _ 3 } + 2{\gamma _ 2}{\gamma _ 3}\tilde \alpha _ r^\parallel } } { { { { \tilde \varepsilon } _ 3}{\gamma _ 2 } + { { \tilde \varepsilon } _ 2}{\gamma _ 3 } + 2{\gamma _ 2}{\gamma _ 3}\tilde \alpha _ r^\parallel } } } \right];}\\ { f_{\bf{k}}^{{\rm{te}}}\left ( \omega \right ) } \\ { = 1 - { e^ { - 2{\gamma _ 2}kd}}\left [ { \frac{{{\gamma _ 2 } - { \gamma _ 1 } + 2{{\left ( { \omega /ck } \right)}^2}\tilde \alpha _ l^ \bot } } { { { \gamma _ 2 } + { \gamma _ 1 } - 2{{\left ( { \omega /ck } \right)}^2}\tilde \alpha _ l^ \bot } } } \right]\left [ { \frac{{{\gamma _ 2 } - { \gamma _ 3 } + 2{{\left ( { \omega /ck } \right)}^2}\tilde \alpha _ r^ \bot } } { { { \gamma _ 2 } + { \gamma _ 3 } - 2{{\left ( { \omega /ck } \right)}^2}\tilde \alpha _ r^ \bot } } } \right ] , } \end{array } \label{equ101}\ ] ] where we assumed there are 2d sheets at both interfaces . these sheets may be different so we have put the subscripts @xmath206 and @xmath207 on the polarizability of the left and right sheet , respectively . we obtain two freestanding 2d sheets from our geometry by letting all three media be vacuum . then the mode condition functions in ( [ equ101 ] ) reduce to @xmath208\left [ { \frac{{{\gamma ^{\left ( 0 \right)}}\tilde \alpha _ 2^\parallel \left ( { k,\omega } \right)}}{{1 + { \gamma ^{\left ( 0 \right)}}\tilde \alpha _ 2^\parallel \left ( { k,\omega } \right ) } } } \right];}\\ { \begin{array}{*{20}{l } } { f_{\bf{k}}^{{\rm{te}}}\left ( \omega \right ) = 1 - { e^ { - 2{\gamma ^{\left ( 0 \right)}}kd}}\left [ { \frac{{{{\left ( { \omega /ck } \right)}^2}\tilde \alpha _ 1^ \bot \left ( { k,\omega } \right)}}{{{\gamma ^{\left ( 0 \right ) } } - { { \left ( { \omega /ck } \right)}^2}\tilde \alpha _ 1^ \bot \left ( { k,\omega } \right ) } } } \right]}\\ { \quad \quad \quad \quad \quad \quad \quad \times \left [ { \frac{{{{\left ( { \omega /ck } \right)}^2}\tilde \alpha _ 2^ \bot \left ( { k,\omega } \right)}}{{{\gamma ^{\left ( 0 \right ) } } - { { \left ( { \omega /ck } \right)}^2}\tilde \alpha _ 2^ \bot \left ( { k,\omega } \right ) } } } \right]}. \end{array } } \end{array } \label{equ102}\ ] ] when a 2d layer is placed next to a substrate there will be induced current densities at the interface . the effect of these densities can be obtained in a much simpler way using mirror images . how these mirror images look like was derived in the section [ images ] for a more complex system where the 2d layer is embedded in a medium . for our system the mirror - image current - densities are @xmath209 when @xmath72 and @xmath210 are the actual longitudinal and transverse , respectively , current densities in the 2d layer . these mirror images are at a distance @xmath86 from the interface and inside the substrate . we have suppressed the argument @xmath211 in all @xmath212-functions . the coupled longitudinal current densities give rise to tm modes and the transverse to te modes . we start with the tm modes . the resulting longitudinal electric fields from a longitudinal current density was derived in ( [ equ41 ] ) and ( [ equ42 ] ) . the field in the 2d layer is @xmath213 . \end{array } \label{equ104}\ ] ] we close the loop by noting that the current density in the 2d layer is the conductivity times the electric field . thus , @xmath214\\ = - { { \tilde \alpha } _ \parallel } \left ( { { \bf{k}},\omega } \right){\gamma ^{\left ( 0 \right)}}\left [ { 1 + { e^ { - 2{\gamma ^{\left ( 0 \right)}}kd}}\frac{{{\gamma _ s } - { { \tilde \varepsilon } _ s}\left ( \omega \right){\gamma ^{\left ( 0 \right)}}}}{{{\gamma _ s } + { { \tilde \varepsilon } _ s}\left ( \omega \right){\gamma ^{\left ( 0 \right ) } } } } } \right]{k_\parallel } \left ( { { \bf{k}},\omega } \right ) , \end{array } \label{equ105}\ ] ] where we have used ( [ equ103 ] ) in the last step to eliminate the mirror density . we may now use the first and last parts of this equation to find @xmath215 } \right\}\\ \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \times { k_\parallel } \left ( { { \bf{k}},\omega } \right ) = 0 . \end{array } \label{equ106}\ ] ] this has the trivial solution that the current density is zero . there are non - trivial solutions , the normal modes , if what is inside the curly brackets is zero . this leads to the mode condition function for tm modes , @xmath216\\ \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \times \left [ { \frac{{{{\tilde \varepsilon } _ s}\left ( \omega \right){\gamma ^{\left ( 0 \right ) } } - { \gamma _ s}}}{{{{\tilde \varepsilon } _ s}\left ( \omega \right){\gamma ^{\left ( 0 \right ) } } + { \gamma _ s } } } } \right ] , \end{array } \label{equ107}\ ] ] where we have chosen as reference system the system when the 2d layer is at infinite distance from the substrate . now , we proceed with the te modes . the resulting transverse electric fields from a transverse current density was derived in ( [ equ41 ] ) and ( [ equ42 ] ) . the field in the 2d layer is @xmath217 . \label{equ108}\ ] ] we close the loop by noting that the current density in the 2d layer is the conductivity times the electric field . thus , @xmath218\\ = { { \tilde \alpha } _ \bot } \left ( { { \bf{k}},\omega } \right)\frac{{{{\left ( { \omega /ck } \right)}^2}}}{{{\gamma ^{\left ( 0 \right)}}}}\left [ { 1 + { e^ { - 2{\gamma ^{\left ( 0 \right)}}kd}}\frac{{{\gamma ^{\left ( 0 \right ) } } - { \gamma _ s}}}{{{\gamma ^{\left ( 0 \right ) } } + { \gamma _ s } } } } \right]{k _ \bot } \left ( { { \bf{k}},\omega } \right ) , \end{array } \label{equ109}\ ] ] where we have used ( [ equ103 ] ) in the last step to eliminate the mirror density . we may now use the first and last parts of this equation to find @xmath219 } \right\}\\ \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \times { k _ \bot } \left ( { { \bf{k}},\omega } \right ) = 0 . \end{array } \label{equ110}\ ] ] this has the trivial solution that the current density is zero . there are non - trivial solutions , the normal modes , if what is inside the curly brackets is zero . this leads to the mode condition function for te modes , @xmath220 where we have chosen as reference system the system when the 2d layer is at infinite distance from the substrate . we obtain a freestanding 2d sheet above a substrate from our geometry in section [ meret22d ] by letting media @xmath26 and @xmath27 be vacuum and medium @xmath221 be the substrate . then the mode condition functions in ( [ equ101 ] ) reduce to @xmath222\left [ { \frac{{{{\tilde \varepsilon } _ s}\left ( \omega \right){\gamma ^{\left ( 0 \right ) } } - { \gamma _ s}}}{{{{\tilde \varepsilon } _ s}\left ( \omega \right){\gamma ^{\left ( 0 \right ) } } + { \gamma _ s } } } } \right];}\\ { f_{\bf{k}}^{{\rm{te}}}\left ( \omega \right ) = 1 - { e^ { - 2{\gamma ^{\left ( 0 \right)}}kd}}\left [ { \frac{{{{\left ( { \omega /ck } \right)}^2}{{\tilde \alpha } _ \bot } \left ( { { \bf{k}},\omega } \right)}}{{{\gamma ^{\left ( 0 \right ) } } - { { \left ( { \omega /ck } \right)}^2}{{\tilde \alpha } _ \bot } \left ( { { \bf{k}},\omega } \right ) } } } \right]\left [ { \frac{{{\gamma ^{\left ( 0 \right ) } } - { \gamma _ s}}}{{{\gamma ^{\left ( 0 \right ) } } + { \gamma _ s } } } } \right ] . } \end{array } \label{equ112}\ ] ] if retardation effects are neglected there are no te modes and the mode condition function for tm modes is reduced further to @xmath223\left [ { \frac{{{{\tilde \varepsilon } _ s}\left ( \omega \right ) - 1}}{{{{\tilde \varepsilon } _ s}\left ( \omega \right ) + 1 } } } \right ] , \label{equ113}\ ] ] which is fully in line with what we obtained starting from the non - retarded theory . we have derived the expressions for the dispersion interaction in systems containing 2d sheets in several different complementary ways , all with a condensed matter physics perspective . the geometries considered were two free standing 2d sheets and one 2d sheet above a substrate . the derivations were either performed using many - body theory with diagrammatic perturbation expansions based on feynman diagrams or were based on electromagnetic normal modes . the normal modes were generated in several different ways . the zero - point energy is a very important concept in the normal mode derivation of the dispersion interaction . it is very interesting to note that the same results are obtained using many - body theory without invoking zero - point energy . 10 g. d. mahan 1990 _ many - particle physics _ ( new york and london : plenum press ) a. l. fetter , and j. d. walecka 1971 _ quantum theory of many - particle systems _ ( new york : mcgaw - hill ) bo e. sernelius 2001 _ surface modes in physics _ ( berlin : wiley - vch ) bo e. sernelius 2012 _ graphene _ * 1 * 21 bo e. sernelius 2011 _ epl _ * 95 * 57003 bo e. sernelius 2012 b * 85 * 195427 m. bostrm , and bo e. sernelius 2012 a * 85 * 012508 bo e. sernelius 2012 _ int . * 14 * 531 g. l. klimchitskaya , v. m. mostepanenko , and bo e. sernelius 2014 b * 89 * 125407 m. bordag , i. v. fialkovsky , d. m. gitman , and d. v. vassilevich 2009 b * 80 * 245406 ignat v. fialkovsky , valery n. marachevsky , and dmitri v. vassilevich 2011 b * 84 * 035446 jalal sarabadani , ali naji , reza asgari , and rudolf podgornik 2011 b * 84 * 155407 v. svetovoy , z. moktadir , m. elwenspoek , and h. mizuta 2011 _ epl _ * 96 * 14006 d. drosdoff , and lilia m. woods 2010 b * 82 * 155459 bo e. sernelius , and k .- f . beggren 1980 surf . sci . * 98 * 191 bo e. sernelius , and p. bjrk 1998 b * 57 * 6592 j. d. jackson 1999 _ classical electrodynamics _ ( new york : john wiley&sons , inc.)p156 f. stern 1967 546

small variations of the entanglement entropy and the expectation value of the modular hamiltonian are computed holographically for circular entangling curves in the boundary of , using gravitational perturbations with general boundary conditions in spherical coordinates . agreement with the first law of thermodynamics , , requires that the line element of the entangling curve remains constant . in this context , we also find a manifestation of electric - magnetic duality for the entanglement entropy and the corresponding modular hamiltonian , following from the holographic energy - momentum / cotton tensor duality . entanglement entropy , holography , duality , arxiv : 1503.00627

1503.00627

it has been known for a long time that there is a striking similarity between black hole physics and thermodynamics @xcite , which suggests a deep connection between gravity and thermodynamics . this led to several attempts to understand einstein s equations as effective equations emerging from the thermodynamics of underlying degrees of freedom , such as @xcite . on the other hand , ads / cft correspondence @xcite provides a broad framework allowing the description of gravitational theories with ads asymptotics in @xmath4 dimensions as emergent from strongly coupled conformal field theories in @xmath5 dimensions . a fair question in the ads / cft framework is whether the similarities between gravity and thermodynamics can be explained by considering einstein s equations as thermodynamic relations for the conformal field theory degrees of freedom @xcite . more recently , it has also been suggested that the connection between gravity and thermodynamics should not be attributed to thermal statistics , but rather to quantum statistics related to quantum entanglement physics @xcite . more specifically , it has been conjectured that the entanglement entropy , which is a measure of entanglement between subsystems of a composite quantum system and it is defined for a given entangling surface that separates the degrees of freedom of the corresponding conformal field theory into two subsystems , is directly connected to the area of an open extremal hypersurface in the emergent asymptotically ads geometry whose boundary is the entangling surface . this conjecture , which is named after ryu - takayanagi @xcite , provides a quantitative tool to understand how gravitational dynamics emerges from thermodynamics related to entanglement in the boundary conformal field theory . so far , this programme has been advanced by comparing the variation of entanglement entropy to the variation of the expectation value of the so called modular hamiltonian for any given entangling surface @xcite . the latter can be expressed in terms of the holographic energy - momentum tensor when spherical entangling surfaces are taken in poincar coordinates @xcite , while the former is provided by the ryu - takayanagi formula @xcite . enforcing the first law of thermodynamics for entanglement through holography , imposes constraints for the metric perturbations around ads space , which to linear order turn out to be einstein s equations satisfying dirichlet boundary conditions @xcite . in this context , it is also known that all solutions of the linearized einstein s equations with dirichlet boundary conditions satisfy the first law of thermodynamics . in this paper , we work out the holographic realization of the first law of thermodynamics for small perturbations of spherical @xmath2 space - time having axial symmetry and satisfying general boundary conditions . the general framework for field equations satisfying general boundary conditions in @xmath2 is provided in reference @xcite . the entangling surfaces are now curves , since the boundary conformal field theory is @xmath6 dimensional , and they are taken to be circular , bounding a polar cap region , so that they respect the axial symmetry of the bulk geometry . our general result is that the first law of thermodynamics for entanglement is realized holographically in all cases , hereby extending previous works beyond dirichlet boundary conditions , provided that the line element of the entangling curve is inert to the perturbations . in this context , we also examine the role of gravitational electric - magnetic duality to entanglement physics . it is well known that small perturbations of maximally symmetric spaces exhibit a rank-2 generalization of electric - magnetic duality , interchanging the linearized einstein equations with the bianchi identities in four space - time dimensions . this symmetry was originally discussed for gravitons in minkowski space @xcite , but it was subsequently generalized in the presence of cosmological constant by considering metric perturbations of @xmath7 @xcite and @xmath2 space - time @xcite . there , it was also found that electric - magnetic duality in @xmath2 has a holographic manifestation as energy - momentum / cotton tensor duality . these considerations provide the gravitational analogue of the holographic interpretation of electric - magnetic duality in theories with @xmath8 gauge symmetry @xcite , but their physics in the space of boundary three - dimensional conformal theories still remain largely unexplored . duality acts on metric perturbations by interchanging their boundary conditions . hence , our interest in the holographic description of gravitational perturbations satisfying general boundary conditions in the spirit of reference @xcite . extending the applications of gravitational duality to entanglement entropy and related issues may shed new light into this interesting subject . the material of this paper is organized as follows : in section 2 , we present a review of the notions of entanglement entropy and modular hamiltonian , together with their holographic description , and include various formulae that will be used in the computations . in section 3 , we discuss the general theory of gravitational perturbations of @xmath2 space - time and formulate the linearized einstein equations as an effective schrdinger problem , splitting the perturbations into two distinct classes with opposite parity . the presentation is made general , encompassing arbitrary boundary conditions . in section 4 , we compute holographically the variations of the entanglement entropy and the modular hamiltonian and compare the two expressions for general boundary conditions . the first law of thermodynamics for entanglement is verified in all cases , while describing the subtleties that go into the computation . in section 5 , we address the role of electric - magnetic duality in holography and study its implications for the first law of thermodynamics for entanglement . finally , section 6 , contains our conclusions and a short discussion of open problems . there are also three appendices containing various technical details and formulae that are used in the main text . we present a brief account of the notions of entanglement entropy and modular hamiltonian together with their holographic description in terms of bulk space geometry . we also derive some general formulae that will be used later for gravitational perturbations of @xmath2 space - time satisfying general boundary conditions . consider a composite quantum system comprising of several subsystems . even if the composite system lies in a pure state , with density matrix @xmath9 , this may not be true for its subsystems , which are hereby described by a density matrix equal to the partial trace of @xmath9 over the degrees of freedom of the complementary subsystem , @xmath10 when the complementary subsystems @xmath11 and @xmath12 are not entangled , the reduced density matrix @xmath13 also describes a pure state . entanglement between systems @xmath11 and @xmath12 is encoded to the spectrum of the reduced density matrix @xmath13 under the implicit assumption that the composite system lies in a pure state . the entanglement entropy is defined as the von neumann entropy associated to the reduced density matrix @xmath13 , @xmath14 the density matrix @xmath13 is hermitian and positive semi - definite , leading to the definition of the corresponding modular hamiltonian as @xmath15 then , the entanglement entropy can be rewritten in terms of the modular hamiltonian as @xmath16 small variations in the pure state of the overall system or the region @xmath11 generate variations of the density matrix , @xmath17 , and , thus , the entanglement entropy will also be perturbed . we have @xmath18\\ & = - { \mathrm{tr}}\left ( { \ln { \rho _ a}\delta { \rho _ a } } \right ) - { \mathrm{tr}}\left ( { \delta { \rho _ a } } \right)\\ & = { \mathrm{tr}}\left ( { { h_a}\delta { \rho _ a } } \right ) = \delta \left\langle { { h_a } } \right\rangle , \end{split}\ ] ] since the trace of the density matrix is normalized to one and thus , @xmath19 . thus , the variations of the entanglement entropy and the expectation value of the corresponding modular hamiltonian are equal @xmath20 this equation is the direct analog of the first law of thermodynamics for entanglement physics @xcite that leads our work . the ryu - takayanagi conjecture @xcite connects the entanglement entropy of a region @xmath11 defined by the entangling boundary surface @xmath21 in the boundary field theory to the area of an extremal co - dimension two open surface in the bulk gravitational dual theory with boundary @xmath21 . specifically , the entanglement entropy is given by @xmath22 where @xmath23 is the corresponding extremal co - dimension two surface in the bulk . in the following , without loss of generality , we set newton s gravitational constant @xmath24 . these expressions are applicable to all holographic models . for @xmath2 , which is of interest here , @xmath21 is a closed curve and @xmath23 is two - dimensional . then , the area of the extremal surface , which itself will be denoted by @xmath23 in the following , is expressed in terms of the induced metric as @xmath25 where @xmath26 and @xmath27 here , @xmath28 is the bulk metric , @xmath29 are coordinates parametrizing the extremal surface , @xmath30 are the parametric equations of the extremal surface in the bulk and @xmath31 is the induced metric on the extremal surface . extremal surfaces in ads space - times have been mostly studied in poincar patch coordinates . in those coordinates , the space - time line element of @xmath2 with unit scale is @xmath33 then , for a disc region @xmath11 in the boundary plane described by @xmath34 , the corresponding extremal surface @xmath23 in the ads bulk is given by @xmath35 without loss of generality we may take the disc centered at @xmath36 . passing from poincar coordinates @xmath37 to global coordinates @xmath38 with the aid of the coordinate transformation @xmath39 the space - time metric takes the form @xmath40 whereas the extremal surface corresponding to the choice @xmath41 is given in global coordinates by @xmath42 , \ , \varphi \in \left [ { 0,2\pi } \right ) \ , , \label{eq : extremal_surface_global_r}\ ] ] where @xmath43 and @xmath44 are specific functions of @xmath45 and @xmath46 ( see [ sec : coordinate_transformation ] for more details ) . equivalently , we have parametrization @xmath47 the region @xmath11 becomes a polar cap in global coordinates and the complementary cap is the region @xmath12 . figure 1 depicts the regions @xmath11 and @xmath12 on the spherical boundary of space - time . has opening angle @xmath43 . the entangling curve @xmath21 is the thick line.,title="fig:",scaledwidth=60.0% ] ( -112,111)@xmath43 ( -65,175)region @xmath11 ( -65,15)region @xmath12 ( -35,140)@xmath21 it is easy to confirm that the surface is extremal , obeying the particular conditions derived by minimizing the area functional , under the assumption that the surface is rotationally symmetric , i.e. , @xmath48 and @xmath49 , since @xmath50 where prime denotes differentiation with respect to @xmath51 . the extremal surface emanates from a polar cap boundary region described by @xmath52 . an introduction of non - vanishing parameters @xmath53 or @xmath54 would rotate the entangling curve so that its symmetry axis would not anymore coincide with the axis corresponding to the azimuthal angle @xmath55 . having said that , we stick to the choice @xmath56 from now on . figure 2 depicts the extremal surface @xmath57 that emanates from a polar cap region in the boundary of space - time in global coordinates and extends in the interior of space - time . the radial coordinate in the plot is proportional to the so called tortoise coordinate , @xmath58 . space - time for various choices of @xmath43.,scaledwidth=60.0% ] parametrizing the extremal surface with the coordinates @xmath55 and @xmath51 , the induced metric turns out to be diagonal with elements @xmath59}},\label{eq : unperturbed_induced_metric_rr}\\ { \gamma _ { \varphi \varphi } } & = { r^2}\left [ { 1 - \left ( { 1 + \frac{1}{{{r^2 } } } } \right){{\cos } ^2}{\theta _ 0 } } \right].\label{eq : unperturbed_induced_metric_phiphi}\end{aligned}\ ] ] the determinant of the induced metric is @xmath60 and the area follows from equation , @xmath61 the first term is the divergent `` area law '' term , while the second one is universal independent of the uv cutoff in the boundary of @xmath62 is given by @xmath63 where @xmath64 is the uv cutoff . the first term is the `` area law '' term . for @xmath5 even , the logarithmic term is universal and connected to the conformal anomaly @xcite . for @xmath5 odd , which is the case of interest here ( @xmath65 ) , the constant term is universal and it obeys a holographic `` c - theorem '' @xcite . ] @xcite . the area functional depends on the background metric @xmath28 as well as on the embedding variables @xmath66 , and , of course , it also depends implicitly on the entangling curve @xmath21 curving the region @xmath11 in the boundary field theory . perturbations of the bulk metric induce changes of the minimal surface . the variation of its area is given in general by @xmath67 as both the metric and the embedding equations of the extremal surface vary around their unperturbed values @xmath68 and @xmath69 . when the perturbation obeys dirichlet boundary conditions , the region @xmath11 and its boundary @xmath21 remain fixed . then , the second term vanishes , as the original surface described by equations @xmath69 extremizes the area functional with the given loop @xmath21 . when the metric perturbations satisfy general boundary conditions , computing the variation of the area of the minimal surface is more subtle . for this , we suppose that the boundary loop @xmath21 specifying the extremal surface is determined by a set of parameters @xmath70 , which may also vary as metric perturbations are turned on . then , the area functional has the particular form @xmath71 and its variation breaks down as follows , @xmath72 the second term vanishes , as it corresponds to the variation of the area for variations of the extremal surface with a fixed boundary that is provided by the entangling loop @xmath21 . the last term can be simplified if one considers the area as functional of the metric as well as the parameters @xmath70 specifying the entangling loop . as a result , the total variation of the area takes the form @xmath73 hereby defining the individual contributions @xmath74 and @xmath75 . the first term follows by varying the relation and the result is written in terms of the unperturbed induced metric @xmath76 and @xmath77 as @xmath78 where @xmath79 the second term can be easily calculated for polar caps , which only depends on a single parameter @xmath70 that is taken to be @xmath43 . then , @xmath75 follows by varying the relation with respect to the cap parameter @xmath43 . the individual terms @xmath74 and @xmath75 will be explicitly computed later for small perturbations of the @xmath2 metric satisfying the linearized einstein equations with general boundary conditions . unlike the ryu - takayanagi formula for expressing the entanglement entropy in terms of the bulk gravitational theory , there is no similar expression for the modular hamiltonian . in general , the modular hamiltonian is a non - local operator and there is no way to this day to find an expression for the modular hamiltonian for a general boundary state and region @xmath11 . in some cases , however , the modular hamiltonian generates a geometric flow leading to a local expression . such an example is provided by taking a disk region @xmath11 of radius @xmath45 in a minkowski space boundary @xcite , in which case the modular flow in the cauchy development of the disk can be connected with the modular flow in rindler space through a conformal transformation ; the details are provided in [ sec : modular_hamiltonian_disc ] . then , the modular hamiltonian is expressed in terms of the holographic energy - momentum tensor and a conformal killing vector that leaves invariant the entangling curve @xmath21 and its causal development , as @xmath80 where @xmath81 is a space - like surface with boundary @xmath21 and @xmath82 is the differential volume form on @xmath81 @xcite . if @xmath81 is selected to be a constant time slice , it will coincide with the region @xmath11 . for a disk in minkowski space , this conformal killing vector @xmath83 is a linear combination of the killing vectors corresponding to the time translation and a special conformal transformation ; it is selected so that @xmath83 vanishes on @xmath21 and its causal development . using the coordinate transformation , the conformal killing vector in question takes the following form in global coordinates ( for more details see [ sec : coordinate_transformation ] ) , @xmath84 , \label{eq : dmh_killing_vector}\ ] ] which indeed satisfies the conformal killing vector equation @xmath85 and at the same time vanishes at the entangling surface @xmath86 and @xmath87 . the conformal killing vector @xmath83 is the boundary limit of a conformal killing vector field @xmath88 in the bulk , @xmath89 , \label{eq : dmh_mh_killing_bulk}\end{gathered}\ ] ] which also vanishes at the entangling surface . note in passing that if one considers generalized theories of gravity ( i.e. , higher derivative corrections ) an appropriate generalization of ryu - takayanagi conjecture will be required . such generalizations can be obtained through the iyer - wald theorem that involves bifurcate killing horizons generated by the conformal killing vector @xmath88 given by @xcite . formula can be evaluated for any space - like surface with boundary identical to the entangling surface . selecting the @xmath87 surface and using the specific form of the killing vector , the modular hamiltonian for the polar cap region @xmath52 is written as @xmath90 this completes the presentation of the general formulae for the quantities of interest . in section [ sec : ee_perturbed_ads4_calculation ] @xmath1 will be calculated and compared to @xmath0 for all different kinds of @xmath2 perturbations satisfying general boundary conditions . now we come to the theory of gravitational perturbations of @xmath2 space - time , allowing for general boundary conditions . it is convenient to split the perturbations into two complementary sets with opposite parity and reduce the linearized einstein equations to an affective schrdinger problem , which is exactly solvable . then , we write down the holographic energy - momentum tensor for all such perturbations , which will be used later . the material we present here is based on the discussion found in @xcite ( but see also @xcite for an overview of the subject ) and references therein . we consider einstein gravity in four space - time dimensions with negative cosmological constant @xmath91 , @xmath92 @xmath2 is the maximally symmetric solution whose metric takes the following form in spherical coordinates , @xmath93 where @xmath94 we set the ads scale @xmath95 equal to one to simplify the presentation . for later use , we define the tortoise coordinate @xmath96 as @xmath97 which is given explicitly as @xmath98 the tortoise coordinate is an angular variable ranging from @xmath99 to @xmath100 as @xmath51 varies from @xmath99 to @xmath101 . next , we consider linear perturbations around the @xmath2 metric satisfying einstein s equations . the gravitational perturbations fall in two complementary classes with opposite parity , namely axial and polar perturbations . without loss of generality we restrict attention to axially symmetric deformations , which are described in terms of legendre polynomials @xmath102 instead of more general spherical harmonics @xmath103 . the metric perturbations of this class are parametrized by two functions @xmath104 and @xmath105 as follows , @xmath106 up to reparametrizations . it turns out that einstein s equations are equivalent to the effective schrdinger problem with respect to the tortoise coordinate , @xmath107 where the functions @xmath104 and @xmath105 are expressed in terms of the solutions as @xmath108 the metric perturbations of this class are parametrized by three functions @xmath109 , @xmath110 and @xmath111 as follows , @xmath112 , \label{eq : polar_perturbations_metric}\end{gathered}\ ] ] up to reparametrizations . similarly to the previous case , einstein s equations for polar perturbations turn out to be equivalent to the effective schrdinger problem with respect to the tortoise coordinate , @xmath113 which is identical to the one describing the axial perturbations . the functions @xmath109 , @xmath110 and @xmath111 are expressed in terms of the solutions of the effective schrdinger problem as @xmath114 thus , the spectrum and the eigen - functions of the operator @xmath115 in the closed interval @xmath116 $ ] completely determine the gravitational perturbations of @xmath2 space - time . the general solution of the linearized einstein equations is written as linear combination of the axial and polar solutions . the solution of effective schrdinger problem can be expressed in terms of hypergeometric functions . the normalizable solution , which vanishes at @xmath117 , is @xmath118 the other independent solution of the effective schrdinger problem is @xmath119 but it diverges as @xmath120 and is not normalizable . the boundary conditions as @xmath121 can be systematically described by expanding the normalizable solutions in powers of @xmath122 , as @xmath123 using properties of the hypergeometric functions , it turns out that @xmath124 and @xmath125 are expressed in terms of gamma - functions as @xmath126 the coefficients of all other terms are expressed in terms of @xmath124 and @xmath125 , but the details do not really matter . we only note here , for later use , that @xmath127 the boundary conditions imposed as @xmath121 are solely described in terms @xmath124 and @xmath125 , since @xmath128 thus , solutions , characterized by @xmath129 obey dirichlet boundary conditions for the effective schrdinger problem , while solutions characterized by @xmath130 obey neumann boundary conditions . solutions with @xmath124 and @xmath125 taking more general values correspond to more general ( mixed ) boundary conditions determined by the ratio @xmath131 . in the following , we will denote by @xmath132 the coefficients of the large @xmath51 expansion of @xmath133 associated to axial perturbations and @xmath134 for the polar perturbations , for distinction . for given boundary conditions , the spectrum is discrete ; for example , for dirichlet boundary conditions we find @xmath135 whereas for neumann boundary conditions the spectrum is @xmath136 more general boundary conditions give rise to intermediate ( generally not equidistant ) frequencies . in all cases , the spectrum is implicitly determined by ratios of gamma - functions fixed by the ratio @xmath137 or @xmath138 . the energy - momentum tensor is divergent in asymptotic ads spaces and appropriate renormalization is required to make sense of it , using counter - terms . here , we summarize the results of holographic renormalization @xcite for the boundary space - time metric and the corresponding energy - momentum tensor for gravitational perturbations of @xmath2 space - time satisfying general boundary conditions , following @xcite . thus , we have per sector the following results , setting newton s constant @xmath24 : the boundary metric after conformal rescaling takes the form @xmath139 whereas the non - vanishing components of the energy momentum tensor for small perturbations are @xmath140 \ , .\end{aligned}\ ] ] in this sector , the boundary metric after conformal rescaling takes the form @xmath141 whereas the non - vanishing components of the energy - momentum tensor for the corresponding small perturbations are @xmath142 in all cases , the energy - momentum tensor is traceless and covariantly conserved , @xmath143 with respect to the boundary metric , as required on general grounds . equations and imply that for axial perturbations dirichlet boundary conditions for the effective schrdinger problem correspond to dirichlet conditions for the metric , while for polar perturbations neumann boundary conditions for the effective schrodinger problem correspond to dirichlet conditions for the metric . these results will be particularly useful while computing the modular hamiltonian as an appropriately weighted integral of the energy density for all kind of perturbations . it can already be seen that @xmath1 always vanishes for axial perturbations , since @xmath144 irrespective of boundary conditions , whereas @xmath1 does not vanish for polar perturbations satisfying boundary conditions other than @xmath145 . we are now in position to compute holographically the variation of the entanglement entropy and the modular hamiltonian for gravitational perturbations of @xmath2 space - time satisfying general boundary conditions . first , we compute the variation of the area of the extremal surface emanating from the loop @xmath21 , as @xmath146 , then we compute @xmath1 , and , finally , we compare the results with the first law of thermodynamics for the entanglement entropy , @xmath3 . we find that the deformations of the entangling curve @xmath21 should be isoperimetric , when the boundary conditions of the metric allow for fluctuating geometries on the boundary of @xmath2 for , otherwise , the first law of thermodynamics would not be realized holographically . our results generalize earlier work on the subject , going from dirichlet to more general boundary conditions for the metric . first , we work out the contribution @xmath147 to the variation of the area of the minimal surface . naturally , we consider the effect of axial and polar perturbations separately . for axial perturbations , the variation of the induced metric on the minimal surface , which is generally given by equation , turns out to be @xmath148 and so the corresponding variation of the area , given by equation , is @xmath149 for polar perturbations , the variation of the induced metric turns out to be @xmath150 then , using equations , and , we find after some algebra that @xmath151 { e^ { - i\omega t_0}}{p_l}\left ( { \cos { \theta _ 0}\sqrt { 1 + \frac{1}{{{r^2 } } } } } \right ) .\end{gathered}\ ] ] trading @xmath51 with the tortoise coordinate @xmath96 and expressing @xmath152 and @xmath153 in terms of the @xmath154 , as given by equations and , we obtain @xmath155 { e^ { - i\omega t_0}}{p_l}\left ( { \frac{{\cos { \theta _ 0}}}{{\sin x } } } \right ) .\end{gathered}\ ] ] the effective schrdinger problem can be used further to eliminate @xmath156 , leading to the following expression , @xmath157 } \right . } \\ { + l\left ( { l + 1 } \right)\frac{{\cos { \theta _ 0}\cot x}}{{\tan { \theta _ 0}\sin x } } \ , { \psi _ { \rm{pol } } } } \right]{e^ { - i\omega t_0}}{p_l}\left ( { \frac{{\cos { \theta _ 0}}}{{\sin x } } } \right).\end{gathered}\ ] ] it is convenient at this point to employ the following identities among legendre polynomials , @xmath158 to perform integration by parts and get @xmath159_{\frac{\pi } { 2 } - { \theta _ 0}}^{\frac{\pi } { 2}}\\ - \pi { e^ { - i\omega t_0 } } \left [ { \cot { \theta _ 0 } \ , \frac{{l\left ( { l + 1 } \right)}}{{2l + 1 } } \left ( { { p_{l + 1}}\left ( { \frac{{\cos { \theta _ 0}}}{{\sin x } } } \right ) - { p_{l - 1}}\left ( { \frac{{\cos { \theta _ 0}}}{{\sin x } } } \right ) } \right ) } { \psi _ { \rm{pol } } } \right]_{\frac{\pi } { 2 } - { \theta _ 0}}^{\frac{\pi } { 2 } } .\end{gathered}\ ] ] the contribution from the end - point @xmath160 vanishes for both bracketed terms , and , thus , the result is @xmath161 . \label{eq : dag_formula_limit_x}\end{gathered}\ ] ] going back to the coordinate @xmath51 and using the asymptotic expansion of the schrdinger wave - functions together with equation , we arrive at the final expression @xmath162 . \label{eq : dag_formula_limit_r}\end{gathered}\ ] ] note that @xmath74 is divergent when the perturbations of the metric do not satisfy dirichlet boundary conditions , @xmath163 . since we are considering axially symmetric perturbations of the bulk metric , the deformations of the entangling curve are taken to be independent of the angular coordinate @xmath55 to conform with the symmetry . thus , the variations of the parameter @xmath43 in the polar cap are taken to be independent of @xmath55 , i.e. , @xmath165 consequently , @xmath75 can be derived directly by varying with respect to @xmath43 , leading to the following result @xmath166 for all kind of metric perturbations . note that @xmath164 also appears to be divergent on general grounds . it is important to see how these divergences combine with those appearing in @xmath147 to form @xmath167 . summing up the two contributions , we obtain the net variation of the area of the minimal surface , @xmath167 , and , hence , the variation of entanglement entropy , @xmath168 . we present the result for axial and polar perturbations , setting newton s constant @xmath24 . thus , for axial perturbations , we have @xmath169 whereas for polar perturbations the result reads @xmath170 \\ + \frac{\pi } { 2}\mathop { \lim } \limits_{r \to \infty } r\cos { \theta _ 0 } \ , \delta { \theta}^{{\rm{polar}}}\left ( t_0 \right ) . \label{eq : entanglement_entropy_result_polar}\end{gathered}\ ] ] the variance of the entropy appears to be divergent , which is unphysical . specific conditions should be imposed on @xmath171 to make them cancel , but these will be discussed later , while comparing @xmath168 with @xmath172 , together with their geometric meaning . next , we compute the variation of the modular hamiltonian by specializing formula to axial and polar perturbations satisfying general boundary conditions . as before , newton s constant is set equal to 1 . since @xmath173 for the axial perturbations , we immediately obtain @xmath174 in all cases , irrespective of boundary conditions . for polar perturbations with @xmath175 , the component @xmath176 of the holographic energy - momentum tensor does not vanish , and , hence , the corresponding variance of the modular hamiltonian differs from zero . first , substituting the expression for @xmath176 into equation , yields the following expression for the variance of the modular hamiltonian , @xmath177 then , we perform integration by parts and make use of formula twice to find @xmath178 .\end{gathered}\ ] ] next , using the following identity of legendre polynomials , @xmath179 to eliminate @xmath180 and @xmath181 , we arrive at the expression @xmath182 , \end{gathered}\ ] ] which can be conveniently rewritten as @xmath183 .\end{gathered}\ ] ] using once more the identity yields the final result , @xmath184 . \label{eq : modular_hamiltonian_result_polar}\end{gathered}\ ] ] we see that in all cases the variance of the modular hamiltonian is finite . the first law of thermodynamics @xmath3 for the entanglement entropy and the modular hamiltonian is a tautology from the point of view of the boundary theory . its realization through the ryu - takayanagi formula and the holographic energy - momentum tensor is not automatic , but it is a consistency check of the whole scheme . comparing the results derived above , we find that @xmath185 for any polar cap region a , provided that the entangling curve at the boundary also deforms , where appropriate , in a certain way . the boundary is not held fixed when general boundary conditions are imposed on the metric , and , thus , the polar cap region @xmath11 and the entangling curve @xmath21 also vary . thus , @xmath186 should be adjusted accordingly to achieve agreement with the first law of thermodynamics . as will be seen shortly this requirement has natural geometric interpretation . it also gets rid of the divergences appearing in the computation of @xmath168 . first , we consider the case of axial perturbations and compare the expressions for @xmath168 and for @xmath172 . the first law is realized holographically provided that @xmath187 irrespective of boundary conditions . thus , axial perturbations do not require adjustment of the polar cap region . a way to understand this is to consider the boundary metric of the perturbed @xmath2 space - time , prior to conformal scaling , and work out the two - dimensional induced metric on the corresponding polar cap region . for axial perturbations it reads @xmath188 the polar cap is placed on a round sphere that remains inert to perturbations irrespective of boundary conditions . next , we consider the case of polar perturbations and compare the corresponding expressions for @xmath168 and for @xmath172 . the first law of thermodynamics for entanglement entropy is realized holographically provided that the entangling curve in the polar cap region varies according to the equation @xmath189\tan { \theta _ 0 } \ , { p_l}\left ( { \cos { \theta _ 0 } } \right ) . \label{eq : delta_theta_0_polar}\ ] ] this also takes care of the unphysical divergences appearing in @xmath168 . note at this end that a subleading term proportional to @xmath190 is included in the variation @xmath171 ; it contributes a finite term to @xmath168 which should also be removed to match it with @xmath172 . in this case , perturbations of the bulk metric require adjusting the entangling curve in the boundary . to understand what is happening , let us consider the boundary metric of the perturbed @xmath2 space - time , prior to conformal scaling , and work out the two - dimensional induced metric on the corresponding polar cap region @xmath11 . for polar perturbations , the induced metric on the spherical slices of the boundary is @xmath191 where @xmath192 is the radial function appearing in the general parametrization of the bulk space metric . the asymptotic form for @xmath111 follows from the large @xmath51 expansion of the schrdinger wave function @xmath154 , via equation , and it reads @xmath193 it implies , in particular , that the condition on @xmath194 takes the following equivalent form @xmath195 up to irrelevant subleading terms . the deformation of the entangling curve is such that its line element is inert to the perturbation . using the spherical metric , we find that the induced line element of the boundary curve of a polar cap region with parameter @xmath196 is @xmath197 here , we also take the large @xmath51 limit as the computations take place on the boundary of space - time , prior to rescaling . setting @xmath198 , where @xmath186 is provided by , we have to linear order that @xmath199 thus , although the polar cap region and its boundary undergo perturbations , the line element of the entangling curve remains constant , @xmath200 summarizing , we see that for all kind of gravitational perturbations of @xmath2 space - time ( axial or polar ) and for all kind of boundary conditions , the holographic realization of the first law of thermodynamics requires that the line element of the entangling curve remains invariant . region @xmath11 , whose entanglement entropy is provided by formula , deforms with time in such a way that the line element of the entangling curve remains constant . this requirement also takes care of the divergencies that otherwise would have inflicted the variance of the entanglement entropy . since the first law of thermodynamics is trivially valid from the cft point of view , the above statement should be viewed as addendum to ryu - takayanagi prescription . although the laws of entanglement thermodynamics are not yet completely understood , it has been shown that for spherical regions the reduced density matrix @xmath13 is identical to a thermal density matrix with effective inverse temperature equal to the perimeter of the entangling circular curve @xcite . through this connection , our results can be thought to provide the holographic interpretation of an isothermal process of entanglement thermodynamics . an interesting question arises by comparing our calculation to previous ones @xcite . in those papers , the first law of thermodynamics for entanglement was realized holographically using gravitational perturbations satisfying dirichlet boundary conditions , without making reference to the contribution @xmath164 . in our case , even for dirichlet boundary conditions for the polar perturbations of the metric , @xmath201 , @xmath164 is non - vanishing . its contribution is absolutely necessary for validating the first law of thermodynamics for entanglement . the resolution to this ostensible contradiction is that each one of the terms arising in @xmath147 and @xmath164 depend on the choice of coordinates , but their net sum is , in fact , coordinate independent when the line element of the entangling curve remains constant . the variation @xmath164 vanishes for dirichlet boundary conditions only in some coordinate systems , which include the fefferman - graham coordinates used in references @xcite . more technical details about this issue are provided in [ sec : coordinate_dependence ] . the demand for isometric deformations of the entangling curve is also strongly connected with renormalization issues of the entanglement entropy . in odd dimensional conformal field theories , as in our case , the finite contribution to the entanglement entropy can be contaminated by the divergent terms , using alternative definitions of the uv cutoff scale @xmath64 . this issue has already been noticed in the literature @xcite . the universal constant term can be isolated using an appropriate renormalized entanglement entropy , which in @xmath6 conformal field theories ( and at least for special choices of the entangling curve ) takes the form @xcite @xmath202 where @xmath203 is the length of the entangling curve , resolving the above issue . this is also in agreement , via the ryu - takayanagi conjecture , with the definition of renormalized area in the bulk theory , provided by graham and witten @xcite ; in that work the same contamination arises from the divergent terms , resulting in a specific choice of radial coordinate to cure the problem . when the metric perturbations of @xmath2 are taken in the form or , the corresponding radial coordinate @xmath51 does not belong to that special class , calling for a renormalized definition of holographic entanglement entropy , as in equation . in this language , the term @xmath147 corresponds to the variation of @xmath204 , whereas the term @xmath164 corresponds to the variation of the counter - term . then , as it is implied by our results , the inclusion of the counter - term is equivalent to the statement that the unrenormalized entanglement entropy , which is provided by the ryu - takayanagi formula , is computed for a region @xmath11 bounded by an entangling curve that deforms isometrically . in this section , we review briefly the electric - magnetic duality exhibited by linearized gravity , together with its holographic manifestation , following @xcite , and examine the implications to entanglement entropy . linearized gravity around maximally symmetric backgrounds exhibits a rank-2 generalization of electric - magnetic duality in four space - time dimensions @xcite . it is best described in terms of the weyl tensor @xmath205 and its dual counterpart , @xmath206 as follows : the weyl tensor vanishes identically for maximally symmetric backgrounds , @xmath207 , @xmath7 or @xmath2 , which are solutions of the full non - linear einstein equations with cosmological constant @xmath91 . linearized gravity around the maximally symmetric solution , whose choice depends upon the sign of the @xmath91 , exhibits a symmetry under the interchange of the electric and magnetic components of the corresponding weyl tensor . at the level of equations , it amounts to exchanging the role of the linearized field equations and the bianchi identities . two metric perturbations @xmath208 and @xmath209 are said to be dual to each other if @xmath210 so that the electric components of the one weyl tensor are the magnetic components of the other and vice - versa . although these relations are highly non - local , they are naturally resolved by decomposing the gravitational perturbations into two distinct classes of opposite parity , namely axial and polar perturbations . it turns out that @xmath211 provided that the solutions of the corresponding effective schrdinger problems are interrelated as @xmath212 since the schrdinger problems of the axial and polar perturbations are identical , though the precise form of the effective potential depends on the cosmological constant @xmath91 , the relation among @xmath213 and @xmath214 enforces the two classes of perturbations to have the same frequency @xmath156 and the same boundary conditions at spatial infinity . specializing the discussion to gravitational perturbations of @xmath2 space - time , which is the subject of the present work , one finds that electric - magnetic duality has a holographic manifestation @xcite . quite remarkably , the holographic energy - momentum tensor @xmath215 and the cotton tensor @xmath216 of the metric at the boundary of space - time follow from the electric and magnetic components of the bulk weyl tensor ( with respect to the radial adm decomposition of the metric ) as @xmath217 the cotton tensor is traceless , symmetric and covariantly conserved , as for the energy - momentum tensor , and it measures the deviations of the boundary metric from conformal flatness due to the boundary conditions one imposes on the perturbations of the metric . the last equations are general and applicable to the full non - linear regime of @xmath2 gravity , but when specialized to the linearized theory they give rise to the so called energy - momentum / cotton tensor duality that is resolved in sectors as @xmath218 the extremal surface @xmath219 , whose area provides the entanglement entropy , is defined with respect to a given bulk metric @xmath28 for a given boundary region . let us define @xmath220 as the extremal surface taken with respect to the dual metric @xmath221 , following , @xmath222 the corresponding entanglement entropy is taken to be @xmath223 since the unperturbed @xmath2 space is self dual , @xmath224 , the unperturbed extremal surface is also self - dual . the variation of the dual entanglement entropy for axial and polar perturbations , which are dual to each other , can be calculated in the same way as the variation of entanglement entropy . as a direct consequence of equations , we obtain @xmath225 similarly , one can define the variation of the dual modular hamiltonian using the cotton tensor of the original boundary metric , as @xmath226 as direct consequence of equations and , it turns out that @xmath227 it is true that @xmath3 for all kind of perturbations , and , thus , it is also true that @xmath228 showing that the first law of thermodynamics for entanglement is self - dual under gravitational duality . note in this context that the boundary metric is not identical to the dual boundary metric , since the duality transformation interchanges dirichlet and neumann boundary conditions for the metric ( it only preserves the boundary conditions of the effective schrdinger problems ) . since the line element of the entangling curve remains inert to perturbations , the entangling curve is not identical to the dual entangling curve . likewise , the polar cap region @xmath11 is not the same as its dual counterpart @xmath229 . these differences should be taken into account in the definitions of the dual thermodynamic quantities above . in this paper we studied the variation of the entanglement entropy and the expectation value of the modular hamiltonian for small gravitational perturbations of @xmath2 space - time satisfying general boundary conditions . we found that the first law of thermodynamics for entanglement is realized holographically through the ryu - takayanagi formula , provided that the line element of the entangling curve on the boundary remains constant , hereby generalizing previous results on the subject . it should be viewed as addendum to the ryu - takayanagi prescription . in this context , we also noted that the perimeter of the entangling curve is the `` area '' appearing in the `` area law '' term for entanglement entropy . thus , our demand that the line element of the entangling curve remains constant is equivalent to the statement that the `` area law '' term is inert to metric perturbations for all different kind of boundary conditions . in turn , this appears to be equivalent to using a certain renormalized form for the holographic entanglement entropy . although our presentation is technically limited to metric perturbations and entangling curves that are axially symmetric , choosing polar cap regions with circular boundary , we expect it to generalize in the absence of axial symmetry . it will be interesting to see how exactly this happens . we also examined the implications of electric - magnetic duality of linearized gravity for the entanglement entropy and the associated first law of thermodynamics . we defined a dual entanglement entropy and a dual modular hamiltonian and found that they also satisfy the first law of thermodynamics for appropriately related dual regions @xmath11 and @xmath229 at the boundary of @xmath2 . these results too can be thought to provide an additional consistency check of the ryu - takayanagi conjecture . typically , electric - magnetic duality connects the dual degrees of freedom of the theory in a non - trivial and non - local way . in the absence of excitations , it reduces to a trivial local relation , but this is not any more so in the presence of excitations . if one defines the entanglement entropy in a region by tracing out the degrees of freedom in the complement , the same degrees of freedom will correspond to a different region in the dual description . thus , the dual entanglement entropy should be defined for a different entangling curve . our results , showing that the holographic realization of the first law of thermodynamics is achieved only when the line element of the entangling curve remains constant , are in accord with the different boundary conditions satisfied by dual gravitational excitations so that the dual boundary metrics differ . it would be interesting to see if such considerations can also be applied to other backgrounds such as @xmath2 black holes . perturbations of black - holes exhibit a duality relation which is best described in terms of supersymmetric partner potentials for the effective schrdinger problems of the axial and polar sectors . although this duality has no direct interpretation in terms of the electric and magnetic components of the weyl tensor in the bulk , as in @xmath2 space - time , it has holographic manifestation as energy - momentum / cotton tensor duality for the hydrodynamic modes of black holes @xcite . we hope to address the related thermal effects of black holes elsewhere . this research is supported and implemented under the aristeia action of the operational programme for education and long life learning and is co - funded by the european union ( european social fund ) and national resources of greece . a preliminary account of the results was presented by one of us ( g.p . ) at the workshop aspects of fluid / gravity correspondence " held in thessaloniki , greece , 16 - 20 february 2015 . we thank the organizers for their kind invitation and the participants for fruitful discussions . it is known that in poincar coordinates , the extremal surface corresponding to a disk region with radius @xmath45 centered at @xmath230 is given by the `` hemisphere '' , @xmath231 changing to global spherical coordinates , using the coordinate transformation , the surface takes the form , @xmath232 where @xmath43 and @xmath44 are given by @xmath233 \left [ { { { \left ( { r - { \tau _ 0 } } \right)}^2 } + 1 } \right ] } } } \ , , \\ \cos { t_0 } & = \frac{{1 + { r^2 } - { \tau _ 0}^2}}{{\sqrt { \left [ { { { \left ( { r + { \tau _ 0 } } \right)}^2 } + 1 } \right ] \left [ { { { \left ( { r - { \tau _ 0 } } \right)}^2 } + 1 } \right ] } } } \ , .\end{aligned}\ ] ] to express the variation of the modular hamiltonian for the polar cap region in spherical coordinates , we needed the expression for the conformal killing vector in the boundary , @xmath83 , that leaves the entangling curve invariant . this is the limit of a conformal killing vector in the bulk , @xmath88 , which in poincar coordinates and with the appropriate normalization takes the following form @xcite , @xmath234 .\end{gathered}\ ] ] in our analysis , the entangling curve is taken to be axially symmetric , as for the metric perturbations . this selects @xmath56 , so that the center of the disc is placed at the north pole . furthermore , since the @xmath2 metric is static , we may choose a specific instant of time to simplify the calculations . the choice @xmath235 implies that @xmath236 at that instant , the conformal killing vector @xmath88 in the bulk takes the form @xmath237 .\end{gathered}\ ] ] converting to spherical coordinates by the transformation , we find @xmath238,\end{gathered}\ ] ] which can be rewritten as @xmath239 .\end{gathered}\ ] ] then , the boundary limit of this expression provides the desired conformal killing vector @xmath83 , which reads @xmath240\ ] ] and is used to compute the variation of the modular hamiltonian in section [ subsec : de_formula ] . it is not known how to express the modular hamiltonian for a general region and state . this difficulty stems from the fact that , in general , the modular hamiltonian is a non - local operator . exceptionally , in some cases , the modular hamiltonian generates a geometric flow and can be expressed as a local operator . here , we outline how this is actually realized for a disk region in minkowski space - time . the modular hamiltonian defines a symmetry of the system . more specifically , the symmetry group is provided by the unitary operators @xmath241 and it is called the modular group , @xmath242 we define @xmath243 . when the modular hamiltonian is not a local operator , this flow is non - local . if one extends the modular group to imaginary parameters , there will be a periodicity relation in imaginary time , @xmath244 thus , the state @xmath9 is thermal with respect to the time evolution of the modular symmetry with temperature @xmath245 . let us consider minkowski space - time in coordinates @xmath246 . the rindler space @xmath247 is defined as the causal development of the half plane @xmath248 . the causal development of a region @xmath11 is the set of space - time points whose causal space - time curves necessarily cross region @xmath11 . the modular transformations map the local operators @xmath249 in rindler space @xmath247 to themselves . more specifically , the modular flow acts on the coordinates as follows , @xmath250 where @xmath251 and the latin indices @xmath252 run from 2 to @xmath253 . the modular hamiltonian for @xmath247 can be calculated explicitly . transforming to the usual rindler coordinates @xmath254 , the metric becomes @xmath255 corresponding to a thermal state with temperature @xmath256 . thus , the density matrix can be expressed as @xmath257 with modular hamiltonian @xmath258 the result for rindler space can be subsequently used to obtain the modular hamiltonian for a disc @xmath259 of radius @xmath45 . the conformal transformation @xmath260 with @xmath261 yields @xmath262 with @xmath263 letting @xmath264 and @xmath265 , we see that this conformal transformation maps the half plane @xmath266 to the disk @xmath259 , @xmath267 , and the rindler space @xmath268 to the causal development of the disk @xmath269 , @xmath270 , where @xmath271 . for later use , we express the new radial coordinate as @xmath272 also , the inverse coordinate transformation is @xmath273 where @xmath274 we want to determine the form of the modular flow in the new coordinate system . for this purpose , we first make use of equations and to find the flow of the conformal factor @xmath275 , @xmath276\\ & = \omega \left ( { \frac{1}{2}\frac{{{r^2 } + { r^2 } - { t^2}}}{{{r^2 } } } + \frac{1}{2}\cosh 2\pi s\frac{{{t^2 } + { r^2 } - { r^2}}}{{{r^2 } } } - \sinh 2\pi s\frac{t}{r } } \right)\\ & = \frac{\omega } { { { r^2}}}\left ( { r\cosh \pi s - { x^ + } \sinh \pi s } \right)\left ( { r\cosh \pi s + { x^ - } \sinh \pi s } \right ) . \end{split}\ ] ] the radial coordinate @xmath51 , which is given by equation , also flows through its dependence on the conformal factor , @xmath277 similarly , the modular flow of the new time coordinate can be expressed as @xmath278 which in turn implies that @xmath279 finally , we obtain @xmath280 the modular flow in the new coordinates yields by differentiation @xmath281 clarifying the action of the modular hamiltonian on local operators in the new coordinates . restricting attention to the time slice @xmath282 , one recovers the usual expression for the modular hamiltonian for a disk of radius @xmath45 @xcite , namely @xmath283 the individual contributions to the variation of the area , @xmath147 and @xmath164 , which appear in section [ sec : ee_perturbed_ads4_calculation ] , are coordinate dependent , unlike their sum which is coordinate independent when the line element of the entangling curve remains constant . furthermore , the term @xmath164 only vanishes for dirichlet boundary conditions provided that one works in coordinate systems , unlike ours , where specific metric elements ( to be made more precise later ) contain no contributions that fall like @xmath122 close to the boundary . these statements are substantiated below by performing a small perturbative coordinate transformation altering the @xmath122 asymptotics of the metric elements , @xmath284 the area of the extremal surface in @xmath2 is provided by equation . writing the result in terms of the @xmath285 coordinate , we have @xmath286 the change of coordinate introduces a change in the area of the minimal surface in @xmath2 , which functionally resembles @xmath147 , @xmath287 this seems to be odd at first sight , because the area should be independent of the choice of coordinates , but @xmath219 is infinite and the result depends on way one is taking the limit . ultimately we are interested in finding a result that is independent of the limiting process . on the other hand , the metric element @xmath288 of @xmath2 space in the @xmath289 coordinate is expressed as @xmath290 demanding that the length of the entangling curve remains invariant , we obtain @xmath291 which , in turn , gives a non - vanishing contribution to the area of the minimal surface , which resembles @xmath292 , @xmath293 thus , demanding that the line element of the entangling curve remains constant , we arrive at the following result , @xmath294 this provides the desired coordinate independent prescription for taking the limit . the variation of the area of the minimal surface in @xmath2 should be zero under changes of coordinates . likewise , for perturbations of @xmath2 space - time , the combined variation @xmath295 , which is not zero , does not depend on the choice of coordinates , unlike the individual terms @xmath147 and @xmath164 that depend upon it . as an illustrative example of the effect that coordinate transformations may have to the components of the variation of the area of the minimal surface in the presence of gravitational perturbations of @xmath2 space - time , we consider the case of polar perturbations . using the change of coordinate with the choice @xmath296 we find @xmath297\end{gathered}\ ] ] and @xmath298 one can readily verify that @xmath299 , using the results of section [ sec : ee_perturbed_ads4_calculation ] , as advertised . choosing dirichlet boundary conditions for the metric , @xmath201 , we find that @xmath300 vanishes , whereas @xmath301 is finite , equal to @xmath302 . note at this end that the asymptotic behavior of @xmath288 in @xmath289 coordinate is @xmath303 summarizing , the term @xmath164 can only be neglected for dirichlet boundary conditions and in coordinate systems where the metric elements corresponding to elementary lengths parallel to the entangling surface do not contain subleading terms that fall like @xmath122 close to the boundary . this is precisely the content of the calculations reported in earlier works @xcite , using dirichlet boundary conditions and writing the metric in fefferman - graham coordinates which obey the restrictions stated above . if any one of these restrictions do not hold , the variance @xmath164 will become relevant in the calculation and the need for invariance of the line element of the entangling surface will come into play , as in our calculations . as emphasized above , the individual terms @xmath304 and @xmath292 contain gauge artifacts , but not their sum , which is coordinate independent . j. d. bekenstein , black holes and entropy " , phys . d * 7 * , 2333 ( 1973 ) . j. m. bardeen , b. carter and s. w. hawking , the four laws of black hole mechanics " , commun . * 31 * , 161 ( 1973 ) . s. w. hawking , particle creation by black holes " , commun . * 43 * , 199 ( 1975 ) [ erratum - ibid . * 46 * , 206 ( 1976 ) ] . j. m. maldacena , the large n limit of superconformal field theories and supergravity " , int . j. theor . phys . * 38 * , 1113 ( 1999 ) [ adv . * 2 * , 231 ( 1998 ) ] [ hep - th/9711200 ] . s. s. gubser , i. r. klebanov and a. m. polyakov , gauge theory correlators from noncritical string theory " , phys . b * 428 * , 105 ( 1998 ) [ hep - th/9802109 ] . e. witten , anti - de sitter space and holography " , adv . theor . math . * 2 * , 253 ( 1998 ) [ hep - th/9802150 ] . s. ryu and t. takayanagi , holographic derivation of entanglement entropy from ads / cft " , phys . rev . lett . * 96 * , 181602 ( 2006 ) [ hep - th/0603001 ] . s. ryu and t. takayanagi , aspects of holographic entanglement entropy " , jhep * 0608 * , 045 ( 2006 ) [ hep - th/0605073 ] . v. e. hubeny , m. rangamani and t. takayanagi , a covariant holographic entanglement entropy proposal " , jhep * 0707 * , 062 ( 2007 ) [ arxiv:0705.0016 [ hep - th ] ] . t. nishioka , s. ryu and t. takayanagi , holographic entanglement entropy : an overview " , j. phys . a * 42 * , 504008 ( 2009 ) [ arxiv:0905.0932 [ hep - th ] ] . m. van raamsdonk , comments on quantum gravity and entanglement " , arxiv:0907.2939 [ hep - th ] . m. van raamsdonk , building up spacetime with quantum entanglement " , gen . * 42 * , 2323 ( 2010 ) [ int . j. mod . d * 19 * , 2429 ( 2010 ) ] [ arxiv:1005.3035 [ hep - th ] ] . t. takayanagi , entanglement entropy from a holographic viewpoint " , class . * 29 * , 153001 ( 2012 ) [ arxiv:1204.2450 [ gr - qc ] ] . n. lashkari , m. b. mcdermott and m. van raamsdonk , gravitational dynamics from entanglement thermodynamics " , jhep * 1404 * , 195 ( 2014 ) [ arxiv:1308.3716 [ hep - th ] ] . t. faulkner , m. guica , t. hartman , r. c. myers and m. van raamsdonk , gravitation from entanglement in holographic cfts " , jhep * 1403 * , 051 ( 2014 ) [ arxiv:1312.7856 [ hep - th ] ] . r. g. leigh and a. c. petkou , gravitational duality transformations on ( a)ds(4 ) " , jhep * 0711 * , 079 ( 2007 ) [ arxiv:0704.0531 [ hep - th ] ] . s. de haro , dual gravitons in ads(4 ) / cft(3 ) and the holographic cotton tensor " , jhep * 0901 * , 042 ( 2009 ) [ arxiv:0808.2054 [ hep - th ] ] . i. bakas , duality in linearized gravity and holography " , class . quant . * 26 * , 065013 ( 2009 ) [ arxiv:0812.0152 [ hep - th ] ] . i. bakas , dual photons and gravitons " , publ . beograd * 88 * , 113 ( 2010 ) [ arxiv:0910.1739 [ hep - th ] ] .

clusters of galaxies produce negative features at wavelengths mm in cmb maps , by means of the thermal sz effect , while point radio sources produce positive peaks . this fact implies that a distribution of unresolved sz clusters could be detected using the negative asymmetry introduced in the odd - moments of the brightness map ( skewness and higher ) , or in the probability distribution function ( pdf ) for the fluctuations , once the map has been filtered in order to remove the contribution from primordial cmb fluctuations from large scales . this property provides a consistency check to the recent detections from cbi and bima experiments of an excess of power at small angular scales , in order to confirm that they are produced by a distribution of unresolved sz clusters . however it will require at least 1.5 - 2 times more observing time than detection of corresponding power signal . this approach could also be used with the data of the planned sz experiments ( e.g. act , ami , amiba , apex , 8 m south pole telescope ) . [ firstpage ] cosmology : cosmic microwave background cosmology : observations galaxies : clusters : general .

astro-ph0211430

fluctuations in the cosmic microwave background ( cmb ) radiation can provide information about hot gas in galaxy clusters over a wide range of redshifts ( sunyaev & zeldovich 1972,1980 ( hereafter sz ) , and @xcite ) . on arcminute angular scales and smaller , the thermal sz contribution to the cmb anisotropy is expected to dominate that of the primary anisotropies ( @xcite , @xcite ( hereafter swh ) , @xcite ) . a new generation of experiments is measuring the cmb sky at these angular scales . in particular , two recent experiments , bima @xcite and cbi @xcite , both observing at frequencies around 30 ghz , have detected an excess of power in the multipole region @xmath1 , where the sz power is expected to be dominant over the cmb signal . nevertheless , at these observing frequencies , radio point sources are known to also produce a significant contribution to the power @xcite if they are not subtracted properly from the cmb maps . the reported detections of power have argued that this point - source contamination is not a problem , thus suggesting that the signal could be due to the sz effect @xcite . these arguments are based on analytical models or simulations of what we would expect to measure . thus , it would be interesting to explore , in a model - independent way , the nature of these contributions . the importance of this topic has been stressed recently by @xcite , who suggested to use a cross - correlation of cmb maps with maps of the large scale structure . this idea has been applied for this purpose to other datasets with larger angular resolutions ( e.g. @xcite , to the cobe data , or @xcite , to the tenerife data ) . here , we propose a general model - independent method to determine if the measured power excess in a _ single - frequency _ map is ( mainly ) due to point sources or sz clusters . to this end , we use the fact that for frequencies below @xmath2 ghz ( @xmath0 mm ) , the thermal sz effect produces negative features in the maps , while the point sources produce positive peaks . we illustrate this fact with figure [ typical ] , where we show two simulated one - dimensional maps , one of sz clusters observed at @xmath3 ghz , and the other one of point sources . dotted lines show the original ( without sources of any kind ) zero level of fluctuations , while dashed lines show the observed ( average ) zero level once the mean of the map has been subtracted . with the same level of fluctuations ( same rms at the observed scale ) , a power spectrum analysis is not able to distinguish these two cases , so we need to use an statistic carrying information about the sign of the subjacent signal ( e.g. the skewness ) to suggest the nature of the objects producing this excess of power . the existence of negative skewness at @xmath0 mm , while positive skewness at @xmath4 mm , is a clear prediction for sz clusters . full - width half - maximum , and no noise . dashed lines show the average ( zero ) level of fluctuations , while dotted lines show the original zero level before subtracting the mean to the map . with the same level of fluctuations , the @xmath5 analysis does not permit to distinguish between these two cases . therefore , we need to use the skewness , or to proceed with an analysis of the asymmetry of the @xmath6 curve in order to separate these two cases.,title="fig : " ] full - width half - maximum , and no noise . dashed lines show the average ( zero ) level of fluctuations , while dotted lines show the original zero level before subtracting the mean to the map . with the same level of fluctuations , the @xmath5 analysis does not permit to distinguish between these two cases . therefore , we need to use the skewness , or to proceed with an analysis of the asymmetry of the @xmath6 curve in order to separate these two cases.,title="fig : " ] we investigate here the discrimination between positive and negative sources using the probability distribution function ( pdf ) for the observed flux . from a given map , _ the pdf function can be obtained easily as an histogram of the ( normalised ) number of pixels within a given flux interval_. this tool has been widely used in radio astronomy when studying the statistical properties of a background of point sources @xcite , because in that case the shape of this function is strongly related with the statistical properties of the sources ( i.e. theirs spatial distribution ) . in this context , this function is known as the deflection probability distribution , or the @xmath6 curve . this @xmath6 formalism has been successfully applied to study the diffuse x - ray background @xcite , as well as to determine the contribution of discrete point sources to cmb maps @xcite . for the cmb , if we assume the standard inflationary scenario , then the primordial fluctuations are gaussian , so the @xmath6 itself is a gaussian , as well as for the standard instrumental noise . however , the main characteristic of this @xmath6 curve for point sources ( @xcite ) or for sz clusters ( @xcite ) is its non - gaussianity . typical curves for a @xmath6 distribution of point sources or sz clusters will exhibit long tails ( see figure [ figura_bonita ] ) . the point is that at @xmath0 mm , sources will produce a positive tail , while sz clusters will give a negative one . it is important to mention that at @xmath4 mm , both agns and sz - clusters will produce positive tails . then it is necessary to use other characteristics of both populations ( frequency spectra , etc ) to distinguish them . as an illustration , figure [ compara_pd_sz ] demonstrates p(d ) for sz sources at four frequencies , @xmath7 = 107 and 150 ghz ( where clusters are giving negative signal ) and 270 and 520 ghz ( where the signal from clusters is positive , and exactly opposite in sign to the previous cases ) . ) this curve will be explained in detail in section [ sec : general ] . ] function ( see equation ( [ s2:g(x ) ] ) ) , and four frequencies ( @xmath8 107 , 150 , 270 and 520 ghz ) where this function ( and so the flux density ) takes the same absolute value . second panel shows the p(d ) function for these four cases , using a simple truncated power - law to model the cluster source counts ( see section 4 ) , with values @xmath9 = 28 ( s/1jy)@xmath10 sr@xmath11 jy@xmath11 at 150 ghz , truncating at @xmath12 mjy , and with angular resolution @xmath13 . the p(d ) function is presented relative to its average value , so the distribution for the cases @xmath14 ghz and 150 ghz is symmetric around zero respect to the other two cases.,title="fig : " ] function ( see equation ( [ s2:g(x ) ] ) ) , and four frequencies ( @xmath8 107 , 150 , 270 and 520 ghz ) where this function ( and so the flux density ) takes the same absolute value . second panel shows the p(d ) function for these four cases , using a simple truncated power - law to model the cluster source counts ( see section 4 ) , with values @xmath9 = 28 ( s/1jy)@xmath10 sr@xmath11 jy@xmath11 at 150 ghz , truncating at @xmath12 mjy , and with angular resolution @xmath13 . the p(d ) function is presented relative to its average value , so the distribution for the cases @xmath14 ghz and 150 ghz is symmetric around zero respect to the other two cases.,title="fig : " ] it is possible to consider clusters of galaxies as `` extended sources '' with a peculiar spectrum given by @xcite @xmath15 where @xmath16 is the optical depth for thompson scattering , @xmath17 is the dimensionless frequency , with @xmath18 and @xmath19 , so that @xmath20 does not depend on redshift @xmath21 , and @xmath22 \label{s2:g(x)}\ ] ] is the spectral shape factor . note that this shape factor includes the term from @xmath23 , the planck function , and @xmath24 and @xmath25 are the formulae for the cmb spectrum distortions due to comptonization from @xcite . from here , two different but equivalent approaches can be used to estimate the spectral luminosity of the cluster . we can obtain the spectral luminosity by just integrating the change of the cmb intensity due to scattering by individual electrons of temperature @xmath26 over the cluster volume @xcite @xmath27 this expression , for the case of isothermal intergalactic gas , is proportional to the total amount of electrons in the cluster , because in that case @xmath28 , where @xmath29 is the mass of the hot intergalactic gas , and @xmath30 is the mean molecular weight per electron . the important point here is that clusters increase rapidly their spectral luminosities with redshift ( @xmath31 ) . taking into account the luminosity distance to the source @xmath32 , we can obtain the spectral flux as @xmath33 . on the other hand , according to @xcite , we can use the central value of the comptonization parameter for the cluster @xmath34 for a given @xmath35 , the surface brightness of the cluster does not depend on redshift . then , the flux from the cluster is equal to @xmath36 where @xmath37 corresponds to the angular dimension of the cluster core radius , and the @xmath38 function takes into account the angular dependence of @xmath39 over the cluster image . one important conclusion from this point of view of the problem is that clusters with given physical parameters should have a minimum flux at some redshift , due to the well - known redshift dependence of the angular dimension of the cluster with a given core radius . in the universe with @xmath40 we have minimum angular dimension at @xmath41 , and @xmath42 , this happens at @xmath43 . ] and at higher @xmath21 , both angular distance and the flux will increase : clusters with given physical parameters have minimum flux where its angular dimension is minimum . it is important to note that 1 . for an experiment with an angular resolution @xmath44 larger than the core radius of clusters ( @xmath45 ) , they will be unresolved objects , and therefore they will appear as point sources for us ; 2 . according to the dependence of @xmath46 from x , these point sources will have `` positive '' flux at @xmath4 mm , and `` negative '' flux at @xmath0 mm . 3 . for given physical parameters , clusters show a minimum flux at that redshift @xmath21 where the angular dimension is minimal . on the other hand , the observed source counts depend on the luminosity function . for instance , if we use the press - schechter formalism ( @xcite , hereafter ps ) , then we have a divergence of the ( comoving ) number density of objects at low masses ( i.e. low fluxes ) . nevertheless , we know that cooling and feedback play an important role in the sz predictions , so we do not expect to find bright sz clusters with masses below few times @xmath47 . then , it is necessary to introduce a low mass cutoff in ps formalism in order to derive realistic sz predictions , as it has been done by several authors ( e.g. @xcite ) . thus , we expect to observe a minimum flux and a minimum angular dimension for sz clusters . several authors have studied the contribution of the sz effect to the power spectrum of cmb fluctuations at small scales , both theoretically ( e.g. @xcite ) or using simulations ( e.g. swh , @xcite ) . in any case , the main observational emphasis is put on the power spectrum ( @xmath5 ) , because it is easier to measure than , for example , the bispectrum . when determining if the excess of power at low scales detected by cbi and bima is due to sz clusters , the comparison has been done in terms of the power spectrum @xcite . however , when we are working with the power spectrum , we are losing information about the sign of the fluctuations , so we need to find the skewness , or to proceed with an analysis of the asymmetry of the @xmath6 function . the formalism relating the ( differential ) source counts and the pdf ( or @xmath6 function ) of the observed deflection @xmath48 at a given point , due to a population of poissonian - distributed unresolved sources , was first discussed by @xcite , and extended by @xcite . this analytical approach will permit us to understand much more deeply the observed shapes of the @xmath6 curves , and to relate them with the underlying source counts . using the standard notation from radio astronomy , let @xmath9 be the differential counts per solid angle , at a given frequency @xmath7 , and let @xmath49 the response of a radio telescope to a point source ( normalised to 1 at the peak ) . let @xmath50 be the response of the instrument to a source of flux density @xmath51 located at a given distance @xmath52 of the beam centre . then , the mean number of source responses of flux between @xmath53 and @xmath54 in the beam , @xmath55 , is given by @xmath56 \frac{d\omega}{b(\theta , \phi)}\ ] ] the relationship between @xmath9 ( or @xmath55 ) and the @xmath6 function , for the case of a pencil beam antenna , is given in terms of the characteristic functions of @xmath55 and @xmath6 , which can be written as @xmath57 and @xmath58 , respectively . the equations are @xmath59 @xmath60\ ] ] and the @xmath6 function can be obtained as the inverse fourier transform @xmath61 this relation can be also employed in the case of tracking interferometers ( @xcite ) , using the cleaned map . for the case of a phase switch interferometer , the above relations still holds , but replacing fourier transforms by bessel transforms @xcite . it should be noted that these expressions are general , and therefore valid for the case of negative sources . indeed , from the previous equations it follows that _ if a population of positive sources @xmath9 , with @xmath62 , is described by the function @xmath6 , then the population of sources @xmath63 , with @xmath64 , is described by @xmath65_. thus , in this section , we will restrict ourselves to study distributions of positive sources ( @xmath66 for @xmath64 ) , given that we can obtain the corresponding distribution for negative sources with the transformation @xmath67 . all the equations described through this section are implicitly assuming non - resolved objects . @xcite have studied the modifications introduced by extended sources , showing that the condition for unresolved objects applies until @xmath68 , where @xmath69 is the typical size of the source , and @xmath44 the beam size . let us consider here several particularly simple but useful cases for the @xmath9 function , which can be treated analytically . these cases will be used later . for a more general case of source counts , we can use a monte carlo method to work out the p(d ) distribution ( e.g. @xcite ) . first of all , we will consider a power - law shape , @xmath70 , with @xmath62 . as a second case , we will also consider a truncated power - law at a certain flux density @xmath71 , i.e. @xmath72 for @xmath73 , and @xmath70 , for @xmath74 . hereafter , we will assume that @xmath51 is given in @xmath75 , so we are implicitly writing @xmath76 , and the units for @xmath77 are jy@xmath11 sr@xmath11 . we will also assume a gaussian antenna pattern , described as @xmath78 , where @xmath79 is the width of the beam , and @xmath80 its full - width half maximum ( fwhm ) . in these cases , the @xmath55-function can be analytically obtained . following @xcite , we define the effective solid angle ( @xmath81 ) as @xmath82^{\beta-1 } d\omega = \frac{\pi \theta_b^2}{4 ( \beta-1 ) \log 2}\ ] ] we then obtain @xmath83 for a pure power - law , and @xmath84 for a power - law truncated below @xmath71 and the gaussian beam . from these equations , it is straight - forward to obtain numerically the @xmath6 function . these problems have been studied by @xcite and @xcite for the power law case , which is valid for @xmath85 , so in these cases the numerical calculation is even easier . ] , and by @xcite ( analytically ) and @xcite ( numerically ) for the truncated case . finally , it is also interesting to consider the case of a power - law source counts with an upper cut - off in flux , @xmath86 . this is the expression to use in order to compute the @xmath6 function from a map where the brightest sources have been subtracted down to the flux @xmath86 . therefore , if @xmath70 for @xmath87 , and @xmath88 elsewhere , then @xmath89 apart from the @xmath6 function , it is also interesting to characterise the properties of the source population by the moments of the @xmath55 distribution . for a pure power - law expression for the number - flux - density relation , it is clear that the @xmath90th - moment of the r - distribution is @xmath91 where @xmath92 represents the cutoff value for point - source subtraction . for the truncated power - law case , the @xmath90th - moment of this distribution is @xmath93 the second moment of the @xmath55 function ( @xmath94 ) has been extensively used to characterise the confusion noise , i.e. the noise due to the presence of faint unresolved sources inside the beam @xcite . normally , the adopted criterion for the detection of sources is such as the intensity @xmath95 times the sigma of the map , so the minimum subtraction threshold can be written as @xmath96 , being @xmath97 the usual values . thus , inserting this condition in equation ( [ s^n_powerlaw ] ) , we obtain for the power - law case : @xmath98^{1/(\beta -1 ) } \label{ec : sigmacq}\ ] ] where we explicitly write that @xmath99 depends on @xmath95 for this choice of the subtraction threshold . note that this threshold can be decreased if we use measurements at higher angular resolution . we will discuss this in section [ sec : roleofradiosources ] . for the power - law source counts , it is easy to show that the flux at which we have one source every @xmath100 beam areas , @xmath101 , can be written as @xmath102 from these last two equations , we immediately see that for the flux level @xmath103 , we have one source every @xmath104 beam areas . for the particular case of @xmath105 and @xmath106 , this expression takes the value @xmath107 , so we recover the well - known result that if we have more than one source of a given flux every 3040 beam areas ( it depends on @xmath108 ) , we will be limited by confusion noise . we should remind here that the @xmath6 function provides information ( roughly ) down to the flux @xmath109 at which we have one source per beam area @xcite . to conclude this section , we will derive the scaling of @xmath6 with frequency for sz clusters . this scaling can be derived from the one for the source counts , which is given by @xmath110 where @xmath111 . this equation is similar to the one obtained by @xcite for the scaling of the differential source counts of point sources with power - law spectra ( @xmath112 ) . , where @xmath113 does not depends on @xmath7 . ] from here , we obtain @xmath114 we see that clusters of galaxies should be described by a single @xmath6 function , which is the same at all frequencies ( but rescaled ) if @xmath44 is the same , and which is equivalent to the pdf for the @xmath39 parameter . however , our description permits the use of the main characteristic of the effect , the existence of negative sources . therefore , if we compare data from two frequencies , one above @xmath115 ghz and the other one below , then in the first case the @xmath6 for clusters will exhibit a positive tail while in the second case , a negative tail , being in both cases described by the same ( rescaled ) @xmath6 function . if we now use this expression to derive the moments of the @xmath6 function , we obtain @xmath116^n < d^n ; \nu_0 > , \qquad n = 1,2,3, ... \ ] ] therefore , we explictily see that the normalised moments of a map of thermal sz clusters ( with no noise ) , @xmath117 , are exactly the same in magnitude for all frequencies , but we have a change of sign in all the ( normalised ) odd - moments when we cross the frequency @xmath115 ghz . as we have seen in the last section , the shape of the @xmath6 function for ( positive / negative ) sources provides a means of determining the underlying source counts . therefore , we will discuss here which are the typical source counts both for radio sources and sz clusters . several authors ( e.g. @xcite ) have studied the contribution to the confusion limits from different populations of extragalactic sources . in the context of cmb measurements at frequencies close to @xmath118 ghz , radio sources are known to produce the main contribution to the confusion noise . because of this reason , recent experiments have used a source subtraction technique , and thus have produced source counts for these radio sources . these curves are well fitted by power - laws in the flux density region around a few mjy . typical values for the @xmath77 and @xmath108 parameters , at frequencies around 30 ghz , are @xmath119 sr@xmath11 jy@xmath11 , for @xmath120 mjy ( from the vsa experiment , @xcite ) , and @xmath121 sr@xmath11 jy@xmath11 , for @xmath122 mjy ( from the cbi group , @xcite ) . apart from these source counts , we can also extrapolate the @xmath123jy source counts at 8.4 ghz from vla observations @xcite up to 30 ghz . using the spectral index @xmath124 ( @xmath125 ) , we obtain @xmath126 sr@xmath11 jy@xmath11 . all these source counts are summarised in figure [ review_scounts ] . and @xmath127 ( see section [ sec_scounts ] for details ) , as well as the source counts from the paper of swh ( squares ) using hydrodynamic simulations . both source counts for clusters compare well in shape around 1 mjy , although the hydrodynamic simulations show a @xmath12820% less objects at these fluxes , and do not show an strong cutoff at low fluxes . for comparison , we also present here the differential source counts for radio sources at 30 ghz from several experiment : vsa ( taylor et al . 2002 ) , at 34 ghz ; cbi ( mason et al . 2002 ) at 31 ghz ; and vla ( fomalont et al . 2002 ) at 8.4 ghz . the source counts for this last experiment have been extrapolated up to 30 ghz using their mean spectral index @xmath129 . from the simple observation of these source counts , we expect that experiments with high angular resolution are going to be dominated by radio sources , if they do not consider a source subtraction strategy . ] we should mention that the contribution of radio sources to the power ( and thus to the skewness ) is decreasing with frequency in power @xmath130 , where @xmath131 is the spectral index of radio sources . therefore , the contribution of radio sources to the observed map becomes less important at higher frequencies ( @xmath132 ) , while the contribution of clusters does not depend on frequency in the raileigh - jeans ( rj ) region of the spectrum . it is important to remind that other populations of sources ( e.g. thermal dust emission from galaxies ) contribute to the source counts at higher frequencies . however , we expect that future experiments like alma will show us those populations very precisely . given these source counts , the shapes of the corresponding distribution functions are characterised by long positive tails , as we have seen in the previous section . inclusion of the clustering effects of sources @xcite broadens the shape of the @xmath6 , but the important point here is that the long positive tail is still maintained . @xcite discussed count curves for thermal sz clusters , and showed that they differ strongly from the case of radio sources . @xcite have derived the source counts for the thermal sz effect using the press - schechter mass function and assuming unresolved single - type clusters . for the scaling of the temperature with the mass of the cluster , they use @xmath133 , so @xmath134 . these numbers are in agreement with those obtained from recent x - ray observations @xcite , and have been shown to fit simultaneously optical and x - ray cluster data @xcite . @xcite show that typical curves can be well fitted by euclidean power - laws @xmath135 down to a few @xmath136 , with typical values of @xmath137 sr@xmath11 jy@xmath11 ( extrapolated down to 30 ghz @xmath138 1 cm ) , and introducing a low - flux cutoff of @xmath139 mjy . swh , using hydrodynamical simulations , came to similar conclusions , although their results do not show an exact power - law behaviour , and the low - flux cutoff is one order of magnitude smaller . for illustration , we show in figure [ review_scounts ] these source counts , and our derived source counts for ps clusters , together with the radio source counts described in the last subsection . the qualitative behaviour is the same pointed above . these curves will be described in detail in section [ sec : general ] . here , we will point out two general aspects of any modelling of sz clusters . first , we should stress that the value for @xmath71 in each model ( semi - analytic or numeric ) depends on the chosen mass cutoff , @xmath140 , i.e. the minimum mass of an object contributing to the sz effect ( see discussion at the end of section [ sec:2 ] ) . this minimum mass ( @xmath140 ) is related to the minimum flux ( @xmath141 ) observed in a given cosmology . using equation ( [ di_sz ] ) , and assuming that the gas in the cluster is isothermal , we obtain the expression for the total sz decrement @xmath142 for a galaxy cluster , as a function of its mass @xmath143 where @xmath144 stands for the angular diameter distance , and @xmath145 is the gas mass fraction . using the scaling @xmath146 pointed above , we can derive for the rayleigh - jeans region of the spectrum that @xmath147 @xmath148 from here , it is straight - forward to infer the minimum flux for a given @xmath140 . as an example , if we use the standard values @xmath149 , @xmath150 , @xmath42 , and we assume a constant value @xmath151 , then we find that for a given mass we have the minimum flux at @xmath152 , and its value is @xmath153 at @xmath154 . typical values for @xmath141 can be seen in table [ table0 ] . @xcite were considering @xmath155 independent on redshift , so @xmath141 in their case was reached simultaneously with the minimum of angular diameter . here we use a dependence of @xmath155 with redshift ( @xmath156 ) , and hence the minimum is reached when the function @xmath157 takes its minimum . .dependence of the observed flux cutoff for clusters with the minimum mass of a cluster , for @xmath8 30 ghz ( see details in text ) . [ cols="^,^",options="header " , ] we consider an sz simulated map following ps prescription , covering 3 square degrees with a pixel size of @xmath158 and a beam size of @xmath159 , with @xmath160 , and we add different noise levels . the obtained p(d ) are shown in figure [ pd_plus_noise ] , while we present here the observed power ( @xmath161 ) in the map , the infered signal due to sz ( estimated as @xmath162 = \sqrt{rms^2 - \sigma_{noise}^2}$ ] ) , and the measured skewness in the map ( without applying any filtering technique ) , with its variance ( @xmath163 ) . we illustrate these facts by computing the @xmath6 functions and the skewness for the complete set of 15 sz maps based on ps modelling , observed with a gaussian beam of @xmath159 , and adding different noise levels per pixel ( @xmath164 ) , quoted in terms of the power due to clusters ( @xmath165 ) . when then consider the following cases for @xmath166 : 0.3 , 1 , 2 , 5 , 10 and 20 . we will also assume that the noise is gaussian and uncorrelated , as a first approximation . the resulting p(d ) curves are shown in figure [ pd_plus_noise ] , while the measured rms in the map , the recovered excess of power , and the observed skewness , are quoted in table [ table1 ] . we can see that the tail of the p(d ) function has practically disappeared at @xmath167 , although is still possible to measure skewness in the map at a high significance . when we go to @xmath168 , the skewness is no longer visible , although it is possible to detect an excess of power . at @xmath169 , anything can be detected . . we add to this map several realizations of white gaussian noise , with the amplitude quoted in the figure , and then we compute the p(d ) function . the measured excess of power and skewness in each particular case are quoted in table [ table1 ] . ] if we now include sources in our maps , the determination of how deep we need to integrate in order to get information about the skewness can be done in the following simple way . we define here the signal - to - noise ratio per pixel , @xmath170 , as the quotient @xmath171 . we quote this quantity because it is easy to infer from the map . in addition , it is also straight - forward to convert these values into integration time @xmath172 for a given experiment , because we usually have @xmath173 . in the previous expression , @xmath174 corresponds to the power introduced by clusters and sources all together . then , if we want a @xmath95-sigma detection of skewness , it is straight - forward to derive from equation ( [ sum_skew ] ) that we need a signal - to - noise ratio better than @xmath175^{-1/2}\ ] ] where @xmath176 is the combined skewness of sources and clusters ( [ sum_skew ] ) , and @xmath177 is the number of pixels in the map , i.e. @xmath178 , with @xmath179 the fraction of sky covered , and @xmath180 the pixel solid angle . for the typical values of skewness found in our simulations , the required signal - to - noise ratio for the detection of skewness due to both point sources and sz clusters is around 25 - 40% times larger than the required ratio for the detection of an excess of power . in other words , we need ( the well - known result of ) 1.5 - 2 times more integration time to detect skewness than to detect an excess of power . we will mention here that these values have been obtained directly from the maps , without applying any filtering . these numbers can be improved if we apply to our maps an appropriate filter to enhance the contribution of sources over the noise level , previously to the skewness or p(d ) computation . we will discuss this issue in the next section , where we also consider the effect of primary anisotropies on the maps . the computations in the last two sections were done assuming that the primordial cmb gives a negligible contribution compared with that of the sz signal and the noise . however , this is only true at arcminute scales , where the primordial anisotropy is damped to very small amplitudes , well - below the level of the expected sz anisotropy ( see the top panel of figure [ bispectrum_press ] ) . therefore , only those experiments which are observing at the high multipole region of the angular power spectrum are going to directly observe the cluster contribution . this is the case of interferometers with long baselines ( e.g. amiba , see @xcite , or bima @xcite , which has a primary beam of @xmath181 ) , or experiments with small fields of view . as it is well - known , if we are observing small patches on sky , we are sensible to multipoles greater than @xmath182 , being @xmath183 the size of our map ( see e.g. @xcite for a detailed study ) . therefore , for small fields of view we are implicitly filtering out the lowest multipoles , and we can directly apply the formalism described in the last section . however , if our experiment is sensitive to low multipoles ( @xmath184 ) , then it is necessary to add a new step in our pipeline , which consists in filtering these low multipoles , if we want to remove the contribution of primordial anisotropies . this will be the case for experiments which will cover thousands of square degrees in sky ( e.g. 8 m south pole telescope ) , or experiments covering the whole sky ( e.g planck satellite ) . this step is common to all techniques which are aimed to extract sources / clusters from a map using a single frequency observation . many filters can be found in literature , going from matched filters ( e.g. @xcite ) , pseudo - filters ( e.g. @xcite ) or wavelets ( e.g. @xcite ) . any of these methods can be used to pre - process the maps , reducing the relative contribution of primary anisotropies , and also that of the noise . the important point is that under linear filters , gaussian data will still be gaussian , and a distribution of sources still will show an skewed shape . for illustrating this issue , we have performed a new simulation of sz clusters , covering a larger area ( @xmath185 pixels of 0.5side each , so the map covers @xmath186 square degrees ) , using the same parameters as for the previous simulations , but with @xmath187 . the reason of considering such a large area is because future experiments will cover hundreds ( or thousands ) of square degrees on sky . we add to this map a cmb realisation following the power - spectra plotted in the top panel of figure [ bispectrum_press ] . the rms of the simulated sz map and the cmb realisation are @xmath188 and @xmath189 , respectively . finally , we also add white gaussian noise with an amplitude @xmath190 ( per pixel of 0.5 ) , and we smooth the resulting map with a gaussian beam of @xmath191 . using this map , we have examined two different filters : * a hanning ( or high - pass ) filter , which removes the contribution of all multipoles below a certain value @xmath192 , but linearly increasing from 0 to 1 in the range @xmath193 $ ] to avoid the ringing associated to a sharp cut . we have used here @xmath194 , and @xmath195 . * a matched filter optimised to detect point - source like objects @xcite , so @xmath196 . here , @xmath197 and @xmath198 are the coefficients in a legendre polynomial expansion of the filter and of the beam , respectively , and @xmath199 is the sum of the power spectrum of the components to be removed ( noise and cmb in this case ) . , for four cases : ( a ) adding cmb primary anisotropies , and white gaussian noise of 10 @xmath200 per pixel ( solid line ) ; ( b ) filtering the previous map with the optimal matched - filter described in haehnelt & tegmark ( 1996 ) ( dashed line ) ; ( c ) filtering the map with a hanning ( high - pass ) filter ( dotted line ) ; ( d ) considering the sz map alone ( dot - dashed line ) . bottom : same as the previous panel , but without including noise . in order to use the same scale for plotting all curves , we have divided each map for its @xmath161 prior to the p(d ) computation . in this case , @xmath201 , @xmath202 and @xmath203 . , title="fig : " ] , for four cases : ( a ) adding cmb primary anisotropies , and white gaussian noise of 10 @xmath200 per pixel ( solid line ) ; ( b ) filtering the previous map with the optimal matched - filter described in haehnelt & tegmark ( 1996 ) ( dashed line ) ; ( c ) filtering the map with a hanning ( high - pass ) filter ( dotted line ) ; ( d ) considering the sz map alone ( dot - dashed line ) . bottom : same as the previous panel , but without including noise . in order to use the same scale for plotting all curves , we have divided each map for its @xmath161 prior to the p(d ) computation . in this case , @xmath201 , @xmath202 and @xmath203 . , title="fig : " ] in figure [ filters ] we show the predicted @xmath6 function from our simulation when adding cmb primary anisotropies , with and without noise , and the predicted @xmath6 after filtering the maps with the previous filters . for comparison , it is also shown the @xmath6 function for the sz map alone . in order to use the same scale for plotting all the curves , we have divided each map by its rms prior to the @xmath6 computation . we can see that a simple high - pass filter is able to remove the main contribution from the primordial cmb , so the @xmath6 function becomes dominated by the contribution of sz clusters , and the negative tail is visible . however , an small residual signal of the cmb fluctuations still persists , even in the case where we do not consider noise . nevertheless , this residual can be modelled as an extra gaussian noise : a @xmath6 analysis does not distinguish two signals if both of them are gaussian . when using the optimal filter , the @xmath6 function shows its tail more clearly . therefore , when pre - processing our maps with a filter , we are able to reduce the problem to the case discussed in the previous section , where we only have sz signal , sources and noise , and hence the negative tail is showing us the presence of clusters . for completeness , we report the number of pixels which are above 3 sigmas in these maps , to illustrate the significance of the `` detection '' of the tail . for the case which includes noise , these numbers are 7500 , 21503 , 40475 and 76345 for the non - filtered , the hanning filtered , the optimal filtered and the pure sz maps , respectively . these numbers have a sampling error of 106 pixels . for a gaussian distribution , we should expect to have 11324 pixels ( 0.27% for @xmath204 ) out of @xmath205 in total . , for the cases ( a ) , ( b ) and ( d ) discussed in figure [ filters ] . we can see that the observed p(d ) function is distorted due to the sample variance contribution of the primordial cmb anisotropies . the dotted line corresponds to a gaussian with a width equal to the observed rms in the map , i.e. the p(d ) function that we would observe if all the signals in the map were gaussian and we had no sample variance . bottom : same as previous panel , but averaging 15 realisations . in this case , the sample variance is reduced and a negative tail is marginally detected even without filtering . in this case , @xmath206 and @xmath207 in the top panel and @xmath208 and @xmath209 in the bottom one.,title="fig : " ] , for the cases ( a ) , ( b ) and ( d ) discussed in figure [ filters ] . we can see that the observed p(d ) function is distorted due to the sample variance contribution of the primordial cmb anisotropies . the dotted line corresponds to a gaussian with a width equal to the observed rms in the map , i.e. the p(d ) function that we would observe if all the signals in the map were gaussian and we had no sample variance . bottom : same as previous panel , but averaging 15 realisations . in this case , the sample variance is reduced and a negative tail is marginally detected even without filtering . in this case , @xmath206 and @xmath207 in the top panel and @xmath208 and @xmath209 in the bottom one.,title="fig : " ] it is also interesting to demonstrate what we would expect if a given survey is covering an small area on sky , but large enough to have an important contribution from primary anisotropies . this would be the case of an hypothetic experiment covering 1-side patch on sky , so we have used here the simulations described in section [ subsec : ps ] . in this case , we have the additional problem of the sample variance : the observed p(d ) function will show deviations from the expected gaussian curve . therefore , filtering in this case is absolutely necessary . we illustrate this point with figure [ filters2 ] , where we show the p(d ) function from a single 1-side map , containing sz signal , cmb and no noise , and also the same plot for the average of 15 maps of the same type . we can see that , although the p(d ) function for the map without filtering is marginally showing a negative tail when we average 15 maps , the sample variance is completely deleting this tail when we consider a single map . this point can also be illustrated if we use those 15 @xmath210-side simulations to compute the uncertainty introduced by the sample variance in the number of pixels above the 3@xmath211 threshold . this number was found to be 0.5% , whilst the expected number of pixels for a gaussian distribution above 3@xmath211 is 0.27% . however , even in this case the filtering is able to remove the main contribution of the primary anisotropies , and hence the negative tail is seen . , but adding a cmb realisations to each one of them , following the plotted @xmath212cdm model . here we do not apply any filtering to the maps . top : the recovered power spectrum from the 15 realisations ( diamonds ) . error bars correspond to the field - to - field variance . at low @xmath213-values , the power spectrum traces that of the cmb , while in the high @xmath213 region , the spectrum traces that of the sz maps ( dot - dashed line ) . bottom : bispectrum for the same 15 realisations ( solid line ) . dotted lines show the 1-sigma field - to - field variance . we see that at high @xmath213 , the bispectrum follows that of the sz signal ( dot - dashed line ) , but a low-@xmath213 ( large scales ) its goes to zero because the power becomes dominated by that of the primordial cmb , which has zero skewness . , title="fig : " ] , but adding a cmb realisations to each one of them , following the plotted @xmath212cdm model . here we do not apply any filtering to the maps . top : the recovered power spectrum from the 15 realisations ( diamonds ) . error bars correspond to the field - to - field variance . at low @xmath213-values , the power spectrum traces that of the cmb , while in the high @xmath213 region , the spectrum traces that of the sz maps ( dot - dashed line ) . bottom : bispectrum for the same 15 realisations ( solid line ) . dotted lines show the 1-sigma field - to - field variance . we see that at high @xmath213 , the bispectrum follows that of the sz signal ( dot - dashed line ) , but a low-@xmath213 ( large scales ) its goes to zero because the power becomes dominated by that of the primordial cmb , which has zero skewness . , title="fig : " ] , with ( solid ) and without ( dot - dashed ) adding primordial cmb anisotropies . when the cmb is added , the maps are filtered , prior to the skewness computation , using a hanning filter with @xmath214 , as described in the text . although the skewness in real map is diluted due to the residual power from the cmb in the filtered map , it is still detectable . ] we conclude this section showing the expected skewness and bispectrum from our 15 ps simulations taking into account a cmb component in the maps . it is important to notice here that for the bispectrum is not necessary to use any filter , because it directly gives the skewness at each angular scale . these results for the power - spectrum and the bispectrum are shown in figure [ bisp_sz_cmb ] . from here , we can conclude that if the relative contribution of sz clusters to the observed bispectrum is larger than the one from radio sources , then we expect to directly see their negative signal in the bispectrum without pre - processing the maps . to obtain the skewness in real - space , we proceed as in section [ sec:7.3 ] , but pre - processing the maps with any of the filters described in the last subsection . the obtained results from averaging over the same 15 ps maps are shown in figure [ skew_sz_cmb ] . the main conclusion is that the residual cmb power in the filtered map is diluting the sz skewness , although it is still clearly seen . this is exactly what we would expect , using equation ( [ sum_skew ] ) . from simple inspection of the top panel of figure [ bisp_sz_cmb ] , we expect to have a residual cmb power after filtering with a hanning filter of @xmath214 which is similar , or slightly smaller than the sz power . therefore , we would expect the skewness to be reduced ( roughly ) by a factor @xmath215 , as we observed from the simulations . in this paper we have discussed five statements : * the contribution of sz clusters to the map noise at @xmath0 mm does not depend on the wavelength , and has a strong and peculiar non - gaussian negative tail in the @xmath6 function . * this contribution has characteristic negative skewness ( or bispectrum ) at @xmath0 mm . this fact can be used by current single - frequency experiments , such as cbi or bima , or by future experiments , such as act , ami , amiba , apex , or the 8-m south pole telescope , to distinguish if the detected excess of power at small angular scales is due to sz clusters . in addition , the detection of skewness only requires a factor 1.5 or 2 more integration time than the detection of an excess of power . once the skewness is detected , the @xmath6 function starts to show an asymmetry . * any multi - frequency experiment would have noise due to clusters of galaxies with @xmath6 at @xmath4 mm equal to @xmath65 at @xmath0 mm . * skewness and bispectrum will have different signs in these two spectral regions . * when dealing with real maps where primordial cmb fluctuations are present , it is necessary to use filters to remove the contribution of large angular scales . only in that case we can detect the presence of clusters / sources in the @xmath6 function . we would like to thank volker springel for permission to use the maps obtained in @xcite . we also thank s.d.m . white , c. hernndez - monteagudo , a. banday , m. bartelmann , s. zaroubi and p. carreira for useful comments . jarm acknowledges support from the cmbnet european network . we would like to thank also the anonymous referee for his suggestion of including a discussion about the p(d ) function in the presence of primary cmb anisotropies . 99 atrio - barandela , f. & m " ucket , j. p. 1999 , apj , 515 , 465 barcons , x. 1992 , apj , 396 , 460 banday , a. j. , gorski , k. m. , bennett , c. l. , hinshaw , g. , kogut , a. , & smoot , g. f. 1996 , apjl , 468 , l85 birkinshaw , m. 1999 , phys . , 310 , 97 bond , j. r. et al . , 2002 , apj , submitted ( astro - ph/0205386 ) bower , r. g. 1997 , mnras , 288 , 355 carlstrom , j. e. , holder , g. p. , & reese , e. d. 2002 , ara&a , 40 , 643 cavaliere , a. , friedland , a. , gursky , h. , & spada , g. 1973 , apj , 182 , 405 cavaliere , a. & setti , g. 1976 , a&a , 46 , 81 cay ' on , l. et al . 2000 , mnras , 315 , 757 cole , s. & kaiser , n. 1988 , mnras , 233 , 637 condon , j. j. 1974 , apj , 188 , 279 condon , j. j. & dressel , l. l. 1978 , apj , 222 , 745 condon , j. j. 1984 , apj , 287 , 461 cooray , a. r. , grego , l. , holzapfel , w. l. , joy , m. , & carlstrom , j. e. 1998 , aj , 115 , 1388 cooray , a. 2000 , phys . d , 62 , 103506 cooray , a. 2001 , phys . d , 64 , 63514 cooray , a. & melchiorri , a. 2002 , phys . d , accepted ( astro - ph/0204250 ) dawson , k. s , holzapfel , w. l. , carlstrom , j. e. , laroque , s. j. , miller , a. , nagai , d. & joy , m. 2002 , apj , submitted ( astro - ph/0206012 ) de luca , a. , desert , f. x. , & puget , j. l. 1995 , a&a , 300 , 335 diego , j. m. , martnez - gonz ' alez , e. , sanz , j. l. , cay ' on , l. , & silk , j. 2001 , mnras , 325 , 1533 eke , v. r. , cole , s. , & frenk , c. s. 1996 , mnras , 282 , 263 ettori , s. , de grandi , s. , & molendi , s. 2002 , a&a , 391 , 841 fabian , a. c. 1975 , mnras , 172 , 149 ferreira , p. g. , magueijo , j. , & gorski , k. m. 1998 , apj , 503 , l1 fischer , m. l. & lange , a. e. 1993 , apj , 419 , 433 franceschini , a. , toffolatti , l. , danese , l. , & de zotti , g. 1989 , apj , 344 , 35 fomalont , e. b. , kellermann , k. i. , anderson , m. c. , weistrop , d. , wall , j. v. , windhorst , r. a. , & kristian , j. a. 1988 , aj , 96 , 1187 fomalont , e. b. , kellermann , k. i. , partridge , r. b. , windhorst , r. a. , & richards , e. a. 2002 , aj , 123 , 2402 fomalont , e. b. , partridge , r. b. , lowenthal , j. d. , & windhorst , r. a. 1993 , apj , 404 , 8 haehnelt , m. g. & tegmark , m. 1996 , mnras , 279 , 545 hewish , a.1961 , mnras , 123 , 167 hobson , m. p. & magueijo , j. 1996 , mnras , 283 , 1133 holder , g. p. 2002 , apjl , submitted ( astro - ph/0207633 ) holder , g. p. 2002b , apj , accepted ( astro - ph/0205467 ) jenkins , a. , frenk , c. s. , white , s. d. m. , colberg , j. m. , cole , s. , evrard , a. e. , couchman , h. m. p. , & yoshida , n. 2001 , mnras , 321 , 372 kesden , m. , cooray , a. & kamionkowski , m. 2002 , phys . d , submitted ( astro - ph/0208325 ) kneissl , r. , jones , m. e. , saunders , r. , eke , v. r. , lasenby , a. n. , grainge , k. , & cotter , g. 2001 , mnras , 328 , 783 komatsu , e. & kitayama , t. 1999 , apj , 526 , l1 komatsu , e. & seljak , u . 2002 , mnras , submitted ( astro - ph/0205468 ) komatsu , e. , wandelt , b. d. , spergel , d. n. , banday , a. j. , & g ' orski , k. m.2002 , apj , 566 , 19 korolev , v. a. , sunyaev , r. a. , & yakubsev , l. a. 1986 , soviet astronomy letters , 12 , 339 lin , h. , chiueh , t. , & wu , x. 2002 , astro - ph/0202174 longair , m. s. , & sunyaev , r. a. 1969 , nature , 223 , 719 luo , x. 1994 , apj , 427 , l71 mason , b. s. et al . , 2002 , apj , submitted ( astro - ph/0205384 ) mohr , j. j. , mathiesen , b. , & evrard , a. e. 1999 , apj , 517 , 627 molnar , s. m. & birkinshaw , m. 2000 , apj , 537 , 542 ostriker , j. p. & steinhardt , p. j. 1995 , nature , 377 , 600 padmanabhan , t. 1993 , structure formation in the universe . cambridge , uk : cambridge university press peebles , p. j. e. 1993 , principles of physical cosmology . princeton series in physics , princeton , nj : princeton university press . press , w. h. & schechter , p. 1974 , apj , 187 , 425 rowan - robinson , m. & fabian , a. 1974 , mnras , 167 , 419 rubi~ no - martn , j. a. , atrio - barandela , f. , & hern ' andez - monteagudo , c. 2000 , apj , 538 , 53 santos , m. g. et al . 2002 , physical review letters , 88 , 241302 sanz , j. l. , herranz , d. , & martnez - g ' onzalez , e. 2001 , apj , 552 , 484 sheth , r. k. & tormen , g. 1999 , mnras , 308 , 119 scheuer , p. a. g. 1957 , proc.camb.phil.soc , 53 , 764 scheuer , p. a. g. 1974 , mnras , 166 , 329 schulz , a. e. & white , m. 2002 , apj submitted ( astro - ph/0210667 ) seljak , u. ; , burwell , j. , & pen , u. 2001 , phys.rev.d , 63 , 63001 springel , v. , white , m. , & hernquist , l.2001 , apj , 549 , 681 [ erratum : 2001 , apj , 562 , 1086 ] sunyaev , r. a. 1980 , soviet astronomy letters , 6 , 213 sunyaev , r. a. & zeldovich , y. b. 1970 , ap&ss , 7 , 3 sunyaev , r. a. & zeldovich , y. b. 1972 , comments on astrophysics , 4 , 173 sunyaev , r. a. & zeldovich , i. b. 1980 , ara&a , 18 , 537 taylor , a. c. et al . 2002 , mnras , submitted ( astro - ph/0205381 ) tegmark , m. & de oliveira - costa , a. 1998 , apj , 500 , l83 toffolatti , l. , argueso gomez , f. , de zotti , g. , mazzei , p. , franceschini , a. , danese , l. , & burigana , c. 1998 , mnras , 297 , 117 viana , p. t. p. & liddle , a. r. 1999 , mnras , 303 , 535 zhang , p. , pen , u. , & wang , b. 2002 , apj , 577 , 555 zeldovich , y. b. & sunyaev , r. a. 1969 , ap&ss , 4 , 301 the scaling relation for the gas temperature used in this work was obtained from fittings to x - ray observations . however , we could also have used that one derived from the spherical collapse model , which takes the form @xcite : @xmath217 where @xmath218 is the ratio of the mean density inside the virial radius to the critical density , and @xmath108 is the ratio of the specific galaxy kinetic energy to specific gas thermal energy . this scaling has been used by other authors ( e.g. @xcite ) . we will consider this scaling here to recompute the simulations , and we will show that the qualitative behaviour of the p(d ) function and the skewness remains unchanged . we will adopt here @xmath219 , the same parameters describing the cluster ( @xmath220 mpc , @xmath221 mpc , and @xmath222 @xmath223 ) , and the same cosmological model as in the main text ( @xmath150 , @xmath224 , @xmath225 , and @xmath226 ) . we integrate the ps mass function in the same range as in the main text . concerning the core evolution , in this work we have used an entropy - driven model with @xmath227 , which is described as : @xmath228 but without using the @xmath218 factor . therefore , we now repeat the computation taking it into account . a detailed study of how the power spectrum changes when assuming different core evolution models ( a self - similar collapse , or an entropy - driven model with different values for the @xmath229 parameter ) , can be found in @xcite . ) , but different scaling relations for the temperature and the core radius with the mass and redshift of the cluster ( see details in text ) . the curves were obtained averaging 15 simulations ( 1side ) for each case . bottom : same as before , but for the bispectrum . , title="fig : " ] ) , but different scaling relations for the temperature and the core radius with the mass and redshift of the cluster ( see details in text ) . the curves were obtained averaging 15 simulations ( 1side ) for each case . bottom : same as before , but for the bispectrum . , title="fig : " ] for definiteness , we will use the notation model a for the scaling relations used in the main text ( derived from fittings to x - ray observations ) , and model b for the new scaling relations considered in this appendix . figure [ compara_bp_il ] shows the resulting power spectrum ( @xmath230 ) and bispectrum ( @xmath231 ) for these two models . we can see that the qualitative behaviour of the bispectrum is the same . figure [ compara_pd_2scalings ] shows the p(d ) function for these two models . despite the different width of these curves ( due to the different total power in the spectrum , see figure [ compara_bp_il ] ) , when rescaling them by their @xmath211 ( @xmath232 ) , the shape of both curves is very similar . , smoothing the maps with a gaussian beam of size @xmath233 . bottom : same figure , but rescaling the curves by their @xmath211 ( obtained as @xmath232 . , title="fig : " ] , smoothing the maps with a gaussian beam of size @xmath233 . bottom : same figure , but rescaling the curves by their @xmath211 ( obtained as @xmath232 . , title="fig : " ] finally , we have explored how the shape of the @xmath6 is sensitive to the @xmath216 value . we then obtain 15 simulations for each one of the following values of @xmath216 : 0.6 , 0.7 , 0.8 , 0.9 , 1.0 and 1.1 , keeping the same cosmological model ( @xmath150 , @xmath224 and @xmath225 ) . for the scaling of the temperature and core radius with mass and redshift , we adopt the values of the model b described above . each curve has been obtained from averaging 15 realizations of 1size ( error bars are not shown ) . for comparison , it is also shown the corresponding primordial cmb power spectrum , and the current observational measurements from cbi and bima . bottom : bispectrum of the sz effect for the same values of @xmath216 considered above . , title="fig : " ] values ) . each curve has been obtained from averaging 15 realizations of 1size ( error bars are not shown ) . for comparison , it is also shown the corresponding primordial cmb power spectrum , and the current observational measurements from cbi and bima . bottom : bispectrum of the sz effect for the same values of @xmath216 considered above . , title="fig : " ] it is well - known that the sz angular power spectrum scales as @xmath234 ( see , e.g. @xcite ) . in our particular case , our power spectra derived from the simulations shows an scaling @xmath235 . figure [ bp_sigma8 ] shows the obtained power spectra and bispectrum for these values . the shapes of the bispectrum curves change slightly when changing the value of @xmath216 . however , we can find an scaling for a given multipole . given that sz power spectrum peaks around @xmath236 , we have done the study for @xmath237 , obtaining that the bispectrum scales as @xmath238 . the corresponding @xmath6 functions are shown in figure [ pd_sigma8 ] , using a beam size of @xmath233 . given that we are particularly interested in the intermediate asymptotic region , we obtained the scaling of the afactor in that region ( @xmath239 ) . from our simulations , we have @xmath240\ ] ] for the region @xmath241 and @xmath242 . this is precisely the scaling with @xmath216 that we would expect at first order , given that the shape of the @xmath6 function is roughly proportional to the @xmath211 of the sz map , and the latter is proportional to @xmath243 . for illustration , in the lower panel of figure [ pd_sigma8 ] we show the different @xmath6 curves , rescaled by @xmath244 , for the case @xmath233 . as we can see , their asymptotics become similar . . the maps have been smoothed with a @xmath245 fwhm gaussian beam . bottom : same p(d ) curves , but rescaled by @xmath246 . we can see that now the asymptotic regions are parallel.,title="fig : " ] . the maps have been smoothed with a @xmath245 fwhm gaussian beam . bottom : same p(d ) curves , but rescaled by @xmath246 . we can see that now the asymptotic regions are parallel.,title="fig : " ]

internal gravity waves play a primary role in geophysical fluids : they contribute significantly to mixing in the ocean and they redistribute energy and momentum in the middle atmosphere . until recently , most studies were focused on plane wave solutions . however , these solutions are not a satisfactory description of most geophysical manifestations of internal gravity waves , and it is now recognized that internal wave beams with a confined profile are ubiquitous in the geophysical context . we will discuss the reason for the ubiquity of wave beams in stratified fluids , related to the fact that they are solutions of the nonlinear governing equations . we will focus more specifically on situations with a constant buoyancy frequency . moreover , in light of recent experimental and analytical studies of internal gravity beams , it is timely to discuss the two main mechanisms of instability for those beams . i ) the triadic resonant instability generating two secondary wave beams . ii ) the streaming instability corresponding to the spontaneous generation of a mean flow . internal waves , instability , mean - flow

1702.07762

internal gravity waves play a primary role in geophysical fluids @xcite : they contribute significantly to mixing in the ocean @xcite and they redistribute energy and momentum in the middle atmosphere @xcite . the generation and propagation mechanisms are fairly well understood , as for instance in the case of oceanic tidal flows @xcite . by contrast , the dissipation mechanisms , together with the understanding of observed energy spectra resulting from nonlinear interactions between those waves , are still debated @xcite . several routes towards dissipation have been identified , from wave - mean flow interactions to cascade processes , but this remains a fairly open subject from both theoretical @xcite and experimental points of view @xcite . the objective of this review is to present important recent progress that sheds new light on the nonlinear destabilization of internal wave beams , bridging part of the gap between our understanding of their generation mechanisms based mostly on linear analysis , and their subsequent evolution through nonlinear effects . until recently , most studies were focused on plane wave solutions , which are introduced in classical textbooks @xcite . strikingly , such plane waves are not only solutions of the linearized dynamics , but also of the nonlinear equations @xcite . however , spatially and temporally monochromatic internal wave trains are not a satisfactory description of most geophysical internal gravity waves @xcite . indeed , oceanic field observations have rather reported internal gravity beams with a confined profile @xcite . in the atmosphere , gravity waves due to thunderstorms also often form beam - like structures @xcite . oceanic wave beams arise from the interaction of the barotropic tide with sea - floor topography , as has been recently studied theoretically and numerically @xcite , taking into account transient , finite - depth and nonlinear effects , ignored in the earlier seminal work by @xcite . the importance of those beams has also been emphasized recently in quantitative laboratory experiments @xcite . from these different works , it is now recognized that internal wave beams are ubiquitous in the geophysical context . the interest for internal gravity beams resonates with the usual pedagogical introduction to internal waves , the saint andrew s cross , which comprises four beams generated by oscillating a cylinder in a stratified fluid @xcite . thorough studies of internal wave beams can be found in @xcite . moreover , @xcite have realized that an inviscid uniformly stratified boussinesq fluid supports time - harmonic plane waves invariant in one transverse horizontal direction , propagating along a direction determined by the frequency ( and the medium through the buoyancy frequency ) , with a general spatial profile in the cross - beam direction . these wave beams are not only fundamental to the linearized dynamics but , like sinusoidal wavetrains , happen to be exact solutions of the nonlinear governing equations . remarkably , @xcite showed that the steady - state similarity linear solution for a viscous beam @xcite is also valid in the nonlinear regime . in light of the recent experimental and analytical studies of those internal gravity wave beams , it is thus timely to study their stability properties . the structure of the review is the following . first , in section [ sectionintroductive ] , we introduce the subject by presenting concepts , governing equations and approximations that lead to the description of gravity waves in stratified fluids . we dedicate a special emphasis on the peculiar role of nonlinearities to explain why internal gravity wave beams are ubiquitous solutions in oceans and middle atmospheres . then , in section [ triadicresonanceinstability ] , we discuss the classic triadic resonant instability that corresponds to the destabilization of a primary wave with the spontaneous emission of two secondary waves , of lower frequencies and different wave vectors . in addition to the simple case of plane waves , we discuss in detail the generalization to wave beams with a finite width . section [ streaminginstability ] is dedicated to the streaming instability , the second important mechanism for the instability of internal gravity waves beams through the generation of a mean flow . finally , in section [ conclusionsperspectives ] , we draw some conclusions and discuss main future issues . let us consider an incompressible non rotating stratified boussinesq fluid in cartesian coordinates ( @xmath0,@xmath1,@xmath2 ) where @xmath2 is the direction opposite to gravity . the boussinesq approximation amounts to neglecting density variations with respect to a constant reference density @xmath3 , except when those variations are associated with the gravity term @xmath4 . the relevant field to describe the effect of density variations is then the buoyancy field @xmath5 , with @xmath6 the full density field , = ( @xmath7,@xmath8,@xmath9 ) the space coordinates and @xmath10 the time coordinate . let us call @xmath11 the density of the flow at rest , with buoyancy frequency @xmath12 . the corresponding buoyancy profile @xmath13 is denoted @xmath14 . the buoyancy frequency @xmath15 varies in principle with the depth @xmath9 . in the ocean , @xmath15 is rather large in the thermocline and weaker in the abyss . for the sake of simplicity , however , @xmath15 will be taken constant in the remainder of the paper . in some studies , to ease greatly the theoretical analysis , this approximation that looks drastic at first sight can be relaxed when @xmath15 changes smoothly by relying on the wkb approximation . the equations of motion can be written as a dynamical system for the perturbed buoyancy field @xmath16 and the three components of the velocity field = ( @xmath17,@xmath18,@xmath19 ) : @xmath20 with @xmath21 the pressure variation with respect to the hydrostatic equilibrium pressure @xmath22 , and @xmath23 the kinematic viscosity . we have neglected the molecular diffusivity , which would imply a term @xmath24 in the right - hand side of equation ( [ eq : cons_masse ] ) , with @xmath25 the diffusion coefficient of the stratifying element ( molecular diffusivity for salt , thermal diffusivity for temperature ) . the importance of the dissipative terms with respect to the nonlinear ones are described by the reynolds @xmath26 and the peclet numbers @xmath27 , with @xmath28 and @xmath29 typical velocity and length scales , or equivalently by the reynolds number and the schmidt number @xmath30 . in many geophysical situations , both reynolds and peclet numbers are large , and molecular effects can be neglected at lowest order . in such cases , the results do not depend on the schmidt number . in laboratory settings , the peclet is often also very large , at least when the stratification agent is salt , in which case @xmath31 m@xmath32s@xmath33 . however , the viscosity of water is @xmath34 m@xmath32s@xmath33 , and the corresponding reynolds numbers are such that viscous effects can play an important role , as we will see later . let us first consider the simplest case of two - dimensional flow , which is invariant in the transverse @xmath8-direction . the non - divergent two - dimensional velocity field is then conveniently expressed in terms of a streamfunction @xmath35 as @xmath36 . introducing the jacobian @xmath37 , the dynamical system ( [ eq : div_u ] ) , ( [ eq : ns_strat ] ) and ( [ eq : cons_masse ] ) is expressed as @xmath38 differentiating equation ( [ equationenpsi ] ) with respect to time and equation ( [ equationenrho ] ) with respect to the spatial variable @xmath7 , and subtracting the latter from the former , one gets finally @xmath39 describing the nonlinear dynamics of non - rotating non - diffusive viscous stratified fluids in two dimensions . in the linear approximation , assuming vanishing viscosity , the right - hand side of equation ( [ equationenpsietrho ] ) immediately vanishes leading to the following wave equation for the streamfunction @xmath40 this equation is striking for several reasons . first , its mathematical structure is clearly different from the traditional dalembert equation . indeed , the spatial differentiation appears at second order in both terms . time - harmonic plane waves with frequency @xmath41 , wave vector @xmath42 and wavenumber @xmath43 are solutions of equation ( [ eq_disp_gravity_non_viscous ] ) , if the dispersion relation for internal gravity waves @xmath44 is satisfied . @xmath45 is the angle between wavenumber @xmath46 and the vertical . the second important remark is that contrary to the usual concentric waves emitted from the source of excitation when considering the dalembert equation , here four different directions of propagation are possible depending on the sign of @xmath47 and @xmath48 . this is an illustration of the anisotropic propagation due to the vertical stratification . the third remarkable property is that the dispersion relation features the angle of propagation rather than the wavelength , emphasizing a clear difference between internal waves and surface waves . this is also a crucial property for this review since it will allow us to define beams with a general profile , rather than with a single wavenumber . it is striking and pretty unusual that plane waves are solutions of the inviscid nonlinear equation ( [ equationenpsietrho ] ) even for large amplitudes . indeed , the streamfunction of the plane wave solution is a laplacian eigenmode , with @xmath49 . consequently , the first jacobian term vanishes in equation ( [ equationenpsietrho ] ) . equation ( [ equationenpsi ] ) leads therefore to the so - called polarization relation @xmath50 , with @xmath51 the polarization prefactor . consequently , the second jacobian in ( [ equationenpsietrho ] ) vanishes : @xmath52 . to conclude , both nonlinear terms in equation ( [ equationenpsietrho ] ) vanish for plane wave solutions , that are therefore solutions of the nonlinear equation , for any amplitude . since the frequency @xmath41 is independent of the wavenumber , it is possible to devise more general solutions , time - harmonic with the same frequency @xmath41 , by superposing several linear solutions associated to the same angle of propagation , but with different wavenumbers @xmath53 @xcite . introducing the along - beam coordinate @xmath54 , defined along the direction of propagation , and the cross - beam coordinate @xmath55 ( see * figure [ profilselonetabb ] * ) , the plane wave solution can be written as @xmath56 since @xmath57 and @xmath58 . if one introduces @xmath59 , one obtains the velocity field @xmath60 and the buoyancy perturbation @xmath61 one can actually obtain a wider class of solutions by considering an arbitrary complex amplitude @xmath62 . indeed , the fields @xmath63 and @xmath64 do not depend on the longitudinal variable @xmath65 . consequently , after the change of variables , the jacobians , which read @xmath66 , simply vanish , making the governing equations linear . as discussed in @xcite , note that uni - directional beams , in which energy propagates in one direction , involve plane waves with wavenumbers of the same sign only : @xmath67 or @xmath68 . we see that the class of propagating waves that are solutions of the nonlinear dynamics in a boussinesq stratified fluid is much more general than plane wave solutions : there is a whole family of solutions corresponding to uniform plane waves in the longitudinal direction @xmath65 , but with a general profile in the cross - beam direction @xmath69 , as represented in * figure [ profilselonetabb]*. and @xmath69 , of the angle of inclination @xmath45 , and finally of the group and phase velocities @xmath70 and @xmath71 . ( b ) geometry of a uniform ( along @xmath65 ) internal wave beam inclined at an angle @xmath45 with respect to the horizontal . the beam profile varies in the cross - beam @xmath69 direction , and the associated flow velocity is in the along - beam direction @xmath65 . the transverse horizontal direction is denoted by @xmath8 . , title="fig:",scaledwidth=70.0% ] -.5truecm @xcite have generalized those results by computing asymptotic solutions for a slightly viscous nonlinear wave beam with amplitude slowly modulated along @xmath65 and in time . after considerable manipulation , it turns out that all leading - order nonlinear advective - acceleration terms in the governing equations of motion vanish , and a uniform ( along @xmath65 ) beam , regardless of its profile ( along @xmath69 ) , represents an exact nonlinear solution in an unbounded , inviscid , uniformly stratified fluid . this result not only extends the validity of the @xcite steady - state similarity solution to the nonlinear regime , but emphasizes how nonlinearity has only relatively weak consequences . this has profound and useful outcomes on the applicability of results obtained with linear theory , for comparisons with field observations , laboratory experiments or numerical simulations . the vanishing of the nonlinear contributions is really unexpected and results from the combination of numerous different terms . @xcite noticed , however , that the underlying reason for the seemingly miraculous cancellation of the resonant nonlinear terms was the very same one that had been already pointed out by @xcite . after lengthy calculations , in both cases , the reason is a special case of the jacobi identity @xmath72+j\left[c , j(a , b ) \right ] + j\left[j(a , c),b \right]=0 $ ] . @xcite were studying near - critical reflection of a finite amplitude internal wave on a slope to heal the singularity occurring in the solution of @xcite . using matched asymptotic , they took a distinguished limit in which the amplitude of the incident wave , the dissipation , and the departure from criticality are all small . at the end , although the reconstructed fields do contain nonlinearly driven second harmonics , they obtained the striking and unusual result that the final amplitude equation happens to be a linear equation . the underlying reason was already this jacobi identity . to conclude , the effects of nonlinearities on plane waves or wavebeams exhibit very peculiar properties . there are two important points to keep in mind . first , plane waves and internal wave beams are solutions of the full equation . second , identifying a solution does not mean that it is a stable one . this remark is at the core of the present review : we will focus in the following on the behavior of wave beams with respect to the triadic resonant and the streaming instabilities . it was first realized fifty years ago that internal gravity plane waves are unstable to infinitesimal perturbations , which grow to form temporal and spatial resonant triads @xcite . this nonlinear instability produces two secondary waves that extract energy from a primary one . energy transfer rates due to this instability are now well established for plane waves @xcite . the instability was observed in several laboratory experiments @xcite and numerical experiments on propagating internal waves @xcite or reflecting internal tides on a horizontal or sloping boundary @xcite . oceanic field observations have also confirmed the importance of this instability , especially close to the critical latitude , where the coriolis frequency is half of the tidal frequency @xcite . recent experiments by @xcite , however , followed by a simple model and numerical simulations by @xcite as well as a theory by @xcite have shown that finite - width internal gravity wave beams exhibit a much more complex behavior than expected in the case of interacting plane waves . this is what will be discussed in this section . the classic triadic resonant instability corresponds to the destabilization of a primary wave through the spontaneous emission of two secondary waves . the frequencies and wave vectors of these three waves are related by the spatial , @xmath73 , and the temporal , @xmath74 , resonance conditions , where the indices 0 and @xmath75 refer respectively to the primary and secondary waves . in the inviscid case , the most unstable triad corresponds to antiparallel , infinitely long secondary wave vectors associated with frequencies which are both half of the primary wave frequency : @xmath76 . because of the direct analogy with the parametric oscillator , this particular case defines the parametric subharmonic instability ( psi ) . this special case applies to many geophysical situations , and especially for oceanic applications . in laboratory experiments , viscosity plays an important role and the two secondary wave frequencies are different . by abuse of language , some authors have sometimes extended the use of the name psi to cases for which secondary waves do not oscillate at half the forcing frequency . to avoid confusion , in the general case , it is presumably more appropriate to use the acronym tri . looking for solutions of the basic equations ( [ equationenpsi ] ) and ( [ equationenrho ] ) as sum of three plane waves as follows @xmath77 and @xmath78 , with @xmath79 for the primary wave and @xmath80 for the secondary ones , and denoting @xmath81 the derivative of the amplitude @xmath82 , one gets ( see for example @xcite ) @xmath83 e^{i \left(\mbox{\boldmath $ k$}_j \cdot \mbox{\boldmath $ r$ } - \omega_j t\right ) } + c.c . & = & - j(\nabla^2\psi , \psi)\,.\label{eqpsi2}\\ \sum_{j}^ { } [ \dot r_j - i \omega_j r_j - i n^2 \ell_j \psi_j ] e^{i \left(\mbox{\boldmath $ k$}_j \cdot \mbox{\boldmath $ r$ } - \omega_j t\right ) } + c.c . & = & -j(b , \psi)\,,\label{eqrho2}\end{aligned}\ ] ] the left - hand sides represent the linear parts of the dynamics . neglecting the nonlinear terms , as well as the viscous terms and the temporal evolution of the amplitudes , one recovers the polarization expression @xmath84 and the dispersion relation @xmath85 . this linear system is resonantly forced by the jacobian nonlinear terms on the right - hand side when the waves fulfill a spatial resonance condition @xmath86 and a temporal resonance condition @xmath87 the jacobian terms in equations ( [ eqpsi2 ] ) and ( [ eqrho2 ] ) can then be written as the sum of a resonant term that will drive the instability , plus some unimportant non resonant terms . introducing this result into equation ( [ eqpsi2 ] ) , one obtains three relations between @xmath88 and @xmath89 for each mode @xmath90 $ ] with @xmath91 or @xmath92 . one gets @xmath93\ , , \label{eq : rpm}\end{aligned}\ ] ] where @xmath94 . here , one traditionally uses the `` pump - wave '' approximation , which assumes that over the initial critical growth period of the secondary waves , the primary wave amplitude , @xmath95 , remains constant and that the amplitude varies slowly with respect to the period of the wave ( @xmath96 ) . differentiating the polarization expression , cumbersome but straightforward calculations @xcite lead to first order to @xmath97/({2\omega_\pm k_\pm^2})$ ] . differentiating equation ( [ equation1z ] ) , one gets @xmath98 the general solution is @xmath99 with @xmath100 and @xmath101 . in conclusion , a vanishingly small amplitude noise induces the growth of two secondary waves by a triadic resonant mechanism . since their sum gives the primary frequency ( see equation ( [ temporalcondition ] ) ) , @xmath102 and @xmath103 are subharmonic waves . the growth rate of the instability depends on the characteristics of the primary wave , namely its wave vector , its frequency and its amplitude @xmath95 , but also on the viscosity @xmath23 . using the dispersion relation for internal waves , the temporal resonance condition leads to @xcite @xmath104 whose solutions are presented in * * figure [ dessindelacacouete]**a . once the primary wave vector @xmath105 is defined , any point of the solid curve corresponds to the tip of the @xmath106 vector , while @xmath107 is obtained by closing the triangle . the choice between the labels + and - is essentially arbitrary and this leads to the symmetry @xmath108 in * * figure [ dessindelacacouete]**a . without loss of generality , we will always call @xmath106 the largest wavenumber . satisfying equation ( [ equation_k1m1 ] ) once the primary wave vector @xmath109@xmath110 is given . two examples of vector triads ( @xmath105 , @xmath106 , @xmath107 ) are shown . the dotted curve is defined by @xmath111 . the solid green curves correspond to the central branch , while the dashed and dash - dotted black curves correspond to the external branch . ( b ) and ( c ) corresponding growth rates @xmath112 as a function of the normalized wave vector modulus @xmath113 . ( b ) presents the inviscid case while ( c ) presents a viscous case corresponding to @xmath114 . , title="fig : " ] -.5truecm one can observe two distinct parts of this resonance locus , characterized by the position of @xmath113 with respect to 1 . the wavelength of the secondary waves generated by the instability can be * both smaller than the primary wavelength : this case corresponds to the external branch of the resonance locus and implies an energy transfer towards smaller scales ( represented by black curves in * figure [ dessindelacacouete ] * ) . * one larger and the other one smaller : this case corresponds to the central branch of the resonance locus and implies an energy transfer towards smaller and larger scales ( represented by solid green curves in * figure [ dessindelacacouete ] * ) . among the different possible solutions on the resonance locus , the one expected to be seen experimentally or numerically is the one associated with the largest growth rate . in the inviscid case , the most unstable growth rate occurs for @xmath115 , with essentially @xmath116 , and therefore @xmath117 . this ultraviolet catastrophe is healed in the presence of viscosity , which selects a finite wavelength for the maximum growth rate @xcite as shown in figure [ dessindelacacouete]c . for typical laboratory scale experiments , the values of @xmath118 corresponding to significant growth rates are of the same order of magnitude as the primary wavenumber @xmath119 , as can be seen in * figure [ dessindelacacouete]c * , with @xmath120 and @xmath121 . thus , tri corresponds to a direct energy transfer from the primary wave to small scales where viscous effects come into play , without the need of a turbulent cascade process . the fact that viscosity has a significant effect on the selection of the excited resonant triad , preventing any large wave number secondary wave to grow from the instability , has been observed by @xcite in laboratory experiments on wave beams . however , they also found a different type of triads than those predicted by the previous theoretical arguments . this will be discussed in more detail in the following sections . the expression for the growth rate @xmath122 implies that the amplitude of the stream function has to be larger than the critical value @xmath123 to get a strictly positive growth rate @xcite . the threshold for the instability is thus given by the global minimum of this function of several variables . let us focus on the particular case where @xmath106 tends to @xmath105 by considering the following description of the wave vector components @xmath124 and @xmath125 where @xmath126 , @xmath127 , while @xmath128 and @xmath129 are positive quantities . using the dispersion relation , the temporal and spatial resonance conditions , @xcite have shown that @xmath130 is the only acceptable value to balance the lowest order terms . plugging these relations into the expression of @xmath131 , one gets @xmath132 and @xmath133 , which leads to @xmath134 . the minimum of this positive expression being zero , it shows that there is no threshold for an infinitely wide wave beam , even when considering a viscous fluid . plane wave solutions are thus always unstable to this triadic resonant instability . the above theory for the tri does not take into account the finite width of the experimental beam . qualitatively , the subharmonic waves can only extract energy from the primary wave if they do not leave the primary beam before they can extract substantial energy @xcite . the group velocity of the primary wave is aligned with the beam , but the group velocity of the secondary waves is definitely not , and these secondary waves eventually leave the primary wave beam , as illustrated in * figure [ dessindelavitessedegroupe]*. this is a direct consequence of the dispersion relation , which relates the direction of propagation to the frequency : a different frequency , smaller for subharmonic waves , will lead to a shallower angle . three comments are in order : \i ) the angles between primary and secondary waves strongly influence the interaction time , and thus the instability . \ii ) secondary waves with small wave vectors , having a larger group velocity @xmath135 , leave the primary wave beam more rapidly and have less time to grow in amplitude . such solutions will therefore be less likely to develop , opening the door to stabilization of the primary wave by the finite width effect . this clarifies why experiments with the most unstable secondary waves on the internal branch ( small wave vector case ) of the resonance locus ( * figure [ dessindelavitessedegroupe ] * ) were found to be stable @xcite contrary to the prediction for plane waves . this decisive role of the group velocity of the short - scale subharmonic waves was identified long ago by @xcite . \iii ) at the other end of the spectrum , small wavelengths are more affected by dissipation and will also be less likely to be produced by tri . consequently , only a window of secondary wavelengths is possibly produced by tri . . ( c ) the vector triad with the three arrows representing the primary wave vector @xmath105 ( black ) and the two secondary waves vectors @xmath106 ( red ) and @xmath107 ( blue ) . from this triad , it is possible to deduce the orientation of the group velocities of the three different waves as shown in panels ( a ) and ( b ) . , title="fig:",scaledwidth=99.0% ] -.5truecm a simple energy balance proposed by @xcite makes possible an insightful and more quantitative estimate for the most unstable triad . we introduce the tilted rectangle shown in * figure [ dessindelavitessedegroupe ] * as control area ( denoting @xmath136 the perpendicular beam width ) and we neglect the spatial attenuation of the primary wave in this region ( `` pump - wave '' approximation ) . since secondary waves do not propagate parallel to the primary beam , they exit the control area from the lateral boundaries without compensation . equation ( [ equation1z ] ) is thus modified as follows @xmath137 the first term represents the interaction with the other plane waves of the triadic resonance , the second term is due to viscous damping while the third one accounts for the energy leaving the control area . one finds here also exponentially growing solutions with a positive growth rate slightly modified as @xmath138 in which the effective viscous term now reads @xmath139 . the finite width of the beam is responsible for a new term characterizing the transport of the secondary waves energy out of the interaction region . for infinitely wide wave beams ( @xmath140 ) , one recovers the growth rate @xmath122 obtained in the plane wave case . in contrast , when the beam becomes narrow ( @xmath141 ) , the growth rate decreases to zero , leading to a stabilization . the finite width of a wave beam increases therefore its stability , owing to the transport of the secondary waves out of the triadic interaction zone of the primary wave beam before they can extract substantial energy . this interaction time scales directly with the perpendicular beam width , @xmath136 as can be seen from the expression of @xmath142 . a beautiful weakly nonlinear asymptotic analysis of the finite width effect on tri has been recently proposed by @xcite . mostly interested by oceanic applications , they look for subharmonic perturbations in the form of fine - scale , nearly monochromatic wavepackets with frequency close to one half of the primary frequency ; in this limit , usually called parametric subharmonic instability ( see the sidebar distinguishing tri vs. psi ) , @xmath143 , that defines the angle @xmath144 with the vertical of the wave vectors @xmath145 of opposite directions . the key ingredient in the analytical derivation is to take advantage of the scale difference between the width of the primary beam @xmath136 and the very small carrier wavelength of the subharmonic wave packets @xmath146 . the small amplitude expansion is thus characterized by the small parameter @xmath147 . they consider an expansion with not only the underlying wave beam but also the superimposed subharmonic wavepackets that appear to order @xmath148 . the derivation of the wave - interaction equations leads to six coupled equations for the two primary beam envelopes ( two because of the two phases ) and the four ( @xmath149 ) subharmonic wavepacket envelopes . fortunately , this system can be reduced at leading order to only three coupled equations with three unknowns . taking a distinguished limit in which not only the amplitude of the primary wave , but also the nonlinear , dispersive and viscous terms are all small , they obtain a reduced description of the dynamics . the strategy is as usual to choose the scaling in order to get comparable magnitudes of the different terms . interestingly , as only the quadratic interaction is potentially destabilizing for the primary beam , they compare it with the advection term for subharmonic waves : the former being smaller , this confirms that the resonant interaction can not feed the instability in the limited time during which perturbations are in contact with the underlying beam . beams with a general profile of finite width are thus stable to tri . next , they consider the case of beams with profiles in the form of a monochromatic carrier with @xmath150 wavelength , modulated by a confined envelope . functions of the cross - beam direction @xmath69 ( see * figure [ profilselonetabb ] * ) and of the appropriate slow time @xmath151 , the complex envelopes @xmath152 and @xmath153 , of the primary and secondary waves are thus generalizations of the plane wave solutions considered in section [ derivequplanewaves ] . we recover these solutions with an envelope function independent of the cross - beam coordinate @xmath69 , while an internal wave beam ( such as the one in * figure [ dessindelavitessedegroupe ] ) * will correspond to @xmath154 for @xmath155 and zero otherwise . introducing the appropriate change of variables and rescaling of the relevant variables , the beam envelope @xmath95 and the complex subharmonic envelopes @xmath156 are linked through the following three coupled dimensionless equations @xmath157 where @xmath158 is the renormalized viscous dissipation and @xmath159 a rescaled wavenumber modulus . @xmath160 and @xmath161 are two geometrical parameters , while @xmath151 is the appropriate slow time for the evolution of the subharmonic wave packets and @xmath69 the appropriate cross - beam coordinate . in the appropriate distinguished limit identified by @xcite , the nonlinear term is balanced as in section [ energyapproach ] by the viscous term , but also by the transport term . the coupling between the evolution equations occurs through the nonlinear terms , which allow energy exchange between the underlying beam and subharmonic perturbations . the subsequent behavior of the complex envelopes @xmath156 determines the stability of the beam : if they are able to extract energy via nonlinear interaction with @xmath95 at a rate exceeding the speed of linear transport and viscous decay , the beam is unstable . from this system , it is in principle possible to study the stability of any profile : 1 . for example , a time independent beam @xmath162 is a steady state solution of this system of three equations . the study of its stability relies on looking for the normal mode solutions @xmath163 . for a plane wave , one obtains the growth rate @xmath164 . a subtle point was carefully emphasized by @xcite . the above expression of @xmath122 seems independent of the wave vector disturbance @xmath165 in the inviscid limit , but the derivation has extensively used the hypothesis of fine - scale disturbances , that will of course break down for @xmath166 . the maximum growth rate is indeed attained for finite but small @xmath165 . uniform beams ( internal plane waves with a confined envelope ) are unstable if the beam is wide enough . 2 . @xcite provide also the solution of the initial value problem for a beam with @xmath167 tending towards zero as @xmath69 tends to infinity . they show the existence of a minimum value for the unstable perturbation wavenumber @xmath168 , corresponding to a maximum wavelength . therefore , the possible spatial scale window for secondary wavelengths shrinks towards smaller scales as the beam is made narrower . outside this range , no instability is possible even in the inviscid case . they derive also analytically the minimum width explicitly for the top - hat profile used in the experiments by @xcite ( @xmath154 for @xmath155 and zero otherwise as shown in * figure [ dessindelavitessedegroupe ] * ) . they argue that the existence of a minimum width is valid for a general profile . this minimum is dependent on the beam shape . to summarize , internal plane waves with a confined envelope are unstable if the beam is wide enough , while weakly nonlinear beams with a general but confined profile ( i.e. without any dominant carrier wavenumber ) are stable to short - scale subharmonic waves . @xcite have recently studied theoretically and numerically the triadic resonant instability of an internal gravity beam in presence of a mean advective flow . they keep constant the local wave vector and wave frequency in the frame moving with the fluid in order to encompass both tidal flows and lee waves . their main result is that , by impacting the group velocity of the primary and secondary waves , the mean advection velocity modifies the most unstable triads . they have predicted and confirmed numerically that a strong enough advective flow enhances the instability of the central branch ( leading to large scale mode since one secondary wavelength is larger than the primary one ) with respect to the external branch . however , the model is not able to explain the existence of an interesting stable region , at intermediate velocity in their numerical simulations . to go beyond , it would be necessary to take into account the spatial growth of the secondary waves within the internal wave beam . such a theory relying on the extension of the classical absolute or convective instability is still to be derived . when one includes coriolis effects due to earth s rotation at a rate @xmath169 , the dispersion relation of internal waves is modified and this has of course consequences on the group velocity , which we showed to be intimately associated to the stability of internal wave beams . @xcite have reported experimental signatures of triadic resonant instability of inertial gravity beams in homogenous rotating fluid . it is thus expected that tri will also show up when considering stratified rotating fluid . assuming invariance in the transverse @xmath8 direction , the flow field may be written as @xmath170 with @xmath171 the streamfunction of the non - divergent flow in the vertical plane , and @xmath172 the transverse velocity . introducing the coriolis parameter @xmath173 , where @xmath174 is the latitude , the dynamics of the flow field is given by the following system of three equations @xmath175 in which the equation for the buoyancy perturbation is not modified , while equation ( [ equationenpsi ] ) has been modified and is now coupled to the dynamics of the transverse velocity @xmath18 . as previously , one can study beams of general spatial profile , corresponding to the superposition of time - harmonic plane waves with a dispersion relation ( [ eq_disp_gravity_theta ] ) modified into @xmath176 the next step is again to look for subharmonic perturbations in the form of fine - scale ( with respect to the width of the beam ) , nearly monochromatic wavepackets with frequency close to half the primary frequency . it is straightforward to see that subharmonic waves propagate with an inclination @xmath144 given by @xmath177 that vanishes when @xmath178 , i.e. at the critical latitude @xmath179 @xcite . the modulus of the group velocity of subharmonic waves @xmath180 will thus also vanish at this latitude . the rotation reducing dramatically the ability of subharmonic waves to escape , it may seriously reinforce the instability . @xcite have shown that it is possible to reproduce the asymptotic analysis of the triadic resonant instability with the inclusion of earth s rotation ( see also @xcite ) . one ends up with an unchanged equation ( [ equationforq ] ) for the dynamics of the primary wave , while the coupled dynamics of subharmonic waves is modified as follows @xmath181 with @xmath182 and @xmath183 two parameters depending on the coriolis parameter @xmath184 , which vanish when @xmath184 tends to zero . the important modification is the appearance , on the right - hand side , of the second linear term due to dispersion . it is important here because the first one may disappear since the projection of the respective group velocity @xmath185 of subharmonic envelopes @xmath156 on the across beam direction @xmath186 may vanish . @xcite consider first weakly nonlinear sinusoidal wavetrains , emphasizing two interesting limits : the case far from the critical latitude allows one to recover the results of section [ theoryinthenearlyinviscidlimit ] in which there is no preferred wavelength of instability in the inviscid limit . on the other hand , when the group velocity of perturbations vanishes at the critical latitude , energy transport is due solely to second - order dispersion . this process of energy transport leads to the selection of a preferred wavenumber , independent of damping effects , which may suppress the instability for a sufficiently large damping factor , permitting the underlying wave to survive the instability . they obtained an expression for the growth rate identical to the result in the inviscid limit of @xcite . it explains that , at the critical latitude , additional physical factors , such as scale - selective dissipation , must become important . this result has been shown numerically by @xcite in agreement with in situ measurements @xcite . for beams , there is always a competition between energy extraction from the beam , which varies with beam profile , and the proximity to the critical latitude , without forgetting the viscous effects on the fine - scale structure of disturbances . relying on numerical computations , it is possible to predict the stability properties for a given profile . in general , it turns out that rotation plays a significant role in dictating energy transfer from an internal wave to fine - scale disturbances via tri under resonant configurations . until now , the only laboratory experiment that studied the influence of rotation on the triadic instability of inertia - gravity waves , in a rotating stratified fluid , was performed recently by @xcite . in this study , the set - up that was used by @xcite was placed on a rotating platform , with a range of rotation rates from 0 to 2.16 rpm , allowing the dimensionless coriolis parameter , @xmath187 , to vary in a range from 0 to 0.45 . one of their main findings is the observation that the tri threshold in frequency is lowered ( by about 20% ) , compared to the non - rotating case . an extension of the energy approach developed in section [ energyapproach ] to the rotating case confirms this observation , by showing that the finite - size effect of the beam width is reduced when rotation increases . this enhancement of tri only applies to a limited range of rotation rate since , when the rotation rate overcomes half the primary wave frequency , tri is forbidden due to the dispersion relation not allowing the existence of the lowest - frequency secondary wave . the competition between this limit and the finite - size effect reduction by rotation results in a minimum value for the frequency threshold . the position of this minimum , observed around @xmath188 in the experiment , depends on the reynolds number , defined as @xmath189 . the transposition of this result to high reynolds number situations like in the ocean shows that the tri enhancement is then localized in a narrow coriolis parameter range , with @xmath190 , thus recovering the critical latitude phenomenon . when global rotation is applied to the fluid , another interesting feature is that it creates an amplitude threshold for tri . indeed , it was discussed in section [ threshold ] that in the absence of rotation , plane waves were always unstable . however , as shown in this section , the instability at very low amplitude occurs when @xmath191 tends towards @xmath192 , which implies @xmath193 and therefore @xmath194 . but when there is rotation in the system , a zero - frequency subharmonic wave is no longer allowed , hence the appearance of a threshold in amplitude , which increases with @xmath187 . another important mechanism for the instability of internal gravity waves beams is the generation of a mean flow , also called streaming instability by @xcite . @xcite and @xcite noticed in early times that internal gravity wave beams share several properties with acoustic wave beams . in particular , both kinds of waves may be subject to streaming in the presence of dissipative effects . streaming refers here to the emergence of a slowly evolving , non - oscillating , eulerian flow forced by nonlinear interactions of the oscillating wave - beam with itself @xcite . as reviewed in @xcite , it is now recognized that streaming occurs actually in a variety of flow models ; it remains an active field of research for both theoretical @xcite and experimental @xcite points of view . the fact that dissipative effects are required to generate irreversibly a mean flow through the nonlinear interactions of a wave beam with itself can be thought of as a direct consequence of `` non - acceleration '' arguments that came up in the geophysical fluid dynamics context fifty years ago with important contributions from @xcite , among others . @xcite used those ideas to propose an idealized model for the quasi - biennal oscillation ( qbo ) , together with an experimental simulation of the phenomenon @xcite . the oscillations require more than one wave beam , but @xcite discussed first how a single wave beam propagating in a vertical plane could generate a mean flow . he predicted the vertical shape of this mean flow , emphasizing the important role played by the wave attenuation through dissipative effects . the experiment by @xcite may be thought of as the first quantitative observation of streaming in stratified fluids . those examples correspond , however , to a very peculiar instance of streaming , with no production of vertical vorticity . by contrast , most applications of acoustic streaming since the earlier works of @xcite and @xcite involve the production of vorticity by an irrotational wave . as far as vortical flows are concerned , @xcite noticed important analogies between acoustic waves and internal gravity waves : in both cases , vortical flows and propagating waves are decoupled at a linear level in the inviscid limit , and steady streaming results from viscous attenuation of the wave amplitude . in particular , @xcite noticed that streaming could generate a flow with vertical vorticity . however , experimental observation of the emergence of a vortical flow in stratified fluids through this mechanism remained elusive until recently . while studying the internal wave generation process via a tidal flow over 3d seamounts in a stratified fluid , @xcite observed a strong flow in the plane perpendicular to the oscillating tidal flow . for low forcing , this flow was found to be proportional to the square of the forcing amplitude . that led them to invoke nonlinear interactions , either between the internal wave beam and itself , or between internal waves and the viscous boundary layer . the analysis was not pursued further , and the sign of the vorticity generated , opposite to the one discussed in next subsections , remains puzzling . a few years later , studying the reflection of an internal wave beam on a sloping bottom , grisouard and his collaborators have also discovered this mean - flow generation in experiments @xcite . the basic configuration was an uniform beam reflecting onto a simple slope in a uniformly stratified fluid . as predicted @xcite , the interaction between the incident and reflected waves produced harmonic waves , thereby reducing the amplitude of the reflected wave . however , more surprisingly , they found that the reflected wave was nearly absent because a wave - induced mean flow appeared in the superposition region of the incident and reflected waves , progressively growing in amplitude . comparing two- and three - dimensional numerical simulations , they showed that this mean flow is of dissipative origin and three - dimensional . its presence totally modifies the two - dimensional view considered in the literature for reflection of internal waves . indeed , there has been many interesting theoretical studies of internal gravity waves - mean flow interactions @xcite , but none of them considered the effect of dissipation in three dimensions . the complete and theoretical understanding of the generation of a slowly evolving vortical flow by an internal gravity wave beam was possible using an even simpler set - up that we describe in the following section . @xcite reported observations of a strong mean flow accompanying a time - harmonic internal gravity beam , freely propagating in a tank significantly wider than the beam . we describe below in detail the experimental set - up and the observations , together with two related theories by @xcite and @xcite , which describe well the experimental results , by providing the spatial structure and temporal evolution of the mean flow and illuminating the mechanism of instability . those approaches bear strong similarities with the result obtained by @xcite , who used generalized lagrangian mean theory , in order to describe the emergence of a vortical flow in the presence of an oscillating flow of a barotropic tide above topography variations . @xcite have studied an internal gravity wave beam of limited lateral extent propagating along a significantly wider stratified fluid tank . previously , most experimental studies that were using the same internal wave generator @xcite were quasi - two - dimensional ( beam and tank of equal width ) and therefore without significant transversal variations . * figure [ setupmanipguilhem]a * presents a schematic view of the experimental set - up in which one can see the generator , the tank and the representation of the internal wave beam generated ( see @xcite for additional details ) . the direct inspection of the flow field shows an unexpected and spontaneously generated pair of vortices , emphasized in * figure [ setupmanipguilhem]b * by the tracer particles dispersed in the tank to visualize the flow field using particle image velocimetry . this structure is actually a consequence of the generation of a strong mean flow . this experiment provides therefore an excellent set - up to carefully study the mean - flow generation and to propose a theoretical understanding that explains the salient features of the experimental observations . for the primary wave ( top plots , obtained by filtering @xcite the velocity field at the forcing frequency ) and the mean flow ( bottom plots , obtained by low - pass filtering the velocity field ) as reported respectively in @xcite and @xcite . the four left panels present the side view , while the right ones show the top view . the wave generator is represented in grey with its moving part in black . ] these observations are summarized in * figure [ resultmanipguilhem ] * , which shows side and top views not only of the generated internal wave beam but also of its associated mean flow . one sees that the wave part of the flow is monochromatic , propagating at an angle @xmath45 and with an amplitude varying slowly in space compared to the wavelength @xmath195 . these waves are accompanied by a mean flow with a jet - like structure , in the direction of the horizontal propagation of waves , together with a weak recirculation outside the wave beam . initially produced inside the wave beam , this dipolar structure corresponds to the spontaneously generated vortex shown in * figure [ setupmanipguilhem]b*. moreover , the feedback of the mean - flow on the wave leads to a transverse bending of wave beam crests that is apparent in * figures [ resultmanipguilhem]e * and * [ resultmanipguilhem]f*. taking advantage of the physical insights provided by the experiments , @xcite have proposed an approximate description that uses a time - harmonic wave flow with a slowly varying amplitude in space . the problem contains two key non - dimensional numbers , the froude number @xmath196 and the ratio @xmath197 between the wavelength @xmath195 and the attenuation length scale of the wave beam due to viscosity , @xmath198 @xcite . for analytical convenience , they considered a distinguished limit with the small parameter @xmath199 , together with the scaling @xmath200 where @xmath201 . as usual , the appropriate scaling in the small parameter @xmath128 is deduced from a mix of physical intuition and analytical handling of the calculations . in their case , they were looking for a regime with small nonlinearity and dissipation . in terms of the velocity components @xmath202 , the buoyancy @xmath64 and the vertical vorticity @xmath203 , the governing dimensionless equations in this three - dimensional setting read now @xmath204 @xmath205 @xmath206 @xmath207 in which the index @xmath208 in @xmath209 and @xmath210 reduces to the horizontal velocity field , gradient or laplacian operator . introducing rescaled spatial and time coordinates , a multiple scale analysis is now at hand . looking for a flow field in perturbation series @xmath211 for @xmath212 , @xmath8 or @xmath9 with a priori @xmath213 as suggested by the structure of the beam , together with the vertical vorticity field @xmath214 a tedious but straightforward application of the multiple scale framework ( with @xmath215 and @xmath216 ) gives then , to the first three orders , the structure of the beam : the first order @xmath217 provides the expressions for @xmath218 and @xmath219 , the second order @xmath220 gives the expression for @xmath221 and finally order @xmath222 shows that @xmath219 does not depend on the slow timescale @xmath223 . nonlinear terms contribute a priori for the first time to order @xmath224 , but one interestingly finds again ( see section [ nlterms ] ) that they vanish to this order . to order @xmath225 , one obtains that the term independent of the slow time @xmath226 vanishes and , thus , nonlinear terms do not induce a mean flow to this order either . it is only to order @xmath227 that nonlinear terms directly contribute to the mean - flow generation . the governing equation of the vortical flow induced by the mean flow is then given in the original dimensional units by @xmath228 where the overline stands for the filtering over one period and @xmath229 is the amplitude of the wave envelope . several conclusions can be directly inferred from this analysis : \i ) as emphasized by the first term on the right - hand side , nonlinear terms are crucial as a source of vertical vorticity . note that one recovers that the amplitude of the mean flow is proportional to the square of the wave amplitude as has been invoked from experimental @xcite or theoretical @xcite results . \ii ) the variations of the wave field in the @xmath8-direction ( implying @xmath230 ) are necessary for nonlinearities to be a source of vertical vorticity . this illuminates why three - dimensional effects are crucial and therefore why no mean - flow generation was noticed in two dimensions @xcite . \iii ) finally , the viscous attenuation of the wave field in the @xmath7-direction ( implying @xmath231 ) is also necessary to produce vertical vorticity . in actual experiments , variations of the amplitude in the @xmath7-direction can also come from finite size effects , but are not sufficient to generate a mean flow . one drawback of the above approach , however , is that it does not describe the feedback of the mean flow on the waves . for this reason , the approach becomes inconsistent at long time in the far field region . the above combined experimental and analytical proof of the key role played by viscous attenuation and lateral variation of the wave beam amplitude in the generation of the observed mean flow has therefore motivated a more careful asymptotic expansion by @xcite , taking into account the two - way coupling between waves and mean - flow . this two - way coupling accounts for the horizontal bending of the wave mean - field in @xcite experiments , as explained in section [ completemodel ] . initially , @xcite were interested in three dimensional perturbations of internal wave beams . specifically , they studied the stability of uniform beams subject to oblique modulations which vary slowly in the along - beam and the horizontal transverse directions . results turned out to be fundamentally different from that of purely longitudinal modulations considered in @xcite . because of the presence of transverse variations , a resonant interaction becomes possible between the primary beam and three dimensional perturbations . moreover , their analysis revealed that three - dimensional perturbations are accompanied by circulating horizontal mean flows at large distances from the vicinity of the beam . they studied the linear stability of uniform internal wave beams with confined spatial profile by introducing infinitesimal disturbances to the basic state , in the form of normal modes , not only in the along - beam direction @xmath65 but also in the horizontal transverse direction @xmath8 ( see * figure [ profilselonetabb ] * ) . they used an asymptotic approach , valid for long wavelength perturbations relative to the beam thickness . the boundary conditions combined with the matching conditions between the solution near and far from the beam ensure that the primary - harmonic and mean - flow perturbations are confined in the cross - beam direction . the analysis brings out the coupling of the primary - harmonic and mean - flow perturbations to the underlying internal wave beam : the interaction of the primary - harmonic perturbation with the beam induces a mean flow , which in turn feeds back to the primary harmonic via interactions with the beam . whether this primary - harmonic - mean flow interaction mechanism can extract energy from the basic beam , causing instability , depends upon finding modes which remain confined in the cross - beam direction . in a second stage , @xcite have derived a complete matched asymptotic analysis of the experiment performed by @xcite for a 3d boussinesq weakly nonlinear viscous fluid uniformly stratified . from their prior experience @xcite , they have chosen stretched along - beam spatial coordinate as @xmath232 , slow time as @xmath233 and transverse variations as @xmath234 so that along - beam and transverse dispersions are comparable , together with variations in the cross - beam direction @xmath69 ( see * figure [ profilselonetabb ] * ) . combining this choice with a small nonlinearity scaling as @xmath235 and a weak viscous dissipation @xmath236 that carry equal weight , they were able to fully analyze the mean flow , separately near and far from the beam , before matching both solutions . they derived a closed system of two coupled equations linking the amplitude of the primary time harmonic @xmath237 and the mean - flow component @xmath238 of the cross - beam velocity field . the latter appears to be necessary for matching with the mean flow far from the beam . the equation governing the dynamics of the mean flow reads @xmath239 where @xmath240 stands for the hilbert transform in the transverse coordinate @xmath241 . this immediately shows that transverse variations @xmath242 of the beam are essential for having a nonzero source term . since the generated mean vertical vorticity is given at leading order by @xmath243 , a direct comparison with equation ( [ equationmeanflowguilhem ] ) is possible . the first term in equation ( [ equationformeanflowinit ] ) , which involves derivatives in both horizontal coordinates corresponds to the term identified by @xcite , while this more complete analysis sheds light on an additional term deriving from purely transverse variations . using an intermediate equation , @xcite end finally with the alternative and more elegant form @xmath244\right).\label{equationformeanflow}\ ] ] moreover , they show that to match inner and outer solutions , this induced mean flow turns out to be purely horizontal to leading order and also dominant over the other harmonics . the comparison of this theoretical description agrees very well with the experimental results as beautifully emphasized by the different panels presented in * figure [ resultmanipguilhem]*. as far as a comparison with the experimental results of @xcite is concerned , a common caveat of the predictions by @xcite and @xcite is the assumption of small wavelength compared to the length scale of the wave envelope , which is only marginally satisfied in the experiments . one may for instance wonder if the horizontal structure of the observed waves is primarily due to the feedback of the mean flow on the wave , or to the sole diffraction pattern of the wave due to this absence of scale separation . this needs to be addressed in future works . using an analysis based on the generalized - lagrangian - mean ( glm ) theory , @xcite have also studied the role of dissipating oceanic internal tides in forcing mean flows . for analytical convenience , they model wave dissipation as a linear damping term @xmath245 in the buoyancy equation ( [ eq : cons_masse ] ) , and neglect the viscous term in the momentum equation ( [ eq : ns_strat ] ) . within this framework , they discuss in detail the range of situations in which a strong , secularly growing mean - flow response can be expected . their principal results include the derivation of an expression for the effective mean force exerted by small - amplitude internal tides on oceanic mean flows . at leading order , taking into account the background rotation and using a perturbation series in small wave amplitude , they derive the following explicit expression @xmath246 for the average over the tidal period of the vertical vorticity . it is remarkable that one recovers in the presence of rotation a forcing term on the right - hand side that is analogous to the forcing terms obtained by @xcite and @xcite in the non - rotating case . in inviscid rotating flows , vortical modes are at geostrophic equilibrium , and there is a frequency gap separating those geostrophic modes from inertia - gravity waves . this frequency gap generally precludes interactions between geostrophic modes and wave modes . the work of @xcite , however , shows that the combination of nonlinear and dissipative effects allows for a one - way energy transfer from inertia - gravity wave modes to geostrophic modes , through a genuinely three dimensional mechanism . using equation ( [ equationbulhlergrisouard ] ) , @xcite compute the effective mean force numerically in a number of idealized examples with simple topographies . although a complete formulation with dissipative terms in the momentum equation is necessary , the conclusion of this important work by @xcite is that energy of inertia - gravity waves in rotating fluids can be transferred to a horizontal mean flow by a similar resonance mechanism as described in the experiment by @xcite . one understands therefore that mean flows can be generated in regions of wave dissipation , and not necessarily near the topographic wave source . we have presented several recent experimental and theoretical works that have renewed the interest of internal wave beams . after emphasizing the reason for their ubiquity in stratified fluids they are solutions of the nonlinear governing equations this review has presented the two main mechanisms of instability for those beams : \i ) triadic resonant instability . we have shown that this instability produces a direct transfer of energy from large scales ( primary waves ) to smaller scales ( subharmonic ones ) for inviscid plane waves , but that it is no longer true for internal wave beams since the most unstable triad may combine subharmonic waves with larger and smaller wavelength . moreover , the effects of the finite size and envelope shape for the onset of triadic resonant instability have been overlooked . these features have to be taken into account to safely reproduce the complete nonlinear transfer of energy between scales in the ocean interior or in experimental analog @xcite , and therefore to find its stationary state , the so - called garrett and munk spectrum @xcite or its possible theoretical analog , the zakharov spectrum for the wave turbulence theory @xcite . \ii ) streaming instability . now that the mechanism underlying streaming instability and the conditions for its occurrence have been identified , several other examples will probably be reported in the coming years . for example , such a mean - flow generation has also been observed in a recent experiment @xcite for which the reflection of internal gravity waves in closed domains lead to an internal wave attractor . two lateral stokes boundary layers generate indeed a fully three - dimensional interior velocity field that provides the condition for the mean flow to appear . with a perturbation approach , @xcite confirmed this theoretically and showed that the generated 3d velocity field damps the wave beam at high wave numbers , thereby providing a new mechanism to establish an energetic balance for steady state wave attractors . @xcite have also recently studied experimentally the generation of a mean flow by a progressive internal gravity wave in a simple two - dimensional geometry , revisiting an experimental analog of the quasi - biennial oscillation @xcite . they study the feedback of the mean flow on the wave , an essential ingredient of the quasi - biennial oscillation . which is the dominant mechanism ? @xcite have recently suggested that streaming instability are central to three - dimensional internal gravity wave beam dynamics in contrast to the tri of sinusoidal wave train relevant to uniform beams , the special case of a internal plane wave with confined spatial profile . this review reinforces therefore the need for more three - dimensional experiments studying wave - induced mean flow . in particular , the conditions that favor mean - flow generation with respect to triadic resonant interaction remains largely unknown . angles of propagation ? three - dimensionality ? this is an important question that needs to be addressed . in an incompressible non - rotating linearly stratified boussinesq fluid , 1 . plane waves are solutions of the linear and nonlinear equations for any amplitude . 2 . internal wave beams , which correspond to the superposition of plane waves with wave vectors of different magnitude but pointing in the same direction , are solutions of the linear and nonlinear equations . 3 . plane waves solutions are always unstable by tri . general localized internal wave beams are stable while ( quasi ) spatial - harmonic internal wave beams are unstable if the beam is wide enough . 5 . in presence of rotation , beams of general spatial profile are more vulnerable to tri especially close to the critical latitude where nearly - stationary wavepackets remain in the interaction region for extended durations , facilitating energy transfer . 6 . internal gravity wave beams with confined spatial profile are linearly unstable to three - dimensional modulations . when the wave beam is attenuated along its direction of propagation and when the wave - envelop varies in the transverse horizontal direction , nonlinear interactions of the wave beam with itself induce the emergence of a horizontal mean - flow with vertical vorticity . the authors are not aware of any biases that might be perceived as affecting the objectivity of this review . this work was supported by the labex imust ( anr-10-labx-0064 ) of universit de lyon , within the program `` investissements davenir '' ( anr-11-idex-0007 ) operated by the french national research agency ( anr ) . this work has achieved thanks to the resources of psmn from ens de lyon . we acknowledge the contributions of g. bordes , b. bourget , c. brouzet , p .- cortet , e. ermanyuk , m. le bars , p. maurer , f. moisy , j. munroe , h. scolan , and a. wienkers to our research on this topic . we thank t. akylas , v. botton , n. grisouard , h. karimi , t. kataoka , l. maas , t. peacock , c. staquet , b. voisin for helpful discussions . alford m. , peacock t , mackinnon ja , nash jd , buijsman mc , centuroni lr , chao sh , chang mh , farmer dm , fringer ob , fu kh , gallacher pc , graber hc , helfrich kh , jachec sm , jackson cr , klymak jm , ko ds , jan s , johnston tms , legg s , lee ih , lien rc , mercier mj , moum jn , musgrave r. , park jh , pickering ai , pinkel r , rainville l , ramp st , rudnick dl , sarkar s , scotti a , simmons hl , st laurent lc , venayagamoorthy sk , wang yh , wang j , yang yj , paluszkiewicz t , tang ty . the formation and fate of internal waves in the south china sea . _ nature _ 521:65 - 69 alford m. , mackinnon ja , zhao z. , pinkel r. , klymak j. , peacock t. 2007 . internal waves across the pacific . _ geophys . _ 34:l24601 beckebanze f , maas lrm . damping of 3d internal wave attractors by lateral walls _ proceedings of the viii international symposium on stratified flows _ , san diego , 29 august-1 september 2016 , editor : university of california at san diego . callies j , ferrari r , & bhler o. 2014 . transition from geostrophic turbulence to inertia gravity waves in the atmospheric energy spectrum . _ proceedings of the national academy of sciences _ 111:1703317038 kataoka t , akylas tr . 2016 three - diensional instability of internal gravity beams . _ proceedings of the viii international symposium on stratified flows _ , san diego , 29 august-1 september 2016 , editor : university of california at san diego . leclair m , grisouard n , gostiaux l , staquet c , auclair f. 2011 . reflexion of a plane wave onto a slope and wave - induced mean flow . _ proceedings of the vii international symposium on stratified flows _ , rome , 2226 august 2011 , editor : sapienza universit di roma pairaud i , staquet c , sommeria j , mahdizadeh m. 2010 . generation of harmonics and sub - harmonics from an internal tide in a uniformly stratified fluid : numerical and laboratory experiments . in iutam symposium on turbulence in the atmosphere and oceans ( ed . d. dritschel ) , vol . 5162 . springer

in this paper , we study phase transitions in a slender circular cylinder composed of a compressible hyperelastic material with a non - convex strain energy function . we aim to construct the asymptotic solutions based on an axisymmetrical three - dimensional setting and use the results to describe the key features ( in particular , instability phenomena ) observed in the experiments by others . the difficult problem of the solution bifurcations of the governing nonlinear partial differential equations ( pde s ) is solved through a novel approach . by using a methodology involving coupled series - asymptotic expansions , we derive the normal form equation of the original complicated system of nonlinear pde s . by writing the normal form equation into a first - order dynamical system and with a phase - plane analysis , we manage to deduce the global bifurcation properties and to solve the boundary - value problem analytically . the asymptotic solutions ( including post - bifurcation solutions ) in terms of integrals are obtained . the engineering stress - strain curve plotted from the asymptotic solutions can capture the key features of the curve measured in a few experiments ( e.g. , the stress drop , the stress plateau , and the small stress valley ) . it appears that the asymptotic solutions obtained shed certain light on the instability phenomena associated with phase transitions in a cylinder , in particular the role played by the radius - length ratio . also , an important feature of this work is that we consider the clamped end conditions , which are more practical but rarely used in literature for phase transition problems . * key words * : phase transformation ; hyperelastic material ; asymptotic analysis ; cylinder ; bifurcations of pde s

0805.4047

applications of phase - transforming materials such as shape memory alloys ( smas ) and shape memory polymers are very broad . for example , they have been used to make satellite dampers , golf club heads and snake - like robots and so on . in particular , these materials have been used to design many minimal surgery devices ( pelton _ et al . a deep and thorough understanding of the behaviour of this type of phase - transforming material is essential in the manufacturing and designing of these devices . systematic experiments on uniaxial extensions of superelastic niti alloy ( a kind of shape memory alloys and also one kind of phase - transforming materials ) wires and strips ( shaw @xmath0 kyriakides 1995 , 1997 , 1998 ) showed the measured engineering stress - strain curves have the key features : the nucleation stress occurs at a local maximum which is significantly larger than the maxwell stress ; following the nucleation stress there is a sharp stress drop ; and afterwards there is a stress plateau . these features were also observed in experiments done by others ( sun _ et al . _ 2000 , tse @xmath0 sun 2000 , favier _ et al . _ 2001 and li @xmath0 sun 2002 ) . theoretically , solid - solid phase transitions have also been studied for a long time in the context of both continuum and lattice theories . the seminal work of ericksen ( 1975 ) , which considered a continuum one - dimensional stress problem , made clear that for a non - convex strain energy function the solution with two phases can arise and there are multiple solutions . based on the lattice model for a two - phase martensitic material , it is possible to deduce in the related continuum model the strain - energy function has double wells ; see ball @xmath0 james ( 1992 ) . in general , it is now understood that in the continuum scale for a material whose strain - energy function is non - convex phase transitions can take place ( see also , abeyaratne _ et al . _ 2001 ) . with a proper choice of a strain - energy function , phase transitions can be modeled through a continuum theory . still , to justify this point of view , it is desirable ( probably necessary ) to compare the analytical solutions based on this type of energy functions with experimental results . but , the difficulty is that we lack mathematical theories for mixed type equations that typically arise from non - convex energy functions . analytical solutions for boundary - value problems are very few . in the classical paper of ericksen ( 1975 ) , analytical solutions were constructed for a static problem based on a pure one - dimensional stress model , which neglects the effect from other dimensions . however , a one - dimensional model appears to be not sophisticated enough to capture some key features observed in experiments . we give the explanations below . in the experiments ( shaw @xmath0 kyriakides 1995 , 1997 ; li @xmath0 sun 2002 ) , it was observed that after the nucleation stress was reached the nucleation process began and accompanying with it there was a radial contraction ( necking ) . also , after the two - phase state was formed , the deformation was inhomogeneous , one part being thin and one part being thick . as pointed out by chang _ et al . _ ( 2006 ) that the axial extent of the transformation front is of the order of the radius . thus , to model the nucleation process and the inhomogeneous deformation of different thicknesses , one should consider the radial deformation . theoretically , if one treats the problem as a one - dimensional stress one , there is a discontinuous interface between two phases with two different strain values . the thicknesses of the cylinder at two phases are different , thus the shear strain @xmath1 ( @xmath2 is the radial displacement and @xmath3 is the axial coordinate ) is nonzero ( infinite at the phase boundary ) . the traction - free boundary conditions require the two stress components @xmath4 ( depending on @xmath5 up to the second order ) and @xmath6 ( depending on @xmath1 ) to be zero at the lateral surface , which can not be satisfied . thus , it appears that to model phase transitions by a one - dimensional stress approximation violates the traction - free boundary conditions . in the papers of dai @xmath0 cai ( 2006 ) and cai @xmath0 dai ( 2006 ) , phase transitions in a slender cylinder composed of a _ special _ _ incompressible _ elastic material was considered . a novel series - asymptotic approach is utilized to reduce the field equations . a proper asymptotic model equation is derived , which takes into account the influences of the radial deformation and traction - free boundary conditions . the solutions for two boundary - value problems are obtained , and they could capture the key features observed in experiments . in this paper , we study the phase transitions in a slender cylinder composed of a _ general _ _ compressible _ elastic material due to tension / extension . also , different from the previous studies on _ incompressible _ materials , here we further consider the coupling effect of the material nonlinearity and geometrical size . another important new feature is that we consider the more practical clamped end conditions instead of the natural boundary conditions used in cai @xmath0 dai ( 2006 ) . we consider the problem in a three - dimensional setting , different from the one - dimensional stress problem studied by ericksen ( 1975 ) . however , the strain - energy function is assumed to have the same property as that in ericksen ( 1975 ) , i.e. , for a one - dimensional stress problem the stress - strain curve has a peak - valley combination ( cf . figure 1 ) . we aim at constructing the asymptotic solutions and using them to explain the experimental results . mathematically , to deduce the analytical solutions for the present problem is a very difficult task . one need to deal with coupled nonlinear partial differential equations ( pde s ) together with complicated boundary conditions . further , the existence of multiple solutions ( corresponding to the instability phenomena ( e.g. , stress drop ) observed in experiments ) makes the problem even harder to solve . here , the analysis is carried out by a novel methodology developed earlier ( dai @xmath0 huo 2002 ; dai @xmath0 fan 2004 ; dai @xmath0 cai 2006 ) , which is capable of treating the global bifurcation problem of nonlinear pde s and obtaining the post - bifurcation solutions . we construct the solutions and extract from them important information on the deformation configurations , the nucleation stress , the instability phenomena and the transformation front . comparisons with experimental results are made , which show that the asymptotic solutions can capture the key features of the experimental engineering stress - strain curves and the instability phenomena as observed in experiments . the qualitative agreements give supporting evidence that a non - convex strain energy function can be used to describe solid - solid phase transitions and the instability phenomena in phase transformations are mainly due to the non - convexity of the strain energy function . this paper is arranged as follows . in section 2 , we formulate the field equations by treating the slender cylinder as a three - dimensional object . in section 3 , we carry out a non - dimensionalization process to extract the important small variable and two small parameters which characterize this problem . then we derive the normal form equation of the original governing nonlinear pde s in section 4 , through series and asymptotic expansions . in section 5 , we show that the euler - lagrange equation can also lead to the same normal form equation , which justifies our method in deriving this equation . in section 6 , we propose the clamped boundary conditions . in section 7 , we construct the asymptotic solutions for both a force - controlled problem and a displacement - controlled problem . we also use the solutions obtained to explain some experimental results . finally , some conclusions are drawn . we consider the axisymmetric deformations of a slender elastic cylinder subject to a static axial force at two ends . the lateral surface is traction - free and the end conditions will be considered later . in an undeformed state , the radius of the cylinder is @xmath7 and the total length is @xmath8 . it is assumed that @xmath9 . we take the cylindrical polar coordinate system and denote @xmath10 and @xmath11 the coordinates of a material point of the cylinder in the reference and current configurations , respectively . the finite radial and axial displacements can be written as @xmath12 we introduce the orthonormal bases associated with the cylindrical coordinates and denote these by @xmath13 , @xmath14 , @xmath15 and @xmath16 , @xmath17 , @xmath18 in the reference and current configurations , respectively . then the deformation gradient tensor @xmath19 is given by @xmath20 for an hyperelastic material , the strain energy function @xmath21 is a function of the three invariants @xmath22 , @xmath23 and @xmath24 of the left cauchy - green strain tensor @xmath25 ; that is , @xmath26 . we suppose that @xmath21 is non - convex in a pure one - dimensional stress problem such that phase transition can take place . the nominal stress tensor @xmath27 is given by @xmath28 if the strains are small , it is possible to expand the nominal stress components in term of the strains up to any order . the formula containing terms up to the third order material nonlinearity is ( cf . fu @xmath0 ogden 1999 ) @xmath29 where @xmath30 is the components of the tensor @xmath31 , @xmath32 , @xmath33 and @xmath34 are incremental elastic moduli , which can be calculated once a specific form of the strain - energy function is given , and the formulas can be found in appendix a. from the formula ( 2.4 ) , we could obtain the nominal stress components @xmath35 . for example , @xmath36 where @xmath37 , @xmath38 and @xmath39 are some elastic moduli , whose formulas are given in appendix a. the other non - zero stress components can also be obtained but we omit their lengthy expressions for brevity . owing to the complexity of calculations , we shall only work up to the third - order material nonlinearity . in the experiments , the maximum strain is less than @xmath40 , and such an approximation is sufficiently accurate . @xmath27 satisfies the following field equations : @xmath41 which yields the following equations @xmath42 @xmath43 we consider the case that the lateral surface of the cylinder is traction - free . thus , we have the boundary conditions @xmath44 we shall derive the asymptotic solutions of ( 2.7 ) and ( 2.8 ) under ( 2.9 ) and some end conditions for a given external force . equations ( 2.7 ) and ( 2.8 ) together with ( 2.4 ) provide the governing equations for two unknowns @xmath2 and @xmath45 . the formal two are very complicated nonlinear partial differential equations ( pde s ) and the boundary conditions ( 2.9 ) are also complicated nonlinear relations ( cf . ( 3.3)-(3.6 ) ) . to describe the instability phenomena during phase transitions , one need to study the bifurcations of this complicated system of nonlinear pde s . here , we shall adopt an alternative approach involving coupled series - asymptotic expansions to tackle this bifurcation problem . a similar methodology has been developed to study nonlinear waves and phase transitions in incompressible materials ( see dai @xmath0 huo 2002 , dai @xmath0 fan 2004 , dai @xmath0 cai 2006 , cai @xmath0 dai 2006 ) . first , we shall nondimensionalize this system to identify the relevant small variable and small parameters . we first introduce a very important transformation @xmath46 then do the following scalings : @xmath47 where @xmath48 is the length of the cylinder , @xmath49 is a characteristic axial displacement , and @xmath50 is regarded to be a small parameter ( equivalent to a small engineering strain ) . substituting ( 3.1 ) and ( 3.2 ) into ( 2.7 ) and ( 2.8 ) , we obtain @xmath51+\cdots=0,\ \ \eqno(3.3)\ ] ] @xmath52 here and hereafter , we have dropped the tilde for convenience . the full forms of ( 3.3 ) and ( 3.4 ) are very lengthy and can be found in appendix b , here we just present the first few terms . substituting ( 3.1 ) and ( 3.2 ) into the traction - free boundary conditions ( 2.9 ) , we obtain @xmath53 \\&+\epsilon^2[(\frac{1}{6}\theta_1+\frac{2}{3}\theta_2+\frac{1}{2}\theta_3)v^3+(\frac{1}{2}\theta_2+\frac{3}{2}\theta_4)v^2w_z+(\frac{1}{2}\theta_3+\frac{1}{2}\theta_4)vw_z^2+\frac{1}{6}\theta_2w_z^3 \\&+s((\frac{1}{2}\theta_5+\frac{1}{2}\theta_8)vv_z^2+\frac{1}{2}\theta_7v_z^2w_z+(\theta_1 + 2\theta_2+\theta_3)v^2v_s+(2\theta_2 + 2\theta_4)vw_zv_s \\&+\theta_3w_z^2v_s+(2\theta_{12}+2\theta_{15})vv_zw_s+2\theta_{14}v_zw_sw_z+(2\theta_5 + 2\theta_8)vw_s^2 + 2\theta_7w_zw_s^2 ) \end{aligned } \eqno(3.5)\ ] ] @xmath54|_{s=\nu}=0 , \end{aligned}\ ] ] @xmath55+\epsilon^2[(\frac{1}{2}\theta_{12}+\frac{1}{2}\theta_{13}+\theta_{15})v^2v_z+(\theta_{14}+\theta_{15})vv_zw_z \\&+\frac{1}{2}\theta_{12}v_zw_z^2+(\theta_5+\theta_6 + 2\theta_8)v^2w_s+(2\theta_7 + 2\theta_8)vw_zw_s+\theta_5w_z^2w_s \\&+s(\frac{1}{6}\theta_{16}v_z^3+(2\theta_{12}+2\theta_{15})vv_zv_s+2\theta_{14}v_zw_zv_s+\theta_{17}v_z^2w_s+(4\theta_5 + 4\theta_8)vv_sw_s \\&+4\theta_7w_zv_sw_s+2\theta_{16}v_zw_s^2+\frac{4}{3}\theta_9w_s^3)+s^2(2\theta_{12}v_zv_s^2 + 4\theta_5v_s^2w_s)]|_{s=\nu}=0 , \end{aligned } \eqno(3.6)\ ] ] where @xmath56 ( square of the diameter - length ratio ) is a small parameter for a slender cylinder . we note that in the above equations the dependence on @xmath57 and @xmath58 is entirely through @xmath59 and @xmath56 , respectively . equations ( 3.3)-(3.6 ) comprise a new system of complicated nonlinear pde s with complicated boundary conditions , which is still very difficult to solve directly . however , it is characterized by a small variable @xmath60 ( @xmath61 ) and two small parameters @xmath50 and @xmath62 , which permit us to use expansion methods to proceed further . from equations ( 3.3)-(3.6 ) , we can see that the two unknowns @xmath63 and @xmath64 depend on the variable @xmath65 and the small variable @xmath60 and the small parameters @xmath50 and @xmath62 ; that is @xmath66 to go further , we seek series expansions in terms of the small variable @xmath60 : @xmath67 @xmath68 substituting ( 4.2 ) and ( 4.3 ) into ( 3.3 ) , the left hand becomes a series in @xmath60 and all the coefficients of @xmath69 ( @xmath70=@xmath71 , @xmath72 , @xmath73 , @xmath74 ) should vanish . equating the coefficients of @xmath75 and @xmath76 to be zero yields that @xmath77 @xmath78 similarly , substituting equations ( 4.2 ) and ( 4.3 ) into ( 3.4 ) and equating the coefficient of @xmath75 to be zero yields that @xmath79 the expressions of @xmath80 , @xmath81 and @xmath82 are very lengthy , which are omitted for brevity . substituting ( 4.2 ) and ( 4.3 ) into the traction - free boundary conditions ( 3.5 ) and ( 3.6 ) , we obtain @xmath83 @xmath84 where the lengthy expressions for @xmath85-@xmath86 are omitted , and interested readers can find the expressions for @xmath87 ( @xmath88 @xmath89 @xmath72 , @xmath74 , @xmath90 ) in appendix c. in obtaining the above equations , we have neglected terms higher than @xmath91 . now the governing equations ( 3.3)-(3.4 ) are changed into a one - dimensional system of differential equations ( 4.4)-(4.8 ) for the unknowns @xmath92 , @xmath93 , @xmath94 , @xmath95 and @xmath96 . by using a regular perturbation method , we can express @xmath93 by @xmath92 and @xmath95 from ( 4.4 ) , then substitute @xmath93 into ( 4.6 ) , we can also express @xmath96 by @xmath92 and @xmath95 . similarly , we substitute @xmath93 and @xmath96 into ( 4.5 ) to obtain the expression of @xmath94 in terms of @xmath92 and @xmath95 through the regular perturbation method . the results are given below : @xmath97 @xmath98 @xmath99 where @xmath100 @xmath101 are constants related to material constants , whose expressions are omitted for brevity . substituting @xmath93 , @xmath96 and @xmath94 into ( 4.7 ) and ( 4.8 ) and omitting the higher order terms yield the following two equations with only two unknowns @xmath92 and @xmath95 : @xmath102 and @xmath103 where @xmath104 @xmath105 are also constants related to material constants , whose expressions are omitted for brevity . ( note : interested readers can contact the corresponding author for the formulas for @xmath100 and @xmath104 . ) by using the above two equations , we can express @xmath95 in terms of @xmath92 and its derivatives with respect to @xmath65 . from ( 4.12 ) , we have @xmath106 . \end{aligned } \eqno(4.14)\ ] ] from ( 4.12 ) and ( 4.13 ) , we also have @xmath107+\cdots , \eqno(4.15)\ ] ] and @xmath108+\cdots.\ \ \eqno(4.16)\ ] ] substituting @xmath109 into ( 4.15 ) and ( 4.16 ) , then substituting ( 4.16 ) into @xmath110 of ( 4.14 ) and ( 4.15 ) into the other part of ( 4.14 ) , we obtain @xmath111 where @xmath112 @xmath113,\\ \alpha_3=&\frac{1}{64\xi_1(\xi_1+\xi_2)^3}[13\xi_1 ^ 4 + 4\xi_2 ^ 3(4\eta_1 - 3\eta_2)-\xi_1 ^ 2\xi_2(23\xi_2+\eta_1 + 13\eta_2 - 26\eta_3 ) \\&+4\xi_1\xi_2 ^ 2(3\eta_1 - 10\eta_2 + 4\eta_3)+\xi_1 ^ 3(10\xi_2 - 3\eta_1 - 7\eta_2 + 6\eta_3)],\\ \alpha_4=&\frac{1}{16\xi_1(\xi_1-\xi_2)(\xi_1+\xi_2)^4}[3\xi_1 ^ 6 + 2\xi_2 ^ 5(-\eta_1+\eta_2)+\xi_1 ^ 3\xi_2 ^ 2(-51\xi_2 + 10\eta_2 - 6\eta_3 ) \end{aligned}\ ] ] @xmath114 . \end{aligned}\ ] ] by substituting ( 4.16 ) and ( 4.17 ) into ( 4.13 ) , we obtain @xmath115 where @xmath116 , \end{aligned } \eqno(4.19)\ ] ] @xmath117 , \end{aligned } \eqno(4.20)\ ] ] @xmath118 . \end{aligned } \eqno(4.21)\ ] ] integrating ( 4.18 ) once , we obtain @xmath119 where @xmath120 is the integration constant . it is important to find the physical meaning of @xmath120 , since to capture the instability phenomena observed in the experiments , one needs to study the global bifurcation as the physical parameters vary . for that purpose , we consider the resultant force @xmath121 acting on the material cross - section that is planar and perpendicular to the cylinder axis in the reference configuration , and the formula is @xmath122 by using ( 4.2 ) , ( 4.3 ) and the expressions of @xmath95 , @xmath96 and @xmath93 in terms of @xmath92 in ( 2.5 ) , it is possible to express @xmath123 in terms of @xmath92 . then , carrying out the integration in ( 4.23 ) , we find that @xmath124,\ \ \ \eqno(4.24)\ ] ] where @xmath125 is the young s modulus . comparing ( 4.22 ) and ( 4.24 ) , we have @xmath126 . if we retain the original dimensional variable and let @xmath127 , we have @xmath128 where @xmath129 is the dimensionless engineering stress . since ( 4.25 ) is derived from the three - dimensional field equations , once its solution is found , the three - dimensional strain and stress fields can also be found . also , it contains all the required terms to yield the leading - order behavior of the original system . therefore , we refer ( 4.25 ) as the normal form equation of the system of nonlinear pde s ( 3.3 ) and ( 3.4 ) together with the boundary conditions ( 3.5 ) and ( 3.6 ) under a given axial resultant . * remark * : although the final form of the normal form equation is one dimensional , it is different from other one - dimensional phenomenological models in literature ( see coleman 1983 , tong _ et al . _ 2001 , shaw 2002 , chang _ _ 2006 ) in the sense that it is derived from the three - dimensional field equations in a mathematically consistent manner and all three - dimensional quantities can be calculated once its solutions are found . thus , we can provide the three - dimensional current configuration in a two - phase state , which can not be deduced in the existing one - dimensional models in literature . this equation contains a higher - order derivative term which plays the role of regularization . trunskinovsky ( 1982 , 1985 ) pioneered the idea of the regularization augmentation for solid - solid phase transitions which involves adding terms ( like a strain gradient ) into the usual constitutive stress - strain relation ( for the gradient approach , see also aifantis @xmath0 serrin 1983 and triantafyllidis @xmath0 aifantis 1986 ) . different from these gradient theories , the gradient term ( which represents the influence of the radial deformation and traction - free boundary conditions ) in our equation is derived and its coefficient is explicitly given . the normal form equation ( 4.25 ) is derived in the previous section based on the equilibrium equations ( 2.7 ) and ( 2.8 ) . now , we consider the energy and show that the same equation can be derived through the variational principle . the calculations are tremendously complicated in the present method ( we use mathematica for all the symbolic computations ) , and if the two derivations of the normal form equation agree with each other , it also supports the correctness of our calculations . we expand the strain energy function up to the fourth - order nonlinearity ( which implies that the stress components are up to the third - order nonlinearity ) , and as a result we obtain @xmath130 the strain energy per unit length is given by @xmath131 by using the same manipulations in section 3 and section 4 , we can get the average strain energy over a cross section @xmath132 , \end{aligned } \eqno(5.3)\ ] ] where @xmath133 ( @xmath134 ) are some constants , whose expressions can be found in appendix a. by further using ( 4.12 ) and ( 4.13 ) , we can reduce the above equation as @xmath135 \\=&e [ \frac{1}{2}v^2+\frac{1}{3}d_1v^3+\frac{1}{4}d_2v^4-\frac{1}{8}a^2 vv_{zz}+a^2(h_1vv_z^2+h_2v^2v_{zz } ) ] , \end{aligned } \eqno(5.4)\ ] ] where @xmath136 @xmath137 . \end{aligned } \eqno(5.6)\ ] ] the total potential energy is then given by @xmath138 further by the variational principle , from the euler - lagrange equation we obtain @xmath139=\gamma . \eqno(5.8)\ ] ] it can be seen from ( 5.5 ) , ( 5.6 ) and ( 4.21 ) that @xmath140 . so the equation ( 5.8 ) is exactly the equation ( 4.25 ) . the expressions of @xmath141 and @xmath142 are themselves important . when there are multiple solutions , the smallest energy criterion can be used to judge the preferred solution ( configuration ) . now , we conduct a detailed analysis on the normal form equation with clamped end boundary conditions . we suppose that the strain energy function is non - convex in a homogeneous constant strain state such that phase transition can take place . this requires @xmath143 we rewrite equation ( 4.25 ) as a first - order system @xmath144 the critical points of this system are determined by @xmath145 and @xmath146 in this paper , we always choose @xmath147 , @xmath148 . we can see that with these material constants , the critical stress values @xmath149 , @xmath150 , @xmath151 ( cf . ( 6.4 ) ) and the corresponding strain values ( cf . figure 1 ) are close to the experimental results ( tse @xmath0 sun 2000 and li @xmath0 sun 2002 ) . the @xmath152 curve corresponding to equation ( 6.3 ) is shown in figure 1 . + the peak stress value @xmath150 , the valley stress value @xmath149 and the maxwell stress value @xmath151 are @xmath153 the natural boundary conditions @xmath154 at the two ends have been used by many authors ( e.g. , ericksen 1975 ; tong _ et al . _ 2001 ; cai @xmath0 dai 2006 ) . however , such conditions are difficult to be realized in practice . in experiments , usually the two ends are clamped into the rigid bodies . to better describe the realistic situation , in this paper we will consider the case that the two ends of the cylinder are clamped into rigid constraints . in this case , we can propose the conditions according to the fact that there is no radial displacement for the point at the lateral surface of the ends , i.e. , @xmath155 by using ( 3.1 ) , ( 3.2 ) and ( 4.2 ) , neglecting the @xmath156 terms , we obtain @xmath157 in deriving equation ( 4.18 ) , @xmath95 and @xmath96 have been expressed in terms of @xmath92 . using such relations in equation ( 6.6 ) and with the help of equation ( 4.25 ) , we obtain @xmath158 where @xmath159 , @xmath160 and @xmath161 are material constants . notice that in deriving ( 6.7 ) we have neglected the terms higher than @xmath162 . it s clear that if we choose different values of @xmath159 , @xmath160 and @xmath161 , the properties of the boundary condition ( 6.7 ) may also have a lot of differences . here we will only consider a simple case . we will choose @xmath159 , @xmath160 and @xmath161 such that the cubic equation ( 6.7 ) has only one unique positive root @xmath163 when @xmath164 , and this root is much smaller than the positive roots of equation ( 6.3 ) . in this paper , we will choose @xmath165 , @xmath166 and @xmath167 to get the following graphic result : + thus the boundary conditions at the two ends are @xmath168 we note that @xmath163 depends on the value of the engineering stress @xmath169 . * remark * : although the normal form equation is one - dimensional , it is derived from a three - dimensional setting , and as a result we can use the end conditions by considering the quantities in the radial dimension . if one directly introduces a one - dimensional model , such an option is not available . we now consider the solutions for a given engineering stress @xmath169 under the clamped boundary conditions ( 6.8 ) . in the following , without loss of generality , we take @xmath170 . then in equation ( 4.25 ) , @xmath7 is equivalent to the diameter - length ratio . due to symmetry , we only consider the part @xmath171 and boundary conditions ( 6.8 ) can be replaced by @xmath172 we regard the engineering stress @xmath169 as the bifurcation parameter . as @xmath169 varies , there are seven types of phase planes , which are shown in figure 3 . @xmath173{figure-3a.eps}\\ \small\textcolor[rgb]{0.00,0.00,0.00}{(a ) } \end{aligned}\ \ \ \ \ \ \ \ \ \begin{aligned } \includegraphics[width=46mm , height=30mm]{figure-3b.eps}\\ \small\textcolor[rgb]{0.00,0.00,0.00}{(b ) } \end{aligned}\ ] ] @xmath173{figure-3c.eps}\\ \small\textcolor[rgb]{0.00,0.00,0.00}{(c ) } \end{aligned}\ \ \ \ \ \ \ \ \ \begin{aligned } \includegraphics[width=46mm , height=30mm]{figure-3d.eps}\\ \small\textcolor[rgb]{0.00,0.00,0.00}{(d ) } \end{aligned}\ ] ] @xmath174{figure-3e.eps}\\ \small\textcolor[rgb]{0.00,0.00,0.00}{(e ) } \end{aligned}\ \ \ \ \ \ \ \ \ \begin{aligned } \includegraphics[width=45.5mm , height=30mm]{figure-3f.eps}\\ \small\textcolor[rgb]{0.00,0.00,0.00}{(f ) } \end{aligned}\ \ \ \ \ \ \ \ \ \begin{aligned } \includegraphics[width=45.5mm , height=30mm]{figure-3g.eps}\\ \small\textcolor[rgb]{0.00,0.00,0.00}{(g ) } \end{aligned}\ ] ] + every trajectory in the phase planes of figure 3 is a solution of the system ( 6.2 ) . however , only those trajectories that satisfy equation ( 7.1 ) represent the physical solutions for the present problem . it can be seen that for a trajectory to be the physical solution a necessary and sufficient condition is that it contacts the @xmath175-axis once ( so that @xmath176 ) and contacts the vertical line @xmath177 once ( so that @xmath178 ) and the z - interval for this segment of the trajectory is exactly equal to 1 . to deduce the solution , we integrate ( 4.25 ) once to obtain @xmath179 where @xmath180 is the integration constant . if one trajectory contact the @xmath175-axis at point @xmath181 , @xmath182 must be a real root of the equation @xmath183 by using ( 7.2 ) and consider the case @xmath184 , we can get @xmath185 with @xmath186 . here , for simplicity , we choose @xmath187 such that @xmath188 for graphic results . suppose that @xmath189 , we can obtain @xmath190 where @xmath191 to satisfy the end boundary condition ( 7.1 ) , we need to choose @xmath182 such that @xmath192 based on the phase planes in figure 3 and equation ( 7.5 ) and ( 7.6 ) , we can construct all the possible solutions for a given @xmath169 . case ( a ) @xmath193 ( cf . figure 3(a)-3(d ) ) in this case , from figure 3(a)-3(d ) , we can see that only the trajectories located between the left saddle point @xmath96 and the vertical line @xmath177 ( notice that @xmath194 , cf . figure 2 ) can be the possible solutions . for any given @xmath169 , @xmath163 and @xmath96 can be calculated immediately from the two equations ( 6.3 ) and ( 6.7 ) . by using the equation ( 7.6 ) , we can determine the value of @xmath182 . once @xmath182 is determined , through numerical integration , the corresponding solution can be obtained from ( 7.5 ) . for any @xmath169 satisfies @xmath193 , we found that there is only one solution ( denoted by @xmath195 ) , which is very close to the saddle point @xmath96 . case ( b ) @xmath196 ( cf . figure 3(e ) ) in this case , for some given @xmath169 , there are multiple solutions . from figure 3(e ) , we can see that the trajectories located between the left saddle point @xmath96 and the vertical line @xmath177 can be the possible solutions . through some calculations , we found that for any @xmath196 , there exist one solution ( also denoted by @xmath195 ) , which is represented by a trajectory to the left of ( also very close to ) the saddle point @xmath96 . we can see that the trajectories located between the homoclinic orbit and the right saddle point @xmath197 can also be the possible solutions . through some calculations , we found that there is a critical engineering stress @xmath198 which satisfies @xmath199 . when @xmath200 , there exist another two solutions . one solution ( denoted by @xmath201 ) is represented by a trajectory outside ( but very close ) to the homoclinic orbit . the other solution ( denoted by @xmath202 ) is represented by a trajectory outside the second one and to the left of ( very close to ) the right saddle point @xmath197 . thus the critical engineering stress @xmath203 can be considered as a bifurcation point . when @xmath204 , there is only one solution . when @xmath200 , there are three solutions . we found that @xmath203 is a monotonically increasing function of the diameter - length ratio @xmath7 , e.g. , @xmath205 , @xmath206 . case ( c ) @xmath207 ( cf . figure 3(f ) ) in this case , there are three solutions . the first solution ( also denoted by @xmath195 ) is represented by a trajectory to the left of ( also very close to ) the cusp point @xmath208 . the second solution ( also denoted by @xmath201 ) is represented by a trajectory to the right of ( also very close to ) the cusp point @xmath208 . the third solution ( also denoted by @xmath202 ) is represented by a trajectory to the left of ( also very close to ) the saddle point @xmath197 . case ( d ) @xmath209 ( cf . figure 3(g ) ) in this case , there exist another critical engineering stress @xmath210 which is a little bit larger than @xmath150 . when @xmath211 , there are three solutions . the first two solutions ( also denoted by @xmath195 and @xmath201 ) are represented by two trajectories ( very close to each other ) located in the middle part of the interval @xmath212 . the third solution ( also denoted by @xmath202 ) is represented by a trajectory to the left of ( very close to ) the saddle point @xmath197 . when @xmath213 , there is only one solution ( also denoted by @xmath202 ) , which is represented by a trajectory to the left of ( very close to ) the saddle point @xmath197 . thus the critical engineering stress @xmath214 can be considered as another bifurcation point . we found that @xmath214 is also a monotonically increasing function of the diameter - length ratio @xmath7 , e.g. , @xmath215 , @xmath216 . for the radius @xmath217 , the axial strain distribution @xmath218 corresponding to the above four cases are plotted in figure 4 . @xmath219{figure-4a.eps}\\ \small\textcolor[rgb]{0.00,0.00,0.00}{(a ) } \end{aligned}\ \ \ \ \ \ \ \ \ \begin{aligned } \includegraphics[width=55mm , height=41mm]{figure-4b.eps}\\ \small\textcolor[rgb]{0.00,0.00,0.00}{(b ) } \end{aligned}\ ] ] @xmath219{figure-4c.eps}\\ \small\textcolor[rgb]{0.00,0.00,0.00}{(c ) } \end{aligned}\ \ \ \ \ \ \ \ \ \begin{aligned } \includegraphics[width=55mm , height=41mm]{figure-4d.eps}\\ \small\textcolor[rgb]{0.00,0.00,0.00}{(d ) } \end{aligned}\ ] ] + from the above results , it can be seen that there are multiple solutions for @xmath220 . we suppose that the preferred solution should be the one which has the smallest total potential energy value . to determine which solution is the most preferred one , we consider the total potential energy values of all the possible solutions . from ( 5.7 ) and due to symmetric , we can get @xmath221 without loss of generality , we choose @xmath222 and @xmath223 for graphic results . for the radius @xmath217 , we plot the total potential energies for all the solutions in figure 5 . + from figure 5 , it can be seen that there is a critical stress value @xmath224 ( e.g. , @xmath225 ) . for @xmath226 , the first solution @xmath195 is the preferred solution , and for @xmath227 , the third solution @xmath202 is the preferred solution . thus , in a loading process , after @xmath228 , the configuration of the cylinder may jump from solution @xmath195 towards solution @xmath202 , which is corresponding to the phase transition process . * remark * : the existence of @xmath229 ( @xmath230 ) is significant and it implies that under clamped end conditions the phase transition can only happen at a stress value larger than the maxwell stress . we now consider the case that the total elongation @xmath231 is given . since we have taken @xmath48 to be @xmath72 , @xmath232 is actually the engineering strain . the governing equation is still equation ( 4.25 ) , but now @xmath169 is an unknown parameter . in the previous section , we have obtained all the solutions for a given @xmath169 . if for a given @xmath232 we can find the corresponding value @xmath169 , then we can obtain the solutions for a displacement - controlled problem . we first plot the @xmath233 curves corresponding to the solutions we obtained in the previous section for four different values of @xmath7 in figure 6 . @xmath234{figure-6a.eps}\\ \small\textcolor[rgb]{0.00,0.00,0.00}{(a ) } \end{aligned}\ \ \ \ \ \ \ \ \ \begin{aligned } \includegraphics[width=60mm , height=38mm]{figure-6b.eps}\\ \small\textcolor[rgb]{0.00,0.00,0.00}{(b ) } \end{aligned}\ ] ] @xmath234{figure-6c.eps}\\ \small\textcolor[rgb]{0.00,0.00,0.00}{(c ) } \end{aligned}\ \ \ \ \ \ \ \ \ \begin{aligned } \includegraphics[width=60mm , height=38mm]{figure-6d.eps}\\ \small\textcolor[rgb]{0.00,0.00,0.00}{(d ) } \end{aligned}\ ] ] + we can see that the @xmath233 curves shown in figure 6 have some important features . first , these curves obtained from the asymptotic solutions capture the main characteristics of the engineering stress - strain curves measured in a number of experiments ( shaw @xmath0 kyriakides 1995 , sun _ _ 2000 , tse @xmath0 sun 2000 , favier _ et al . _ 2001 and li @xmath0 sun 2002 ) , i.e. , there is a stress peak , which is followed by a sharp stress drop , and then there is a stress plateau ( cf . figure 6(c ) ) . second , it can be seen that there is a snap - back when the diameter - length ratio @xmath7 is small ( cf . figure 6(a ) and 6(b ) ) . as the diameter - length ratio become larger , this snap - back phenomenon disappear ( cf . figure 6(c ) and 6(d ) ) . third , as the diameter - length ratio @xmath7 increases , the whole @xmath233 curve moves towards left , especially in the high - strain region . it is clear that this feature is due to the boundary effect . with these curves , we are ready to convert the solutions of a force - controlled problem into those of a displacement - controlled problem . as we pointed out that if the diameter - length ratio @xmath7 is large enough , there is no snap - back on the corresponding @xmath233 curve ( cf . figure 6(c ) and 6(d ) ) . thus , in this case , for a given displacement @xmath232 , we can only get one corresponding stress value @xmath169 from the @xmath233 curve . for this @xmath169 , there is a unique phase plane . there are maybe more than one trajectories in this phase plane which satisfy the end boundary conditions , however only one of them can give the required displacement value , which then represents the unique solution for the given @xmath232 . as we pointed out before that the normal form equation ( 4.25 ) is derived from the three - dimensional field equations . with the solutions of the normal form equation , we can recover the three - dimensional strain fields . here , we take @xmath235 as an example . corresponding to the four points @xmath236 shown in figure 7 , + we draw the shapes of the cylinder in figure 8 ( the radial deformation has been enlarged for clearness ) . + figure 8(a ) represents a low - strain state , figure 8(b ) represents a transition state , figure 8(c ) represents a two - phase state , and figure 8(d ) represents a high - strain state . as the end displacement increases these four states appears consecutively . here , the asymptotic solutions we obtained can describe the whole deformation process . we know that the dimensionless engineering stress @xmath169 represent the stress density acting on the cross section of the cylinder in the reference configuration . now , based on the solutions obtained , we can also calculate the true stress density @xmath237 acting on the cross section of the cylinder in the current configuration . figure 9 shows the distributions of the true stress @xmath237 corresponding to the four points @xmath236 . we see , very interesting , the true stress actually decreases from a to b and to c. we also point out that a pure one - dimensional model can not yield the true stress distribution . + if the diameter - length ratio @xmath7 is relatively small , we know that there is a snap - back on the corresponding @xmath233 curve ( cf . figure 6(a ) and 6(b ) ) . thus , for the total elongation @xmath232 located in some special region , there may exist multiple corresponding stress values @xmath169 , i.e. , the solutions are not unique . then we need to determine which solution is the preferred one . we take @xmath238 as an example . from figure 10 , we can see that when @xmath239 , there are three possible solutions , which are labeled as @xmath240 , @xmath241 and @xmath242 , respectively . + to choose the preferred solution from @xmath240 , @xmath241 and @xmath242 , we calculate the total potential energy values of these solutions . denote @xmath243 the total potential energy of the cylinder in the displacement - controlled problem and we have @xmath244 here we still choose @xmath222 , @xmath223 . in figure 11 , we plot the differences of the total potential energies between solutions @xmath241 , @xmath242 and solution @xmath240 for @xmath239 . + from figure 11 , we can see that for @xmath245 , solution @xmath240 is the preferred solution and for @xmath246 , solution @xmath242 is the preferred solution . the engineering stress - strain curve corresponding to the preferred solution is shown in figure 12 . + it is worth to note that there exists a jump for the stress @xmath169 at @xmath247 . this curve also captures the main features of the experimental results in the loading process ( shaw @xmath0 kyriakides 1995 , sun _ _ 2000 , tse @xmath0 sun 2000 , favier _ et al . _ 2001 and li @xmath0 sun 2002 ) . figure 13 shows the true stress - strain ( @xmath237-@xmath232 ) curves ( occurring at @xmath248 and @xmath249 ) corresponding to the preferred solutions for @xmath238 . @xmath250{figure-13a.eps}\\ \small\textcolor[rgb]{0.00,0.00,0.00}{(a ) } \end{aligned}\ \ \ \ \ \ \ \ \ \begin{aligned } \includegraphics[width=63mm , height=43mm]{figure-13b.eps}\\ \small\textcolor[rgb]{0.00,0.00,0.00}{(b ) } \end{aligned}\ ] ] + it is worth to note that for the case occurring at @xmath248 , the drop for @xmath237 become very small and after that there is also a stress plateau ( with the true stress value larger than the maxwell stress ) . for the case occurring at @xmath249 , the drops for @xmath237 become larger . but now the stress plateaus have been divided into two parts , which corresponding to the fact that phase transition first start at the middle part of the cylinder and transform gradually towards the two ends . we consider phase transitions induced by axial tension / extension in a slender circular cylinder composed of a general compressible hyperelastic material with a non - convex strain energy function under clamped end conditions . in order to capture the macroscopic phenomena observed in the experiments by others , the problem is formulated in a three - dimensional setting and as a result it is governed by a system of coupled nonlinear pde s with complicated nonlinear boundary conditions . a novel approach involving coupled series - asymptotic expansions is developed to derive the normal form equation ( nfe ) of the original system . by conducting the phase - plane analysis on this nfe , we manage to deduce the global bifurcation properties of its solutions . the solutions ( including those for post - bifurcation solutions ) for both a force - controlled problem and a displacement - controlled problem are obtained . these asymptotic solutions demonstrate the essential features of phase transitions in a cylinder and are consistent with the experimental data qualitatively . specifically , the engineering stress - strain curve is shown to have the same features as observed in experiments . these qualitative agreements with experiments give strong supporting evidence that a non - convex strain energy function can be used to describe phase transitions and the non - convexity of the strain energy function is the main cause of the instability phenomena associated with phase transitions in a slender cylinder . an important finding is that under the clamped end conditions , there exists a critical stress value @xmath229 , which depends on the radius - length ratio and larger than the maxwell stress , such that phase transition can only happen when the engineering stress is larger than this value . this is different from the case of the nature boundary condition that phase transition can happen as long as the engineering stress is larger than the maxwell stress . * acknowledgments * the work described in this paper was fully supported by two grants from the research grants council of the hksar , china ( project numbers : cityu 100804 and cityu 100807 ) . @number#1 addtoreset#1sectionnamedefthe#1 . @nostar@section .@em @empty@section @numberequation@numberfigure@numbertable @star@section@empty@section@section @numberequation@numberfigure@numbertable for initially isotropic material , in the case that there are no prestresses , @xmath21 should be a function of the principle stretches @xmath251 , @xmath252 and @xmath253 , namely @xmath21=@xmath254 . denote by @xmath255 , then @xmath256 should vanish since there are no prestresses . the non - zero first order incremental elastic moduli can be written as @xmath257 there are only two independent constants among @xmath37 . the non - zero second order incremental elastic moduli can be written as @xmath258 there are only three additional independent constants among @xmath259 . the non - zero third order incremental elastic moduli can be written as @xmath260 @xmath261 there are only four additional independent constants among @xmath262 . the expressions of the constants @xmath133 ( @xmath134 ) in equation ( 5.3 ) are given below : @xmath263 @xmath264 @xmath265 the full forms of the non - dimensional field equations ( 3.3 ) and ( 3.4 ) are given below : @xmath266+\epsilon^2[(\theta_2 + 3\theta_4+\theta_{12}\\ & + \theta_{13}+2\theta_{15})v^2v_z+(2\theta_3 + 2\theta_4 + 2\theta_{14}+2\theta_{15})vv_zw_z+(\theta_2+\theta_{12})v_zw_z^2+(\theta_3+\theta_4)v^2w_{zz}\\ & + 2\theta_2vw_zw_{zz}+\frac{1}{2}\theta_1w_z^2w_{zz}+(2\theta_5 + 2\theta_6 + 4\theta_8)v^2w_s+(4\theta_7 + 4\theta_8)vw_zw_s+2\theta_5w_z^2w_s\\ & + s((\frac{2}{3}\theta_{16}+\frac{1}{2}\theta_7+\frac{1}{2}\theta_8)v_z^3+(\theta_7+\theta_8)vv_zv_{zz}+\theta_5v_zw_zv_{zz}+\frac{1}{2}\theta_5v_z^2w_{zz}\\ & + ( 10\theta_{12}+2\theta_{13}+12\theta_{15}+2\theta_2 + 6\theta_4)vv_zv_s+(10\theta_{14}+2\theta_{15}+2\theta_3 + 2\theta_4)v_zw_zv_s\\ & + ( 2\theta_3 + 2\theta_4)vw_{zz}v_s+2\theta_2w_zw_{zz}v_s+(2\theta_{14}+2\theta_{15}+4\theta_{17})v_z^2w_s+(2\theta_{14}+2\theta_{15})vv_{zz}w_s\\ & + 2\theta_{12}w_zv_{zz}w_s+2\theta_{12}v_zw_{zz}w_s+(20\theta_5 + 4\theta_6 + 24\theta_8)vv_sw_s+(20\theta_7 + 4\theta_8)w_zv_sw_s\\ & + ( 8\theta_{16}+2\theta_7 + 2\theta_8)v_zw_s^2 + 2\theta_5w_{zz}w_s^2+\frac{16}{3}\theta_9w_s^3+(\theta_{12}+\theta_{13}+2\theta_{15}+\theta_2 + 3\theta_4)v^2v_{sz}\\ & + ( 2\theta_{14}+2\theta_{15}+2\theta_3 + 2\theta_4)vw_zv_{sz}+(\theta_{12}+\theta_2)w_z^2v_{sz}+(4\theta_{14}+4\theta_{15})vv_zw_{sz}+2\theta_5w_z^2w_{ss}\\ & + 4\theta_{12}v_zw_zw_{sz}+(8\theta_7 + 8\theta_8)vw_sw_{sz}+8\theta_5w_zw_sw_{sz}+(2\theta_5 + 2\theta_{6}+4\theta_8)v^2w_{ss}\\ & + ( 4\theta_7 + 4\theta_8)vw_zw_{ss})+s^2(2\theta_7v_zv_{zz}v_s+(16\theta_{12}+4\theta_{15}+2\theta_2 + 2\theta_4)v_zv_s^2 + 16\theta_7v_sw_sw_{sz}\\ & + 2\theta_3w_{zz}v_s^2 + 4\theta_{14}v_{zz}v_sw_s+(\theta_{16}+\theta_7)v_z^2v_{sz}+(4\theta_{12}+4\theta_{15}+4\theta_2 + 4\theta_4)vv_sv_{sz}\\ & + ( 4\theta_{14}+4\theta_3)w_zv_sv_{sz}+(4\theta_{14}+4\theta_{17})v_zw_sv_{sz}+(4\theta_{16}+4\theta_7)w_s^2v_{sz}+8\theta_{14}v_zv_sw_{sz}\\ \end{aligned}\ ] ] @xmath267=0 , \end{aligned } \eqno(b1)\ ] ] @xmath268+\epsilon^2[(\frac{5}{2}\theta_5+\frac{1}{2}\theta_6 + 3\theta_8)vv_z^2\\ & + ( \frac{5}{2}\theta_7+\frac{1}{2}\theta_8)v_z^2w_z+(\frac{1}{2}\theta_5+\frac{1}{2}\theta_6+\theta_8)v^2v_{zz}+(\theta_7+\theta_8)vw_zv_{zz}+\frac{1}{2}\theta_5w_z^2v_{zz}\\ & + ( \theta_7+\theta_8)vv_zw_{zz}+\theta_5v_zw_zw_{zz}+(4\theta_1 + 8\theta_2 + 4\theta_3)v^2v_s+(8\theta_2 + 8\theta_4)vw_zv_s\\ & + 4\theta_3w_z^2v_s+(8\theta_{12}+8\theta_{15})vv_zw_s+8\theta_{14}v_zw_zw_s+(2\theta_{14}+2\theta_{15})vw_{zz}w_s+2\theta_{12}w_zw_{zz}w_s\\ & + ( 6\theta_5 - 2\theta_6 + 4\theta_8)vw_s^2+(6\theta_7 - 2\theta_8)w_zw_s^2+(\theta_{12}+\theta_{13}+2\theta_{15}+\theta_2 + 3\theta_4)v^2w_{sz}\\ & + ( 2\theta_{14}+2\theta_{15}+2\theta_3 + 2\theta_4)vw_zw_{sz}+(\theta_{12}+\theta_2)w_z^2w_{sz}+s(\frac{1}{2}\theta_9v_z^2v_{zz}+(8\theta_5 + 2\theta8)v_z^2v_s\\ & + ( 2\theta_5 + 2\theta_8)vv_{zz}v_s+2\theta_7w_zv_{zz}v_s+2\theta_7v_zw_{zz}v_s+(14\theta_1 + 16\theta_2 + 2\theta_3)vv_s^2\\ & + ( 14\theta_2 + 2\theta_4)w_zv_s^2 + 2\theta_{16}v_zv_{zz}w_s+(28\theta_{12}+4\theta_{15})v_zv_sw_s+4\theta_{14}w_{zz}v_sw_s\\ & + 2\theta_{17}v_{zz}w_s^2 + 24\theta_5v_sw_s^2+(4\theta_5 + 4\theta_8)vv_zv_{sz}+4\theta_7v_zw_zv_{sz}+(8\theta_{12}+8\theta_{15})vw_sv_{sz}\\ & + 8\theta_{14}w_zw_sv_{sz}+(\theta_{16}+\theta_7)v_z^2w_{sz}+(4\theta_{12}+4\theta_{15}+4\theta_2 + 4\theta_4)vv_sw_{sz}\\ & + ( 4\theta_{14}+4\theta_3)w_zv_sw_{sz}+(4\theta_{14}+4\theta_{17})v_zw_sw_{sz}+(4\theta_{16}+4\theta_7)w_s^2w_{sz}\\ & + ( 2\theta_1 + 4\theta_2 + 2\theta_3)v^2v_{ss}+(4\theta_2 + 4\theta_4)vw_zv_{ss}+2\theta_3w_z^2v_{ss}+(4\theta_{12}+4\theta_{15})vv_zw_{ss}\\ & + 4\theta_{14}v_zw_zw_{ss}+(8\theta_5 + 8\theta_8)vw_sw_{ss}+8\theta_7w_zw_sw_{ss})+s^2(2\theta_5v_{zz}v_s^2+(\frac{40}{3}\theta_1+\frac{8}{3}\theta_2)v_s^3\\ & + 8\theta_5v_zv_sv_{sz}+16\theta_{12}v_sw_sv_{sz}+(4\theta_{12}+4\theta_2)v_s^2w_{sz}+2\theta_5v_z^2v_{ss}+(8\theta_1 + 8\theta_2)vv_sv_{ss}\\ & + 8\theta_2w_zv_sv_{ss}+8\theta_{12}v_zw_sv_{ss}+8\theta_5w_s^2v_{ss}+8\theta_{12}v_zv_sw_{ss}+16\theta_5v_sw_sw_{ss})\\ & + 8s^3\theta_1v_s^2v_{ss}]=0 . \end{aligned } \eqno(b2)\ ] ] 99 abeyaratne , r. , bhattacharya , k. , knowles , j.k . , strain - energy functions with multiple local minima : modeling phase transformations using finite thermoelasticity . in : fu , y. , ogden , r. ( eds . ) , nonlinear elasticity : theory and applications , cambridge : cambridge university press , pp . 433 - 490 . cai , z.x . , dai , h.h . , 2006 . phase transitions in a slender cylinder composed of an incompressible elastic material . ii . analytical solutions for two boundary - value problems , proc . ser . a math . 462 , 419 - 438 . dai , h.h . , fan , x.j . , 2004 . asymptotically approximate model equations for weakly nonlinear long waves in compressible elastic rods and their comparisons with other simplified model equations , math . solids 9 , 61 - 79 . dai , h.h . , cai , z.x . , phase transitions in a slender cylinder composed of an incompressible elastic material . i. asymptotic model equation , proc . . lond . ser . a math . 462 , 75 - 95 . favier , d. , liu , y. , orgeas , l. , rio , g. , 2001 . mechanical instability of niti in tension , compression and shear . in : sun , q.p . ( eds . ) , proc . iutam symp . on mechanics of martensitic phase transformation in solids , boston ; kluwer academic publishers , pp . 205 - 212 . sun , q.p . , li , z.q . , tse , ken k.k . , 2000 . on superelastic deformation of niti shape memory alloy micro - tubes and wires band nucleation and propagation . in : gabbert u. , tzou h.s . ( eds . ) , proc . iutam symp . on smart structures and structure systems , magdeburg , germany : kluwer academic publishers , pp . 1 - 8 . tong , p. , lam , c.c . , sun , q.p . phase transformation of thin wires in tension . in : sun q.p . ( eds . ) , proc . iutam symp . on mechanics of martensitic phase transformation in solids , hong kong , 11 - 15 june 2001 , boston : kluwer academic publishers , pp . 221 - 232 .

we report new 350 m polarization observations of the thermal dust emission from the cores surrounding the low - mass , class 0 ysos l1527 , ic348-smm2 and b335 . we have inferred magnetic field directions from these observations , and have used them together with results in the literature to determine whether magnetically regulated core - collapse and star - formation models are consistent with the observations . these models predict a pseudo - disk with its symmetry axis aligned with the core magnetic field . the models also predict a magnetic field pinch structure on a scale less than or comparable to the infall radii for these sources . in addition , if the core magnetic field aligns ( or nearly aligns ) the core rotation axis with the magnetic field before core collapse , then the models predict the alignment ( or near alignment ) of the overall pinch field structure with the bipolar outflows in these sources . we show that if one includes the distorting effects of bipolar outflows on magnetic fields , then in general the observational results for l1527 and ic348-smm2 are consistent with these magnetically regulated models . we can say the same for b335 only if we assume the distorting effects of the bipolar outflow on the magnetic fields within the b335 core are much greater than for l1527 and ic348-smm2 . we show that the energy densities of the outflows in all three sources are large enough to distort the magnetic fields predicted by magnetically regulated models .

1103.4370

only a few percent of the mass of a molecular cloud is typically converted into stars ( mckee & ostriker 2007 ; goldsmith et al . the rate of star formation inferred for molecular clouds in general is significantly less than that expected from free - fall gravitational collapse of the gravitationally bound mass in these regions . consequently , there must be some mechanism which regulates the rate of star formation within molecular clouds . two mechanisms have been proposed ; magnetic support ( e.g. , shu , adams & lizano 1987 ; mouschovias & ciolek 1999 ) and super - alfvenic turbulence ( e.g. , mac low & klessen 2004 ) . in addition , observations ( goldsmith & arquilla 1985 ; goodman et al . 1993 ) show that the rotation rates of cores ( density @xmath1 10@xmath2 @xmath3 ; size @xmath4 0.1 pc ) are less than expected for cores condensed from their less dense background clouds . in the models favoring magnetic support , this slow core rotation is the result of magnetic braking ( mestel 1985 ; basu & mouschovias 1994 ) . mouschovias & paleologou ( 1979 , 1980 ) showed analytically that the braking timescale for a cloud rotating with its rotation axis perpendicular to the magnetic field is typically an order of magnitude smaller than for a cloud rotating with its axis of rotation parallel to the magnetic field . this results in core rotation rates reduced in amplitude and core rotation axes tending to align with the cloud magnetic field immediately in the vicinity of cores . in this paper , we use our submillimeter polarimetry results to test gross features of the magnetically regulated core - collapse and star - formation models in the literature . a key feature of many such magnetically regulated dynamical collapse models ( shu , adams & lizano 1987 ; allen , shu & li 2003 ; allen , li & shu 2003 ; galli & shu 1993a , b ; shu , li & allen 2004 ; tomisaka 1998 ) is the formation of a flattened inner core ( or `` pseudo - disk '' ) of several thousand au in extent with its symmetry axis parallel to the core magnetic field . this pseudo - disk is the inner part of the infall region of the core surrounding a yso ; it is not formed by rotation , but by the geometry of the magnetic field . the pseudo - disk is a dynamically collapsing entity , accreting onto the protostar and its associated keplerian disk ( size @xmath4 100 au ) . signatures of magnetic regulation include a fairly uniform magnetic field outside the infall region of a core , a pinched magnetic field line structure within this region , and the overall direction of the magnetic field parallel to the axis of symmetry of the pseudo - disk . the level of uniformity of the field outside the infall region depends on the quasi - static evolution prior to the dynamical collapse of the core . in the galli & shu ( 1993a , b ) models , the field lines outside the infall region are uniform , but in the allen , shu & li ( 2003 ; hereafter asl03a ) and allen , li & shu ( 2003 ; hereafter als03b ) models , they are already inclined towards a gentle pinch configuration through quasi - static contraction of the core as a whole before the onset of the dynamical inside - out collapse of the inner core . various mechanisms have been proposed to remove angular momentum from the gas falling onto a protostar through a keplerian disk within the inner core ; many use the observed , ubiquitous , bipolar outflows which turn on during the class 0 phase of star formation ( konigl & pudritz 2000 ; shu et al . 2000 ; tomisaka 1998 ) . the general expectation from the above magnetically regulated collapse and outflow models is that the outflow axis will be parallel or nearly so to the core magnetic field , since in these models the core rotation axis is aligned or nearly aligned with the core magnetic field , and so the keplerian disk and bipolar outflow axes are also parallel or nearly parallel to the core magnetic field lines . the cartoon and caption in figure 1 summarize the above magnetically regulated scenario . the basic scenario depicted in figure 1 has not been observationally verified ; so far observations tailored to test parts of this scenario have not provided irrefutable evidence for the general existence of pseudo - disks , let alone the alignment of the mean magnetic fields with respect to the symmetry axis of these disks . recent polarimetry studies by ward - thompson et al . ( 2009 ) and tassis et al . ( 2009 ) imply mean magnetic field directions more aligned with the short axis of observed flattened cores than not ( consistent with figure 1 ) , but by no means aligned . the statistical analysis of tassis et al . ( 2009 ) of their 24 high - mass star forming regions assumed random line - of - sight angles for the cores and magnetic fields , resulting in a best - fit model consisting of a thick oblate core with a magnetic field on average having an angle of 24@xmath5 to the short axis of the core . tassis et al . ( 2009 ) state that more observations are required in order to reduce the inaccuracies caused by unknown line - of - sight projection effects , but their results do provide some validity to the pseudo - disk scenario , albeit with slightly mis - aligned magnetic field configurations . in regards to the near alignment of the outflow to the magnetic field shown in figure 1 , mnard & duchne ( 2004 ) used polarization measurements of background stars to study the relative orientations of the jets from classical t tauri stars ( ctts ) embedded in the taurus - auriga molecular cloud and the magnetic field in their vicinity . typically the background stars used in this study were separated by about 0.5@xmath5 or more from the evolved ctts being examined , so the magnetic field being probed in this way is not that of the field of a remnant core surrounding the ctts , but the magnetic field of the molecular cloud in the vicinity of the ctts . this study concluded that the jets of embedded ctts in the molecular cloud taurus - auriga are not aligned with the large - scale molecular cloud magnetic fields . indeed the jets within this one molecular cloud are not aligned with each other . large scale turbulence may play a role - possibly producing angular momentum and magnetic structures within the cores which are not aligned with the major axis of symmetry or the large - scale magnetic field of the much larger parent molecular cloud . possible evidence for this can be seen in taurus - auriga in the two pre - stellar cores observed by kirk , ward - thompson , & crutcher ( 2006 ) , for which the polarized 850 @xmath0 m emission indicates changes in magnetic field directions in the cores relative to the large - scale magnetic field in the cloud . it is the magnetic field in these cores , not the large - scale magnetic field , that may align the direction of future outflows from that core if magnetic regulation occurs within the cores . hence the results of mnard & duchne ( 2004 ) do not necessarily disprove the magnetically regulated scenario . valle , bastien & greaves ( 2000 ) , henning et al . ( 2001 ) , matthews & wilson ( 2002 ) , valle , greaves & fiege ( 2003 ) , wolf , launhardt & henning ( 2003 ) , girart , rao & marrone ( 2006 ) , kwon et al . ( 2006 ) and attard et al . ( 2009 ) , among others , undertook observational studies of magnetic field structures within cores containing low - mass , embedded ysos with bipolar outflows using polarized submillimeter continuum emission . low - mass ysos are good objects to study because they tend to be embedded in simple , relatively isolated regions . the results of the above studies taken as a whole do not show a clear case for the alignment of outflows with core magnetic fields . however , most of the ysos in this combined sample are binaries , are very distant , are not class 0 , or do not have their outflow axis lying close to the plane - of - the - sky . the last criterion is an important one because a pinched magnetic field structure , which is symmetric about an axis of an outflow with a large line - of - sight component , would produce polarization vectors with a large variation in position angles . to minimize projection effects when testing the theory of alignment between outflows and core magnetic fields , the axis of the outflow should lie close to the plane - of - the - sky . we have begun a survey of low - mass , isolated ( single ) , nearby ( @xmath6 400 pc ) , class 0 ysos that have well defined bipolar outflows which lie nearly in the plane - of - the - sky . our study will provide maps of the 350 @xmath0 m polarization vectors within a 10,000 au radius around each embedded yso in our survey . we will also provide interferometric maps of the outflows within the cores for each of our survey ysos where such maps do not already exist . this is so we can better determine the orientation of the outflow and its interaction with the core gas . we are using 350 @xmath0 m rather than longer wavelength polarization because we want to obtain the best spatial resolution using sharp ( see section 2 ) , and because 350 @xmath0 m emission is weighted towards the warmer emission of the cores immediately surrounding a yso rather than towards the extended cooler core envelope . this paper gives a summary of our results on three of the ysos in our study , l1527 , b335 , and ic348-smm2 . section 2 outlines our observations and summarizes our results , section 3 compares these results to magnetically regulated models , and section 4 gives our conclusions . the observations presented here were carried out at the caltech submillimeter observatory ( cso ) using the sharp polarimeter . sharp is a fore - optics module that provides polarimetric capability to sharc - ii , the cso s 32 @xmath7 12 pixel submillimeter camera . sharp divides the incident submillimeter radiation into two orthogonally polarized beams that are then reimaged onto the two ends of the `` long and skinny '' sharc - ii detector . a half - wave plate upstream of the polarization - splitting optics is rotated every few minutes during data collection , and the two orthogonally polarized 12 @xmath7 12 pixel `` sub - images '' acquired for four different half - wave plate angles are combined in software to determine the total flux as well as the linear polarization state of the radiation . sharc - ii and sharp are described respectively by dowell et al . ( 2003 ) and li et al . the 350 @xmath0 m observations described here were made with a beam size of 10@xmath8 ( fwhm ) . the data were collected in chop - nod mode . this involves rapidly modulating the tilt of the cso secondary mirror in cross - elevation ( `` chopping '' ) while more slowly `` nodding '' the entire telescope back and forth , thereby making near simultaneous measurements of the source and two nearby sky reference positions . a chop - nod observation is carried out at each of the four half - wave plate angles ( 0@xmath5 , 22.5@xmath5 , 45@xmath5 , and 67.5@xmath5 ) thereby forming a `` half - wave plate cycle '' . as described by dowell et al . ( 1998 ) and kirby et al . ( 2005 ) , the net effect is that the total source flux and a component of the linearly polarized flux are measured for each half - wave plate position in a cycle , while removing the spatially extended `` background '' that includes atmospheric and telescope emission as well as any galactic emission covering a sky area large compared to the chopper throw . for the observations described here , the chopper throw ranged from 120@xmath8 ( for b335 and l1527 ) to 300@xmath8 ( for ic348-smm2 ) and the chopping frequency was @xmath41 hz . because of sky rotation , the reference beams rotate about the main beam on the sky as the source is tracked . our three targets do have extended flux , so it was necessary to avoid making observations at hour angles for which significant source flux contaminates the reference beams . each sharp half - wave plate cycle requires about 7 minutes of elapsed time . in between successive half - wave plate cycles , the pointing position is dithered by about 10 - 20@xmath8 . data were obtained for b335 during the nights of 28 , 29 , and 30 april 2007 ( @xmath480 cycles ; zenith submillimeter opacity @xmath9 1.0 - 1.8 ) , for ic348-smm2 on the night of 10 november 2007 ( @xmath420 cycles ; @xmath9 1.0 ) , and for l1527 during the nights of 9 and 10 november 2007 and 6 and 10 september 2008 ( @xmath485 cycles , @xmath9 0.8 - 1.2 ) . sharp data analysis is carried out in two steps . in the first step , each individual half - wave plate cycle is processed to obtain three 12 @xmath7 12 pixel maps , one for each of the stokes parameters @xmath10 , @xmath11 , and @xmath12 . ( parameter @xmath10 corresponds to the total flux while @xmath11 and @xmath12 fully characterize the linearly polarized flux . ) a detailed discussion of the techniques used during this first step can be found in section 3.4 of hildebrand et al . ( 2000 ) and equations 2 , 3 , and 4 of attard et al . corresponding error - maps are also obtained for each of the three stokes parameters . this is done by first using the variance in the individual total and polarized flux measurements to estimate the uncertainties in this time stream ( as in section 4.1 of dowell et al . 1998 ) and then propagating these uncertainties to the stokes parameter maps . at this point , we remove from the stokes parameter maps any pixels that have abnormally high errors . the second step of the analysis involves combining these single - cycle 12 @xmath7 12 pixel maps of @xmath10 , @xmath11 , and @xmath12 , to form final maps of @xmath10 , @xmath11 , and @xmath12 . we account for the dithering and the sky rotation by interpolating the single - cycle maps onto a regular equatorial - coordinate grid , which causes a modest loss of angular resolution ( houde & vaillancourt 2007 ) . errors in the final maps are propagated from the corresponding errors in the single - cycle maps . corrections for changing atmospheric opacity ( kirby et al . 2005 ) as well as for instrumental polarization and polarimetric efficiency ( li et al . 2008 ) are also made during this second analysis step . an important question that we can ask during the second step of the analysis is whether the single - cycle @xmath11 and @xmath12 maps are consistent with one another within their nominal errors . recall that these nominal errors are computed during the first step of the analysis . a quantitative answer to this question is provided by the reduced chi squared , @xmath13 . averaging together the @xmath13 values found over the @xmath11 and @xmath12 maps of each source , we obtain mean @xmath13 values of 1.7 for l1527 , 1.5 for ic348-smm2 , and 2.1 for b335 . we do not know the origin of the extra errors that cause these elevated @xmath13 values . however , for each of our three data sets ( one for each source ) we were able to verify that these errors occur mainly on time scales that are short compared to the total duration of the data set . ( this duration is several hours for ic348-smm2 , and several days or even longer for the other two sources . ) thus , it seems reasonable to treat the extra errors as if they are statistical in nature . accordingly , we inflate our nominal errors by the square root of @xmath13 . finally , the degree @xmath14 and angle @xmath15 of polarization and their associated errors @xmath16 and @xmath17 are computed for each sky position via standard techniques ( see section 3.4 of hildebrand et al . the uncertainties @xmath16 and @xmath17 are affected by both the polarized flux errors ( @xmath18 , @xmath19 ) and the total flux errors ( @xmath20 ) . the latter have a negligible effect , however , because @xmath21 . we consider sky positions for which @xmath22 to be polarization detections and sky positions for which @xmath23 to be low upper limits on @xmath14 . @xmath14 is then corrected downward to account for polarization bias ( as described in section 4.2 of hildebrand et al . 2000 and in vaillancourt 2006 ) . we keep track of the sky positions where the polarimetric significance drops below @xmath24 after this bias correction , as discussed below . note that we have chosen to set our detection threshold at @xmath22 rather than applying the more conservative criterion @xmath25 used for previous sharp observations ( e.g. , attard et al . when using the latter , more conservative threshold , the @xmath26 uncertainties @xmath27 in the angles of polarization ( which translate into @xmath26 uncertainties in the magnetic field angles ) are below @xmath28 ( serkowski 1962 ) . with our more lenient threshold , these @xmath26 error bars range up to almost @xmath29 ( see table 1 ) . because our goal in this paper is to test the gross predictions of the magnetically regulated collapse models rather than its fine details , we believe that this degree of uncertainty is acceptable . however , it should be kept in mind when interpreting our polarimetry results . our polarization detections for l1527 , ic348-smm2 , and b335 are presented in table 1 . figures 2a , 3a , and 4a illustrate the results listed in table 1 . contours indicate the total intensity , i , measured by sharp . the highest contour level corresponds to 90% of the peak flux , and each subsequent contour represents a decrement of 10% of the peak flux . except for very narrow strips at the map edges , all sky positions mapped have total flux errors well below 10% of the peak flux . the morphologies seen in our total intensity contour maps thus represent real structures , with the exception of a few very small flux peaks and flux holes seen at the map edges ( e.g. , southwest edge of figure 3a , and northeast and northwest edges of figure 4a ) . all three sources have been mapped at 850 @xmath0 m ( chandler & richer 2000 , hatchell et al . 2005 , jrgensen et al . 2007 ) , and the morphologies seen at this longer wavelength are similar to those seen in our sharp maps . the bars in figures 2a , 3a , and 4a indicate polarization detections . the length of each bar is proportional to @xmath14 ( see key at bottom left of each figure ) and its orientation indicates the direction of the e - vector of the polarized radiation . the thickness of each bar shows its significance : sky points having greater than @xmath30 significance after bias correction are thickest , points with greater than @xmath24 post - correction significance but less than @xmath30 have intermediate thickness , and sky points for which the significance drops below @xmath24 after bias correction are shown using the thinnest bars . for two sky positions , we obtained @xmath24 upper limits below 1.0% . these correspond to the peak of b335 , where we find a @xmath24 limit of 0.6% and a position offset from the peak of l1527 by ( @xmath31 ) = ( + 9.5@xmath8 , -9.5@xmath8 ) , where our @xmath24 limit is 0.9% . although we do not make use of these upper limits in this paper , we include them in our figures since they represent sensitive measurements that may one day be useful . the two limits are indicated with open circles in figures 2a and 4a . comparing our polarization results for b335 with the 850 @xmath0 m polarimetry of this source presented by wolf , launhardt & henning ( 2003 ; hereafter wlh03 ) , we find reasonable agreement . the wlh03 map is discussed further in section 3.2.2 below . similarly , the 850 @xmath0 m polarization map of l1527 that is presented with no interpretation in the scupol archive paper ( matthews et al . 2009 ) agrees with our 350 @xmath0 m polarization map of this source . in making these comparisons , we refer only to the measured angles of polarization since the degree of polarization of thermal dust emission is known to generally have considerable wavelength dependence ( vaillancourt et al . 2008 , and references therein ) . the b335 350 @xmath0 m emission is very compact ( see figure 4a ) . to date , outflow maps for b335 have not been made with sufficient spatial resolution to determine the structure and kinematics of the outflow within the compact core region of b335 . as part of our survey , we measured the outflow within the b335 core using the carma interferometer . the carma observations of b335 were made with the 15-element array in the d - configuration in july 2008 . the total duration of the observations was 8.4 hours . the qso 1925 + 211 was the phase calibrator ; its flux was determined to be 1.2 jy , using both uranus and neptune as flux calibrators . observations of 3c454.3 were used to calibrate the passband structure . the correlator was configured to place the co 1 - 0 line in the upper sideband ; with 63 channels across a 7.6 mhz bandwidth , the velocity resolution at 115.271 ghz is approximately 0.32 kms@xmath32 . phase calibration was carried out on observations of the 2.7 mm continuum emission in two 500 mhz bands . the gains determined for these wide ( continuum ) bands were then applied to the narrow ( line ) bands . the data were calibrated and mapped using the miriad software package . the fourier transform of u , v visibilities was taken , constrained by an image cell size of 1@xmath8 and a natural weighting function . the size of the synthesized beam is 4.4@xmath8 @xmath7 3.8@xmath8 . the peak intensity in the continuum emission ( figure 4b , grayscale ) is 16.8 mjy / beam ; the rms noise is 1.8 mjy / beam . the blue and red contours in figure 4b show the co 1 - 0 line emission integrated over the velocity ranges 5.4 - 8.0 kms@xmath32 and 8.6 - 11.2 kms@xmath32 , respectively . the contours are at intervals of 3@xmath33 , where @xmath33 is 174 mjy / beam @xmath7 kms@xmath32 . the integrated intensity peak in the blueshifted outflow is 5.6 jy / beam @xmath7 kms@xmath32 ; the peak in the redshifted outflow is 4.0 jy / beam @xmath7 kms@xmath32 . magnetically regulated models predict that a yso will be surrounded by a pseudo - disk a few thousand au in size , and that the pseudo - disk symmetry axis will be aligned or nearly so with that of the yso bipolar outflow . in this sub - section we will explore the extent to which this is the case for our three sources . in the case of l1527 , hogerheijde & sandell ( 2000 ) used a two gaussian model to separate the 450 and 850 @xmath0 m emission of an inner compact core immediately surrounding the embedded yso from the more extended emission . their table 3 shows a deconvolved fwhm compact core size of 10@xmath8 @xmath7 5@xmath8 ( 1400 au @xmath7 700 au ) with a position angle of the major axis @xmath4 30@xmath5 east of north . however , the interferometric c@xmath34o map of this core by ohashi et al . ( 1997 ; see figure 5a ) , made with slightly higher spatial resolution and with the advantage that the extended emission is automatically removed , implies the major axis of the core is oriented north - south , but that the outer regions of the flattened core are distorted possibly by the bipolar outflow thus giving an impression of a tilt away from the north - south orientation when mapped with slightly poorer resolution and extended emission confusion . the bipolar outflow measured by hogerheijde et al . ( 1998 ) in @xmath35co and by zhou , evans & wang ( 1996 ) in @xmath36co has a position angle of 90@xmath5 east of north . thus l1527 is observed to have a flattened core on the scale of a few thousand au , consistent with an edge - on pseudo - disk with its symmetry axis aligned with the outflow axis . in the case of b335 , chandler & sargent ( 1993 ) observed a highly flattened core at 2.7 mm that is resolved only along the major axis , which has a size of @xmath37 ( 2000 au ) with a position angle of 20@xmath5 east of north ( see figure 7a ) . the bipolar outflow measured in @xmath35co by ourselves in this paper , hirano et al . ( 1988 ) and others has a position angle of 90@xmath5 east of north . thus b335 is observed to have a flattened core on the scale of a few thousand au , consistent with an edge - on pseudo - disk with its symmetry axis slightly tilted relative to the outflow axis by about 20@xmath5 . the core emission from ic 348-smm2 has not previously been observed in detail . we use our data to study the morphology of the core surrounding this source . figure 3a shows the underlying 350 @xmath0 m continuum emission intensity from ic348-smm2 which we measured with sharp . the 50% 350 @xmath0 m contour of this source has a major axis with a position angle of 60@xmath5 east of north . the diameter of this major axis is 27@xmath8 , compared to the perpendicular minor axis diameter of 20@xmath8 . however , the 50% contour probably does not represent the fwhm of the core surrounding ic348-smm2 . a more realistic estimation of the fwhm of the core would be the 65% contour , which represents the 50% contour level of the core after the subtraction of a more extended cloud background set at about the 30% contour level in figure 3a . the 65% contour major axis has a position angle of 56@xmath5 east of north , and the diameters of the major and minor 65% contours are 21@xmath8 and 15@xmath8 , respectively . if we assume an elliptical gaussian core , we can deconvolve these fwhm estimates with a 10@xmath8 fwhm gaussian beam to get a core size of 18@xmath8 @xmath7 11@xmath8 ( 5500 au @xmath7 3300 au ) at a position angle of 56@xmath5 . the size measured here is larger than for a typical pseudo - disk quoted in the literature by about a factor of two , but at our spatial resolution we are probably not resolving the true fwhm of the pseudo - disk , we are more likely measuring the outer extended region . the bipolar outflow measured by tafalla , kumar & bachiller ( 2006 ) in @xmath35co has a position angle of 17@xmath5 west of north . thus ic348-smm2 is observed to have a flattened core on the scale of several thousand au , consistent with the outer regions of an edge - on pseudo - disk with its symmetry axis tilted within 20@xmath5 relative to the outflow . in summary , we do see evidence for pseudo - disks and a tendency for alignment , within 20@xmath5 , of the pseudo - disk symmetry axis and the outflow axis for each of the ysos in our sample . our submillimeter polarization measurements displayed in figures 2a , 3a , and 4a do not give a measure of the strength of the magnetic field , but they do give an indication of the net magnetic field direction ( lazarian 2007 ) and the level of uniformity ( hildebrand et al . 2009 ) along a given line - of - sight . the net magnetic field direction along a given line - of - sight is perpendicular to the submillimeter polarization vector measured for that line - of - sight ( lazarian 2007 ) . figures 2b , 3b , and 4b display the direction of the magnetic fields surrounding l1527 , ic348-smm2 and b335 , respectively , as determined by the polarimetry @xmath38-vectors rotated by 90 degrees . we compare these field directions with our observations of the cores density structure ( i.e. , pseudo - disk ) and velocity structure ( i.e. , outflows , infall radii ) . we note , however , that our polarization data for the three cores likely contain contributions from both the magnetic field associated with the core under study as well as that associated with the larger surrounding cloud . to see this , recall ( section 1 ) that the maps shown by kirk et al . ( 2006 ) reveal large changes in field direction as one moves away from the center of a core into the surrounding cloud material ; and note that these changes occur at positions located about one arcminute ( @xmath48000 au ) from the centers of the kirk et al . ( 2006 ) cores , while our own maps extend out to distances of 7000 - 15,000 au from the central yso . thus a significant fraction of the emission we have observed may originate from the cloud at large , not the core under study . one way to account for such contamination is to flag polarization measurements made at positions having lower flux density . such measurements could be contaminated to a large degree by line - of - sight emission associated with the larger parent cloud . the fluxes we measure at the edges of our three maps range from @xmath45% to @xmath440% of the respective peak flux ( see contours in figures 2a , 3a , and 4a ) . we have chosen to set a `` contamination threshold '' at 25% of peak flux ; we will consider polarization measurements obtained at positions having flux below this threshold to be at risk of large amounts of contamination by polarized signals originating in the larger cloud . this choice is somewhat arbitrary , but it has the benefit of flagging polarization measurements coming from regions of very low flux while preserving most of our measurements ; we flag as unreliable just four measurements out of 22 . all of our @xmath39-vectors shown in figures 2b , 3b , and 4b are drawn with the same length and thickness , and those associated with total intensity levels below our contamination threshold are shown using dashed lines . figures 2b and 4b each contain one such suspect @xmath39-vector , and figure 3b contains two . also shown in figures 2b , 3b , and 4b are the bipolar outflow morphologies in these sources ( see the figure captions for references ) . in addition , these figures show circles depicting the measured outer limits of the infall regions in l1527 and b335 based on inverse p - cygni line profiles ( see captions for references ) . no measurement of the infall radius for ic348-smm2 has been made to date , but such an infall region is likely to have a radius range of 20@xmath8 to 30@xmath8 for an infall age similar to the ages of l1527 and b335 ( see table 2 ) . below we will assume an infall radius of 25@xmath8 when we compare our observational results to models . the models we will compare our data to are those of asl03a and als03b . these models consist of self - similar , self - gravitating , singular , isothermal toroids with various amounts of rotation and magnetization . the rotation speeds of the cores range from 0 to 0.5 times their thermal sound speeds and the magnetic - flux - to - mass - ratios of the cores range from from 0 to 0.5 . all models are supercritical in order for collapse to occur without external pressure . presumably the collapse phase occurs after ambipolar diffusion has occurred , producing the supercritical state in the core . als03b note , however , that even in the relatively weakened state of the fields , these fields are responsible for the formation of pseudo - disks , considerable transport of angular momentum , and the resulting size of the centrifugally supported keplerian disk during the collapse phase , and so can not be ignored . in figures 5 , 6 , and 7 we compare our data on l1527 , ic348-smm2 , and b335 , respectively , to the model displayed in figure 8c of als03b , a model with intermediate rotation and magnetic field strengths . it should be noted here that , except for the case where there is no magnetic field to flatten the core and provide polarization , the spatial resolution of our data precludes us from discerning between the various models presented in asl03a and als03b ( see figures 7 and 8 in als03b ) . als03b show ( their figure 7 ) that rotation has only a minor effect on the gas and magnetic field geometry at the spatial scales we are measuring here . however , the aim of our current study is not to test the finer points of magnetic regulated collapse models , but the gross predictions represented in figure 1 and evidence for magnetic field pinches . in addition to our spatial resolution constraints , it is important to bear in mind that our results represent an integration of polarizations along each line of sight , whereas the magnetic field given in figure 8c of als03b represents only the cross - section of the poloidal field on the plane of the sky for an edge - on pseudo - disk . if this cross - section were rotated round the symmetry axis of the pseudo - disk , an integration along a line - of - sight would result in a weakening of the pinch geometry to a more uniform field aligned with the symmetry axis . scrutiny of the results displayed in figures 2 & 3 reveals that the field structures in l1527 and ic348-smm2 are generally consistent with the magnetically regulated dynamical collapse models cited in the introduction in that they show : ( 1 ) pinched field line vectors on the scale of the measured or inferred infall regions for these cores ; and ( 2 ) field line vectors ( with a few exceptions discussed below ) that are basically aligned with the bipolar outflows ( once the distortion of a pinch is subtracted by eye using figures 5 & 6 ) . the exceptions mentioned in ( 2 ) are the three polarization vectors in the low - flux region to the south of ic348-smm2 , which imply an east - west magnetic field , and a vector immediately east of the emission peak of l1527 . a possible explanation for this latter vector is given later in this section . however , the east - west field lines in ic348-smm2 can not be explained in the context of a magnetically regulated model . it is possible these vectors may not be associated with the core of ic348-smm2 , since two of the three reside in a region which has emission less than 25% of the peak emission for the source , and the third resides in a region with emission that is 25.5% of the peak . although in general the polarization results agree with magnetically regulated models , in detail we see significant discrepancies beyond the exceptions mentioned above . the scale of the pinched structure in magnetically regulated models depends on the size ( i.e. , age ) of the infall region . outside the infall region , the magnetic fields should be uniform or nearly uniform , depending on the details of the pre - collapse , quasi - static contraction of the core . only within the infall regions should pinches of the field lines be significant . figure 6 shows how our @xmath39-vectors in ic348-smm2 ( with our assumed infall radius and ignoring the three @xmath39-vector exceptions to the south ) show a remarkable agreement with figure 8c from als03b if the symmetry axis of the pseudo - disk has a 17@xmath40 tilt with respect to the outflow as we measure for this source . the agreement of the predicted field geometry from the als03b model with our measured vector position angles is somewhat surprising since als03b figure 8c gives a cross - section of the poloidal field on the plane of the sky for an edge - on pseudo - disk , while our measurements are an integration along the line of sight . such an integration should smooth out the pinched structure to some degree . the fit by eye between the model and our measurements is not as good if the pseudo - disk symmetry axis in the model is instead aligned with the outflow axis . the fact that the magnetic field fit is better when the model is aligned to the measured pseudo - disk axis , rather than the outflow axis , is consistent with the hypothesis that the pinched magnetic field structure and the pseudo - disk are both products of magnetized contraction and collapse , whereas the outflow is probably a product of rotation . in ic348-smm2 , the rotation axis may not be exactly aligned with the overall magnetic field axis for the core . figure 5 shows that in l1527 our @xmath39-vectors broadly agree with a pinched magnetic field structure aligned with the pseudo - disk axis as shown in figure 8c of als03b , but the figure also shows that there are significant distortions beyond the uncertainties ( @xmath4111@xmath5 ) in the angles of the vectors . we see these distortions close to the edge and in one case beyond the infall boundary . a possible explanation for this additional distortion could be the bipolar outflow in this source which overlaps a significant portion of the region containing our measured vectors ( see figure 2 ) . the effects of bipolar outflows are not included in the models to which we are comparing our results ( asl03a ; als03b ; galli & shu 1993a , b ) . the bipolar outflow in l1527 may also be responsible for the exceptional vector identified previously , which lies immediately east of the l1527 emission peak . this vector implies a magnetic field direction that is almost perpendicular to the outflow . in this scenario , the outflow pushes core material , and therefore also core magnetic field lines , into two polar cones surrounding the bipolar outflow near the emission peak in l1527 , thereby giving rise to both the additional distortions near the edge of the infall boundary as well as the exceptional vector just east of the peak . evidence for conical cavities such as would be required in this scenario has been obtained via mid - ir scattered light observations by tobin et al . ( 2008 ) as well as interferometric measurements of hco@xmath42 by hogerheijde et al . the latter is shown in figure 2b . in addition , the submillimeter maps of hogerheijde & sandell ( 2000 ) and our 350 @xmath0 m map ( figure 2a ) show evidence of an x - like structure in the extended background emission about l1527 that overlaps the observed outflow in that region , implying that significant submillimeter emission ( and any polarization of such emission ) must be coming from the surface of the outflow conical cavities . could the bipolar outflow observed in l1527 have enough energy to distort the magnetic field structure linked to the gas it entrains ? the diagrams in figure 8 in als03b show the pinched magnetic field structures for a number of rotating , dynamically - collapsing toroids of different initial cloud magnetic field strengths . these diagrams also show the contours of @xmath43 ( the ratio of the thermal to magnetic energy density ) . close to the axis of symmetry @xmath43 is much less than one ( @xmath4 0.1 ) , further from the axis of symmetry @xmath43 increases to values above one . these values of @xmath43 can be used together with estimates of the thermal energy within a core to determine the energy density required of an outflow to distort the magnetic field within that core . the core thermal energy density for l1527 can be approximated by @xmath44 , where @xmath45 is given by the mass density of the 10@xmath46 @xmath3 molecular gas measured for the l1527 core about 10@xmath8 away from the center ( hogerheijde & sandell 2000 ) and c@xmath47 is the sound speed which is approximately 0.25kms@xmath32 ( zhou , evans & wang 1996 ) . the lower limit to the outflow energy density in l1527 is expressed as @xmath48 divided by the observed volume of the outflow lobes , where the measured values for @xmath49 ( the mass entrained in the outflows ) and @xmath50 ( the velocity of the entrained gas in the outflows ) are given in table 2 from hogerheijde et al . the volume of both outflow lobes can be approximated by @xmath51 , where @xmath52 , the extent of one lobe in l1527 , is given in table 2 and @xmath53 0.14 sr based on figure 6 in hogerheijde et al . ( 1998 ) showing the extent of the @xmath35co outflow in l1527 . the above imply that the energy density in the outflow is greater than the thermal energy density in the region 10@xmath8 from the peak of l1527 by about a factor of 100 . since in the als03b models the magnetic energy density is less than the thermal energy density away from the axis of symmetry , this implies that the outflow does have the energy required to distort the magnetic field where it disturbs the gas in the core away from the axis of symmetry . indeed , with the energy density of the outflow observed , even the magnetic field very close to the axis of symmetry could be distorted . the ratio of the outflow energy density to core thermal energy density in ic348-smm2 is similar to the ratio in l1527 , if one assumes similar gas densities and sound speeds for ic348-smm2 as measured for l1527 and the outflow parameters in table 2 for ic348-smm2 . but the outflow in ic348-smm2 does not overlap with our measured polarization vectors to a very great degree , so the outflow can not affect the alignment of the magnetic fields which are inferred by these polarization measurements . where there is overlap , the distortion would be minimal for field lines already parallel to the axis of the outflow if the opening angle of the outflow at that location is small as it is for ic348-smm2 ( @xmath4115@xmath5 ; see figure 3 ) . in summary , if one assumes that the low - flux exceptions in ic348-smm2 are not part of the core and one includes the effects of bipolar outflows on field alignment , then our observations of l1527 and ic348-smm2 are consistent with magnetically regulated models . at first glance , our two polarization vectors in b335 imply a magnetic field more perpendicular than parallel to the outflow axis . however , our two inferred @xmath39-vectors are not too different from what would be expected in the region just south - east of the b335 center based on figure 8c of als03b when this figure is aligned with the minor axis of the flattened core as measured by chandler & sargent ( 1993 ) ( i.e. , the symmetry axis of the inferred pseudo - disk ) . in this region of the model , the field lines are pinched towards the axis of symmetry ( see figure 7 ) . figure 7 also shows the @xmath39-vectors inferred from the 850 @xmath0 m polarimetry measurements of wlh03 as well as those inferred from our 350 @xmath0 m polarimetry measurements . our results are broadly consistent with the results of wlh03 , in that they agree with four of the six vectors in the south - east quadrant region of the b335 core , but not , however , with the one vector with which our measurements most nearly overlap . the 20 @xmath39-vectors of wlh03 and our 2 all lie inside the infall region outlined in figures 4 and 7 . the wlh03 results imply on average a more n - s magnetic field structure within b335 , but there is considerable distortion evident in the field lines implied by the 20 vectors , since the standard deviation of the position angles of the @xmath39-vectors is about three times the average measurement error for each vector . wlh03 concluded that the average field they measure in b335 is the direction of the field in the core when it collapsed , and that the flattened core seen in b335 is prolate ( rather than oblate ) with its symmetry axis parallel to the magnetic field . if this is the case , then b335 presents a counter example to the results obtained by tassis et al . ( 2009 ) who concluded the core model that best fits their sample of 24 high - mass cloud cores is an oblate core with the mean magnetic field more aligned with the short axis than the long axis . alternatively , the wlh03 data and ours could imply a more toroidal field in b335 than poloidal . but if this were the case then this is inconsistent with the model of als03b which includes rotation . in this model , only a small volume of a dynamically collapsing core contains twisted ( i.e. , toroidal ) magnetic fields . outside this small region the collapsing flow has yet to be spun up , and inside this region @xmath54 so the field lines are rigid . if b335 does contain a toroidal field configuration , then the als03b model fails to describe it ; there appears to be more twisting of the original poloidal magnetic field lines than can be explained in their model . yet another interpretation of the data , and the one we advocate in this paper , is that b335 is an extreme example of what is happening in l1527 ; the outflow in b335 has distorted the field lines in the core of b335 , either directly or indirectly by exciting more turbulence within the core , so the magnetic field seems to align n - s on average . could the outflow in b335 cause large field distortions in the core ? if we use the same analysis we used for l1527 , but with a core thermal speed of 0.23kms@xmath32 ( zhou et al.1993 ) , a core density at about 10@xmath8 from the center of @xmath4 10@xmath55 @xmath3 ( zhou et al . 1990 ; harvey et al . 2003 ) , the b335 values for the outflow parameters given in table 2 , and @xmath56 0.6 sr ( figure 2 in hirano et al . 1988 ) , then we find that the outflow energy density is about a factor of 6 higher than the core thermal energy density ( 10@xmath8 from the peak ) in b335 . our own outflow data shown in figure 4b give a total kinetic energy for the outflow within the region mapped ( i.e. , @xmath57 , summed over the velocity channels in the ranges 5.4 - 8.0 kms@xmath32 and 8.6 - 11.2 kms@xmath32 ) of 1.8 @xmath7 10@xmath58 erg , once a correction for the 10@xmath5 inclination of the outflow to the plane of the sky has been made . dividing this by the volume of the outflow as defined in figure 4b , we get a kinetic energy density of @xmath4 10@xmath59 erg @xmath3 , which is a factor of 3 higher than the core thermal energy density . this outflow energy density although not enough to distort the magnetic field close to the axis of symmetry of the pseudo - disk in b335 , is enough to distort the field further out . the outflow cavity outlined by the co observations in figure 4b is coincident with the region within the b335 core that most likely would have a magnetic field energy density that is greater than the outflow energy density ; beyond this outflow cavity the reverse is likely to be true . the outflow cavity is a region in which the gas and dust densities are low , and so it is a region of low 350 @xmath0 m emission along our line of sight . our polarization measurements are thus weighted away from the outflow cavity region towards those regions along our line of sight where the energy density of the outflow could distort the magnetic field structure . but why is this distortion observed to be so large in b335 compared to the distortion in l1527 ? the difference between b335 and the other two ysos presented here is that the bipolar outflow in b335 is much larger in length , width and apparent age than the outflows in l1527 and ic348-smm2 ( table 2 ) . therefore , the field lines within the b335 core could be highly distorted because the outflow has had time to plow through a greater portion of the core or excite greater gas turbulence in the core . in order to give some quantitative assessment of the magnetic field scenarios discussed above , and how they compare to other possible configurations , we carried out reduced chi - squared tests of our data using a number of different theoretical magnetic field configurations for each source . for each source we compared the @xmath39-vectors implied by our data to : ( 1 ) a uniform magnetic field model where the angle of the field is aligned with the mean @xmath39-vector angle implied by our data for that source ; ( 2 ) a uniform magnetic field model where the angle of the field is aligned with the outflow ; ( 3 ) a uniform magnetic field model where the angle of the field is aligned with the symmetry axis of the observed pseudo - disk for that source ; ( 4 ) a uniform magnetic field model where the angle of the field is aligned with the major axis of the observed flattened core ( or pseudo - disk ) for that source ; and ( 5 ) a pinched magnetic field aligned with the symmetry axis of the observed pseudo - disk as presented in figure 8c of als03b . note that model ( 4 ) also corresponds to the case where the magnetic field is toroidal . table 3 summarizes our results . each number in the table represents the reduced chi - squared value , @xmath60 , for the data specified for that column against the model for that row . the reduced chi - squared is calculated as @xmath61 where @xmath62 are the data representing the angles of the @xmath39-vector at various locations , i ; @xmath63 is the uncertainty in each data angle ; @xmath64 is the angle of the magnetic field at the location of each data point for a particular model ; and @xmath65 is the number of degrees of freedom for the data set . for these calculations , the values of @xmath66 were constrained to be @xmath6 90@xmath67 since our @xmath39-vectors have been derived from our polarization @xmath38-vectors which are invariant under 180 degree rotations . for cases where @xmath68 , a value of @xmath69| \le 90\arcdeg$ ] was substituted . for b335 , we carried out the same calculations using the data of wlh03 based on their figure 1 . @xmath70 should be close to 1 for a good fit . an inspection of table 3 shows that @xmath60 is close to this value only for the als03b model for ic348-smm2 and b335 when we consider only our data minus the exceptions discussed in section 3.2.1 . however the fits are not good for this model if we include the exceptions for ic348-smm2 , and if we include the wlh03 data for b335 . this is in agreement with our qualitative assessment above . for l1527 , although no model gives a good fit to our data , the two most favored are the als03b model and the uniform field aligned with the mean @xmath39-vector of our data for l1527 ( i.e. , 60@xmath67 east of north ) . our data do not differentiate between these two models , since our vectors lie mostly in two diagonal quadrants . if we had more data in the other two quadrants , a @xmath70 test would be able to differentiate between these two models . as it is , the @xmath70 results re - enforce our qualitative assessment above that although our data agree with the als03b model in a broad sense , there are enough deviations to it to require an explanation ; possibly a disturbance to the magnetic field structure due to the outflow in l1527 . further , based on table 3 , our data for l1527 , ic348-smm2 , and b335 clearly favor the als03b model over the toroidal " models , and the als03b model is slightly more favored than the models with a uniform field aligned with the outflow and a uniform field aligned with the pseudo - disk symmetry axis . this implies there is some evidence for a pinch in the magnetic field configuration for these sources . however , the data of wlh03 for b335 , in contradiction , favor the model with a uniform field aligned with the major axis of the flattened core over all the other models , although the fit is not good for this model , thus implying some other underlying magnetic field configuration must be present - possibly a configuration disrupted by the outflow in b335 . in summary , the chi - squared results in table 3 support the qualitative discussions we presented in sections 3.2.1 and 3.2.2 . we also compared our data to a random magnetic field model for each source . we did this by calculating the root - mean - square ( rms ) of the differences between the directions of the @xmath39-vectors at points lying spatially adjacent to one another in each of our sources . for l1527 we obtained 11 pairs of measurements separated by 10@xmath8 or @xmath71 ( see figure 2 ) and for ic348-smm2 we obtained 10 such pairs ( see figure 3 ) for our rms calculations . points separated by @xmath72 ( i.e. , one beam diameter ) or greater represent independent measurements . for l1527 we calculated an rms value for these differences in adjacent @xmath39-vector directions to be 19@xmath67 , and for ic348-smm2 a value of 15@xmath67 . we then carried out the same calculation on 10 randomly generated numbers ranging between -90 and 90 , and did this 50 times . the resulting rms values formed a peaked gaussian distribution with a mean of 51 and a dispersion @xmath73 . ( this mean is close to the theoretical rms of @xmath4 52 for randomly selected numbers between -90 and 90 . ) thus our rms values for l1527 and ic348-smm2 lie 5.5@xmath33 and 6@xmath33 , respectively , below the expected rms for randomly distributed @xmath39-vectors , implying that the magnetic fields in l1527 and in ic348-smm2 are not random in nature . we carried out the same rms calculation on two different sets of 13 pairs of adjacent @xmath39-vector measurements made by wlh03 in b335 ( see figure 7 ) , resulting in rms values of 42@xmath67 and 34@xmath67 . both values are much larger than the average uncertainties ( @xmath74 ) for the measurements reported in wlh03 . these rms values imply that the magnetic field is possibly more randomized in b335 than in the other sources . this latter point is consistent with our suggestion that the magnetic field in b335 is more distorted by its outflow than are the magnetic fields in l1527 and ic348-smm2 . we have used the sharp polarimeter on the cso to obtain 350 @xmath0 m intensity maps and polarization measurements on l1527 , ic348-smm2 , and b335 to test whether or not magnetically regulated models for low - mass star formation are consistent with observations of sources which should have little distortions in their various geometries due to projection effects ( i.e. , sources with bipolar outflows which lie close to the plane - of - the - sky ) . ( 1 ) our data for ic348-smm2 combined with the data of others for l1527 and b335 show flattened cores consistent with edge - on pseudo - disks having symmetry axes that are nearly , but in two cases not exactly , parallel to the bipolar outflows in these sources . \(2 ) there is evidence that the sources l1527 and ic348-smm2 each contains a pinched magnetic field structure with its symmetry axis approximately aligned with the symmetry axis of the inferred embedded pseudo - disk in each . the evidence is strong ( i.e. , the goodness - of - fit to the data is good ) for ic348-smm2 , if certain low - flux polarimetry measurements , which could be associated more with the background cloud than the core , are ignored . in ic348-smm2 , where the outflow and pseudo - disk axes are not exactly aligned , the pinched magnetic field structure fits the data better when it is aligned with the symmetry axis of the pseudo - disk rather than with the outflow axis . this is consistent with the hypothesis that the pinched magnetic field structure and the pseudo - disk are both products of magnetized contraction and collapse , whereas the outflow is only indirectly related to the core magnetic field structure inasmuch as that magnetic field structure has influenced the orientation of the rotation axis of the core . \(3 ) in l1527 , however , the magnetic field structure shows considerable distortion from an ideal pinched field line structure given the measured infall radius of this source . our hypothesis is that this distortion is caused by the bipolar outflow in l1527 . we show that the outflow has sufficient energy density to distort the magnetic field structure in the core of this source . this distortion is not seen in ic348-smm2 , because the only inferred @xmath39-vectors that overlap with the outflow in this source are vectors that would not have been altered by the outflow . \(4 ) the magnetic field structure observed in the b335 core is not aligned with the outflow axis of this source , but our two @xmath39-vectors are consistent with models with pinched field lines through a pseudo - disk , if the pseudo - disk is tilted with respect to the outflow axis by 20@xmath67 ( i.e. , consistent with the pseudo - disk observed for b335 by chandler & sargent ( 1993 ) ) . however , if we combine our data with those of wlh03 , the fit to these pinched field line models is not good . our explanation is that b335 is an extreme example of the bipolar outflow driven field distortion that we are seeing in l1527 . the main difference between l1527 and b335 is that the outflow in b335 is much larger in extent than the one in l1527 , is assumed therefore to be much older , and so has had time to cause a greater degree of distortion of the core magnetic field lines . this is not the interpretation given by wlh03 of their b335 observations . they interpreted their results to imply a near uniform field lying perpendicular to the outflow in this source - in contradiction to the predictions of magnetically regulated collapse as summarized in figure 1 . ( 5 ) more core magnetic field structures need to be mapped to elucidate the overall dynamical collapse story , given the considerable variation in core structures observed in our sample of just three . in short , the gross predictions of the magnetically regulated models ( i.e. , as summarized in figure 1 ) need to be tested further . we are grateful to the caltech submillimeter observatory tac , management , and staff for making this study possible over the many observing runs that it has consumed . we thank darren dowell , megan krejny , woojin kwon , and hiroko shinnaga for help with the observations , and ron taam for valuable discussions . we are grateful to the national science foundation for supporting the caltech submillimeter observatory via grant ast-0838261 , and for supporting the operation of sharp via grant ast-0909030 to northwestern university . astronomical research by p.f.g . and n.l.c . is supported by the jet propulsion laboratory , california institute of technology . lrrrrrr l1527 & 38.0 & 9.5 & 5.6 & 2.1 & -51.0 & 9.5 + & 28.5 & 19.0 & 3.3 & 1.6 & -56.1 & 12.0 + & 19.0 & 9.5 & 2.5 & 0.9 & -63.7 & 9.7 + & 19.0 & 19.0 & 3.5 & 1.1 & -41.7 & 10.1 + & 9.5 & -28.5 & 5.8 & 1.8 & 18.2 & 9.0 + & 9.5 & 0.0 & 1.0 & 0.5 & -66.4 & 12.3 + & 9.5 & 9.5 & 1.6 & 0.5 & -36.5 & 9.6 + & -9.5 & -19.0 & 2.8 & 1.3 & 3.1 & 10.0 + & -9.5 & -9.5 & 1.8 & 0.7 & -13.2 & 10.3 + & -19.0 & -9.5 & 2.2 & 1.2 & -28.4 & 14.5 + ic348-smm2&9.5 & 19.0 & 5.7 & 2.1 & 84.8&10.5 + & 0.0 & 19.0 & 6.3 & 2.4 & 81.6&9.4 + & -9.5 & -28.5 & 9.2 & 3.7 & -9.8&10.0 + & -9.5 & 19.0 & 5.1 & 2.8 & 61.9&12.8 + & -19.0 & -28.5 & 9.5 & 4.4 & -4.7&11.0 + & -19.0 & -19.0 & 5.8 & 3.0 & 4.0&11.8 + & -19.0 & 19.0 & 7.9 & 3.4 & 54.4&13.3 + & -28.5 & -9.5 & 5.9 & 2.9 & 33.7&11.4 + & -28.5 & 0.0 & 9.4 & 3.1 & 32.3&8.2 + & -28.5 & 9.5 & 8.0 & 4.1 & 26.3&12.0 + b335 & 9.5 & -19.0 & 3.2 & 1.5 & 56.5&11.9 + & 9.5 & -9.5 & 1.3 & 0.7&58.8&14.6 + lrrr ra ( 2000 ) & 04:39:53.9 & 03:43:56.9 & 19:37:01.03 + dec ( 2000 ) & 26:03:11 & 32:03:06 & 07:34:11 + distance ( pc ) & 140 & 300 & 250 + infall radius , r@xmath75 ( pc ) & 0.026 ( 38 ) & & 0.04 ( 33 ) + sound speed , c@xmath47 ( kms@xmath32 ) & 0.25 & & 0.23 + infall age ( @xmath4 r@xmath75/c@xmath47 ) ( yr ) & @xmath76 & & @xmath77 + @xmath35co outflow extent , r@xmath78 ( pc ) & 0.095 ( 140 ) & 0.14 ( 95 ) & 0.3 ( 250 ) + pa of outflow axis e from n & 90@xmath40 & @xmath79 & 90@xmath40 + outflow mass , m@xmath78 ( m@xmath80 ) & 0.18 & 0.033 & 0.44 + characistic outflow velocity , & & & + v@xmath78 ( kms@xmath32 ) & 20 & 54 & 13 + outflow age ( @xmath81 r@xmath78/v@xmath78 ) ( yr ) & @xmath82 & @xmath83 & @xmath84 + lcccccc @xmath64 = mean @xmath62 & 8.3 & 8.2 & 10.4 & 5.5 & 0.02 & 13.7 + @xmath64= @xmath85 & 16.4 & 15.0 & 20.9 & 8.1 & 19.4 & 36.9 + @xmath64= @xmath86 & 16.4 & 15.0 & 12.7 & 4.8 & 8.2 & 38.5 + @xmath64= @xmath87 & 30.3 & 33.2 & 31.7 & 42.0 & 16.2 & 14.2 + als03b model & 9.9 & 7.8 & 14.0 & 0.4 & 0.8 & 31.5 + 10,000 au , the diameter of a pseudo - disk is @xmath42,000 au , and the diameter of a keplerian disk is @xmath4100 au . in the magnetically regulated core collapse scenario : the pseudo - disk symmetry axis is aligned with the core magnetic field ; and magnetic braking tends to align the core rotation axis with the magnetic field , but this alignment may not be exact . the pseudo - disk is a dynamically collapsing object formed by the magnetic fields , not rotation . the keplerian disk is an object formed by rotation and so its symmetry axis is aligned with the core s rotation axis , as too is the outflow axis if the outflow is driven by rotation . , width=432 ] m polarization vectors for l1527 . the thickness of each vector indicates the significance , with all vectors having significance greater than @xmath24 before bias correction ( see section 2.2 ) . note the very low polarization ( @xmath88 ) marked by a small open circle near the flux peak ( see section 2.2 ) . the contours indicate the total intensity , i , measured by sharp . the highest contour level corresponds to 90% of the peak flux , and each subsequent contour represents a decrement of 10% of the peak flux . ( b ) inferred magnetic field directions superimposed on a section of a figure from hogerheijde et al . ( 1998 ) showing @xmath35co 3 - 2 observations of the bipolar outflow ( contours ) and a hco@xmath42 map tracing gas on the cavity walls surrounding the outflow ( gray area ) . b - vectors in regions having 350 @xmath0 m flux less than 25% of the peak are shown as dashed lines . the full hogerheijde et al . figure shows the blueshifted outflow ( solid contours ) extending 2@xmath89 east and the redshifted outflow ( dashed contours ) extending 2@xmath89 west from the yso in l1527 . the large dashed circles in ( a ) and ( b ) show the extent of the measured infall region for this object ( myers et al . 1995 ; zhou , evans & wang 1996 ) . , width=576 ]

we analyze the idea that nuclear activity , either agn or star formation , can be triggered by interactions , studying the percentage of active , hii and quiescent galaxies with companions . our sample was selected from the palomar survey , and avoids selection biases faced by previous studies . this sample was split into 5 different groups , seyfert galaxies , liners , transition galaxies , hii galaxies and absorption line galaxies . the comparison between the local galaxy density distributions of the different groups showed that in most cases there is no statistically significant difference among galaxies of different activity types , with the exception that absorption line galaxies are seen in higher density environments , since most of them are in the virgo cluster . the comparison of the percentage of galaxies with nearby companions showed that there is a higher percentage of liners , transition , and absorption line galaxies with companions than seyferts and hii galaxies . however , we find that when we consider only galaxies of similar morphological types ( ellipticals or spirals ) , there is no difference in the percentage of galaxies with companions among different activity types , indicating that the former result was due to the morphology - density effect . also , only small differences are found when we consider galaxies with similar h luminosities . the comparison between hii galaxies of different h luminosities shows that there is a significantly higher percentage of galaxies with companions among hii galaxies with l(h) erg s , than among those with l(h) erg s , indicating that interactions increase the amount of circumnuclear star formation , in agreement with previous results . the fact that we find that galaxies of different activity types have the same percentage of companions , suggests that interactions between galaxies is not a necessary condition to trigger the nuclear activity in agns . we compare our results with previous ones and discuss their implications .

astro-ph0110235

two of the major concerns in the study of agns are the origin of the gas which fuels the nuclear black hole and the mechanisms responsible for making the gas lose angular momentum and move from galactic scales down to the inner @xmath41 pc region of the galaxy . in spiral galaxies , gas in their disks can be a natural source to fuel the agn . several mechanisms have been suggested to explain how it is possible to transport gas from the disk of a spiral galaxy to its nucleus , like interactions ( gunn 1979 ; noguchi 1987,1988 ; hernquist 1989 ; barnes & hernquist 1992 ) , or bars ( schwartz 1981 ; norman 1987 ; shlosman , frank & begelman 1990 ) . in the case of elliptical galaxies the situation is more complicated , because they only have small quantities of gas , which implies that the gas which feeds their nuclei most likely has an origin external from the galaxy . reviews about this subject , with detailed explanations of the possible mechanisms responsible for feeding the nucleus are given by shlosman et al . ( 1990 ) and combes ( 2001 ) . the influence of interactions on the fueling of the nucleus of agns as well as hii galaxies has been the topic of several studies . from the theoretical point of view and n - body simulations , noguchi ( 1987,1988 ) , hernquist ( 1989 ) , barnes & hernquist ( 1992 ) , byrd et al . ( 1986 ) , byrd , sundelius & valtonen ( 1987 ) , lin , pringle & rees ( 1988 ) , hernquist & mihos ( 1995 ) , mihos & hernquist ( 1994 ) , taniguchi & wada ( 1996 ) , among others , have shown that interactions between galaxies , mergers , and minor mergers can be responsible for bringing gas from the disk to the nuclear region , where it is shocked and compressed , producing a starburst and subsequently feeding the nucleus . this idea is qualitatively confirmed with observations , but the models still have to explain fine details . it is a well established fact that luminous galaxies , in particular luminous and ultraluminous infrared galaxies are closely related to interacting systems ( sanders et al . 1988 ; lawrence et al . 1989 ; leech et al . 1994 ; gallimore & keel 1993 ) . it has also been shown by keel et al . ( 1985 ) , wright et al . ( 1988 ) and kennicutt et al . ( 1987 ) that interacting galaxies have a higher level of h@xmath0 , infrared and radio emission , as well as an increased nuclear star formation rate . however , in the case of lower luminosity sources , like seyfert galaxies , the situation is not so clear . there is no consensus on how important is the effect of interactions in the feeding of their nuclei , and if these galaxies have an excess of companions . stauffer ( 1982 ) was one of the first to point out that seyferts usually occur in groups . dahari ( 1984 , 1985 ) made a large compilation of possible companion galaxies around seyferts . he found that these galaxies have an excess of companions relative to normal galaxies . although this result have been confirmed by several other studies ( rafanelli , violato & baruffolo 1995 ; laurikainen et al . 1994 ; mackenty 1989 , 1990 ) , there have been some claims to the contrary . fuentes - williams & stocke ( 1988 ) , bushouse ( 1986 , 1987 ) , ulvestad & wilson ( 1984 , 1985 ) and de robertis , yee & hayhoe ( 1998 ) did not find a detectable excess of companions around seyferts . other studies of seyfert galaxies have shown yet another result , where seyfert 2 s have a larger number of companions when compared to normal galaxies , while seyfert 1 s do not ( laurikainen & salo 1995 and dultzin - hacyan et al . similarly , in a study of galaxies with activity types other than seyfert , pastoriza et al . ( 1994 ) showed that galaxies in high and low density media have similar stellar populations , while telles & maddox ( 2000 ) found that low redshift hii galaxies are , on average , found in slightly lower density environments than normal field galaxies . taylor et al . ( 1995 ) found a different result , but according to telles & maddox ( 2000 ) this can be due to the fact that the progenitors of hii galaxies are different from those of field galaxies . the different results obtained by different papers are most likely due to the way they selected their samples and control samples , as pointed out by heckman ( 1990 ) and osterbrock ( 1993 ) . in order to do a proper comparison between the percentage of active , normal and hii galaxies with companions , it is necessary to have a sample of active galaxies and a control sample with matching host galaxy properties , which may have been a problem in the earlier studies . another problem with these studies is that they usually did not have information about the redshifts of the companion galaxies . in most of the cases , the only information available was the magnitudes of the possible companions and their projected distances to the galaxies being studied . this means that these companions may not necessarily be physically associated . the way around this problem was to apply statistical corrections for the number of background and foreground objects around the galaxies being studied . here we try to solve these problems by studying the percentage of companion galaxies in the palomar survey ( ho , filippenko & sargent 1997a ) . we will address problems related to the selection of the sample , control sample , and the availability of redshifts for the companions galaxies . we will also be able to study the percentage of companion galaxies in galaxies of different activity types , not only seyferts and normal galaxies as done in previous papers . this paper is organized in the following way . section 2 presents the sample and corresponding subsamples , while section 3 gives the techniques and data used . section 4 presents the comparison of local galaxy densities for galaxies of different activity types , and section 5 shows the comparison of the percentage of companions in these galaxies as a function of different properties ( e.g. morphological type , h@xmath0 luminosity ) . section 6 presents the discussion of the results and the summary . as pointed out in the previous section , one of the main problems for the previous works in this subject was the selection of the active and corresponding control samples of galaxies . here we solve this problem by using the galaxies in the palomar survey ( ho et al . this sample is ideal for this kind of study , because it contains the galaxies brighter than b@xmath5 mag in the northern hemisphere . the relative advantages of this sample relative to others is discussed by ho & ulvestad ( 2001 ) . they also point out selection biases which affect other samples commonly used in the literature . here we summarize the advantages of the palomar survey . the most important characteristic of this sample is the fact that it provides homogeneous high - quality spectroscopic measurements of emission line fluxes , as well as activity classification of the galaxies . another of its advantages is the fact that it covers a large range of agn and hii luminosities , as well as a large range of host galaxy parameters , like morphological types and absolute magnitudes , besides the fact of being large enough to ensure a significant comparison between galaxies of different activity types as a function of these parameters . since we are going to compare the percentage of galaxies with companions among different activity types , not only seyferts and normal galaxies , this sample is ideal for this work , because it includes both the agn and control sample in itself , allowing us to determine the frequency of galaxies with companions in the local universe , avoiding the selection effects faced by previous works . the initial sample studied by ho et al . ( 1997a ) contained 486 galaxies . however , we decided to exclude local group galaxies according the list from mateo ( 1998 ) , because it includes a large number of low luminosity dwarf spheroidals . galaxies with b@xmath6 mag , which were included in their survey for historical reasons , or because they were interesting objects , have also been excluded . this gives a total number of 451 galaxies , divided into 46 seyferts , 193 hii galaxies , 88 liners , 63 transition objects and 61 absorption line galaxies . transition objects are galaxies which present emission line ratios intermediate between those of hii galaxies and liners , and absorption line galaxies are those without any detectable emission lines . other galaxies in this sample had emission line ratios which resulted in different classifications , depending on the emission line ratios used . they were classified by ho et al . ( 1997a ) as l2/s2 , or s2/h for example , and in these cases we used the first of the two classifications . given the small number of seyferts and liners with broad emission lines in the sample , we do not attempt to analyze these galaxies separately . nevertheless , we point out that schmitt et al . ( 2001 ) have shown , using a well defined sample of seyfert galaxies , that there is no significant difference in the percentage of seyfert 1 s and seyfert 2 s with companion galaxies . a statistical comparison of the emission line and host galaxy properties of these galaxies as a function of activity type was presented by ho , filippenko & sargent ( 1997b ) . it should be noticed that most of the galaxies classified as hii nuclei present only low levels of current star formation , with h@xmath0 luminosities similar to those of individual giant hii regions . due to this fact , the lower luminosity hii galaxies should be considered as quiescent galaxies and not as starburst . we used two different techniques to compare the environments of the different types of galaxies . the first one used the local galaxy densities defined by tully ( 1988 ) , @xmath7 , and compared the galaxies in the different activity classes using the kolmogorov - smirnov test ( ks - test ) . @xmath7 is the density of galaxies brighter than m@xmath8 mag in the vicinity of the object of interest . this value was calculated by tully ( 1988 ) , using all galaxies in the local universe with distances smaller than 40 mpc , taking into account completeness effects . the values used in this paper were obtained from table 10 of ho et al . ( 1997a ) , and cover 92% of the sample . the second technique searched for companions around the galaxies using ned and the digitized sky survey images ( dss ) . several parameters have been used to determine if a galaxy has a companion or not . we adopt here a slightly modified version of the one used by rafanelli et al . ( 1995 ) , which was determined based on catalogs of interacting and pair galaxies . we consider a galaxy as a companion if its distance to the galaxy of interest is smaller than 5 times the diameter ( d@xmath9 ) of that galaxy , the difference in brightness between them is smaller than 3 magnitudes ( @xmath10 3 mag ) , and the difference in radial velocities is smaller than 1000 km s@xmath2 ( @xmath11 1000 km s@xmath2 ) . we will call the galaxy around which we search for companions the primary and the companion galaxy secondary . the original parameters used by rafanelli et al . ( 1995 ) considered only secondaries at a distance smaller than 3 times the diameter of the primary . we decided to increase this value to 5 times the diameter of the primary . we will show below that this different criterion does not make a significant difference in the results . we emphasize that these criteria used to search for companions around the galaxies are simply empirical , corresponding to the range of properties found in galaxy pairs . according to charlton & salpeter ( 1991 ) , pairs with velocity differences smaller than 150 km s@xmath2 are likely to be bound . however , the galaxies do not need to be bound in order to interact , the interaction can happen between galaxies which are not gravitationally bound . based on the magnitude criterion alone , we can calculate that , for an m@xmath12 primary galaxy , with m@xmath13 ( binney & merrifield 1998 ) a secondary galaxy 3 mag fainter will have m@xmath14 , a value between that of the lmc and the smc , which have m@xmath15 mag and m@xmath16 mag , respectively . the lmc and smc magnitudes were calculated using the values from mateo ( 1992 ) and the distance moduli obtained by cioni et al . this shows that the search for companions will not be biased towards bright galaxies only . the most luminous galaxy in the sample is ngc4594 , which has m@xmath17 , and the search for companions around this galaxy will consider only the galaxies with absolute magnitudes of the order of m@xmath18 . on the other hand , the faintest galaxy in the sample is ngc404 , which has m@xmath19 and companions to this galaxy could have luminosities as faint as m@xmath20 which is a value typical of dwarf spheroidal galaxies like fornax and leo ( binney & merrifield 1998 ) . most of the galaxies in the sample have m@xmath21 ( ho et al . 1997b ) , which means that our criteria will select only secondaries with absolute magnitudes similar to that of the smc . lower luminosity dwarf / spheroidal and irregular galaxies , like the ones on the local group , will be missed . it was suggested by de robertis et al . ( 1998 ) that mergers of gas - rich disk galaxies with satellite companions ( dwarf galaxies ) may be an important way to trigger the activity in seyfert galaxies . the data and criteria used in this paper does not allow us to test this hypothesis . it may be possible to do such a search in the future , using results from the sloan survey . one of the problems with previous studies was the fact that they were usually not able to obtain redshifts of the supposed companion galaxies . due to this fact they had to rely on statistical methods to determine the probability of two galaxies being companions , as well as to correct for the number of background and foreground galaxies based on galaxy number counts . since we cut our sample at b@xmath5 mag , the faintest secondary galaxies must have magnitudes b@xmath22 mag . in this way we are sure that most of the galaxies will have redshift information . redshift surveys , such as the cfa ( huchra et al . 1983 ) observed galaxies as faint as @xmath23 , while the cfa2 ( falco et al . 1999 ) is observing galaxies up to @xmath24 . approximately 90% of the sample have been covered by redshift surveys . to ensure we did not miss any galaxy without radial velocity information , for those galaxies without known companions , according to our criteria , we checked ned for possible secondary galaxies without radial velocity information around the primaries , and also visually inspected the dss images for secondary galaxies up to 3 mag fainter than the primary galaxies . we did not find any possible secondary galaxy without radial velocity information . we also checked if the search of galaxies on the dss was able to detect low surface brightness galaxies . according to krongold , dultzin - hacyan & marziani ( 2001 ) the dss images have a surface brightness limit of @xmath25 mag pixel@xmath2 , which , for pixels with a size of 1.7@xmath26 , corresponds to @xmath27 mag arcsec@xmath28 . most of the low surface brightness galaxies detected by impey et al . ( 1996 ) have b band central surface brightnesses of the order of @xmath29 mag arcsec@xmath28 or fainter , effective radii r@xmath30 and effective surface brightness @xmath31 mag arcsec@xmath28 . using the equations from impey , bothum & malin ( 1988 ) , we have that the integrated magnitude of an exponential disk galaxy can be calculated by m@xmath32 , which for a galaxy with @xmath33 mag arcsec@xmath28 and r@xmath30 corresponds to m@xmath34 . consequently , most of the low surface brightness galaxies will be fainter than b@xmath3515.5 , which is the faintest magnitude for our companion galaxies . this means that we did not miss a significant number of low surface brightness galaxies in the search for companions on the dss images . table 1 presents the galaxies for which we could find companions . this table is separated in 5 parts , according to activity type . it presents the names of the primaries , the magnitude differences between the primaries and the secondaries , the distances between them in units of the primary diameter , the moduli of the difference between the radial velocities and the names of the secondaries . we also present in this table the h@xmath0 luminosities of the primaries , whether they are ellipticals , s0 s or a later type galaxies and comment if they are members of the virgo cluster . we consider as late types , galaxies which have morphological types sa or later . this morphological separation is based on the fact that elliptical galaxies have very little or no internal gas , s0 s have some gas , while galaxies of morphological type sa and later have gas in their disks which can feed the nucleus or form stars . we point out the fact that two of the galaxies in table 1 , ngc4041 and ngc4564 , have a secondary at a distance larger than 5d@xmath9 . we consider these two galaxies interacting because their secondaries are much larger then them and their distances are smaller than 5 diameters of the secondary . in fact , the secondaries also are in the palomar sample . one possible problem which could exclude secondary galaxies from our sample is the case of edge - on galaxies , which can be highly reddened and thus result in a magnitude difference larger than 3 mag . low mass starbursting galaxies can also create a selection bias , since the starburst would make them brighter and they would be considered as a secondary . however , these effects are expected to affect galaxies of all activity types , and given the size of the sample , they should be smeared out , since adding or subtracting one galaxy with a companion will not change the statistical results significantly . another effect which can influence the results in the case of broad lined seyfert galaxies , is the contribution from the nucleus to the integrated magnitude of their hosts . to correct for this effect we used the observed nuclear b magnitudes given by ho & peng ( 2001 ) . we found that this effect is negligible for most of the galaxies in the sample , with the exception of ngc4151 , where the nucleus increases the integrated magnitude by @xmath36 mag . we consider that this effect is also minimal for broad lined liners , since their nuclear continuum emission at optical wavelengths is weak ( ho 1999a ) . figure 1 presents the distribution of the modulus of the difference between the radial velocities of the primary and secondary galaxies . we can see that although we considered as companions two galaxies with @xmath37 km s@xmath2 , @xmath3870% of the pairs have velocity differences smaller than 300 km s@xmath2 . this number increases to 90% for velocity differences smaller than 500 km s@xmath2 . figure 1 also presents the distribution of radial velocity moduli of the sample of galaxy pairs from karachentsev ( 1987 ) . the visual inspection of the two groups of galaxies shows that they have similar distributions , indicating that most of the galaxies with @xmath37 km s@xmath2 in the palomar are real pairs . the ks test confirms this result , giving a 14% probability that two samples drawn from the same parent population would differ as much as these two samples . the total number of galaxies in each one of the subsamples as well as the number of galaxies with companions , percentage of companions and corresponding 1@xmath39 uncertainties are given in table 2 . in this table we also show these numbers for the cases when we exclude the galaxies which are members of the virgo cluster from the sample , and the results obtained from spliting the activity groups into subgroups of galaxies with different morphologies , or h@xmath0 luminosities . this section presents the comparison of the local galaxy densities , @xmath7 , of galaxies with different activity types . this measurement , defined by tully ( 1988 ) , gives the density of galaxies brighter than m@xmath8 around the galaxy of interest , thus it is an estimate of their environments . these values are tabulated by ho et al . ( 1997a ) . comparing the different activity classes , without excluding any galaxies , the ks test shows that , with the exception of absorption line galaxies , there is no statistically significant difference between their @xmath7 distributions . in the case of absorption line galaxies , the ks test shows that their @xmath7 distribution usually is different from that of other types of galaxies at the 1% significance level . these results can be seen on the top left panel of figure 2 , where we show the percentage of galaxies , of a given activity type , as a function of log(@xmath40 ) . the analysis of the different distributions shows that absorption line galaxies are skewed towards higher density regions , due to the fact that @xmath3850% of them are in the virgo cluster . in fact , once we compare only those galaxies which are not members of the virgo cluster ( top right panel of figure 2 ) , there is no statistically significant difference in the @xmath7 distributions . the same result is obtained if we compare only the galaxies which are members of the virgo cluster . this test was repeated separating the galaxies according to their h@xmath0 luminosities . we found that comparing only the galaxies with h@xmath0 luminosities smaller , or larger than 10@xmath41 erg s@xmath2 gives essentially the same result , that the @xmath7 distributions of most types of galaxies are not significantly different . the corresponding log(@xmath40 ) percentage distributions are presented in the middle panels of figure 2 left for l(h@xmath0)@xmath42 erg s@xmath2 , and right for l(h@xmath0)@xmath1 erg s@xmath2 . notice that absorption line galaxies are not included in this comparison because they do not have emission lines . only in the comparison between liners and hii galaxies with l@xmath43 erg s@xmath2 , we find a difference in the @xmath7 distribution , with the ks - test showing a 2% probability that two samples drawn from the same parent population would differ as much as these two samples . this result is due to the fact that these hii galaxies are found , on average , in slightly lower galaxy density regions than the liners . this result is due to a difference in the morphological types of these two groups of galaxies , since hii galaxies are found only in spirals , while a considerable portion of liners are in ellipticals , which are found in higher density environments . we discuss this effect in more detail in the next section . a similar result was obtained when we separated the galaxies by morphological type , again without a significant difference in the @xmath7 distributions , which can be seen in the bottom left panel of figure 2 for early type galaxies ( ellipticals and s0 s ) and bottom right panel for late type galaxies ( later than s0a ) . only in the comparison between liners and absorption line galaxies in early type systems , we find a 1.4% probability that two samples drawn from the same population would differ as much as these galaxies . this result is due to the fact that almost all absorption line galaxies are early type systems , approximately half of them are in the virgo cluster and consequently have higher local galaxy densities . another test done using the local galaxy density estimator was to compare , for galaxies of the same activity type , if there was any difference in the @xmath7 distributions when we compare galaxies of early and late morphological types , or for l@xmath43 and @xmath42 erg s@xmath2 . again we did not find any significant difference , indicating that galaxies with the same activity type , but different morphological types , or luminosities are found in similar environments . this section shows the results of the comparison between the percentage of galaxies with companions among different activity classes . besides separating the galaxies according to their activity types , we also split them based on membership to the virgo cluster , morphological type and h@xmath0 luminosity . the comparison between the different types of galaxies is done using contingency table analysis ( press et al . 1993 ) , which is a measure of association for two distributions . given the number of galaxies with and without companions in two of our subsamples ( e.g. seyferts and liners ) , this analysis gives the probability of any of these measurements being correlated . figure 3 shows the percentage of galaxies with companions as a function of the distance between the primary and the secondary , in units of the primary diameter ( d@xmath9 ) . although it would be ideal to plot the uncertainties of the percentage of galaxies with companions in the figures , this would result in extremely dense figures , with a large amount of lines superposing each other , which would be very difficult to understad . instead of doing this , we give the percentages , together with the corresponding 1@xmath39 uncertainties in table 2 . each panel in figure 3 shows a different range of magnitude differences between the primary and secondary galaxies . we can see that , besides the fact that the percentage of galaxies with companions is smaller for smaller magnitude differences between the primary and the secondary , the distributions do not differ too much for different @xmath44 values . the most noticeable effect in this plot is the fact that for magnitude differences of @xmath45 mag , there are no seyfert galaxies with companions . the analysis of figure 3 shows that , contrary to what previous papers have shown , seyfert and hii galaxies are the ones with the smaller percentage of companion galaxies . on the other hand , liners , transition and absorption line galaxies have the larger percentage of companions . in particular , the frequency of liners with companions is almost twice that of seyferts and hii galaxies . the results of the comparisons presented below are summarized in table 3 . this table presents the probability that the null hypothesis , that there is no correlation between the two quantities being compared , is correct . this probability is calculated based in @xmath46 statistics . we consider that there is an excess of companions in galaxies with a given activity type relative to another , only if the contingency table analysis gives a probability smaller than 5% . this means that the null hypothesis , that there is no correlation , is rejected at the 2@xmath39 level . the contingency table analysis confirms that seyferts have a smaller percentage of companions when compared to liners , with only a 3.3% probability , that the null hypothesis can be accepted . a similar result is obtained for hii galaxies , which have only a 0.2% probability of having the same percentage of companions as liners , and 1.6% when compared to absorption line galaxies . all other combinations of different subclasses have similar percentages of companion galaxies within the statistical uncertainties . we also tested whether considering galaxies as companions only if their distances are smaller than 3 times the diameter of the primary , the original rafanelli et al . ( 1995 ) criterion , would give a different result . we find that , in most of the cases , it only reduces slightly the probability of two samples having the same percentage of companion galaxies , from 3.3% to 0.6% in the case of the comparison between seyferts and liners , from 0.2% to 0.001% for liners and hii galaxies , from 6.5% to 2.4% for transition and hii galaxies , from 7.7% to 10.5% for seyferts and absorption galaxies , and from 1.6% to 1% for hii and absorption galaxies . this same test was done for all the comparisons presented below , with similar results for 3 and 5 diameters , so , hereafter we will discuss only the results obtained when considering distances smaller than 5 diameters of the primary . as we did in section 4 , for the comparison between the local galaxy density distributions , here we repeat the analysis presented above excluding the galaxies which are in the virgo cluster . we do this because these galaxies are in a higher density environment , and consequently have a higher probability of having close companions . since @xmath3850% of the absorption line galaxies are in virgo , a much higher proportion than for the other activity classes , this can influence the results . figure 4 presents , in a similar way to figure 3 , the percentage of galaxies , which are not members of the virgo cluster , with companions . the results are very similar to those obtained when we do not exclude the virgo cluster members ( figure 3 ) . the only difference is that now liners do not have a significant excess of companions when compared to seyferts ( 6.5% ) , which is also the case for the comparison between hii and absorption line galaxies ( 7.2% ) . we still find a significant excess of companions when we compare liners to hii galaxies , with a probability of only 0.04% that these two subsamples are similar . another test done to the sample was to separate the different activity types into subgroups , based on their host galaxy morphological types . we decided to separate the galaxies into ellipticals , s0 s , and late type galaxies , those with morphological types sa and later . this division was based on the fact that late type galaxies have enough gas in their disks , and an interaction can easily move this gas into the nuclear region . in the case of elliptical galaxies , the amount of gas is much smaller , so the source of gas to feed their nuclei may have to be external to the galaxy . in s0 galaxies the amount of gas is larger than that of ellipticals , so we test whether combining these galaxies with ellipticals or late types makes a difference to the results . in the comparison between the percentage of companions in elliptical galaxies of different activity classes , we do not include hii galaxies and seyferts , because there are no hii galaxies in ellipticals and only a small number of seyferts with this morphology . similarly , when analyzing the late type galaxies we do not include absorption line objects . the results of the separation of the sample into ellipticals , late types , early types ( ellipticals plus s0 s ) , and late types plus s0 s is presented in figure 5 . in the case of early types , there is no significant excess of companions among most of the galaxies studied . the comparison between the late type galaxies shows that all different activity classes have similar percentages of galaxies with companions . when comparing only elliptical galaxies , we find that there is no significant difference between the different activity classes . a similar result is obtained when we add late types and s0 s , with the exception that in this case liners have a slightly higher percentage of companions when compared to hii galaxies , but the probability that the two distributions are similar is 4.8% , only marginally significant . an interesting result which comes out of the analysis of figure 5 , is the fact that liners and transition galaxies in ellipticals have @xmath42 times higher a percentage of companions than the corresponding galaxies in late type systems . in the case of liners , the probability of the two samples being the same is 0.4% . for transition objects this result is not statistically significant , given the small number of transition galaxies in ellipticals . this result is maintained when we compare early ( e@xmath47s0 ) and late type liners , but in this case the probability of the two samples being the same is 4% . the results of separating the sample into galaxies with h@xmath0 luminosities smaller or larger than 10@xmath41 erg s@xmath2 is presented in figure 6 . absorption line galaxies are not used in this analysis . for galaxies with l@xmath48 erg s@xmath2 , there is no difference in the percentage of liners , seyferts and transition galaxies with companions . however , liners and transition galaxies have a significantly higher percentage of companions when compared to hii galaxies , with the null hypothesis that the there is no correlation between the activity type and the percentage of companions being rejected at the 0.3% and 1.5% significance levels , respectively . the comparison of seyferts and hii galaxies does not show a significant difference ( 11% ) . this result is again due to the mixture of ellipticals and spirals in the liners and transition galaxies samples . if we consider only spiral galaxies ( s0 s + late types ) with l(h@xmath0)@xmath3 erg s@xmath2 , we find that there is significant difference between the percentage of companion in these galaxies and hii galaxies . the comparison of galaxies with l@xmath43 erg s@xmath2 shows that there is no statistically significant difference in the percentage of galaxies with companions among different activity classes . the comparison between liners and seyferts shows a small probability that they have similar percentages of companions , 5.9% , but this can be due to the small number of seyferts . we compared if there is a difference in the proportion of galaxies with companions for objects with the same activity class , but h@xmath0 luminosities larger and smaller than 10@xmath41 erg s@xmath2 . there is no significant difference for liners , transition galaxies and seyferts . however , there is a significantly higher percentage of galaxies with companions in the case of hii galaxies with l(h@xmath0)@xmath1 erg s@xmath2 , than in hii galaxies with l(h@xmath0)@xmath42 erg s@xmath2 , rejecting the null hypothesis at the 2.9% significance level . this result agrees with the ones presented by bushouse ( 1986 ) , keel et al . ( 1985 ) and kennicutt et al . it also agrees with the result obtained by ho , filippenko & sargent ( 1997c ) , who found , in this sample , that hii galaxies in barred early type hosts have higher h@xmath0 luminosities . in fact , the percentage of hii galaxies with companions increases to @xmath3865% for galaxies with l(h@xmath0)@xmath49 erg s@xmath2 , a total of 9 out of 14 galaxies with companions . this percentage is significantly higher than that of the rest of the hii galaxies , as well as all the other galaxies in the sample , rejecting the null hypothesis at the 0.007% and 1.5% significance levels , respectively . the previous section shows that a simple analysis , not taking into account morphology or activity level ( h@xmath0 luminosity ) of the galaxies , gives that the percentage of liners , transition and absorption line galaxies with nearby companions is significantly higher than that of seyferts and hii galaxies . this result contradicts all previous works in this field , which either found that seyferts and hii galaxies have the same percentage of companions as other galaxies , or that they have a higher percentage of companions . however , we show that when we take into account the morphological types of the galaxies and split them into subgroups containing only ellipticals and only spirals , the situation is different . there is no difference in the percentage of galaxies with companions among different activity types if we consider only galaxies of similar morphological types . this result is consistent with the one found by fuentes - williams & stocke ( 1988 ) , bushouse ( 1986 , 1987 ) , de robertis et al . ( 1998 ) , and also with more recent results on clustering of low luminosity agns at higher redshifts ( brown et al . 2001 ; schreier et al . 2001 ) . the percentage of elliptical galaxies with companions is approximately two times higher than that of spirals . this explains why liners , transition and absorption line galaxies have higher percentages of companions when all galaxies are considered , since a higher percentage of these galaxies is found in ellipticals when compared to seyferts and hii galaxies . given the fact that most of the hii galaxies in the palomar sample have only small quantities of recent star formation , most of them should be considered as normal , quiescent galaxies . we found that a there is a higher percentage of galaxies with companions among hii galaxies with l(h@xmath0)@xmath1 erg s@xmath2 then in lower h@xmath0 luminosity hii galaxies . the percentage of companions increases even more for hii galaxies with higher h@xmath0 luminosities , which confirms previous results ( kennicutt et al . 1997 , keel et al . 1985 , bushouse 1986 ) . the results we obtained separating the galaxies by morphological types is somewhat expected . it was shown by dressler ( 1980 ) and whitmore , gilmore & jones ( 1993 ) in the study of clusters of galaxies that the percentage of ellipticals increases in higher density environments . although this result was based on clusters , postman & geller ( 1984 ) found that the results also apply for groups of galaxies . furthermore , whitmore et al . ( 1993 ) found that the percentage of ellipticals in close pairs is higher than that of spirals , supporting the results we found . charlton , whitmore & gilmore ( 1995 ) showed that ellipticals are found more often in pairs than spirals in clusters and high density environments , like large groups . we found that @xmath430% of the ellipticals in our sample are in the virgo cluster , while only 20% of the spirals are found in virgo . another @xmath430% of the ellipticals are found in groups of galaxies with 10 or more galaxies garcia ( 1993 ) , suggesting that the higher percentage of ellipticals with companions is in fact due to the morphology - density relation . the study of the local galaxy densities ( section 4 ) shows that , in most of the cases , the distribution of this quantity does not differ for different activity types . in the cases where we find significant differences , they can be explained if we exclude the galaxies which belong to the virgo cluster , or if we separate the galaxies into different morphological types . these results agree with the ones obtained in the comparison between the percentage of companions in galaxies with different activity types . it can be argued that the results found for seyfert galaxies is somewhat questionable , since the palomar survey has a large portion of low luminosity seyfert galaxies , which are usually not observed in other surveys . in this way , the results presented here would be biased towards low luminosity seyferts , which , in a way similar to hii galaxies , may have a higher percentage of companions as the luminosity increases . evidence of this effect is given by heckman , carty & bothun ( 1985 ) , heckman et al . ( 1984 ) and hutchings ( 1983 ) , who showed that radio galaxies and quasars are found in higher density environments . we believe that this is not the case . we compare the results presented here with the ones from schmitt et al . ( 2001 ) for warm infrared seyfert galaxies . the median [ oiii ] luminosity of their seyfert galaxies , which is believed to be an isotropic indicator of the intrinsic luminosity of these galaxies , is l([oiii])@xmath50 erg s@xmath2 , calculated using de grijp et al . ( 1992 ) values . this value corresponds to 70 times the median value of the [ oiii ] luminosity of seyfert galaxies in the palomar survey ( l([oiii])@xmath51 erg s@xmath2 ) , obtained from ho et al . ( 1997a ) . schmitt et al . ( 2001 ) observed that between 19% and 28% of their seyfert galaxies have companions . their criteria were a little different from ours , assuming that galaxies were companions if the distance between the primary and the secondary was smaller than 3 times the diameter of the primary , and the brightness difference between them smaller than 3 mag . since it was not possible for them to find radial velocities for all possible companion galaxies , they could only put limits on the percentage of galaxies with companions . the lower limit is 19% , which corresponds to those galaxies where it was possible to confirm that the companion has a radial velocity difference smaller than 1000 km s@xmath2 , while the upper limit is 28% , corresponding to all galaxies , including those without information about the velocity of the secondary . the uncertainty in these measurements is @xmath45% , given by poisson statistics . the number of seyfert galaxies with companions at distances smaller than 3 diameters in the palomar survey is 8 in a sample of 46 galaxies , a total of 17%@xmath52% . given the uncertainties involved in these measurements , there is no significant difference in the percentage of seyfert galaxies with companions in the palomar and schmitt et al . ( 2001 ) samples , which shows that low and high luminosity seyfert galaxies similar environments . the contingency table analysis gives that there is a 16% probability that the two samples have the same number of companions . the explanation of why other papers found that there is a higher percentage of seyferts with companions than normal galaxies with companions is not very clear . one possibility would be that they mixed galaxies of different morphological types . however , the percentage of seyferts in ellipticals is small and this effect could be smeared out in a larger sample . another explanation is related to the way the previous works selected their samples of seyferts . in the case of laurikainen & salo ( 1995 ) and dultzin - hacyan et al . 1994 ) , their samples included a large portion of seyferts selected from ultraviolet surveys . in seyfert 2 galaxies , the unified model predicts that the ultraviolet emission is either nuclear radiation reflected towards the observer ( antonucci 1993 ) , or a circumnuclear starburst ( cid fernandes & terlevich 1995 ) . if the seyfert 1 s and seyfert 2 s have the same amount of ultraviolet excess , this means that the seyfert 2 s were selected from two orders of magnitude higher in the luminosity function , since only 1% of the nuclear light is believed to be reflected , or that their circumnuclear region is dominated by a luminous starburst which is responsible for most of the ultraviolet emission ( see schmitt et al . 2001 for a discussion about this ) . in the latter case , the fact that these papers observed a higher percentage of seyfert 2 s with companions , but not seyfert 1 s , would be due to the fact that the percentage of starbursts with companions goes up as the starburst luminosity goes up , as shown above . the result that seyfert galaxies have the same percentage of companion galaxies as galaxies with other activity types is intriguing and may have important consequences for the theories of how the gas is moved from kiloparsec scales to the inner parsec region of the galaxy and feeds the nuclear black hole . taken at face value , this result indicates that interactions are not important in this process . however , this result can be interpreted in a different way . since we expect a delay between the time when the interaction occurs and when the gas reaches the nucleus , we may be seeing different stages of this process . it has been shown by several simulations ( e.g. byrd et al . 1986 , 1987 , byrd & valtonen 2001 , lin et al . 1988 , hernquist & mihos 1995 ) that interactions move the gas to the nuclear region of the galaxy , where its density and temperature increases , a starburst is formed and later the gas feeds the black hole . also , taking into account the evidence that many galaxies have black holes in their nuclei , suggesting that this may be the case in all galaxies ( magorrian et al . 1998 , ho 1999b , gebhardt et al . 2000 , ferrarese & merritt 2000 ) , it may be possible that all galaxies pass through a period of activity . the galaxies may also pass through different activity types , where they first present a star formation period , when the gas has just moved into the nuclear region , later they present a period of seyfert activity , when the gas is being accreted by the black hole , and finally they go into a period where they turn into a liner or transition galaxy , when the amount of gas available to fuel the nucleus is small . we would like to thank jim ulvestad , luis ho and the referee for comments and suggestions which significantly improved this manuscript . this research made use of the nasa / ipac extragalactic database ( ned ) , which is operated by the jet propulsion laboratory , caltech , under contract with nasa . we also used the digitized sky survey , which was produced at the space telescope science institute under u.s . government grant nagw-2166 . the national radio astronomy observatory is a facility of the national science foundation operated under cooperative agreement by associated universities , inc . lrrrlrrrl ngc0507 & -1.70 & 0.46 & 602 & ngc0508 & & s0 & & ngc1023 & -2.87 & 0.33 & 111 & ngc1023a & & s0 & & ngc2300 & 0.02 & 2.19 & 454 & ngc2276 & & s0 & & ngc2950 & -2.83 & 3.15 & 21 & ugc5179 & & s0 & & ngc3384 & 0.57 & 1.46 & 154 & ngc3379 & & s0 & & ngc3613 & -0.69 & 4.97 & 512 & ngc3619 & & e & & ngc3640 & -2.78 & 0.64 & 441 & ngc3641 & & e & & ngc4262 & -2.09 & 4.92 & 29 & ic0781 & & s0 & yes & ngc4291 & -0.30 & 3.41 & 399 & ngc4319 & & e & & ngc4340 & 0.08 & 1.68 & 326 & ngc4350 & & s0 & yes & ngc4365 & -2.91 & 1.60 & 444 & ngc4370 & & e & yes & ngc4382 & -1.52 & 1.12 & 200 & ngc4394 & & s0 & yes & ngc4406 & -1.96 & 1.21 & 480 & ngc4402 & & e & yes & ngc4417 & 0.07 & 3.76 & 422 & ngc4424 & & s0 & yes & ngc4473 & -0.27 & 3.37 & 884 & ngc4477 & & e & yes & ngc4478 & -0.67 & 2.47 & 596 & ngc4476 & & e & yes & ngc4503 & -2.56 & 2.12 & 136 & ic3470 & & l & yes & ngc4564 & 1.67 & 10.5@xmath53 & 379 & ngc4579 & & e & yes & ngc4638 & -2.76 & 0.81 & 16 & ngc4637 & & s0 & yes & ngc4649 & -2.11 & 0.35 & 302 & ngc4647 & & e & yes & ngc4754 & 0.32 & 2.57 & 412 & ngc4762 & & s0 & yes & ngc5576 & -1.42 & 0.91 & 171 & ngc5574 & & e & & ngc5638 & -1.53 & 0.76 & 68 & ngc5636 & & e & & ngc7332 & -0.15 & 1.50 & 44 & ngc7339 & & s0 & & ngc7457 & -2.30 & 1.87 & 96 & ugc12311 & & s0 & & ngc7619 & -2.60 & 1.06 & 369 & ngc7617 & & e & & ngc0315 & -1.72 & 1.81 & 130 & ngc0311 & 39.55 & e & & ngc0474 & -0.04 & 0.79 & 11 & ngc0470 & 38.52 & s0 & & ngc1961 & -2.76 & 2.71 & 42 & ugc3342 & 39.81 & l & & ngc2336 & -1.61 & 2.73 & 158 & ic0467 & 38.39 & l & & ngc3166 & -2.87 & 1.00 & 4 & ngc3165 & 39.10 & l & & ngc3169 & -0.05 & 1.76 & 110 & ngc3166 & 39.02 & l & & ngc3190 & -1.67 & 1.07 & 253 & ngc3187 & 38.82 & l & & ngc3193 & 0.29 & 1.92 & 108 & ngc3190 & 38.20 & e & & ngc3226 & 1.27 & 0.73 & 165 & ngc3227 & 38.93 & e & & pair ngc3379 & -0.57 & 1.39 & 185 & ngc3384 & 37.94 & e & & ngc3414 & -2.41 & 2.24 & 220 & ugc5958 & 39.23 & s0 & & ngc3507 & -0.65 & 3.78 & 152 & ngc3501 & 39.39 & l & & ngc3607 & -2.21 & 0.64 & 266 & ngc3605 & 38.93 & s0 & & ngc3608 & 0.90 & 1.79 & 189 & ngc3607 & 38.28 & e & & ngc3623 & 0.52 & 2.07 & 80 & ngc3627 & 37.77 & l & & ngc3718 & -0.72 & 1.44 & 31 & ngc3729 & 38.46 & l & & ngc3998 & -1.97 & 1.13 & 353 & ngc3990 & 40.00 & s0 & & ngc4036 & -0.27 & 4.05 & 164 & ngc4041 & 39.35 & s0 & & ngc4111 & -2.43 & 2.40 & 136 & ngc4117 & 39.40 & s0 & & ngc4261 & -2.47 & 0.87 & 322 & ngc4264 & 39.35 & e & yes & ngc4278 & -1.88 & 0.84 & 410 & ngc4283 & 39.17 & e & & ngc4374 & -2.87 & 1.59 & 395 & ngc4387 & 38.89 & e & yes & ngc4394 & 1.52 & 2.04 & 198 & ngc4382 & 38.33 & l & yes & ngc4438 & -1.12 & 0.53 & 717 & ngc4435 & 39.37 & l & yes & ngc4486 & -2.72 & 1.10 & 67 & ngc4478 & 39.44 & e & yes & ngc4550 & -0.41 & 1.10 & 785 & ngc4551 & 38.41 & s0 & yes & ngc4762 & -0.32 & 1.58 & 416 & ngc4754 & 37.49 & s0 & yes & ngc5077 & -0.91 & 1.61 & 39 & ngc5079 & 39.67 & e & yes & ngc5195 & 1.71 & 0.78 & 22 & ngc5194 & 37.94 & s0 & & pair ngc5353 & -0.36 & 0.62 & 352 & ngc5354 & 39.12 & s0 & & ngc5363 & 0.08 & 3.56 & 103 & ngc5364 & 39.42 & s0 & & pair ngc5485 & -1.26 & 2.79 & 594 & ngc5486 & 38.35 & s0 & & ngc5566 & -2.93 & 0.64 & 267 & ngc5569 & 38.66 & e & & ngc5813 & -2.81 & 2.99 & 395 & ngc5811 & 38.56 & e & & ngc5850 & 0.48 & 2.36 & 770 & ngc5846 & 38.66 & l & & ngc5970 & -2.97 & 2.81 & 52 & ic1131 & 38.06 & l & & ngc5982 & 0.67 & 3.08 & 387 & ngc5985 & 38.46 & e & & pair ngc5985 & -0.67 & 1.35 & 386 & ngc5982 & 38.94 & l & & pair ngc6340 & -2.56 & 1.89 & 39 & ic1251 & 38.50 & l & & ngc7626 & 0.13 & 2.57 & 381 & ngc7619 & 38.81 & e & & ngc0520 & 0.00 & 0.02 & 9 & pair & 39.33 & l & & merging ngc0672 & -0.83 & 1.07 & 83 & ic1727 & 37.81 & l & & pair ngc0697 & -2.03 & 4.61 & 167 & ngc0694 & 39.01 & l & & ngc0812 & -2.99 & 3.64 & 647 & ugc1585 & 40.08 & l & & ngc0877 & -0.99 & 4.59 & 173 & ngc0871 & 39.18 & l & & ngc2146 & -2.01 & 2.94 & 604 & ngc2146a&39.76 & l & & ngc2342 & -0.92 & 1.66 & 44 & ngc2341 & 40.93 & l & & ngc2276 & -0.02 & 2.14 & 454 & ngc2300 & 39.87 & l & & ngc2750 & -0.61 & 0.36 & 2 & mcg+04 - 22 - 012&40.60 & l & & ngc2964 & -1.21 & 2.10 & 246 & ngc2968 & 40.01 & l & & ngc3034 & 1.47 & 3.29 & 250 & ngc3031 & 39.71 & l & & ngc3338 & -2.42 & 3.30 & 85 & ugc5832 & 38.15 & l & & ngc3389 & 1.08 & 2.33 & 595 & ngc3384 & 38.60 & l & & ngc3395 & -0.23 & 0.57 & 7 & ngc3396 & 39.40 & l & & pair ngc3430 & -0.66 & 1.56 & 90 & ngc3424 & 39.04 & l & & ngc3504 & -1.36 & 4.46 & 162 & ngc3512 & 40.81 & l & & ngc3646 & -2.84 & 2.01 & 731 & ngc3649 & 39.42 & l & & ngc3684 & 0.19 & 4.53 & 9 & ngc3686 & 38.67 & l & & ngc3686 & -0.19 & 4.32 & 9 & ngc3684 & 39.80 & l & & ngc3690 & 0.00 & 0.05 & 31 & pair & 40.62 & l & & merging ngc3729 & 0.72 & 4.15 & 30 & ngc3718 & 39.41 & l & & ngc3963 & -0.98 & 3.02 & 181 & ngc3958 & 39.61 & l & & ngc4041 & 0.27 & 5.72@xmath53 & 164 & ngc4036 & 39.46 & l & & ngc4088 & -1.58 & 1.97 & 7 & ngc4085 & 39.00 & l & & ngc4123 & -0.33 & 3.20 & 19 & ngc4116 & 40.35 & l & & ngc4274 & -2.27 & 2.80 & 129 & ngc4283 & 38.47 & l & & ngc4273 & -2.36 & 0.81 & 114 & ngc4277 & 40.26 & l & & ngc4298 & 0.26 & 0.73 & 9 & ngc4302 & 39.00 & l & yes & ngc4380 & -2.58 & 2.62 & 2 & ic3328 & 38.02 & l & yes & ngc4424 & -0.07 & 3.06 & 412 & ngc4417 & 39.02 & l & yes & ngc4469 & -2.54 & 4.20 & 15 & ugc7596 & 38.16 & l & yes & ngc4477 & -2.08 & 1.40 & 472 & ngc4479 & 38.84 & s0 & yes & ngc4485 & 2.41 & 1.53 & 83 & ngc4490 & 37.23 & l & & ngc4490 & -2.41 & 0.55 & 83 & ngc4485 & 37.78 & l & & ngc4496a&0.00 & 0.23 & 6 & ngc4496 & 38.47 & l & yes & merging ngc4517 & -2.70 & 1.63 & 410 & ngc4517a&37.37 & l & & ngc4532 & -2.66 & 4.26 & 31 & holmberg vii&39.69 & l & yes & ngc4535 & -1.83 & 4.25 & 736 & ngc4519 & 39.72 & l & yes & ngc4536 & -2.86 & 1.11 & 51 & ngc4533 & 39.38 & l & yes & ngc4567 & 0.61 & 0.44 & 11 & ngc4568 & 38.78 & l & yes & pair ngc4568 & -0.61 & 0.28 & 11 & ngc4567 & 38.95 & l & yes & pair ngc4618 & -1.72 & 1.99 & 66 & ngc4625 & 38.16 & l & & ngc4631 & -1.49 & 2.08 & 38 & ngc4656 & 37.42 & l & & ngc4647 & 2.11 & 0.87 & 302 & ngc4649 & 38.52 & l & yes & ngc4654 & -1.10 & 3.57 & 29 & ngc4639 & 39.07 & l & yes & ngc4656 & 0.00 & 0.24 & 27 & ngc4657 & 37.95 & l & & merging ngc5112 & -0.94 & 3.34 & 24 & ngc5107 & 38.19 & l & & ngc5364 & -0.08 & 2.10 & 103 & ngc5363 & 38.14 & l & & ngc5775 & -1.25 & 1.01 & 111 & ngc5774 & 38.76 & l & & ngc5905 & -0.07 & 3.27 & 82 & ngc5908 & 40.25 & l & & ngc5962 & -2.50 & 3.34 & 42 & ugc9925 & 39.21 & l & & ngc0777 & -2.05 & 2.86 & 448 & ngc0778 & 38.73 & e & & ngc1068 & -1.25 & 4.34 & 140 & ngc1055 & 41.55 & l & & ngc3031 & -1.47 & 1.34 & 240 & ngc3034 & 37.64 & l & & ngc3227 & -1.27 & 0.40 & 165 & ngc3226 & 40.38 & l & & pair ngc3735 & -2.60 & 3.96 & 87 & ugc6532 & 39.82 & l & & ngc4168 & -2.07 & 1.06 & 424 & ngc4165 & 37.60 & e & yes & ngc4258 & -2.79 & 1.83 & 578 & ngc4217 & 38.35 & l & & ngc4565 & -1.45 & 4.28 & 126 & ngc4494 & 37.97 & l & & ngc4579 & -1.67 & 4.89 & 379 & ngc4564 & 39.44 & l & yes & ngc4725 & -2.37 & 2.26 & 16 & ngc4747 & 38.19 & l & & ngc5194 & -1.71 & 0.40 & 22 & ngc5195 & 38.88 & l & & pair ngc5395 & -1.27 & 0.66 & 15 & ngc5394 & 38.87 & l & & ngc0410 & -1.31 & 2.24 & 352 & ngc0407 & 39.43 & e & & ngc0521 & -1.30 & 2.53 & 411 & ic1694 & 39.16 & l & & ngc0524 & -2.80 & 3.48 & 11 & ngc0516 & 38.57 & s0 & & ngc0660 & -1.47 & 2.55 & 78 & ugc1195 & 38.89 & l & & ic1727 & 0.83 & 1.12 & 83 & ngc0672 & 37.69 & l & & pair ngc1055 & 1.25 & 3.96 & 142 & ngc1068 & 37.92 & l & & ngc1161 & -0.71 & 1.00 & 588 & ngc1160 & 38.70 & s0 & & ngc3245 & -1.59 & 2.91 & 26 & ngc3245a&39.59 & s0 & & ngc3627 & -0.52 & 2.17 & 80 & ngc3623 & 38.50 & l & & ngc3628 & -0.34 & 2.41 & 39 & ngc3623 & 36.87 & l & & ngc3917 & -2.77 & 2.22 & 133 & ngc3917a&37.29 & l & & ngc4145 & -2.70 & 2.16 & 154 & ugc7175 & 37.72 & l & & ngc4216 & -2.75 & 1.45 & 101 & ngc4222 & 38.53 & l & yes & ngc4220 & -0.99 & 4.68 & 250 & ngc4218 & 38.26 & l & & ngc4281 & -2.25 & 1.93 & 186 & ngc4277 & 38.61 & s0 & & ngc4350 & -0.08 & 2.07 & 319 & ngc4340 & 38.26 & s0 & yes & ngc4435 & 1.12 & 1.64 & 710 & ngc4438 & 38.98 & s0 & yes & ngc4459 & -2.64 & 2.51 & 293 & ngc4468 & 38.75 & s0 & yes & ngc4527 & -2.78 & 3.15 & 19 & ngc4533 & 38.86 & l & yes & ngc4552 & -2.20 & 3.33 & 862 & ngc4551 & 38.52 & e & yes & ngc4569 & -2.53 & 3.87 & 430 & ngc4531 & 39.91 & l & yes & ngc5354 & 0.36 & 0.85 & 352 & ngc5353 & 38.71 & s0 & & ngc5746 & -1.87 & 2.36 & 147 & ngc5740 & 38.48 & l & & ngc5846 & -2.58 & 1.75 & 324 & ngc5845 & 38.81 & e & & lrrrrrrrrrr all galaxies & 46&12 ( 26%@xmath548%)&193&51 ( 26%@xmath544% ) & 88&40 ( 45%@xmath547%)&63&24 ( 38%@xmath548%)&61&26 ( 43%@xmath548%)excluding virgo & 37&10 ( 27%@xmath549%)&168&38 ( 23%@xmath544% ) & 73&33 ( 45%@xmath548%)&49&17 ( 35%@xmath548%)&35&13 ( 37%@xmath5410%)ellipticals & 3 & 2 ( 67%@xmath5447% ) & 0 & 0 & 19&13 ( 68%@xmath5419% ) & 5 & 3 ( 60%@xmath5435%)&22&12 ( 55%@xmath5416%)s0 s & 10 & 0 & 9 & 1 ( 11%@xmath5411% ) & 20&12 ( 60%@xmath5417%)&18 & 8 ( 44%@xmath5416%)&34&13 ( 38%@xmath5411%)late types & 33&10 ( 29%@xmath5410%)&184&50 ( 27%@xmath544% ) & 49&15 ( 31%@xmath548%)&40&13 ( 33%@xmath549%)&5&1 ( 20%@xmath5420%)l@xmath55 & 22 & 8 ( 36%@xmath5413%)&112&23 ( 21%@xmath544% ) & 57&24 ( 42%@xmath549%)&52&20 ( 38%@xmath549% ) & 0 & 0 l@xmath43 & 24 & 4 ( 17%@xmath548% ) & 81&28 ( 35%@xmath547% ) & 31&16 ( 52%@xmath5413%)&11&4 ( 36%@xmath5418% ) & 0 & 0 lrrrrrrrrrr all galaxies&44.5%&3.3%&0.2%&77.8%&16.9%&6.5%&65.9%&96.3%&7.7%&1.6%all galaxies distance @xmath563 diameters&13.6%&0.6%&@xmath560.1%&22.5%&16.2%&2.4%&79.8%&82.6%&10.5%&1.0%excluding virgo&24.7%&6.5%&0.04%&42.8%&44.8%&8.7%&81.7%&56.7%&35.8%&7.2%ellipticals&72.2%&&&36.4%&&&82.5%&&&ellipticals @xmath47 s0s&21.0%&&&6.2%&&&79.6%&&&late types&73.5%&97.6%&66.4%&14.5%&84.0%&52.3%&13.0%&73.8%&15.1%&17.0%late types @xmath47 s0s&73.5%&8.2%&4.8%&73.9%&16.3%&14.8%&97.5%&66.8%&20.9%&23.0%l@xmath55&69.8%&64.1%&0.3%&&86.5%&1.5%&&10.8%&&l@xmath43&87.2%&5.9%&62.8%&&19.7%&90.6%&&9.4%&&

using high resolution infrared spectroscopy we have surveyed twenty sightlines for h absorption . h is detected in eight diffuse cloud sightlines with column densities varying from to . this brings to fourteen the total number of diffuse cloud sightlines where h has been detected . these detections are mostly along sightlines concentrated in the galactic plane , but well dispersed in galactic longitude . the results imply that abundant h is common in the diffuse interstellar medium . because of the simple chemistry associated with h production and destruction , these column density measurements can be used in concert with various other data to infer the primary cosmic - ray ionization rate , . values range from s to s with an average of s . where h is not detected the upper limits on the ionization rate are consistent with this range . the average value of is about an order of magnitude larger than both the canonical rate and rates previously reported by other groups using measurements of oh and hd . the discrepancy is most likely due to inaccurate measurements of rate constants and the omission of effects which were unknown when those studies were performed . we believe that the observed column density of h is the most direct tracer for the cosmic - ray ionization rate due to its simple chemistry . recent models of diffuse cloud chemistry require cosmic - ray ionization rates on the order of 10 s to reproduce observed abundances of various atomic and molecular species , in rough accord with our observational findings .

0709.1114

in the past several years h@xmath0 has been detected in diffuse interstellar clouds @xcite where it had been expected to exist in abundances below observable limits . this surprising result raised various questions about the diffuse cloud environment . the rather simple chemistry of h@xmath0 allows for only three variable parameters in determining its abundance when the steady state approximation is used : the h@xmath0-electron recombination rate , the electron to hydrogen ratio , and the cosmic - ray ionization rate . previous work @xcite has shown that the first two of these are relatively well constrained . this leaves the cosmic - ray ionization rate as an unconstrained parameter . because the low energy cosmic - rays responsible for most of the ionization in diffuse clouds can not be directly measured in the solar system , we must rely on molecules to act as tracers of the ionization rate . using h@xmath0 , @xcite found the cosmic - ray ionization rate of molecular hydrogen , @xmath10 , to be much larger along the sightline to @xmath11 per than the canonical value of @xmath12 s@xmath6 . prior to the detection of h@xmath0 in diffuse clouds , oh and hd were the molecules of choice for estimating the cosmic - ray ionization rate there . estimates using these molecules required determining rate constants and modeling various reactions on the pathways to forming oh and hd @xcite . the derived values of the ionization rate tended to agree with the canonical value of @xmath4 , the primary cosmic - ray ionization rate , but differ greatly from the value derived from the recent h@xmath0 measurement toward @xmath11 per ( the relation between @xmath4 and @xmath10 is explained in 4.2 and quantified by equation ( [ eq10 ] ) ) . of the three molecules , the simple chemistry of h@xmath0 provides the most direct determination of @xmath4 @xcite , suggesting that measurements of h@xmath0 should produce more accurate results and be a more reliable tracer of the cosmic - ray ionization rate than oh or hd . the higher ionization rate found by @xcite towards @xmath11 per implies the production of more h@xmath0 , and if generally applicable , could account for the higher than expected column densities found in several diffuse clouds @xcite . however , prior to the present work the enhanced ionization rate was known to exist for certain only along one line of sight , and thus could have been considered an anomaly . to test if an enhanced ionization rate is a general property of the diffuse interstellar medium ( ism ) , we have performed a survey of h@xmath0 in nineteen diffuse cloud sightlines . h@xmath0 is detected in eight of the clouds and the overall results , including analysis of previous observations by our group , support a higher ionization rate . when coupled with further arguments , this strongly suggests that a greatly enhanced ionization rate is a typical property of the diffuse ism . all observations were made using the cgs4 spectrometer @xcite on the united kingdom infrared telescope ( ukirt ) at mauna kea . the spectrometer was used with its echelle grating , 0.6 arcsec wide slit , and long camera to provide a resolving power of 40000 . observations were taken in an abba pattern in which the target star is nodded along the slit so that the spectral image falls alternately on different rows of the array . suitable standard stars were observed throughout each night to account for changing atmospheric conditions and air mass . with the exception of the red rectangle where reddening is due to the ejected envelope of a post asymptotic giant branch star , targets were chosen primarily by three criteria : ( 1 ) sightlines known to pass through diffuse clouds ; ( 2 ) early spectral type ; and ( 3 ) bright l - band magnitude . the complete dataset consists of twenty ( nineteen diffuse cloud ) sightlines which were observed intermittently between may 2001 and march 2005 . relevant information concerning these observations is given in table [ tbl1 ] . two of the sightlines from the dataset were examined in @xcite : hd 20041 and @xmath11 oph . however , both have previously been studied using only the @xmath13 transition at 37154.8 ( vacuum wavelengths are used throughout this paper ) , whereas the new data cover the @xmath14 and @xmath15 transitions at 36680.8 and 36685.2 , respectively ( see @xcite or @xcite for a complete description of the transition notation associated with h@xmath0 ) . @xmath11 per was investigated in @xcite , but here data from three more nights of observations are included . overall , we present twenty sightlines for which new or refined h@xmath0 column densities or upper limits are calculated . the reduction process involves multiple steps and software packages . first , raw data are run through starlink s orac - dr pipeline which processes ukirt data . images are then transformed to fits format using starlink s figaro package , and neighboring images are subtracted from each other in noao s iraf package . this subtraction serves to eliminate atmospheric background and detector bias levels from the image . still using iraf , spectra are extracted from the neighbor - subtracted images with the _ apall _ routine . these spectra are imported to igor pro where we have macros set up to complete the reduction @xcite . during this import , ripples in the spectrum caused by guiding and seeing fluctuations while the cgs4 array is shifted along the spectral direction in successive steps of one - third of a pixel during observing are removed . spectra for each object and standard star are then coadded . some frames may be excluded in this step if they are noisier than normal or happen to have cosmic - ray hits on the rows where the spectrum is located . objects are then ratioed with standards to remove atmospheric absorption lines and the continuum level is set to unity . the ratioing is an interactive process where the user may vary the intensity scaling and/or shift the spectrum in wavelength in order to most effectively remove atmospheric lines and obtain the most reliable ratioed spectrum at the wavelengths where the h@xmath0 lines are expected to appear . a fringing pattern caused by the circular variable order - blocking filter is another artifact of cgs4 that needs to be removed . this is accomplished by transforming a spectrum into fourier space using an igor macro . the user can then find the peak caused by the fringing pattern and interpolate across it . once the peak is removed , an inverse fourier transform is performed to produce a spectrum where the fringing pattern is absent . this method is described in more depth in @xcite . after removing the fringing pattern , spectra are wavelength calibrated using the vacuum rest wavelengths of the atmospheric lines . the accuracy of the wavelength calibration is typically @xmath162 km s@xmath6 . finally , the h@xmath0lines are fit with gaussians and the equivalent widths , column densities , and radial velocities are derived . for targets with observations on multiple nights , the reduction process above is followed through the wavelength calibration step . at that point iraf s _ rvcorrect _ routine is used to calculate the velocity of the earth along a given sightline in the local standard of rest ( lsr ) frame . each spectrum is then shifted to be in the lsr frame . this puts the interstellar absorption lines at the same wavelength for any given date and allows spectra from different nights to be coadded . while each spectrum has its continuum set to unity , they originally had different exposure times and intensities . to weight the final coadded spectrum properly , each spectrum is scaled by its original coadded intensity before being added to the spectra from other nights . once the final multiple night coadded spectrum is produced , it is divided by the sum of the scaling factors to reset the continuum level to unity . after this process , line parameters are again extracted by fitting the absorption lines . spectra covering the h@xmath0 doublet are adversely affected by an atmospheric ch@xmath17 line complex which lies just shortward of 36680.8 , the rest wavelength of the @xmath14 line . if the spectrum of the target is blue - shifted , the overlap with this line can be significant and hinder detection or estimates of line strength . typical examples of spectra before the atmospheric lines are removed via ratioing with a standard star are shown in figure [ fig1 ] . the ch@xmath17 feature here absorbs roughly 50% of the incoming light at 36675.3 in both the standard star @xmath18 per ( middle spectrum ) and the object @xmath11 per ( bottom spectrum ) . a weak telluric hdo absorption line near 36681 that can vary with both time and airmass further complicates the reduction , especially when the water column density above the telescope is high and unstable . when the two spectra are ratioed following the methods in 2.2 , the top spectrum in figure [ fig1 ] is produced . the two lines show the rest wavelength positions of the @xmath14 line at 36680.8 and the @xmath15 line at 36685.2 . two arrows mark the expected positions of the h@xmath0 lines due the earth s orbital motion and the radial velocity of the absorbing cloud along the line of sight towards @xmath11 per . clearly the h@xmath0 absorption lines are much weaker than the atmospheric absorption lines and are barely visible at this scaling factor . this illustrates why a careful multi - step reduction process is necessary to detect h@xmath0 . the fully reduced spectra are shown in figures [ fig2]-[fig5 ] . figures [ fig2 ] and [ fig3 ] contain spectra from sightlines with positive h@xmath0 detections , and figures [ fig4 ] and [ fig5 ] show spectra from sightlines with no h@xmath0 detections or marginal detections . arrows indicate the position of the h@xmath0 doublet expected from previous measurements of the gas velocity along each line of sight . these velocities are given in table [ tbl2 ] along with the atomic or molecular species from which they were determined . in the figures the typical noise level can be judged by the peak - to - peak fluctuations in the continuum well off of the h@xmath0 lines . within several ngstrms of 36675.3 , however , the noise in the reduced spectra is several times larger than elsewhere due to the presence of the strong complex of ch@xmath17 lines , which both reduces the atmospheric transmission and emits excess background . the three spectra in figure [ fig2 ] have the strongest h@xmath0 absorption . in each case the lines match up with the arrows so we are confident in the h@xmath0 detection . in hd 20041 the @xmath15 transition is clearly visible but the @xmath14 transition may have been affected by the aforementioned ch@xmath17 line . both absorption lines are clear and strong in hd 229059 . w40 irs 1a presents a somewhat confusing case because there may be either one or two velocity components . if there is one component , then the two absorption lines to longer wavelengths represent the h@xmath0 doublet and the shorter wavelength line at 36677 is a noise artifact . if there are two absorption components , then the @xmath15 transition from the shorter wavelength doublet overlaps the @xmath14 transition from the longer wavelength doublet . however , the w40 irs 1a spectrum is composed of data from two nights and on only one of those nights does the 36677 feature appear . for this reason we conclude that only the longer wavelength doublet is real . our conclusion is consistent with the observations reported by @xcite in which only one velocity component at about 8 km s@xmath6 lsr was seen in @xmath19co emission , but does not match the 2@xmath162 km s@xmath6 lsr reported by @xcite from @xmath20co absorption . note , however , that if the other line is real then it would correspond to a cloud with a radial velocity of about -30 km s@xmath6 lsr , which disagrees badly with both co measurements . in figure [ fig3 ] the absorption lines are again aligned with the arrows marking previously observed gas velocities . @xmath11 per and hd 21389 have the highest signal to noise ratios ( snr ) and their h@xmath0 lines are easily identified . the doublet in x per is relatively clear , but the velocities found by fitting the individual line profiles differ by 2.5 km s@xmath6 . most likely this is due to noise affecting the absorption feature . in hd 169454 the @xmath15 transition of the main doublet is clear , but the @xmath14 line is rather shallow . as in the case of hd 20041 above , this may be caused by interference from the telluric ch@xmath17 line . however , the velocity we derive for the @xmath15 transition differs from the previously measured cloud velocity by about 3.2 km s@xmath6 . because of the different velocities and the lack of a clear @xmath14 line , we are not as confident as in the previous cases that these features represent h@xmath0 absorption , but still consider it a positive detection . the shorter arrows above the hd 169454 spectrum mark the expected location of the h@xmath0 doublet for a high velocity component at 90 km s@xmath6 reported by @xcite . our spectrum may indicate absorption features at this velocity , but further integration time is needed to determine if the features are real . the absorption features in bd -14 5037 both appear to be double peaked , although the signal - to - noise ratio of each double peak is low . it also seems that there are small absorption features on the shorter wavelength shoulders of the main features . two velocity components reported by @xcite match well with the centers of the double peaked features and the shoulder features , making the detections more believable . there is no definitive evidence for h@xmath0 absorption lines in any spectrum in figure [ fig4 ] , with the possible exception of @xmath21 per . the arrow marking the @xmath15 transition of h@xmath0 in @xmath21 per matches rather well with a statistically significant absorption feature in the spectrum . however , at the expected wavelength of the @xmath14 line the absorption is too weak to attempt a fit given the noise level . both @xmath22 per and @xmath23 per show some amount of absorption near the wavelengths where the h@xmath0 lines are expected , but nothing that can be conclusively identified as a detection . the red rectangle and @xmath11 oph sightlines both have a very high snr . with these clean spectra and no h@xmath0 absorption , it is possible to derive strict upper limits . hd 147889 , 40 per , and @xmath21 sco all have no significant absorption features close to the expected wavelengths . figure [ fig5 ] contains two non - detections in spectra with typical noise levels ( hd 168625 and @xmath24 cep ) and two with high noise levels ( hd 21483 and 62 tau ) . by fitting the absorption lines in the spectra with gaussians we are able to obtain the line of sight velocity , full width at half maximum , equivalent width , and the h@xmath0 column density and its uncertainty . values for these parameters along all of the observed sightlines are given in table [ tbl3 ] . using these values in concert with the steady state approximation and a few reasonable assumptions allows for the calculation of other physical parameters of the diffuse clouds along these lines of sight . below are the three reactions which describe the dominant creation and destruction processes for h@xmath0 . reactions ( [ eq1 ] ) and ( [ eq2 ] ) show the formation process , while reaction ( [ eq3 ] ) shows destruction . @xmath25 @xmath26 @xmath27 first , h@xmath28 is ionized to produce h@xmath29 and an electron . this ionization is assumed to be due to a cosmic - ray because @xcite showed that low energy cosmic - rays will penetrate diffuse clouds while the x - ray flux is attenuated in a thin layer at the cloud exterior . they find that for an x - ray with energy 100 ev the optical depth of the cloud reaches unity at an h@xmath28 column density of @xmath30 @xmath2 . because most of our sightlines have column densities much larger than this we assume that cosmic - rays are the only ionization mechanism operating throughout the majority of the cloud . after being ionized the h@xmath29 ion reacts with h@xmath28 to produce h@xmath0 and h. the second step is many orders of magnitude faster than the first step @xcite , so the formation rate of h@xmath0 is proportional to the product of the ionization rate and h@xmath28 density . in diffuse clouds the primary channel of h@xmath0 destruction is electron recombination , which results in either three h atoms or one h atom and one h@xmath28 molecule . the destruction rate is given by a rate constant times the product of the number densities of h@xmath0 and electrons . the steady state approximation assumes that the formation and destruction rates of h@xmath0 are equal . this approximation yields the equation @xcite @xmath31 where @xmath32(x ) is the number density of species x , @xmath10 is the ionization rate of h@xmath28 , and @xmath33 is the h@xmath0-electron recombination rate constant . in contrast to the steady state approximation , time dependent models developed by @xcite showed that the abundance of h@xmath0 is only weakly dependent on the cosmic - ray ionization rate when a cloud is young . this age is quantified by the ratio of molecular hydrogen to atomic hydrogen @xmath34 where smaller values correspond to younger clouds . the h@xmath0 abundance becomes weakly dependent on the cosmic - ray ionization rate when @xmath35 . this value corresponds to a molecular hydrogen fraction ( defined below in equation ( [ eq6 ] ) ) of @xmath36 . because the h@xmath28 fractions in all of the clouds we observed are more than double this value , we neglect time dependence and use the steady state approximation . assuming that gas is uniformly distributed in the cloud , we can substitute the column density divided by the path length for the number density . by doing this and solving for the ionization rate we obtain @xmath37 we further assume that nearly all hydrogen is either in the atomic or molecular state , and define the molecular hydrogen fraction ( the fraction of hydrogen nuclei in molecular form ) as @xmath38 where the denominator is the number density of hydrogen nuclei , @xmath39 . solving for @xmath40 and plugging the result back into the ionization rate equation we find @xmath41 in this form it is possible to measure or estimate all of the variables on the right hand side of the equation , so we can derive values for the ionization rate of molecular hydrogen . the electron recombination rate constant is given by the equation @xmath42 from @xcite which is valid when the electron temperature , @xmath43 , is between 10 k and 4000 k. while @xmath43 is not directly measured , it can be approximated by the excitation temperature derived from the @xmath44 and @xmath45 levels of molecular hydrogen , @xmath46 . this temperature is calculated from measurements of the column densities of the two levels . in sightlines without these measurements we adopt a value of 60 k. the fact that the observed lines of the @xmath44 and @xmath45 levels of h@xmath28 are saturated indicates that few photons are present in the interior of diffuse clouds to radiatively pump these levels . this means that collisions will dominate the equilibrium between these levels . however , the @xmath44 and @xmath45 levels of h@xmath28 have different nuclear spin configurations and thus require collisions with species such as h@xmath47 or h@xmath0 to interconvert @xcite . if collisions with protons are the dominant factor in determining the relative population of the @xmath44 and @xmath45 levels though , then @xmath46 should represent the proton kinetic temperature and thus the kinetic temperature of the gas in general @xcite . while electrons produced by photoionization may begin with much higher temperatures , they should thermalize quickly via collisions with h@xmath28 @xcite . because @xmath43 and @xmath46 should both nearly equal the kinetic temperature of the gas , we substitute @xmath46 for @xmath43 in equation ( [ eq8 ] ) . assuming that nearly all electrons in diffuse clouds are produced via the ionization of c to c@xmath47 and that nearly all atomic carbon has been singly ionized @xcite , the carbon to hydrogen ratio should approximate the electron to hydrogen ratio . @xcite found this value to be about @xmath48 in multiple diffuse clouds . because of the relative uniformity of this ratio in all six of their sightlines , we adopt a single average value for use in all of our calculations . the molecular hydrogen fraction is dependent on h and h@xmath28 number densities , quantities which surely vary through the cloud , but whose variations are not readily measurable . because fluctuations in number density can not be directly measured , we use column densities in place of the number densities in equation ( [ eq6 ] ) and calculate what @xcite refer to as @xmath49 in clouds where we have measurements of the h and h@xmath28 column densities . however , @xmath49 is most likely an underestimate of @xmath50 in the more molecular regions of the cloud which contain higher concentrations of h@xmath0 . this is because atomic hydrogen is more widely distributed than molecular hydrogen and column densities measure material along the entire line of sight @xcite . since the measurement of @xmath51(h ) includes material not associated with h@xmath28 , @xmath49 underestimates the h@xmath28 fraction in the molecular region . for sightlines where measurements are lacking we use @xmath52 , the value for which the column densities of h and h@xmath28 are equal . when possible , estimates of the number density of hydrogen nuclei , @xmath39 , are adopted from the literature based on various atomic and molecular diagnostics . in eleven of our sightlines @xcite used the observed rotational excitation of c@xmath28 to infer the sum of the h and h@xmath28 number densities . this was done by comparing models with various temperatures and number densities to the measured column densities of all the excited states and choosing the best fit . for sightlines where the average value of @xmath50 is known , they converted @xmath53 to @xmath39 . for sightlines where @xmath50 is not known , our adopted value of @xmath52 is used to perform the conversion . in two additional cases @xcite measured column densities of h and the @xmath54 excited level of h@xmath28 , and with some assumptions estimated the product @xmath55 , where @xmath56 is the rate at which h@xmath28 forms on grains . adopting a typical value for @xmath56 then allowed for the computation of @xmath39 . in one more sightline ( 40 per ) @xcite estimated the thermal pressure from measurements of the @xmath44 , 1 , and 2 fine - structure levels of c for a kinetic temperature of 80 k. using this pressure estimate and the h@xmath28 temperature ( @xmath57 k ) , we calculate @xmath39 . unfortunately , the results obtained for a given sightline by using each of these methods can be significantly different . for example , in the sightline toward @xmath11 oph @xcite , @xcite , and @xcite derived values of 215 @xmath58 , 90 @xmath58 , and 117 @xmath58 , respectively for @xmath39 . because of the uncertainties involved with each method and the different final results , the number densities we use are probably uncertain by about a factor of two . for cases where no number density has been determined , a value of 250 @xmath58 is adopted . again assuming a uniform distribution of gas in each cloud , we divide the total hydrogen column densities by these number densities to obtain pathlengths : @xmath59 in sightlines where no h and h@xmath28 column densities have been determined , the total hydrogen column density is estimated from the color excess and the relation @xmath60 @xmath2 @xcite . as the pathlength is calculated directly from the hydrogen number density , it is also uncertain by about a factor of two . using the above relations and approximations , we calculate the ionization rate of molecular hydrogen , @xmath10 . however , most studies examine the primary cosmic - ray ionization rate per hydrogen atom , @xmath4 . while the ionization efficiencies of h and h@xmath28 are dependent on factors such as the helium abundance and ratio of molecular to atomic hydrogen @xcite , we adopt a more simplified approach and use the conversion factor given by @xcite : @xmath61 this conversion stems from the fact that h@xmath28 contains two hydrogen atoms , so the ionization rate is nearly twice as high . also , @xmath4 only accounts for the initial ( primary ) cosmic - ray ionization while @xmath10 includes ionization from energetic secondary electrons which were created in the first ionization event . with equation ( [ eq10 ] ) we convert our values of @xmath10 to @xmath4 so that they can be directly compared to previous observations . the resulting values for the primary cosmic - ray ionization rate and the specific estimates used for each sightline are shown in table [ tbl4 ] . for completeness , we have performed the same analysis for ten sightlines from @xcite and also included the results in table [ tbl4 ] . with all of the assumptions we have made , it is important to investigate the uncertainties that will propagate to the cosmic - ray ionization rate . substituting equations ( [ eq9 ] ) and ( [ eq10 ] ) into equation ( [ eq7 ] ) gives the primary cosmic - ray ionization rate as @xmath62 . \label{eq11}\ ] ] in this equation , the molecular hydrogen fraction , @xmath50 , and the number density of hydrogen nuclei , @xmath39 , are the two most uncertain parameters ( note that the @xmath39 in the denominator is part of the ratio @xmath63 which is well determined ) . because @xmath4 is directly proportional to @xmath39 , any increase or decrease in @xmath39 produces a corresponding increase or decrease in @xmath4 . as previously mentioned , the uncertainty in @xmath39 is probably about a factor of 2 . on the other hand , @xmath4 is inversely related to @xmath50 . the h@xmath28 fraction is by definition between zero and one , and most of our measured values are around 0.5 . for sightlines with the adopted value of 0.67 , the maximum increase is a factor of 1.5 . we take this factor to be an approximation for the uncertainty in @xmath50 . because @xmath49 should always be an underestimate of @xmath50 in measured sightlines , we only consider increasing the molecular hydrogen fraction for those sightlines . taking into account the uncertainties in both @xmath39 and @xmath50 , the true value of @xmath4 in sightlines with measurements of @xmath49 is likely between one third and twice our derived estimate of @xmath4 . for sightlines with no measured h@xmath28 fraction , we allow @xmath50 to vary both up or down by a factor of 1.5 . this results in a possible cosmic - ray ionization rate between @xmath4/3 and 3@xmath4 . these limits arise when the most extreme variations in both @xmath50 and @xmath39 are substituted into equation ( [ eq11 ] ) . however , @xmath50 tends to be higher when @xmath39 is higher because the rate of h@xmath28 formation scales as the square of @xmath39 . this suggests that it is probable that @xmath50 and @xmath39 will vary in the same way , so that the above analysis most likely overestimates the range of possible ionization rates . another property of the gas to be examined is the excitation temperature , determined from relative populations of the different rotational states of h@xmath0 . for the two lines we have observed the temperature may be determined from the equation @xmath64 taken from @xcite . in this case _ ortho _ refers to the population of the ( 1,0 ) state and _ para _ to the ( 1,1 ) state . the @xmath65 s are statistical weights , @xmath66 is the energy difference between the states , @xmath67 is the boltzmann constant , and @xmath68 is the temperature . excitation temperatures derived from this equation are shown in table [ tbl4 ] . if the rotational ( de-)excitation of h@xmath0 is dominated by collisions with h@xmath28 , then the temperature measurements from both species should be similar . however , this is not the case . most h@xmath28 temperatures are around 60 k while the h@xmath0 temperatures are typically about 30 k. this same discrepancy was described by @xcite . in their model calculation of h@xmath0 thermalization , @xcite have shown that the ( 1,1)/(1,0 ) excitation temperature is always lower than the h@xmath28 temperature because of cooling by fast spontaneous emission from the ( 2,2 ) to ( 1,1 ) state . for the typical cloud conditions in this paper ( @xmath69 @xmath58 , @xmath70 k ) the model of @xcite produces an h@xmath0 excitation temperature of about 50 - 55 k which is significantly higher than the observed values of about 30 k. the source of this discrepancy remains unclear . values of the cosmic - ray ionization rate for diffuse clouds observed here as well as those determined for other diffuse clouds observed previously by us are given in the right hand column of table [ tbl4 ] . the detected values in the lines of sight to fourteen sources cover the range 0.53.2@xmath71 s@xmath6 . upper limits , which are given for fifteen diffuse clouds , are consistent with this range of ionization rates , with the possible exception of hd 168607 . while most of the detections of h@xmath0 are confined to the galactic plane , they are widely dispersed in galactic longitude . we therefore conclude that the values of the cosmic - ray ionization rate listed in table [ tbl4 ] are typical for galactic diffuse interstellar clouds . a few of the sightlines we investigated have been studied previously to derive cosmic - ray ionization rates . all of these studies used column densities of either oh , hd , or both in their calculations . because the formation pathways of oh and hd include the ionization of atomic hydrogen , they can be used to determine the h ionization rate . most of these studies @xcite then derived the primary cosmic - ray ionization rate from the h ionization rate , but @xcite did not because they still considered ionization via x - rays to be important . our values of the primary ionization rate for @xmath11 per , @xmath21 per , @xmath23 per , @xmath22 per , and @xmath11 oph are shown in table [ tbl5 ] along with the rates derived from oh and hd measurements as well as cloud modeling . for @xmath11 per our value is over an order of magnitude larger than those reported by @xcite and @xcite . while the rest of our new measurements in table [ tbl5 ] are only upper limits , these are also typically orders of magnitude larger than previously published values . the only exception is @xmath21 per where both papers cite values of @xmath4 about one fourth to one half our upper limit . various model calculations were performed by @xcite to investigate three of the sightlines that we study here : @xmath11 oph , @xmath11 per , and @xmath21 per . in creating these models they used the most recent measurements of rate constants and the column densities of diagnostic species such as h and h@xmath28 as input parameters . by varying a few uncertain parameters , they would then generate lists of predicted column densities for many atomic and molecular species under slightly different conditions . when their paper was written , it was believed that the h@xmath0-electron recombination rate constant was much lower than the currently accepted value . @xcite reported an upper limit correspnding to @xmath72 @xmath73 s@xmath6 at @xmath74 k , and @xcite lowered the upper limit to @xmath75 @xmath73 s@xmath6 at @xmath76 k. due to the wide range of possible recombination rate constants , @xcite performed calculations using both @xmath72 and @xmath77 @xmath73 s@xmath6 . the cosmic - ray ionization rates from their paper listed in table [ tbl5 ] were computed by determining @xmath4 necessary to reproduce observed oh column densities when @xmath78 @xmath73 s@xmath6 . we choose to compare these ionization rates to ours because we obtain @xmath79 @xmath73 s@xmath6 when @xmath80 k is used as the input temperature in equation ( [ eq8 ] ) . the value of @xmath4 inferred by @xcite is about the same as ours for @xmath11 per , but the lower limits they derived for @xmath11 oph and @xmath21 per are larger than our upper limits for both of those sightlines . for their models that used @xmath81 @xmath73 s@xmath6 , @xcite obtained cosmic - ray ionization rates that are about a factor of 1 to 5 times smaller than ours . from these models they also predicted the column density of h@xmath0 along each sightline . their results are all on the order of @xmath51(h@xmath0 ) @xmath82 @xmath2 , which is a few times larger than the observed column densities or upper limits in any of these sightlines . because @xcite use only a slightly smaller cosmic - ray ionization rate ( corresponding to the formation rate ) but a much smaller recombination rate ( corresponding to the destruction rate ) , their prediction of an h@xmath0 column density similar to observed values seems somewhat serendipitous . in addition to the overestimate of the h@xmath0 column density , a small h@xmath0-electron recombination rate constant may have further consequences . @xcite noted that a small value of @xmath33 may have been responsible for underestimates of the primary cosmic - ray ionization rate in the past . this is because a slower destruction rate requires a slower formation rate to produce a given abundance . in addition to the slow recombination rate , there are some other possible explanations for differences between the cosmic - ray ionization rate inferred from h@xmath0 and those inferred from oh and hd . @xcite pointed out that the rate constant associated with the endothermic charge transfer from h@xmath47 to o varies over the temperatures typically associated with diffuse clouds . this means that the oh production rate is temperature dependent . the ionization rates towards @xmath11 per and @xmath11 oph quoted in @xcite were derived using temperatures of 120 k and 110 k , respectively , for the warm components of the cloud models along each sightline @xcite . as these temperatures are about twice as large as the values determined from h@xmath28 , their oh production is much more efficient . the result is a smaller cosmic - ray ionization rate needed to produce the observed oh column density than if a lower temperature had been used . this problem was addressed by the later models of @xcite where tempertaure and density were varied as functions of cloud depth . @xcite went on to make a comprehensive chemical model of the cloud towards @xmath11 per . they determined the value of @xmath4 that would best reproduce all observed atomic and molecular column densities to be @xmath83 s@xmath6 , which is in good agreement with our estimate of @xmath84 s@xmath6 . the difference in these values may arise because we assume a uniform distribution of gas while @xcite invoke a three phase model which includes diffuse gas , dense gas , and magnetohydrodynamic shocks . two different effects may lead to underestimates of @xmath4 from measurements of hd . the first has to do with an overestimate of the total deuterium to hydrogen ratio @xmath85/@xmath39 . this ratio can be used to estimate the molecular deuterium fraction , @xmath86 . however , the observed values of @xmath86 are about an order of magnitude smaller than those predicted by @xmath85/@xmath39 . to explain this discrepancy , @xcite argued that the atomic deuterium fraction must be larger than the total deuterium fraction . this means that approximating @xmath85/@xmath39 with @xmath87 overestimates the total deuterium to hydrogen ratio . @xcite showed that the cosmic - ray ionization rate is inversely related to the deuterium fraction , so an overestimate of @xmath85/@xmath39 will underestimate @xmath4 . secondly , @xcite emphasized the importance of grain neutralization proposed by @xcite . this process reduces the number of h@xmath47 ions in the gas through charge transfer with small grains . by lowering the abundance of h@xmath47 , the production rate of hd will decrease . this is because hd formation is dependent upon the reaction involving the charge transfer from h@xmath47 to d. since neutralization slows down hd production , a larger value of @xmath4 is needed to create a given abundance than if the effect were not taken into account . @xcite used a model which includes grain neutralization and showed that both h@xmath0 and hd column densities can be reproduced with a single ionization rate of @xmath88 s@xmath6 . since oh formation is dependent on a similar charge transfer reaction , grain neutralization and thus a larger cosmic - ray ionization rate may be necessary in its analysis as well . @xcite studied h@xmath0 in the sightline towards @xmath11 per . using nearly the same analysis as this paper , they inferred a value of @xmath89 s@xmath6 which is equivalent to @xmath90 s@xmath6 shown in table [ tbl5 ] . this higher ionization rate is due to various differences in input parameters . in terms of the parameters in this paper , @xcite used 1.5@xmath33 , 1.2@xmath51(h@xmath0 ) , 1.2@xmath39 , and 0.8@xmath63 for the following reasons . the h@xmath0-electron recombination rate constant differs because they approximated the electron temperature with the h@xmath0 temperature instead of the h@xmath28 temperature in equation ( [ eq8 ] ) . further observations have more than doubled the total integration time so that the spectrum and h@xmath0 column density change slightly between papers . the value of @xmath39 used by @xcite was an average number density computed from various measurements , whereas the value used in this paper comes from the c@xmath28 analysis of @xcite . finally , we have adopted a single value of @xmath63 to be used in all calculations while they used h@xmath28 and c@xmath47 column densities measured towards @xmath11 per . while all of the observations and models above are viable methods for finding the cosmic - ray ionization rate , we believe that the use of h@xmath0 should produce the best results due to its relatively simple chemistry . using either oh or hd to calculate @xmath4 requires more measurements , more assumptions , and more variable parameters than using h@xmath0 . more parameters give the opportunity for a greater uncertainty to accumulate during the calculation . fewer uncertainties coupled with advances in instrumentation lead us to speculate that the cosmic - ray ionization rates inferred from h@xmath0 may be the most accurate to date for diffuse clouds . however , improved estimates of @xmath50 and @xmath39 , the two most uncertain values in our calculations , would make h@xmath0 an even better probe of the cosmic - ray ionization rate . several theoretical calculations of @xmath4 have been performed in the last half - century @xcite . in these papers the authors derived a cosmic - ray ionization rate starting from the observed flux of cosmic - rays in our solar system . unfortunately , there are large uncertainties associated with this method . the cosmic - ray spectrum is well measured above about 1 gev , but lower energy particles are deflected from the inner solar system by the magnetic field coupled to the solar wind . the particles which are most important for ionizing species in diffuse clouds are likely those with energies from about 2 to 10 mev . since this portion of the spectrum can not be directly measured , the flux at low energies must somehow be extrapolated from existing data . @xcite assumed that the power law which applies to the flux of high energy cosmic - rays continues down to 10 mev where the spectrum peaks and then decreases linearly with energy . from these assumptions they derived an ionization rate of @xmath91 s@xmath6 . @xcite , however , fit a curve to measurements of cosmic - rays with energies near 100 mev that also matches the high energy spectrum power law . with this method , their spectrum peaks around 100 mev and falls off for lower energies . the result of using their fit is a lower limit of @xmath92 s@xmath6 . in the same paper they derived an upper limit of @xmath93 s@xmath6 via arguments that low energy cosmic - ray protons are accelerated in type i supernova shells . @xcite used data from the _ pioneer _ and _ voyager _ spacecraft as they travelled outward in the solar system where the weaker solar wind allows for the detection of lower - energy cosmic - rays . these data were then combined with previous observations to infer the interstellar proton spectrum . using this proton spectrum and a heavy nuclei spectrum both with low energy cut - offs at 10 mev and an electron spectrum cut - off below 2 mev , @xcite calculated the primary cosmic - ray ionization rate to be @xmath94 s@xmath6 . our ionization rates fall neatly within the bounds formed by these studies and so are not inconsistent with constraints based on direct cosmic - ray measurements and theoretical particle physics . in contrast to our findings in diffuse clouds , the cosmic - ray ionization rate in dense clouds does seem to agree with the canonical value . observations of h@xmath0 towards dense clouds have found column densities roughly the same as those seen in diffuse clouds @xcite . these measurements have been used to calculate the product @xmath95 . when @xmath10 is taken to be the canonical value of @xmath12 s@xmath6 , the resulting pathlength is on the order of a parsec . this is a typical size for dense clouds as measured by other methods such as extinction mapping . since h@xmath0should be a reliable tracer for the cosmic - ray ionization rate in both environments , there must be some mechanism causing the difference between dense and diffuse clouds . one possibility examined by both @xcite and @xcite is cosmic - ray self - confinement . in this process cosmic - rays generate alfvn waves which can effectively confine the lower energy particles ( @xmath96 mev ) to diffuse material , thus preventing them from entering dense clouds . because cosmic - rays in the 1100 mev range are the most efficient at ionization , self - confinement naturally leads to a higher ionization rate in diffuse clouds than in dense clouds . another possibility is that there is a previously unrecognized high flux of low energy cosmic - rays that can penetrate diffuse but not dense clouds . assuming that typical column densities of diffuse clouds are of order @xmath97 @xmath2 and those of dense clouds are of order @xmath98 @xmath2 , cosmic - rays with energies @xmath99220 mev @xcite would contribute to the ionization rate only in diffuse clouds . as we foresee no observational techniques that would distinguish between these two possibilities , a resolution to this question will depend on more sophisticated theoretical treatments . we have surveyed twenty sightlines and detected h@xmath0 along eight of them . column densities are measured for these eight sightlines , and upper limits set for the remaining twelve . besides a concentration near the galactic plane , there seems to be no clear correlation between location in the sky and detecting h@xmath0 , so it is unlikely that we are observing anomalous regions in the galaxy . instead , finding h@xmath0 in so many sightlines suggests that it is ubiquitous in the diffuse ism . from the h@xmath0 column densities and the steady state approximation , we derive cosmic - ray ionization rates for the nineteen diffuse cloud sightlines in this study along with ten sightlines from @xcite . typical values are on the order of @xmath100 s@xmath6 , which falls within theoretical constraints . while this is an order of magnitude larger than most previously inferred values , there are several possible explanations for the discrepancy . the most likely candidates are rate constants with uncertain measurements and physical or chemical effects not included in past models . newer models that do take into account these factors require cosmic - ray ionization rates very similar to our inferred values . coupled with these models , our widespread detection of h@xmath0 in diffuse clouds supports the idea that the typical cosmic - ray ionization rate in such regions should be revised upward by about an order of magnitude . + the authors would like to thank j. h. black and h. s. liszt for helpful comments and suggestions , and acknowledge the staff of ukirt and the developers of the data reduction packages used in this paper . the united kingdom infrared telescope is operated by the joint astronomy centre on behalf of the u.k . particle physics and astronomy research council . the orac - dr and figaro software packages were provided by the starlink project which is run by cclrc on behalf of pparc . ni and bjm have been supported by nsf grant phy 05 - 55486 . to has been supported by nsf grant phy 03 - 54200 . trg s research is 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 . adams , n. g. , & smith , d. 1987 , in iau symposium 120 , astrochemistry , ed . m. s. vardya & s. p. tarafdar ( dordrecht : reidel ) , 1 black , j. h. , & dalgarno , a. 1977 , , 34 , 405 black , j. h. , hartquist , t. w. , & dalgarno , a. 1978 , , 224 , 448 bohlin , r. c. , savage , b. d. , & drake , j. f. 1978 , , 224 , 132 cardelli , j. a. , meyer , d. m. , jura , m. , & savage , b. d. 1996 , , 467 , 334 chaffee , f. h. , jr . , & white , r. e. 1982 , , 50 , 169 cravens , t. e. , & dalgarno , a. 1978 , , 219 , 750 crutcher , r. m. , & chu , y. h. 1982 , in regions of recent star formation , ed . r. s. roger & p. e. dewdney ( dordrecht : reidel ) , 53 dalgarno , a. , yan , m. , & liu , w. 1999 , , 125 , 237 dalgarno , a. 2006 , pnas , 103 , 12269 federman , s. r. , & lambert , d. l. 1992 , , 104 , 691 federman , s. r. , weber , j. , & lambert , d. l. 1996 , , 463 , 181 geballe , t. r. , mccall , b. j. , hinkle , k. h. , & oka , t. 1999 , , 510 , 251 geballe , t. r. , & oka , t. 1996 , , 384 , 334 glassgold , a. e. , & langer , w. d. 1974 , , 193 , 73 gredel , r. , & mnch , g. 1986 , , 154 , 336 hartquist , t. w. , black , j. h. , & dalgarno , a. 1978a , , 185 , 643 hartquist , t. w. , doyle , h. t. , & dalgarno , a. 1978b , , 68 , 65 hayakawa , s. , nishimura , s. , & takayanagi , t. 1961 , pasj , 13 , 184 hobbs , l. m. , thorburn , j. a. , oka , t. , barentine , j. , snow , t. p. , & york , d. g. 2004 , , 615 , 947 jura , m. 1975 , , 197 , 581 jenkins , e. b. , jura , m. , & loewenstein , m. 1983 , , 270 , 88 lamers , h. j. g. l. m. , de groot , m. , & cassatella , a. 1983 , , 128 , 299 le petit , f. , roueff , e. , & herbst , e. 2004 , , 417 , 993 lepp , s. , dalgarno , a. , van dishoeck , e. f. , & black , j. h. 1988 , , 329 , 418 liszt , h. 2003 , , 398 , 621 liszt , h. s. 2006 , , 452 , 269 liszt , h. s. 2007 , , 461 , 205 mccall , b. j. 2000 , philos . trans . london a , 358 , 2385 mccall , b. j. 2001 , ph.d . thesis , univ . chicago mccall , b. j. , geballe , t. r. , hinkle , k. h. , & oka , t. 1998 , science , 279 , 1910 mccall , b. j. , geballe , t. r. , hinkle , k. h. , & oka , t. 1999 , , 522 , 338 mccall , b. j. , & oka , t. 2000 , , 113 , 3104 mccall , b. j. , et al . 2002 , , 567 , 391 mccall , b. j. , et al . 2003 , , 422 , 500 mccall , b. j. , et al . 2004 , , 70 , 052716 mountain , c. m. , robertson , d. j. , lee , t. j. , & wade , r. 1990 , proc . spie , 1235 , 25 nath , b. b. , & biermann , p. l. 1994 , , 267 , 447 odonnell , e. j. , & watson , w. d. 1974 , , 191 , 89 oka , t. , & epp , e. 2004 , , 613 , 349 padoan , p. , & scalo , j. 2005 , , 624 , l97 pendleton , y. j. , sandford , s. a. , allamandola , l. j. , tielens , a. g. g. m. , & sellgren , k. 1994 , , 437 , 683 rachford , b. l , et al . 2002 , , 577 , 221 rickard , j. j. 1974 , , 31 , 47 savage , b. d. , bohlin , r. c. , drake , j. f. , & budich , w. 1977 , , 216 , 291 schulte , d. h. 1958 , , 128 , 41 shuping , r. y. , snow , t. p. , crutcher , r. , & lutz , b. l. 1999 , , 520 , 149 skilling , j. , & strong , a. w. 1976 , , 53 , 253 smith , d. , & adams , n. g. 1984 , , 284 , l13 snow , t. p. , & mccall , b. j. 2006 , ara&a , 44 , 367 snow , t. p. , jr . , york , d. g. , & welty , d. e. 1977 , , 82 , 113 sonnentrucker , p. , welty , d. e. , thorburn , j. a. , & york , d. g. 2007 , , 168 , 58 spitzer , l. , jr . , & tomasko , m. g. 1968 , , 152 , 971 thorburn , j. a. , et al . 2003 , , 584 , 339 van dishoeck , e. f. , & black , j. h. 1986 , , 62 , 109 webber , w. r. 1998 , , 506 , 329 wegner , w. 1994 , , 270 , 229 welty , d. e. , & hobbs , l. m. 2001 , , 133 , 345 @xmath11 oph & 149757 & 2001 may 24 & @xmath18 lib & 576 + & & 2001 may 25 & @xmath18 lib & 1344 + hd 147889 & 147889 & 2001 may 24 & @xmath101 sco & 5760 + @xmath24 cep & 210839 & 2001 may 24 & @xmath102 lyr & 4320 + hd 169454 & 169454 & 2001 may 25 & @xmath18 lib & 4224 + w40 irs 1a & ... & 2001 may 25 & @xmath103 oph & 1800 + & & 2001 may 26 & @xmath101 sco & 1800 + @xmath21 sco & 147084 & 2001 may 26 & @xmath102 lyr & 960 + hd 168625 & 168625 & 2001 may 26 & @xmath103 oph & 1440 + hd 229059 & 229059 & 2001 may 27 & @xmath102 cyg & 1800 + & & 2001 sept 5 & @xmath102 cyg & 1800 + bd -14 5037 & ... & 2001 may 28 & @xmath103 oph & 2016 + hd 20041 & 20041 & 2001 sept 5 & @xmath18 per & 1728 + hd 21389 & 21389 & 2001 sept 5 & @xmath18 per & 1536 + & & 2002 dec 30 & @xmath101 per & 1152 + & & 2002 dec 31 & @xmath18 per & 1536 + & & 2003 jan 1 & @xmath103 tau & 1152 + @xmath11 per & 24398 & 2001 sept 5 & @xmath18 per & 2304 + & & 2002 dec 30 & @xmath18 per & 1152 + & & 2002 dec 31 & @xmath18 per & 1152 + & & 2003 jan 1 & @xmath101 per & 1440 + @xmath21 per & 23180 & 2002 dec 30 & @xmath18 per & 2304 + & & 2002 dec 31 & @xmath18 per & 1152 + @xmath22 per & 24912 & 2002 dec 30 & @xmath18 per & 1920 + & & 2002 dec 31 & @xmath18 per & 1920 + & & 2003 jan 1 & @xmath101 per & 1536 + x per & 24534 & 2002 dec 31 & @xmath18 per & 1920 + & & 2004 jan 22 & hd 17573 & 3600 + & & 2005 jan 5 & hd 17573 & 4320 + & & 2005 jan 6 & @xmath103 aur & 2880 + & & 2005 mar 3 & @xmath103 aur & 5040 + & & 2005 mar 4 & @xmath103 aur & 5040 + 62 tau & 27778 & 2003 jan 1 & @xmath18 per & 2688 + @xmath23 per & 24760 & 2004 jan 23 & @xmath103 aur & 1800 + & & 2005 jan 5 & @xmath103 aur & 1440 + & & 2005 jan 6 & @xmath103 aur & 1440 + 40 per & 22951 & 2004 jan 23 & hd 17573 & 3600 + & & 2005 jan 6 & @xmath103 aur & 2880 + red rectangle & 44179 & 2005 jan 6 & @xmath104 ori & 720 + hd 21483 & 21483 & 2005 jan 25 & hd 17573 & 5760 + hd 20041 & -1.6 & k i & 1 + hd 21389 & -0.5 & ch & 1 + @xmath11 per & 6.9 & ch & 1 + x per & 6.6 & ch & 1 + hd 169454 & 5.3 & ch & 1 + & 90 & na i & ca ii & 2 + hd 229059 & 4.04 & k i & 3 + bd -14 5037 & 8.2 & c@xmath28 & 4 + & 18.2 & c@xmath28 & 4 + w40 irs 1a & 8 & @xmath19co & 5 + hd 21483 & 10.3 & ch & 1 + 40 per & 6.8 & k i & 3 + @xmath21 per & 7.3 & ch & 1 + @xmath23 per & 2.2 & k i & 6 + @xmath22 per & 0.55 & k i & 6 + 62 tau & 4.8 & ch & 1 + red rectangle & -0.8 & k i & o i & 7 + @xmath21 sco & 2.29 & k i & 6 + hd 147889 & 2.7 & k i & 1 + @xmath11 oph & -1.0 & ch & 1 + hd 168625 & 6 & ca ii & 8 + @xmath24 cep & -1.7 & ch & 1 + hd 20041 & @xmath14 & -1.4 & 8.4 & 0.017 & 0.004 & 0.70 & 0.15 + & @xmath15 & -1.5 & 9.6 & 0.036 & 0.004 & 0.91 & 0.10 + hd 21389 & @xmath14 & -1.9 & 11.9 & 0.009 & 0.002 & 0.39 & 0.07 + & @xmath15 & -0.2 & 15.6 & 0.016 & 0.002 & 0.41 & 0.05 + @xmath11 per & @xmath14 & 7.5 & 9.7 & 0.010 & 0.001 & 0.43 & 0.05 + & @xmath15 & 6.8 & 8.8 & 0.010 & 0.001 & 0.26 & 0.03 + x per & @xmath14 & 8.3 & 11.6 & 0.011 & 0.002 & 0.46 & 0.10 + & @xmath15 & 5.7 & 9.6 & 0.012 & 0.002 & 0.31 & 0.06 + hd 169454 & @xmath14 & 2.6 & 11.1 & 0.005 & 0.002 & 0.21 & 0.08 + & @xmath15 & 2.1 & 10.8 & 0.014 & 0.002 & 0.35 & 0.05 + hd 229059 & @xmath14 & 5.9 & 12.2 & 0.058 & 0.003 & 2.42 & 0.14 + & @xmath15 & 4.4 & 12.4 & 0.059 & 0.003 & 1.48 & 0.09 + bd -14 5037 & @xmath14 & 17.6 & 9.6 & 0.010 & 0.003 & 0.40 & 0.13 + & @xmath15 & 18.6 & 10.4 & 0.009 & 0.003 & 0.23 & 0.08 + w40 irs 1a & @xmath14 & 7.5 & 10.8 & 0.051 & 0.008 & 2.12 & 0.33 + & @xmath15 & 8.2 & 6.6 & 0.050 & 0.006 & 1.26 & 0.16 + hd 21483 & @xmath14 & ... & 10 & @xmath1050.036 & ... & @xmath1051.53 & ... + & @xmath15 & ... & 10 & @xmath1050.036 & ... & @xmath1050.93 & ... + 40 per & @xmath14 & ... & 10 & @xmath1050.015 & ... & @xmath1050.60 & ... + & @xmath15 & ... & 10 & @xmath1050.015 & ... & @xmath1050.36 & ... + @xmath21 per & @xmath14 & ... & 10 & @xmath1050.009 & ... & @xmath1050.42 & ... + & @xmath15 & ... & 10 & @xmath1050.009 & ... & @xmath1050.27 & ... + @xmath23 per & @xmath14 & ... & 10 & @xmath1050.009 & ... & @xmath1050.39 & ... + & @xmath15 & ... & 10 & @xmath1050.009 & ... & @xmath1050.24 & ... + @xmath22 per & @xmath14 & ... & 10 & @xmath1050.009 & ... & @xmath1050.33 & ... + & @xmath15 & ... & 10 & @xmath1050.009 & ... & @xmath1050.21 & ... + 62 tau & @xmath14 & ... & 10 & @xmath1050.045 & ... & @xmath1051.89 & ... + & @xmath15 & ... & 10 & @xmath1050.045 & ... & @xmath1051.14 & ... + red rectangle & @xmath14 & ... & 10 & @xmath1050.006 & ... & @xmath1050.21 & ... + & @xmath15 & ... & 10 & @xmath1050.006 & ... & @xmath1050.15 & ... + @xmath21 sco & @xmath14 & ... & 10 & @xmath1050.009 & ... & @xmath1050.36 & ... + & @xmath15 & ... & 10 & @xmath1050.009 & ... & @xmath1050.21 & ... + hd 147889 & @xmath14 & ... & 10 & @xmath1050.009 & ... & @xmath1050.39 & ... + & @xmath15 & ... & 10 & @xmath1050.009 & ... & @xmath1050.24 & ... + @xmath11 oph & @xmath14 & ... & 10 & @xmath1050.003 & ... & @xmath1050.18 & ... + & @xmath15 & ... & 10 & @xmath1050.003 & ... & @xmath1050.12 & ... + hd 168625 & @xmath14 & ... & 10 & @xmath1050.012 & ... & @xmath1050.54 & ... + & @xmath15 & ... & 10 & @xmath1050.012 & ... & @xmath1050.33 & ... + @xmath24 cep & @xmath14 & ... & 10 & @xmath1050.012 & ... & @xmath1050.54 & ... + & @xmath15 & ... & 10 & @xmath1050.012 & ... & @xmath1050.33 & ... + hd 20041 & 1.6 & 0.18 & 76 & 0.72 & 4.18 & 0.67 & 250 & 5.4 & 2.9 + hd 21389 & 1.0 & 0.08 & 51 & 0.57 & 3.31 & 0.67 & 250 & 4.3 & 1.8 + @xmath11 per & 0.7 & 0.06 & 28 & 0.31 & 1.59 & 0.60 & 215 & 2.4 & 3.2 + x per & 0.8 & 0.17 & 30 & 0.59 & 2.20 & 0.76 & 325 & 2.2 & 3.1 + hd 169454 & 0.6 & 0.09 & 180 & 1.12 & 6.50 & 0.50 & 265 & 7.9 & 0.9 + hd 229059 & 3.9 & 0.16 & 28 & 1.71 & 9.92 & 0.67 & 250 & 13 & 2.9 + bd -14 5037 & 0.6 & 0.16 & 26 & 1.55 & 8.99 & 0.67 & 250 & 12 & 0.5 + w40 irs 1a & 3.4 & 0.37 & 27 & 2.90 & 16.8 & 0.67 & 250 & 22 & 1.5 + wr 104 & 2.3 & 0.25 & 38 & 2.10 & 12.2 & 0.67 & 250 & 16 & 1.4 + wr 118 & 6.5 & 0.18 & 40 & 4.13 & 24.0 & 0.67 & 250 & 31 & 2.0 + wr 121 & 2.2 & 0.28 & ... & 1.68 & 9.74 & 0.67 & 250 & 13 & 1.7 + cyg ob2 12 & 3.8 & 0.36 & 27 & 3.35 & 19.4 & 0.67 & 300 & 21 & 1.8 + cyg ob2 5 & 2.6 & 0.19 & 47 & 1.99 & 11.5 & 0.67 & 225 & 17 & 1.5 + hd 183143 & 2.3 & 0.08 & 31 & 1.28 & 7.42 & 0.67 & 250 & 9.6 & 2.3 + hd 21483 & @xmath1052.2 & ... & ... & 0.56 & 3.25 & 0.67 & 250 & 4.2 & @xmath1055.7 + 40 per & @xmath1050.9 & ... & ... & 0.24 & 1.67 & 0.35 & 80 & 6.7 & @xmath1052.6 + @xmath21 per & @xmath1050.6 & ... & ... & 0.31 & 1.52 & 0.54 & 265 & 1.9 & @xmath1055.0 + @xmath23 per & @xmath1050.5 & ... & ... & 0.09 & 0.35 & 0.19 & 15 & 7.5 & @xmath1052.4 + @xmath22 per & @xmath1050.5 & ... & ... & 0.33 & 1.82 & 0.38 & 300 & 2.0 & @xmath1054.5 + 62 tau & @xmath1052.7 & ... & ... & 0.37 & 2.19 & 0.56 & 280 & 2.5 & @xmath10514 + @xmath21 sco & @xmath1050.5 & ... & ... & 0.73 & 4.23 & 0.67 & 225 & 6.1 & @xmath1050.9 + hd 147889 & @xmath1050.6 & ... & ... & 1.07 & 6.21 & 0.67 & 525 & 3.8 & @xmath1051.6 + @xmath11 oph & @xmath1050.3 & ... & ... & 0.32 & 1.40 & 0.65 & 215 & 2.1 & @xmath1051.5 + hd 168625 & @xmath1050.8 & ... & ... & 1.48 & 8.58 & 0.67 & 250 & 11 & @xmath1050.8 + @xmath24 cep & @xmath1050.8 & ... & ... & 0.57 & 2.80 & 0.49 & 115 & 7.8 & @xmath1051.3 + hd 168607 & @xmath1050.6 & ... & ... & 1.61 & 9.34 & 0.67 & 250 & 12 & @xmath1050.5 + hd 194279 & @xmath1051.2 & ... & ... & 1.22 & 7.08 & 0.67 & 250 & 9.1 & @xmath1051.3 + @xmath106 ori & @xmath1050.7 & ... & ... & 0.44 & 2.55 & 0.67 & 250 & 3.3 & @xmath1052.1 + p cyg & @xmath1050.6 & ... & ... & 0.63 & 3.65 & 0.67 & 250 & 4.7 & @xmath1051.2 + cccccc 1 & 3.2 & @xmath1055.0 & @xmath1052.4 & @xmath1054.5 & @xmath1051.5 + 2 & 0.22 & 2.50 & 0.01 & 0.06 & 0.17 + 3 & 0.17 & 1.30 & ... & @xmath1070.26 & ... + 4 & 1 - 2 & @xmath1088 & ... & ... & @xmath1084 + 5 & 5.2 & ... & ... & ... & ... + 6 & 2.5 & ... & ... & ... & ... +

recent extinction studies of the pipe nebula ( d=130 pc ) reveal many cores spanning a range in mass from 0.2 to 20.4 . these dense cores were identified via their high extinction and comprise a starless population in a very early stage of development . here we present a survey of nh ( 1,1 ) , nh ( 2,2 ) , ccs ( 2 ) , and hcn ( 9,8)emission toward 46 of these cores . an atlas of the 2mass extinction maps is also presented . in total , we detect 63% of the cores in nh ( 1,1 ) , 22% in nh ( 2,2 ) , 28% in ccs , and 9% in hcn emission . we find the cores are associated with dense gas ( 10 ) with 9.5 t 17 k. compared to , we find the nh linewidths are systematically narrower , implying that the nh is tracing the dense component of the gas and that these cores are relatively quiescent . we find no correlation between core linewidth and size . the derived properties of the pipe cores are similar to cores within other low - mass star - forming regions : the only differences are that the pipe cores have weaker nh emision and most show no current star formation as evidenced by the lack of embedded infrared sources . such weak nh emission could arise due to low column densities and abundances or reduced excitation due to relatively low core volume densities . either alternative implies that the cores are relatively young . thus , the pipe cores represent an excellent sample of dense cores in which to study the initial conditions for star formation and the earliest stages of core formation and evolution .

0708.3660

the pipe nebula , at a distance of 130 pc , is one of the closest molecular cloud complexes in the galaxy . with a mass of 10@xmath6@xmath0 , it spans @xmath5 48 degs@xmath10 on the sky but contains little evidence for active star formation . in addition to tracing the larger - scale molecular gas associated with the pipe nebula , a , , and ( 10 ) ( highly undersampled ) survey of the region @xcite detected a small number of dense cores . the densest of these , associated with the dark cloud b59 , appears to be the only site within the pipe nebula where active star formation is occurring . indeed , recent _ spitzer _ observations reveal an embedded cluster of young stellar objects within b59 @xcite . recent extinction studies of the pipe nebula ( @xcite ; alves , lombardi , & lada 2007 ) reveal many compact dust cores that likely represent incipient sites of star formation . these cores were identified using 2mass extinction maps derived from the jhk photometry of over 4.5 million stars ( fig . [ pipe ] ; @xcite ) . the individual compact cores were extracted from the larger - scale , more diffuse lower - column density material associated with the molecular cloud via a wavelet transform technique ( see @xcite ) . in total , 159 cores were identified and span a range in peak a@xmath11 and core mass and are distributed throughout the pipe nebula @xcite . this large sample represents a critical data base for measuring the initial conditions for star formation and testing theories of core formation and evolution . indeed , the dense core mass function ( cmf ) generated from these cores reveals a similarity to the shape of the stellar initial mass function ( imf ; e.g. @xcite ) with the profound implication that the stellar imf is in fact predetermined by the form of the cmf . the origin of the stellar imf may be , therefore , directly linked to the origin of dense cores . while the pipe nebula represents an excellent example of a molecular cloud in a very early stage of star formation , little work has been done to ascertain the dynamical and evolutionary status of its cores . our aim is to study these cores in detail via the combination of their extinction and molecular line properties ; the extinction map reveals measurements of the core masses and sizes while the molecular line observations reveal the kinematics and gas temperatures within the cores . to this end , we are conducting extensive ground- and space - based surveys of the cores located within the pipe nebula . the pipe cores are well suited for studies of core formation and evolution because they are distributed over a large area , are well separated and well defined , and represent a complete sample of dense cores within a single cloud with accurate masses in the mass range of @xmath5 0.2 to 20.4@xmath0 . to begin , we have obtained ( 10 ) observations toward a sample of 94 cores distributed throughout the nebula . the details of these observations and results are presented in @xcite . in brief summary , all cores show strong emission , have measured linewidths , @xmath12v , of @xmath5 0.14 0.61 , and show no size - linewidth correlation . because ( 10 ) emission traces all molecular gas along the line of sight , the measured linewidths will trace both the lower density intercore material in addition to the dense gas associated with the core . to isolate the emission from the dense core , we have also obtained nh@xmath1 observations toward a sample of the pipe cores and report the results of these observations here . because nh@xmath1 requires such high densities to be excited ( @xmath13 10@xmath6@xmath7 ) , only the densest gas will exhibit significant nh@xmath1 emission . thus , nh@xmath1 observations will reveal the fraction of cores in the pipe nebula that contain dense gas . determination of the dense core fraction and how it varies across the core mass function and with kinematical state will provide crucial data for understanding the evolutionary status of these objects . the combination of the extinction and molecular line studies ( and nh@xmath1 ) , and the implications for core formation and evolution , is discussed in @xcite , the third paper in this series on the pipe nebula . by accurately measuring the linewidths from the nh@xmath1 ( 1,1 ) emission in each core , and by combining this with measurements of the core masses , one can determine the fraction of cores that are unbound and the core stabilty , which can provide important clues to their formation and evolution . indeed , @xcite find that the majority of pipe cores are pressure confined , gravitationally unbound objects whose properties are dominated by thermal processes in a pressurized medium . ammonia has been used extensively to study both high- and low - mass star - forming regions ( e.g.@xcite ) . because it is the least depleted species in , and , thus , the most reliable tracer of , cold dense gas within starless cores , we use nh@xmath1 to trace the densest gas in the pipe nebula . the 18 individual hyperfine components of the nh@xmath1 ( 1,1 ) transitions can be used to measure the intrinsic one dimensional velocity dispersion ( @xmath14 ) , optical depth ( @xmath15 ) , and excitation temperature ( t@xmath16 ) . observations of the nh@xmath1 ( 2,2 ) transition were also obtained in order to directly measure the gas temperatures of the brightest cores in the sample . ratios of the nh@xmath1 ( 1,1 ) and ( 2,2 ) lines reveal the kinetic temperature ( t@xmath9 ) and , therefore , density of the cores . to probe the very densest gas and to determine the chemical evolution of the cores , we also obtained ccs ( 2@xmath21@xmath3 ) and hc@xmath4n ( 9,8 ) observations . these molecules are found in very dense starless cores ( e.g. @xcite ) . the abundance of carbon - chain molecules ( e.g. ccs and hc@xmath4n ) compared to nh@xmath1 is thought to trace the chemical evolution of cores ; carbon - chain molecules being more abundant in the very earliest stages in core evolution , compared to nh@xmath1 which is more abundant in the later stages ( e.g. @xcite ) . thus , the combination of these data can reveal the youth and star - forming potential of the cores . in this paper , we present the results of a survey for nh@xmath1 ( 1,1 ) , nh@xmath1 ( 2,2 ) , ccs ( 2@xmath21@xmath3 ) , and hc@xmath4n ( 9,8 ) emission toward a sample of 46 of the extinction cores identified in the pipe nebula . all but one of these cores has been observed as part of the survey @xcite . we find that most cores are indeed associated with dense gas . their measured linewidths are narrow , with several cores having linewidths comparable to their thermal linewidths . the derived properties ( t@xmath9 , intrinsic @xmath12v , non - thermal @xmath12v , radius ) of the cores within the pipe nebula are similar to cores within other low - mass star - forming regions : the only difference is that the pipe cores typically have weaker nh@xmath1 emission and most show no current star formation . the observations were obtained using the 100-m robert c. byrd green bank telescope ( gbt ) from 2006 october to 2007 february . the nh@xmath1 ( 1,1 ) nh@xmath1 ( 2,2 ) ccs ( 2@xmath21@xmath3 ) , and hc@xmath4n ( 9,8 ) transitions were observed simultaneously using the k - band ( upper ) receiver . table [ line - info ] lists the frequencies for each of these transitions . all spectra were obtained in the frequency switched mode ( @xmath17 = 4 mhz ) . the spectrometer was set up with a bandwidth of 12.5 mhz , which produced a frequency resolution of 1.846 khz ( velocity resolution of 0.023 ) . the gbt beam width at these frequencies is @xmath5 30 . due to time constraints , a sub - sample of the complete list of cores identified by @xcite were observed . in total , 46 of the 159 cores were observed and were selected to span a range in peak a@xmath11 , core mass , and location within the pipe nebula . the completeness of our sample based on the original catalog of @xcite is shown in figure [ completeness ] as a function of both peak a@xmath11 and core mass . the peak a@xmath11 for the selected cores ranged from 5.3 to 23.7 magnitudes , while their masses range from 0.4 to 20.4@xmath0 ; all cores in the original sample with peak a@xmath11 @xmath13 14 magnitudes and with masses @xmath13 4@xmath0 were observed . spectra were obtained toward the position of peak extinction for each core ( these positions are listed in table [ table1 ] ) . the cores were originally identified from 2mass extinction maps ( @xmath18 @xmath5 1 ; @xcite ) . because the coordinates listed in the catalog of @xcite refer to the centroid and not necessarily the position of peak extinction , the pointing centers for each core where selected by eye to correspond to the true extinction peak . thus , the coordinates listed here may differ slightly from the positions reported in the catalog of @xcite . toward some of these cores we have also obtained higher - angular resolution extinction maps using the 3.5 m eso new technology telescope ( ntt ; @xmath18 @xmath5 15 ; @xcite ) . in many cases , we found sub - structure at higher - angular resolution . where possible , therefore , we used the higher - angular resolution extinction maps to pinpoint the position of peak extinction . we found that the extinction peaks could differ by as much as @xmath5 1 from the quoted positions in the @xcite catalog . thus , for the cores in which we have obtained these higher - angular resolution extinction maps , we use the coordinates of the peak extinction derived from them . these cores are marked with an asterisk in table [ table1 ] . for the majority of cores , four @xmath5 5 minute integrations were combined for a total of 18 minutes on source . toward the brightest 4 cores only a single 5 minute spectrum was obtained . for three cores showing very faint nh@xmath1 ( 1,1 ) emission , we combined additional data for an on source integration time of over 70 minutes . the integration times , system temperatures ( t@xmath19 ) , source elevation ( el ) , and 1 @xmath20 rms for the spectra are listed in table [ table1 ] . because the system temperatures for the individual frequencies varied by @xmath21 5% , we quote a single system temperature and , hence , 1 @xmath20 rms for each core . the pointing and focus were checked at the beginning of each observing session using a nearby , bright pointing source . the pointing corrections were typically @xmath58 in azimuth and @xmath51 in elevation . the flux density scale was determined from observations of 3c286 ( known flux density of 2.38 jy was taken from @xcite ) . the spectra were converted from the raw data format into fits files using the sdfits package . all subsequent data reduction and analysis was performed using the gbtidl package . the individual polarizations for each spectrum were combined and averaged before opacity and elevation corrections were applied . a first order baseline was removed from all spectra using user - defined baseline windows . all quoted temperatures are t@xmath22 ( k ) . the observations were taken at low elevations ( @xmath5 19@xmath23 ) where the pointing model of the gbt is less reliable . as a result , the temperatures of the observations may be in error , either due to unintended pointing offsets or pointing drifts during the integrations . to quantify the repeatability of the observations , we obtained a spectrum toward the core fest 1 - 457 ( our core 109 ) during each observing session . these data show a variation in peak temperature ( t@xmath22 ) of @xmath5 10% , while the linewidths show a variation of @xmath5 3 % ( these values were estimated by fitting two gaussians to the blended central hyperfine components ) . thus , while the temperatures listed may be uncertain , the linewidths and derived parameters ( e.g. @xmath15 , t@xmath9 , t@xmath16 ) , which are derived from the line ratios and , hence , independent of the absolute scale , will be reliable . the nh@xmath1 ( 1,1 ) , nh@xmath1 ( 2,2 ) , ccs ( 2@xmath21@xmath3 ) , and hc@xmath4n ( 9,8 ) spectra for each core observed in the pipe nebula are shown in the appendix , together with its 2mass extinction map . we find many cores show little or no detectable nh@xmath1 , ccs , nor hc@xmath4n emission ( e.g. fig . many of the remaining cores show only weak nh@xmath1 emission from the central blended hyperfine components ( e.g. fig . a number of cores show weak nh@xmath1 emission , but strong ccs emission ( e.g. fig . only a small fraction of the cores show strong nh@xmath1 emission . toward these cores , we clearly see the satellite components in addition to the central blended hyperfine components ( e.g. figs . 30 and 53 ) . we find 12 cores that show bright nh@xmath1 ( 1,1 ) emission from both the central and satellite hyperfine components . to determine their line parameters , we used a forward - fitting routine ( for a full description see @xcite ) . this forward - fitting routine models the spectrum given input physical properties and then optimizes these values via non - linear least squares minimization . this method is in contrast to the usual way of determinining physical properties directly from the nh@xmath1 ( 1,1 ) to ( 2,2 ) line ratios . the advantage of this method is that the uncertainties for the physical parameters can be determined in addition to the covariance among the parameters . in this model , the gas is assumed to have a slab geometry with uniform properties and that the velocity dispersions and excitation temperatures of the transitions are the same , i.e. the gas is in local thermodynamic equilibrium ( lte ) . assuming the beam filling factor is unity , the spectrum is entirely determined by five parameteres : t@xmath9 , t@xmath16 , @xmath15 , @xmath14 , and v@xmath24 . because the exinction cores are always several times larger than the gbt beam , the assumption that the beam filling factor is unity is valid . the spectra are easily modelled using this approach when the opacity is high . in the case of low optical depth , however , the parameters t@xmath16 and @xmath15 become degenerate and it is not possible to solve for them independently . instead , the fit is optimized for the free parameter and , thus , in this case , the spectrum is determined by four free parameters : t@xmath9 , @xmath25 , @xmath14 , and v@xmath24 . table [ line - parameters-1 ] lists the measured t@xmath22 , v@xmath24 , @xmath14 , and integrated intensity for all cores showing nh@xmath1 ( 1,1 ) emission with detections @xmath13 3@xmath20 . for the 12 bright cores , the v@xmath24 and @xmath14 were determined using the forward - fitting routine discussed above ( these cores are marked with an asterisk in table [ line - parameters-1 ] ) . the quoted t@xmath22 in these cases were measured from a single gaussian fit to the central blended component . the integrated intensity was calculated by summing the emission across the spectrum ( from v=@xmath2623 to 30 ) . toward the remaining 34 cores observed , we find 17 cores with weak nh@xmath1 ( 1,1 ) detections . because the emission is considerably weaker in these cases , only the central blended components were detected ( @xmath13 3@xmath20 ) . thus , to characterize this emission , we fit a single gaussian profile to the observed emission . the t@xmath22 , v@xmath24 , and integrated intensity measured from these fits are included within table [ line - parameters-1 ] . because the central component is a blend of 8 individual hyperfine components , the velocity dispersion determined from fitting a single gaussian profile to the emission will be overestimated . to constrain theories of core formation and evolution , we require accurate estimates of the core velocity dispersions . thus , it was necessary to determine the amount by which this measurement of the velocity dispersion was overestimated . to derive the intrinsic velocity dispersion in the cases where we could not fit the spectra with all 18 hyperfine components , we used a conversion factor calculated from synthetic spectra . the synthetic spectra were generated to cover a range in @xmath14 , t@xmath16 , and @xmath15 . to each of these spectra , we fit a single gaussian profile to the central blended components and then compared the measured velocity dispersion to that which was used to generate the spectrum . to determine the correction factor for all measured velocity dispersions we interpolated the relation to produce a curve converting the measured velocity dispersion to that of the intrinsic velocity dispersion ( our derived relation is similar to that modelled by @xcite ) . thus , the correction factor was determined using this method and then applied to each velocity dispersion measured from a single gaussian fit to the central blended hyperfine components ( 17 cores ) . these corrected velocity dispersions are quoted in table [ line - parameters-1 ] . in total , toward the 46 cores observed , nh@xmath1 ( 1,1 ) emission was detected ( @xmath13 3@xmath20 ) in 29 cases ( @xmath5 63% ) . nh@xmath1 ( 1,1 ) emission was detected toward cores spanning the full range of peak a@xmath11 and core mass within our sample . we find no clear correlation between the measured line parameters and either peak a@xmath11 or core mass . most ( @xmath5 86 % ) of the cores have t@xmath22 @xmath21 1 k. the v@xmath24 measured for the cores range from 3 to 6 consistent with the measured v@xmath24 of the molecular gas associated with pipe nebula ( as traced by ; @xcite ) . this confirms that these cores are indeed located within the large - scale molecular cloud associated with the pipe nebula and not either background or foreground objects . the measured velocity dispersions range from @xmath5 0.05 to 0.3 . for the nh@xmath1 ( 2,2 ) , ccs , and hc@xmath4n emission , we fit a single gaussian profile to all spectra with a @xmath13 3 @xmath20 detection . although the nh@xmath1 ( 2,2 ) emission comprise 21 individual hyperfine components , in all cases only the central 12 blended components were detected . tables [ line - parameters-1 ] and [ line - parameters-2 ] list the peak t@xmath22 , v@xmath24 , @xmath14 , and integrated intensity obtained from these fits . toward the 46 cores observed , nh@xmath1 ( 2,2 ) emission was detected in 10 cores ( @xmath5 22% ) , ccs emission in 13 cores ( @xmath5 28% ) , and hc@xmath4n emission in 4 cores ( @xmath5 9% ) . we find nh@xmath1 ( 2,2 ) detections toward cores spanning the range of peak a@xmath11 and core mass , with the most massive core showing the brightest nh@xmath1 ( 2,2 ) emission . in contrast , we only detect ccs emission for cores with peak a@xmath11 @xmath13 8 magnitudes and hc@xmath4n emission toward cores with peak a@xmath11 @xmath13 12 magnitudes . this is expected given that the critical densities of the ccs and hc@xmath4n transitions ( @xmath5 10@xmath27@xmath7 ) are higher than for nh@xmath1 ( @xmath5 10@xmath6@xmath7 ) and that there may be differences in their relative abundances due to evolutionary effects . the forward - fitting model that was used to derive the @xmath14 and v@xmath24 for the cores showing bright nh@xmath1 emission , also measures the physical properties of the gas . figure [ phys - av - mass ] shows the measured t@xmath9 , @xmath15 , and t@xmath16 for the brightest cores as a function of the peak a@xmath11 and core mass . for the 12 cores showing bright nh@xmath1 emission we find that the mean t@xmath9 is @xmath5 12 @xmath29 2 k. however , for the cores with masses @xmath13 5@xmath0 we find t@xmath9 of @xmath5 10 @xmath29 1 k , while cores with masses @xmath21 5@xmath0 have t@xmath9 of @xmath5 13 @xmath29 3 k ( the errors quoted here correspond to the standard deviation ) . because t@xmath9 is measured from the ratios of the nh@xmath1 ( 1,1 ) to ( 2,2 ) lines , this calculation is independent of assumptions about the beam filling factor . the measured opacities for all the cores range from @xmath5 2 to 10 , with a single core having a @xmath15 of @xmath5 20 ( this is fest 1 - 457 , our core 109 ) . the typical t@xmath16 was @xmath5 5 @xmath29 1 k. in the low optical depth regime , the values of @xmath15 and t@xmath16 become degenerate , so the fitting routine optimizes the value of the parameter @xmath25 ( @xmath30t@xmath16 ) . for the 5 cores that were found to lie in the optically thin regime , the values of @xmath15t@xmath16 are plotted in the lower panel of figure [ phys - av - mass ] . these cores all have masses @xmath21 5@xmath0 and calculated values of @xmath15t@xmath16 ranging from 0.5 to 2 . to verify the values of t@xmath9 output from the forward - fitting model , we have also calculated the rotation temperature , t@xmath31 , which is related to t@xmath9 via @xcite @xmath32 \}^{-1}\ ] ] where t@xmath3 is the energy gap between the two levels ( 41.5 k ) . using this relation we calculate t@xmath31 for the 7 cores with measured @xmath15 values to be @xmath5 10.2 @xmath29 0.9 k. the rotation temperature can also be determined directly from the ratios of the peak t@xmath22 for the nh@xmath1 ( 1,1 ) and ( 2,2 ) emission via the expression @xcite @xmath33 } \ ] ] where @xmath15 is the opacity listed in table [ fitspec ] and t@xmath22(1,1 ) and t@xmath22(2,2 ) are the measured peak temperatures listed in table [ line - parameters-1 ] . using this expression , we derive t@xmath31 of @xmath5 9.5 @xmath29 1.8 k. thus , while we find that the mean t@xmath31 measured directly from the line ratios is slightly lower than the t@xmath31 calculated from the forward - fitting routine , they are consistent within the errors especially considering the peak t@xmath22 are measured from a single gaussian fit to the central blended components . moreover , we find that the measured t@xmath31 is comparable to t@xmath9 within the errors . @xcite show that while t@xmath31 is , in general , lower than t@xmath9 , t@xmath31 is a good indicator of the t@xmath9 for kinetic temperatures @xmath21 20 k. however , t@xmath31 significantly underestimates t@xmath9 for higher temperatures . because our measured t@xmath9 values are always lower than 20 k , the similarity of the derived t@xmath31 and t@xmath9 values is expected . with the determination of t@xmath16 and t@xmath9 , the h@xmath28 volume density , n(h@xmath28 ) , can be calculated via the expression @xcite @xmath34 \left [ 1 + \frac{j_{\nu}(t_{k})}{h\nu / k } \right]\ ] ] where a is the einstein a coefficient ( 1.67@xmath3510@xmath36 s@xmath37 ; @xcite ) , c is the rate coefficient for collision de - excitation ( 8.5@xmath3510@xmath38 @xmath39s@xmath37 ; @xcite ) and @xmath40 is the planck function at the given temperature , where , @xmath41 . table [ column - densities ] lists the n(h@xmath28 ) values for the cores with nh@xmath1 emission sufficiently bright enough to reliably calculate both t@xmath9 and t@xmath16 . for these 7 cores we find the n(h@xmath28 ) values ranged from 0.1 to 2.5 @xmath4210@xmath6 @xmath7 . the measured n(h@xmath28 ) derived from the extinction maps for these 7 cores range from 0.5 to 1.3 @xmath35 10@xmath6@xmath7 . in detail , we find that the n(h@xmath28 ) derived from the nh@xmath1 observations reveal densities typically twice those estimated from the extinction maps . because the n(h@xmath28 ) derived from the extinction maps represents the mean n(h@xmath28 ) within the core , we expect these values to be lower than the n(h@xmath28 ) calculated from the nh@xmath1 emission . with the measured opacity and assuming lte conditions , we calculate the nh@xmath1 column density following the method of @xcite via the expression @xmath43 where z(t@xmath44 ) is the partition function ( 4.417 for t@xmath9 = 10 k ; @xcite ) . for this calculation we assume t@xmath9= 10 k and use the measured values of @xmath15 , t@xmath16 , and @xmath12v ( where @xmath12v = 2.35@xmath14 ) for each core . table [ column - densities ] lists the values obtained for the cores in which we could determine @xmath15 and t@xmath16 ( or @xmath15t@xmath16 ) . for these cores , the n(nh@xmath1 ) was typically @xmath5 4 @xmath35 10@xmath45 @xmath46 . for ccs , hc@xmath4n , and we assume the emission is optically thin and calculate the column density via @xmath47 where j@xmath2 is the rotational quantum number of the lower level , t@xmath48 is the background temperature ( 2.7 k ) , e@xmath49 is the energy of the upper level of the transition , @xmath50 is the dipole moment , and b is the rotational constant ( table [ line - info ] lists the values used for these parameters ) . the quantity @xmath51t@xmath52v is simply the integrated intensity listed from table [ line - parameters-2 ] ( with the appropriate beam efficiency conversion from t@xmath22 to t@xmath53 , which is 0.58 for the gbt at these frequencies ) . for this calculation we assume lte conditions and that t@xmath16=10 k. we find typical n(ccs ) of @xmath55@xmath3510@xmath54 @xmath46 , n(hc@xmath4n ) of @xmath53@xmath3510@xmath54 @xmath46 , and n ( ) of @xmath5 1 @xmath35 10@xmath55@xmath46 ( see table [ column - densities ] ) . the derived column densities will be under estimated if the excitation temperature is lower than the assumed value of 10k . for the 17 cores that were detected in nh@xmath1 ( 1,1 ) but showed no nh@xmath1 ( 2,2 ) emission , t@xmath9 , @xmath15 or t@xmath16 could not be determined using the forward - fitting routine , due to its requirement for emission lines in both spectra . in these cases we also calculate the nh@xmath1 column density assuming that the emission is optically thin using the above expression and that t@xmath16=5 k. this value of t@xmath16 was the typical measured t@xmath16 for the 7 cores for which it could be determined from the nh@xmath1 hyperfine components . for these 17 cores we find the typical n(nh@xmath1 ) was @xmath5 3 @xmath35 10@xmath56@xmath46 . figure [ detection - rates ] shows the detection rate for the nh@xmath1 ( 1,1 ) , nh@xmath1 ( 2,2 ) , ccs , and hc@xmath4n transitions as a function of peak a@xmath11 and core mass . not surprisingly , the detection rate is higher toward those cores with higher peak a@xmath11 and higher mass ; in the case of nh@xmath1 ( 1,1 ) we detected emission toward all cores with a peak a@xmath11@xmath13 15 magnitudes and with masses @xmath13 11@xmath0 . figure [ density ] shows the histogram of the mean h@xmath28 volume densities , n(h@xmath28 ) , for the 46 cores surveyed in nh@xmath1 , obtained from the visual extinction measurements . for this calculation , a spherical volume with radius equal to half the measured size of the core was assumed ( as listed in @xcite ) . the masses were calculated from the background subtracted extinction map ( see @xcite and table [ table1 ] ) . we find the n(h@xmath28 ) for the complete sample of 46 cores range from 0.4 to 1.3 @xmath35 10@xmath6@xmath7 . based on the critical density of nh@xmath1 ( n@xmath57 @xmath5 10@xmath6@xmath7 ) , we crudely estimate that the detection rate should be @xmath5 17% . because these volume densities were derived from the extinction maps , they correspond to the mean volume densities within the cores . the fact that we measure a higher detection rate from our observations ( @xmath5 63% ) is not surprising given that nh@xmath1 traces the densest inner parts of a compact core and that the cores most likely have inwardly increasing density gradients . the detection rate for the nh@xmath1 ( 2,2 ) , ccs , and hc@xmath4n transitions are much lower compared to nh@xmath1 ( 1,1 ) likely due to the combination of the fact that higher densities are typically needed for their excitation and that differences in their chemical evolutionary stages will effect their relative abundances . in the case of ccs , its relative distribution to nh@xmath1 within the cores may differ significantly . in many dense , starless cores the distribution of nh@xmath1 and ccs emission is noticably different ; the nh@xmath1 emission is typically centrally concentrated , while the ccs emission arises from a ring - like structure around the nh@xmath1 core ( e.g. @xcite ) . indeed , morphological differences in the distributions of various depleted molecules , typically carbon bearing species , is clearly seen toward the b68 ( our core 101 ) and fest 1 - 457 ( our core 109 ; @xcite ) . the non - detection of nh@xmath1 emission toward many of the cores could arise because of the short integration time per position or because the spectra were not obtained toward the true density peak of the core . because we are primarily interested in the nh@xmath1 line ratios and velocity dispersions , our results will not be significantly affected by the fact that we may not be pinpointing the peak of the nh@xmath1 emission . this will , however , effect the detection rate as we may be missing some emission that falls below our sensitivity limits given the short integration time per point . this is clearly seen in figure [ comparisons ] which compares an 18 minute integration to a 74 minute integration toward the same position in two cores ( cores 23 and 89 ) . the spectra obtained with the shorter integration time are noiser and , thus , the emission in these cases would not be considered a 3@xmath20 detection ( upper panels ) . the longer integrations ( lower panels ) reveal that indeed nh@xmath1 ( 1,1 ) emission does arise from within these cores , albiet the emission is considerably weaker . given longer integration times per point , the detection rate for these cores would likely increase . because only a single spectrum was obtained toward each core , maps are necessary to determine if many of the non - detections may arise due to either inaccurate coordinates from the extinction maps or pointing errors due to the low observing elevations . for 36 of the cores , higher - angular resolution exinction maps , generated from ntt images ( 15 angular resolution ; @xcite ) , were used to locate the position of peak extinction . for the remaining 13 cores , we used the lower - angular resolution 2mass extinction maps ( 1 angular resolution ; @xcite ) . when comparing the positions of peak extinction between the two maps , we found that the position of peak extinction in the higher - angular resolution maps could differ from that obtained from the lower - angular resolution maps by @xmath5 1 ( @xmath5 2 gbt beams at these frequencies ) . in some cases , the extinction cores identified from the 2mass maps actually consisted of several sub - structures in the higher - angular resolution maps . thus , the positions obtained for the 13 cores from the 2mass exinction map may not reflect the true position of peak extinction , or density , in the core at the angular scale of the gbt beam . however , because the cores are resolved in the 2mass extinction map ( typical sizes of 6 ) , and with respect to the gbt beam , the positions at which the spectra were obtained will still lie within each of the cores . they may not , therefore , necessarily correspond to the true peak of the extinction , density , nor the peak of the nh@xmath1 emission . this will be most obvious if the molecular line emission arises from a very compact , dense core . thus , the non detection in these cases may simply arise due to the fact that we were not pointing at the best position within the core and that the dense cores may be compact relative to the extinction cores . with only a single spectrum toward the cores it is difficult to rule out potential pointing offsets . obtaining nh@xmath1 and ccs maps toward each of the cores will reveal the true density peaks and their relative distributions with respect to the extinction . nevertheless , the majority of these cores appear to be bona fide dense cores . we find that the fraction of dense cores within the pipe nebula , as traced by nh@xmath1 ( 1,1 ) emission , is 100% for all cores with peak a@xmath11 @xmath13 15 magnitudes and mass @xmath13 6@xmath0 . this drops to @xmath5 60% for cores with masses @xmath21 6@xmath0 . deeper observations and mapping could likely raise the detection rate for nh@xmath1 toward these cores . figure [ linewidths ] shows the measured one dimensional velocity dispersions for cores detected in nh@xmath1 ( 2,2 ) , ccs , hc@xmath4n , and as a function of the velocity dispersion measured from the nh@xmath1 ( 1,1 ) emission ( the data are from @xcite ) . for gas with t@xmath9= 10 k the thermal velocity dispersions for these molecules are listed in table [ line - info ] and are marked as dotted lines in figure [ linewidths ] . these data reveal a strong correlation between the nh@xmath1 ( 2,2 ) , hc@xmath4n , , and nh@xmath1 ( 1,1 ) velocity dispersions . in the case of nh@xmath1 ( 2,2 ) , the velocity dispersions are essentially identical to those measured from the nh@xmath1 ( 1,1 ) emission . for the ccs to nh@xmath1 ( 1,1 ) comparison , we see a larger range in the velocity dispersion . we find for hc@xmath4n the velocity dispersions are typically broader than those measured from the nh@xmath1 ( 1,1 ) emission . because hc@xmath4n is heavier than the nh@xmath1 molecule ( and thus has a smaller thermal velocity dispersion ) the fact that the measured velocity dispersion from hc@xmath4n are typically broader than nh@xmath1 , may indicate that this emission is tracing a larger non - thermal component . with only four detections it is difficult to determine this with certainty . we find the velocity dispersions measured from the nh@xmath1 emission are typically narrower than those measured from the : the velocity dispersions are typically 1.4 times broader than nh@xmath1 . this is expected given the emission will trace all the gas along the line of sight , in contrast to nh@xmath1 , which traces only the gas associated with the dense compact core . as a result , the observations will over estimate the non - thermal component of the velocity dispersion . most cores have velocity dispersions which are about twice their thermal velocity dispersions ; the exception being those traced by hc@xmath4n and which show that @xmath14 is @xmath5 3 to 4 times broader than thermal . this suggests that the nh@xmath1 , ccs , hc@xmath4n , and emission is primarily tracing the non - thermal component of the gas . we find that the measured velocity dispersions from the and hc@xmath4n are the comparable for those cores where emission was detected in both transitions . because carbon species are more depleted in the densest inner core regions , it is likely that both these species arise in the outer most core edges . we find three cores that have velocity dispersions comparable to the thermal nh@xmath1 dispersion ( assuming gas at 10 k ) . although these cores have low signal to noise detections , it may be that they are purely dominated by thermal processes . using the unique measurement of t@xmath9 for each of the 12 cores where the nh@xmath1 ( 1,1 ) and ( 2,2 ) lines were sufficiently strong that the spectra could be modelled ( the cores listed in table [ fitspec ] ) , we find that the majority of these cores have measured velocity dispersions @xmath5 1 to 2 times broader than thermal . with this large sample of cores we can also investigate the size - linewidth relation ( e.g. @xcite ) . the data presented here have the advantage of arising from cores within the same cloud which makes comparisons between these properties independent of distance effects . figure [ size - linewidth ] shows the radius versus linewidth plot for cores detected in both and nh@xmath1 ( 1,1 ) . while we see the cores ( marked with open circles , from @xcite ) typically have broader linewidths compared to the nh@xmath1 cores ( marked as filled circles ) , the general trend between these two samples is similar : there appears to be no correlation between the core linewidth and size . as will be discussed in more detail in @xcite , the relatively flat distribution of the data within figure [ size - linewidth ] suggests , therefore , that the cores are dominated by thermal rather than turbulent motions . this is also seen toward dense cores in other star forming regions ( e.g. @xcite ) . past nh@xmath1 surveys of star forming regions such as taurus , ophiuchus , perseus , orion , and cepheus , reveal that the measured properties within these regions showed general trends based on their star formation activity ( see @xcite for a review ) . for instance , the regions dominated by dense young clusters ( e.g. orion and cepheus ) tend to have higher measured t@xmath9 , linewidths , and core masses . in contrast , taurus and ophiuchus have lower values for these properties , however , these cores tend to have higher column densities . perseus , being intermediate in its star - formation activity , also shows intermediate values for the derived nh@xmath1 properties ( see @xcite and referenecs therein ) . many of these studies find strong nh@xmath1 emission and a high detection rate toward the dense cores . however , many of the nh@xmath1 cores within these surveys are associated with _ iras _ sources ; in the case of orion , @xcite found _ iras _ sources coincident with 68% of the nh@xmath1 cores . such a high fraction of cores with obvious infrared emission suggests that star formation is already occurring within these cores . thus , the detection of strong nh@xmath1 emission in a starless core may imply that the star formation process is imminent . this is consistent with chemical evolutionary models that predict nh@xmath1 to be most abundant in the later evolutionary stages of core formation @xcite . our lower nh@xmath1 detection rate , compared to other nh@xmath1 surveys of star - forming regions , is consistent with the fact that the pipe cores have lower densities and most show little signs of active star formation . indeed , the core associated with b59 , the only site within the pipe nebula where current star formation is known to be occurring , is one of the few cores that shows very strong nh@xmath1 emission . although the detected nh@xmath1 emission was typically weak toward our sample of cores , these cores do harbor the densest gas within the pipe nebula . moreover , those cores that show strong ccs emission may be the very youngest within the region . comparing our measured values for t@xmath9 , the intrinsic @xmath12v , the non - thermal @xmath12v , radius , and column density ( fig . [ property - comparisons ] ) to those measured in other star forming regions reveals that the cores in the pipe nebula are most similar to cores within perseus ( see figures within @xcite ) . the obvious difference between the perseus cores and those within the pipe nebula is the derived column densities and the star formation activity : the pipe cores show lower column densities and very little star formation . the lower nh@xmath1 column densities derived for the pipe cores could arise because of differences in the intrinsic extinction , abundance of nh@xmath1 , or excitation conditions compared to other star - forming regions . because the cores within the pipe nebula span a range in peak extinction and core mass similar to other star - forming regions , it seems unlikey that the lower measured column densities arise simply due to differences in their measured extinction . moreover , the measured column densities of the cores show an order of magnitude greater range ( 10@xmath5610@xmath55@xmath46 ) compared to the peak extinction ( 520mags ) . a more likely scenario is that the intrinsic abundance of nh@xmath1 is low within many of these cores , possibly due to their relative chemical youth . it may be , however , that the excitation conditions with these cores are insufficient to significantly populate the levels . it is possible that our derived nh@xmath1 column densities are under estimates and that nh@xmath1 is more abundant within the cores , but simply not excited to the levels required for strong detections . this is consistent with the relatively low mean densities of the cores ( @xmath5 7 @xmath35 10@xmath58@xmath7 ) compared to that necessary to sufficiently populate the observed levels ( @xmath5 10@xmath6@xmath7 ) . detailed radiative transfer modelling is required to distinguish between these scenarios . if we restrict our sample of cores to those with similar column densities as measured toward other star - forming regions ( i.e. n(nh@xmath1 ) @xmath13 10@xmath45@xmath7 ) , we find that the cores within the pipe share similar properties with cores in both perseus and ophiuchus . given that only 6 cores were detected in the pipe with such high column densities , it is difficult to determine the distribution in their properties with any certainty . nevertheless , it appears that the cores in the pipe nebula are similar to those found in other low - mass star forming regions . the most noticable difference is that the majority of pipe cores tend to show weaker nh@xmath1 emission and have little star - formation activity . the chemical characteristics within dense cold cores can reveal their evolutionary stage . because they have different production chemistries , the column densities of nh@xmath1 , ccs , hc@xmath4n , and can reveal the chemical characteristics in dense cores ( e.g. @xcite ) . for instance , models for core evolution @xcite predict an anti - correlation between the abundance of ccs and nh@xmath1 ; with ccs more abundant in the early evolutionary phases . because of the evolutionary effects of depletion and the differences in the molecular distributions within the cores , a single measurement of the column density within the core can only provide limited information on the chemical structure and evolutionary stage . nevertheless , we find that the comparisons between the column densities of the various species observed within the pipe cores are similar to what is seen toward dense cores in taurus and ophiuchus @xcite . in particular , we find that : ( 1 ) the n(nh@xmath1 ) and n(ccs ) show no clear correlation , which is consistent with differences in the densities of the cores ( 0.52.5@xmath3510@xmath6@xmath7 ) ; ( 2 ) the n ( ) is essentially constant for all values of the n(ccs ) , which implies that the ccs is a more reliable tracer of denser gas ; and ( 3 ) that n(ccs ) and n(hc@xmath4n ) are correlated , which is expected given that these carbon - chain molecules are produced under the same chemical conditions within a core @xcite . we also find high abundance ratios of [ nh@xmath1 ] compared to [ ] for the densest cores , confirming that nh@xmath1 is a better tracer of dense gas than and that traces the outer core envelope and is depleted in the core centers . because the chemical abundances trace the evolutionary age of the cores , we can use the fractional abundances of these molecules to estimate the evolutionary ages of the cores . using the model calculations of @xcite , we estimate the cores are at an evolutionary age of @xmath5 6 @xmath35 10@xmath27 years . using the 100 m gbt we have conducted an nh@xmath1 ( 1,1 ) , nh@xmath1 ( 2,2 ) , ccs ( 2@xmath21@xmath3 ) , and hc@xmath4n ( 9,8 ) survey toward a sample of 46 cores within the pipe nebula . these cores were identified via their high extinction and appear to be dense starless cores which likely represent the initial conditions of star formation . in total , we detect nh@xmath1 ( 1,1 ) emission in 29 cores ( @xmath5 63% ) , nh@xmath1 ( 2,2 ) emission in 10 cores ( @xmath5 22% ) , ccs emission in 13 cores ( @xmath5 28% ) , and hc@xmath4n emission in 4 cores ( @xmath5 9% ) . not surprisingly , the detection rate is higher toward those cores with higher peak a@xmath11 and higher mass ; in the case of nh@xmath1 ( 1,1 ) we detected emission toward all cores with a peak a@xmath11@xmath13 15 magnitudes and with masses @xmath13 11@xmath0 . we find that the cores are associated with dense ( @xmath1310@xmath6@xmath7 ) gas . the nh@xmath1 emission revealed n(h@xmath28 ) typically three times greater than the mean densities measured from the extinction maps . this is expected given that the cores likely have inwardly increasing density gradients ; the nh@xmath1 is tracing the densest centers of the cores . using a forward - fitting modelling routine , we estimated the physical properties of the gas ( t@xmath9 , @xmath15 , and t@xmath16 ) for 12 cores . we find that t@xmath9 ranges from 9.5 to 17 k , with a median value of @xmath5 12 @xmath29 2 k. cores with higher mass tend to have lower t@xmath9 values ( 10 @xmath29 1 k ) compared to lower - mass cores ( 13 @xmath29 3 k ) . the measured one dimensional velocity dispersion , @xmath14 , for the cores show that the lines are narrow , with several cores having dispersions close to thermal ( for a 10 k gas ) . a comparison between the linewidths measured for the cores from and nh@xmath1 , reveal that while the general trends are the same , i.e. neither dataset show a size - linewidth relation , the nh@xmath1 linewidths are systematically lower than those measured from the emission . this arises because the traces all emission along the line of sight , whereas nh@xmath1 traces only the densest gas associated with the core . the derived properties of the pipe cores are similar to dense cores within other star - forming regions , i.e.perseus , taurus , ophiuchus . the most obvious difference is that the cores within the pipe nebula typically have weaker nh@xmath1 emission and most show no evidence for current star formation . the weak nh@xmath1 emission may arise due to the low abundance of nh@xmath1 because the cores are chemically young or because the excitation conditions ( densities ) within the gas are insufficient to excite the observed transitions . detailed radiative transfer models are needed to resolve this issue . although we have a limited number of cores , we estimate their evolutionary ages to be @xmath5 6 @xmath35 10@xmath27 years . thus , the cores within the pipe nebula represent an excellent sample of dense cores in which to study the initial conditions of star - formation and the earliest stages of core formation and evolution . we are extremely grateful to erik rosolowsky for the use of the nh@xmath1 forward - fitting routine ahead of publication and for the many informative discussions . we also thank the support staff at the gbt , in particular frank ghigo , for help will the observing setup and data reduction . we acknowledge funding support through nasa origins grant nag-13041 . the 2mass extinction map and nh@xmath1 ( 1,1 ) , nh@xmath1 ( 2,2 ) , ccs ( 2@xmath21@xmath3 ) , and hc@xmath4n ( 9,8 ) spectra for each core are shown in figs [ appendix - fig-12a ] to 55 . labelled in the top left corner of each extinction image is the core number based on the catalog of @xcite . in all cases , the extinction images are 025 @xmath35 025 and are centered on the peak extinction for the core ( as determined from these maps ) . the grey scale is from an a@xmath11 of 0 to 20 magnitudes . the contour levels are 1.2 , 4 , 6.8 , 9.6 , and 12.4 magnitudes ( 3@xmath20 in steps of 7@xmath20 ) . the final image in the appendix , figure 56 , summarises the details of the extinction images and includes the sizes of the 2mass extinction image angular resolution and the gbt beam . each spectrum is centered on the v@xmath24 of the core ( see tables [ line - parameters-1 ] and [ line - parameters-2 ] ) . the nh@xmath1 ( 1,1 ) spectra extend @xmath29 25 while the remaining three spectra extend @xmath29 8 around the core s v@xmath24 . the temperature scale is t@xmath22 . llccccc & & @xmath60 & @xmath50 & e@xmath49 & b & thermal @xmath14 + & & ( ghz ) & ( d ) & ( cm@xmath37 ) & ( ghz ) & ( ) + nh@xmath1 & ( j , k ) = ( 1,1 ) & 23.694 & 1.47 & 16.25 & 298.117 & 0.07 + nh@xmath1 & ( j , k ) = ( 2,2 ) & 22.733 & 1.47 & 45.08 & 298.117 & 0.07 + ccs & j@xmath62 = 2@xmath21@xmath3 & 22.344 & 2.81 & 1.12 & 6.478 & 0.04 + hc@xmath4n & j = 98 & 23.963 & 4.33 & 4.00 & 1.331 & 0.03 + & j = 10 & 109.782 & 0.11 & 3.66 & 54.890 & 0.05 + crrrr core & & & & + & & & & + 12 & 11.3 @xmath29 0.1 & 10.2 @xmath29 0.2 & 7.4 @xmath29 0.0 & + 14 & 12.0 @xmath29 0.4 & 1.0 @xmath29 0.3 & 4.4 @xmath29 0.4 & + 17 & 17.5 @xmath29 1.3 & & & 0.5 @xmath29 0.0 + 20 & 15.2 @xmath29 0.5 & & & 1.7 @xmath29 0.1 + 40 & 10.3 @xmath29 0.5 & 2.4 @xmath29 0.4 & 5.3 @xmath29 0.3 & + 41 & 14.3 @xmath29 0.7 & & & 1.6 @xmath29 0.1 + 42 & 12.4 @xmath29 0.4 & & & 1.9 @xmath29 0.0 + 47 & 12.6 @xmath29 1.0 & 2.5 @xmath29 0.8 & 3.2 @xmath29 0.1 & + 87 & 9.8 @xmath29 0.3 & 6.3 @xmath29 0.3 & 5.4 @xmath29 0.1 & + 101 & 10.4 @xmath29 0.2 & 5.1 @xmath29 0.2 & 4.5 @xmath29 0.1 & + 108 & 14.1 @xmath29 0.8 & & & 1.0 @xmath29 0.0 + 109 & 9.5 @xmath29 0.1 & 19.8 @xmath29 0.3 & 6.6 @xmath29 0.0 & + cccccc core & log[n(h@xmath28 ) ] & log[n(nh@xmath1 ) ] & log[n(ccs ) ] & log[n(hc@xmath4n ) ] & log[n ( ) ] + & ( @xmath63 ) & ( @xmath64 ) & ( @xmath64 ) & ( @xmath64 ) & ( @xmath64 ) + 6 & & 13.7 & 12.4 & & 15.1 + 7 & & 13.2 & & & 15.1 + 8 & & 13.3 & & & 15.1 + 12 & 4.4 & 15.2 & 13.1 & 12.9 & + 14 & 3.7 & 14.0 & 12.4 & & 15.3 + 15 & & 13.4 & & & 15.2 + 17 & & 13.3 & & & 14.5 + 20 & & 13.6 & & & 14.9 + 22 & & 13.2 & & & 14.7 + 23 & & 13.0 & & & 14.8 + 25 & & 13.3 & & & 14.5 + 37 & & & 12.3 & & 15.0 + 40 & 4.0 & 14.3 & 13.0 & 12.4 & 15.1 + 41 & & 13.4 & & & 15.0 + 42 & & 13.5 & 12.2 & & 15.2 + 47 & 3.0 & 14.2 & & & 15.1 + 65 & & 13.6 & & & 15.2 + 70 & & 13.3 & & & 15.2 + 87 & 4.0 & 14.8 & 13.0 & & 15.1 + 89 & & 13.1 & 12.5 & & 15.1 + 91 & & 13.2 & & & 14.9 + 92 & & 13.4 & 12.5 & & 15.2 + 93 & & 13.8 & 12.7 & & 15.2 + 97 & & 13.8 & & & 14.8 + 101 & 3.7 & 14.5 & 12.9 & 12.6 & 14.9 + 102 & & 13.4 & 12.7 & & 14.9 + 108 & & 13.4 & & & 14.6 + 109 & 4.4 & 15.2 & 12.7 & 12.3 & 15.2 + 113 & & 13.2 & & & 14.9 + 132 & & 13.4 & & & 15.1 + the 2mass extinction map of the pipe nebula @xcite . this map was derived from the jhk photometry of over 4.5 million stars . many compact cores are clearly seen in this map in addition to the larger - scale , more diffuse lower - column density material associated with the molecular cloud . @xcite identified 159 compact cores within this region . the 46 cores that comprise our source list are marked with circles . these cores were selected to span a range in peak a@xmath11 , core mass , and location within the pipe nebula . we find that the majority of these are associated with dense gas.,scaledwidth=80.0% ] the completeness of the observed sample of cores given the original catalog of @xcite as a function of peak a@xmath11 ( left ) and core mass ( right ) . all cores from the original catalog of @xcite with peak a@xmath11@xmath13 14 magnitudes and with masses @xmath13 4@xmath0 were observed.,title="fig:",scaledwidth=45.0% ] the completeness of the observed sample of cores given the original catalog of @xcite as a function of peak a@xmath11 ( left ) and core mass ( right ) . all cores from the original catalog of @xcite with peak a@xmath11@xmath13 14 magnitudes and with masses @xmath13 4@xmath0 were observed.,title="fig:",scaledwidth=45.0% ] the derived parameters , t@xmath9 , @xmath15 , and t@xmath16 , output from the forward - fitting modelling of the cores showing strong nh@xmath1 emission as a function of peak a@xmath11 ( left ) and core mass ( right ) . for cores with low opacities , values of @xmath15 and t@xmath16 can not be determined independently . the lower panels shows the values of the parameter @xmath15t@xmath16 which is used to characterize the emission in the case of low optical depths.,title="fig:",scaledwidth=45.0% ] the derived parameters , t@xmath9 , @xmath15 , and t@xmath16 , output from the forward - fitting modelling of the cores showing strong nh@xmath1 emission as a function of peak a@xmath11 ( left ) and core mass ( right ) . for cores with low opacities , values of @xmath15 and t@xmath16 can not be determined independently . the lower panels shows the values of the parameter @xmath15t@xmath16 which is used to characterize the emission in the case of low optical depths.,title="fig:",scaledwidth=45.0% ] the detection rate for each molecular transition as a function of peak a@xmath11 ( left ) and core mass ( right ) . we detect all nh@xmath1 ( 1,1 ) emission toward cores with a peak a@xmath11@xmath13 15 magnitudes and with masses @xmath13 11 @xmath0.,title="fig:",scaledwidth=45.0% ] the detection rate for each molecular transition as a function of peak a@xmath11 ( left ) and core mass ( right ) . we detect all nh@xmath1 ( 1,1 ) emission toward cores with a peak a@xmath11@xmath13 15 magnitudes and with masses @xmath13 11 @xmath0.,title="fig:",scaledwidth=45.0% ] the mean volume h@xmath28 density , n(h@xmath28 ) , calculated from the visual extinction measurements for the 46 cores within our sample ( open histogram ) . the diagonal lines mark the n(h@xmath28 ) measured from the extinction maps for the cores detected in nh@xmath1 ( 1,1 ) . the dotted line marks the critical density of the nh@xmath1 ( 1,1 ) transition . we find @xmath5 17 % of the cores have mean volume densities greater than the critical density for nh@xmath1 . our higher measured nh@xmath1 ( 1,1 ) detection rate ( 63% ) is expected considering that nh@xmath1 ( 1,1 ) traces the densest inner regions within the cores . this is in contrast to the n(h@xmath28 ) measured from the extinction maps which corresponds to the mean volume density within the larger extinction core.,scaledwidth=60.0% ] the comparison between nh@xmath1 ( 1,1 ) spectra obtained with differing integration times ( top : t @xmath5 18 minutes , bottom : t @xmath5 74 minutes ) for two cores ( 23 and 89 , left and right respectively ) . the spectra toward the majority of cores were obtained with an integration time of @xmath5 18 minutes ( c.f . top panels ) . in these two cases , therefore , we would not have detected nh@xmath1 ( 1,1 ) emission . thus , the non - detection of nh@xmath1 ( 1,1 ) emission may simply be due to the short integration times . with longer integration times per point the detection rate for these cores would likely increase.,title="fig:",scaledwidth=45.0% ] the comparison between nh@xmath1 ( 1,1 ) spectra obtained with differing integration times ( top : t @xmath5 18 minutes , bottom : t @xmath5 74 minutes ) for two cores ( 23 and 89 , left and right respectively ) . the spectra toward the majority of cores were obtained with an integration time of @xmath5 18 minutes ( c.f . top panels ) . in these two cases , therefore , we would not have detected nh@xmath1 ( 1,1 ) emission . the non - detection of nh@xmath1 ( 1,1 ) emission may simply be due to the short integration times . with longer integration times per point the detection rate for these cores would likely increase.,title="fig:",scaledwidth=45.0% ] measured one dimensional velocity dispersion for cores with nh@xmath1 ( 2,2 ) , ccs , hc@xmath4n , and emission compared to those measured from the nh@xmath1 ( 1,1 ) emission . all axes are in units of . the solid line traces equal values of the dispersions . the dotted lines trace the thermal velocity dispersion for each molecule ( assuming 10 k gas).,scaledwidth=45.0% ] core radius versus linewidth for all cores detected in ( open circles ) and nh@xmath1(closed circles ) emission . while the linewidths measured from the are typically broader than those measured from nh@xmath1 emission , there is no correlation between the radius and linewidth in either species for the cores within the pipe nebula . the dotted and dashed lines marks the thermal linewidths for nh@xmath1and respectively.,scaledwidth=60.0% ] histograms of the derived properties for the cores within the pipe nebula . comparisions between these histograms and similar ones for other star forming regions ( e.g. taurus , ophiuchus , perseus , orion , and cepheus ) reveal the pipe cores are most similar to cores within perseus and opiuchus . the obvious exception is that the pipe cores have lower column densities and that the region contains very little active star formation.,title="fig:",scaledwidth=45.0% ] + histograms of the derived properties for the cores within the pipe nebula . comparisions between these histograms and similar ones for other star forming regions ( e.g. taurus , ophiuchus , perseus , orion , and cepheus ) reveal the pipe cores are most similar to cores within perseus and opiuchus . the obvious exception is that the pipe cores have lower column densities and that the region contains very little active star formation.,title="fig:",scaledwidth=45.0% ] + histograms of the derived properties for the cores within the pipe nebula . comparisions between these histograms and similar ones for other star forming regions ( e.g. taurus , ophiuchus , perseus , orion , and cepheus ) reveal the pipe cores are most similar to cores within perseus and opiuchus . the obvious exception is that the pipe cores have lower column densities and that the region contains very little active star formation.,title="fig:",scaledwidth=45.0% ] + histograms of the derived properties for the cores within the pipe nebula . comparisions between these histograms and similar ones for other star forming regions ( e.g. taurus , ophiuchus , perseus , orion , and cepheus ) reveal the pipe cores are most similar to cores within perseus and opiuchus . the obvious exception is that the pipe cores have lower column densities and that the region contains very little active star formation.,title="fig:",scaledwidth=45.0% ] + histograms of the derived properties for the cores within the pipe nebula . comparisions between these histograms and similar ones for other star forming regions ( e.g. taurus , ophiuchus , perseus , orion , and cepheus ) reveal the pipe cores are most similar to cores within perseus and opiuchus . the obvious exception is that the pipe cores have lower column densities and that the region contains very little active star formation.,title="fig:",scaledwidth=45.0% ]

deleting a hyperplane from a polar space associated with a symplectic polarity we get a specific , symplectic , affine polar space . similar geometry , called an arises as a result of generalization of the notion of an alternating form to a semiform . some properties of these two geometries are given and their automorphism groups are characterized . mathematics subject classification ( 2010 ) : 51a50 , 51a10 . + key words : affine polar space , symplectic form , alternating map , automorphism . ( # 1,#2)*l*_t ( # 1,#2)*l*_m # 1 # 1 # 1#2_#2 ( # 1,#2) ( # 1,#2) ( # 1,#2) ( # 1,#2) ( # 1,#2) ( # 1,#2) ( # 1,#2) ( # 1,#2) ( # 1,#2 ) # 1c[#1 ] # 1a[#1 ] [ section ] [ thm]lemma [ thm]corollary [ thm]proposition [ thm]fact [ thm]remark [ thm]example [ thm]remark [ thm]problem # 1[#1 ] ) [ 1 ] * example - continuation [ # 1]- * # 1#2[#1]-[#2 ] [ 1 ] * example [ # 1 ] . , a continuation of [ # 1 ] * # 1#2[#1].[#2 ]

1203.2662

in @xcite affine polar spaces are derived from polar spaces the same way as affine spaces are derived from projective spaces , i.e. by deleting a hyperplane from a polar space embedded into a projective space . so , affine polar spaces ( apses , in short ) are embeddable in affine spaces and this let us think of them as of suitable reducts of affine spaces . in general we have two types of affine polar spaces . structures of the first type are associated with polar spaces determined by sesquilinear forms ; one can loosely say : these are stereographical projections of quadrics " . they can also be thought of as determined by sesquilinear forms defined on vector spaces which represent respective affine spaces . these structures and adjacency of their subspaces were studied in @xcite . contrary to @xcite , in this approach minkowskian geometry is not excluded . in particular , the result of @xcite generalizes alexandrov - zeeman theorems originally concerning adjacency _ of points _ of an affine polar space ( cf . @xcite , @xcite ) . the second class of affine polar spaces , which is included in @xcite but is excluded from @xcite , consists of structures associated with polar spaces determined by symplectic polarities . the aim of this paper is to present in some detail the geometry of the structures in this class from view of the affine space in which they are embedded . the position of the class of thus obtained structures let us call them symplectic affine polar spaces is in many points a particular one . firstly , symplectic affine polar spaces are associated with null - systems , quite well known polarities in projective spaces with all points selfconjugate . so , symplectic apses have famous parents . moreover , in each even - dimensional pappian projective space such a ( projectively unique ) polarity exists . thus a symplectic aps is not an exceptional space , but conversely , it is also a canonical " one in each admissible dimension . a second argument refers to the position of the class of symplectic apses in the class of all apses . as an affine polar space is obtained by deleting a hyperplane from a polar space , while the latter is realized as a quadric in a metric projective space , the derived aps appears as a fragment of the derived affine space . if the underlying form that determines the polar space is symmetric then the corresponding aps can be realized _ on _ an affine space in one case only when the deleted hyperplane is a tangent one . and then the affine space in question is constructed _ not _ as a reduct of the surrounding projective space but as a derived space as it is done in the context of chain geometry ( cf . @xcite , @xcite ) . moreover , such an aps can be represented without the whole machinery of polar spaces : it is the structure of isotropic lines of a metric affine space . the only case when the point set of the reduct of a polar space _ is _ the point set of an affine space arises when we start from a null system i.e. in the case considered in the paper . but such an aps is not associated with a metric affine space i.e with a vector space endowed with a nondegenerate bilinear symmetric form . what is a natural analytic way in which a symplectic aps can be represented , when its point set is represented via a vector space ? a way to do so is proposed in our paper : to this aim we consider a `` metric '' , a binary scalar - valued operation defined on vectors . it is not a metric , in particular , it is not symmetric , and it is not invariant under affine translations . nevertheless , it suffices to characterize respective geometry . symplectic apses have famous parents but they have also remarkable relatives . although the `` metric '' : the analytical characteristic invariant of symplectic apses is not a form , it is closely related to forms . loosely speaking , it is a sum of an alternating form @xmath8 defined on a subspace and an affine vector atlas defined on a vector complement of the domain of @xmath8 . immediate generalization with ` an alternating map ' substituted in place of ` an alternating form ' comes to mind . such a definition of a map may seem artificial , the resulting maps , which we call _ semiforms _ , have quite nice synthetic characterization though . their basic properties are established in section [ sec : semiforms ] . a symplectic ` metric ' appears to be merely a special instance of such a general definition and many problems concerning it ( so as to mention a characterization of the automorphism group ) can be solved in this general setting easier . to illustrate and to motivate such a general definition we show in [ exm : prod ] a semiform associated with a vector product , that yields also an interesting geometry . on the other hand , this geometry has close connections ( see ) with a class of hyperbolic polar spaces . a semiform induces an incidence geometry that we call an _ _ ( cf . and ) . it is a @xmath9-space with affine spaces as its singular subspaces ( cf . [ thm : gamma ] ) , and with generalized null - systems comprised by lines and planes through a fixed point ( cf . [ lem : geowiazki:0 ] ; comp . a class with similar properties considered in @xcite ) . in the paper we do not go any deeper into details of neither geometry of semiforms nor geometries other than symplectic apses . we rather concentrate on `` apses and around '' . finally , we pass to our third group of arguments : that geometry of symplectic apses is interesting on its own right . geometry of affine polar spaces is , by definition , an incidence geometry i.e. an aps is ( as it was defined both in @xcite and @xcite ) a partial linear space : a structure with points and lines . from the results of @xcite we get that geometry of symplectic affine polar spaces can be also formulated in terms of binary collinearity of points an analogue of the alexandrov - zeeman theorem . a characterization of apses as suitable graphs is not known , though . the affine polar spaces associated with metric affine spaces ( as it was sketched above ) can be , in a natural consequence , characterized in the `` metric '' language of line orthogonality or equidistance relation inherited from the underlying metric affine structure . it is impossible to investigate a line orthogonality imposed on an affine structure so as it gives rise to a symplectic aps . however , in case of a symplectic aps a `` metric '' mentioned above determines an `` equidistance '' relation which can be used as a primitive notion to characterize the geometry . there is no general commonly accepted axiom system of a weak equidistance relation ( of a congruence of segments , in other words ) . a very natural one , that characterizes metric affine spaces is presented in @xcite . roughly speaking , in accordance with that approach a congruence of segments is an equivalence relation on pairs of points such that bisector hyperplanes are really affine hyperplanes . but these properties are met by our `` symplectic equidistance '' as well . the difference is that a segment and its translate need not be congruent under our equidistance . in this paper we do not intend to give a characterization of symplectic apses in the language of equidistance . nevertheless , we think it is worth to stress on that this is also a possible language for this geometry and to indicate similarities and dissimilarities between our equidistance and that used in metric affine geometry . since our equidistance is not commutative we have two types of bisectors ( cf . ) and thus two types of symmetry under a hyperplane ( cf . [ thm : symmetry ] ) . the first type of a relation of being symmetric wrt . a hyperplane is related to translations and the other type is a central symmetry . recall that the affine space @xmath10 defined over a vector space @xmath11 has the vectors of @xmath11 as its points and the cosets of the @xmath12-dimensional subspaces of @xmath11 as its lines . we write @xmath13 for the ( affine ) translation on the vector @xmath14 , @xmath15 . let @xmath16 be a vector space over a ( commutative ) field @xmath17 with characteristic @xmath18 and let @xmath19 be a nondegenerate bilinear reflexive form defined on @xmath16 . assume that the form @xmath19 has finite index @xmath20 and @xmath21 . we will write @xmath22 for the class of all vector subspaces of @xmath16 and @xmath23 for the class of all @xmath24-dimensional subspaces . in the projective space @xmath25 the form @xmath19 determines the polarity @xmath26 . we write @xmath27 for the class of isotropic subspaces of @xmath16 : @xmath28 assume that @xmath29 . the structure @xmath30 is referred to as the _ polar space _ determined by @xmath31 in @xmath32 . this section may look superfluous from view of symplectic polar spaces but it is used later in an example which justifies our general construction of semiforms . now let @xmath19 be symmetric and @xmath33 be the orthogonality determined by @xmath19 on @xmath34 . set @xmath35 , @xmath36 , and @xmath37\in y\colon u\perp v\}$ ] . then there is a nondegenerate form @xmath38 on @xmath39 such that @xmath40 . note that @xmath41 is a maximal isotropic subspace of @xmath42 and thus the geometry @xmath43 is a , so called , hyperbolic polar space ( or a hyperbolic quadric following ( * ? ? ? 1.3.4 , p. 30 also @xcite ) . now let @xmath44 be a maximal ( i.e. a @xmath45-dimensional ) singular subspace of a hyperbolic polar space @xmath46 of index @xmath47 and let @xmath48 be the structure obtained by deleting the subspace @xmath44 from @xmath46 . in particular we write @xmath49 . [ thm : hippolreduct ] the hyperbolic polar space @xmath46 is definable in its reduct @xmath50 . we need to recover from @xmath50 the points and lines of @xmath44 which are missing to get @xmath46 . let @xmath51 be the family of the maximal singular subspaces of @xmath46 and @xmath52 be the family of maximal singular subspaces of @xmath50 . it is seen that @xmath53 , and thus each element of @xmath52 carries the geometry of a slit space ( cf . @xcite , @xcite ) . write @xmath54 and @xmath55 . so , * @xmath56 consists of the elements of @xmath52 which carry the affine geometry . each @xmath57 determines on @xmath44 a @xmath58-subspace of its `` improper '' points . * @xmath59 consists of the elements of @xmath52 which carry the geometry of a punctured projective space . each @xmath60 determines on @xmath44 a point : its `` improper '' point . next , we need some theory of hyperbolic polar spaces ( cf . 1.3.6 , p. 35 ) ) . in the class @xmath51 we define the relation : @xmath61 iff @xmath62 ; it is an equivalence relation . take @xmath63 , @xmath64 , and @xmath65 , @xmath66 where @xmath67 . set @xmath68 , @xmath69 . then @xmath70 we write @xmath71 when the right - hand of holds . assume first that @xmath72 . then @xmath73 . from assumptions @xmath74 . in case @xmath75 we have @xmath76 , so @xmath77 . hence @xmath78 which means that @xmath79 share a line @xmath80 . clearly , @xmath81 , so @xmath82 is a required line of @xmath50 . in case @xmath83 we get that @xmath84 and thus @xmath85 and the above reasoning can be applied again . conversely , assume that @xmath86 share a line @xmath87 ; then @xmath82 for a line of @xmath46 contained in @xmath79 . take @xmath88 : the suitable improper point . so @xmath89 . on the other hand @xmath90 is the unique improper point of @xmath91 , i.e. @xmath92 . now , consider the relation @xmath93 defined in the class @xmath59 as follows @xmath94 note that this is an equivalence relation and its equivalence classes can be identified with points on @xmath44 . in turn , as @xmath46 is of type d ( i.e. it is hyperbolic ) , the elements of @xmath56 can be identified with the hyperplanes of @xmath44 quite naturally . that way , in terms of @xmath50 , we get an incidence structure with points and hyperplanes of @xmath44 . using standard methods we are able to recover lines of @xmath44 from this incidence structure which makes the proof complete . from now on @xmath19 is a nondegenerate symplectic form of index @xmath20 . then @xmath95 . assume that @xmath29 . the polar space @xmath96 is frequently referred to as a _ null system _ @xcite , ( * ? ? ? since @xmath19 is symplectic , @xmath97 so , the point sets of @xmath46 and of @xmath32 coincide . let @xmath98 be a hyperplane of @xmath46 ( cf . @xcite ) ; then @xmath98 is determined by a hyperplane @xmath99 of @xmath32 ; on the other hand @xmath99 is a polar hyperplane of a point @xmath100 of @xmath32 i.e. @xmath101 . finally , @xmath102 is the set of all the points that are collinear in @xmath46 with the point @xmath100 of @xmath46 . the _ affine polar space @xmath103 derived from ( @xmath46,@xmath100 ) _ is the restriction of @xmath46 to the complement of @xmath99 ; in view of the above the point set of @xmath103 is the point set of the affine space @xmath104 obtained from @xmath32 by deleting its hyperplane @xmath99 . the set @xmath105 of lines of @xmath103 is a subclass of the set @xmath106 of the lines of @xmath104 . moreover , the parallelism of the lines in @xmath105 defined as in @xcite ( two lines are parallel iff they intersect in @xmath98 ) coincides with the parallelism of @xmath104 restricted to @xmath105 . clearly , not every line of @xmath32 that is not contained in @xmath99 and which crosses @xmath99 in a point @xmath107 is isotropic . moreover , none of the lines of @xmath32 through @xmath100 which is not contained in @xmath99 is isotropic . for this reason , in every direction of @xmath104 , except the one determined by @xmath100 , there is a pair of parallel lines in @xmath104 such that one of them is isotropic and the other is not . in this exceptional direction no line is isotropic . in @xcite affine polar spaces determined in metric affine spaces associated with symmetric forms were studied . slightly similar interpretation of @xmath103 can be given here as well . recall that there is a basis of @xmath16 in which the form @xmath19 is given by the formula @xmath108 . we write @xmath109 for the vector subspace spanned by @xmath110 and @xmath111 $ ] for the vector with coordinates @xmath112 ( in some cases @xmath113 , @xmath114 ... may be vectors too ) . let us take @xmath115 ; then @xmath99 is characterized by the condition @xmath116 iff @xmath117 . we write @xmath11 for the subspace of @xmath16 characterized by @xmath118 ; note that the restriction @xmath8 of @xmath19 to @xmath11 is also a nondegenerate symplectic form . we can write @xmath119 and then for scalars @xmath120 and vectors @xmath121 of @xmath11 we have @xmath122,[a_2,b_2,u_2 ] ) = a_1b_2 - a_2b_1 + \eta(u_1,u_2).\ ] ] moreover , @xmath123 where @xmath124 . a point @xmath125 $ ] ( @xmath126 in @xmath17 , @xmath127 ) of can be identified with the subspace @xmath128 of @xmath16 , and the ( affine ) direction of the line @xmath125 + \vgen{b , w}$ ] is identified with the ( projective ) point @xmath129 . [ lem : anal : izolines ] let @xmath130 + \vgen{b , w}$ ] with @xmath131 , @xmath132 be a line of . then @xmath133 the projective completion @xmath134 of @xmath87 in @xmath32 is determined by the pair @xmath135 , @xmath136 of points of @xmath32 . then @xmath137 iff @xmath138 . with we get the claim . for @xmath139,[a_2,u_2]\in{{\mathbb{y}}}$ ] we define @xmath140,[a_2,u_2 ] ) : = \eta(u_1,u_2 ) - ( a_1 - a_2).\ ] ] [ lem : anal : colin ] let @xmath141 $ ] , @xmath142 $ ] with @xmath143 , @xmath144 be a pair of points of . then @xmath145 as in [ lem : anal : izolines ] , we embed given points into @xmath32 ; then @xmath146 corresponds to @xmath147 . since @xmath148 are collinear iff the projective line which joins @xmath149 is in @xmath46 we get that @xmath148 are collinear iff @xmath150 . with we have the claim . in what follows we write @xmath151 if points @xmath152 of are collinear in . from [ lem : anal : izolines ] and [ lem : anal : colin ] we learn that the affine polar space @xmath103 can be defined entirely in terms of a vector space @xmath11 over @xmath17 and a nondegenerate symplectic form @xmath8 on @xmath11 . note that the surrounding affine space is definable in terms of the geometry of . the result is a simple consequence of elementary properties of ( symplectic ) polar spaces , but it is important from the view of `` foundations '' : _ the geometry of symplectic apses can be expressed in the language of the relation @xmath153_. [ prop : afpol2af ] let @xmath154 be two distinct points of @xmath104 . then @xmath155 is the line of @xmath104 that passes through @xmath156 . consequently , the structure is definable in terms of the binary collinearity of @xmath103 and thus it is definable in as well . it suffices to note that @xmath157 is a polar hyperplane of a point @xmath114 and if @xmath158 , then @xmath159 . we can generalize [ lem : anal : izolines ] to @xmath24-dimensional subspaces ( @xmath160 ) . it is seen that @xmath24-dimensional strong subspaces of , i.e. elements of @xmath161 , are obtained by deleting @xmath99 from a strong subspace of @xmath46 ; on the other hand , a @xmath24-dimensional strong subspace @xmath162 of @xmath46 is determined by an element of @xmath163 which , on the other hand , determines a subspace @xmath164 of . an arbitrary @xmath24-dimensional subspace @xmath164 of has the form @xmath125 + y'$ ] , where @xmath165 , and it is obtained from the vector subspace @xmath166 of @xmath16 . then @xmath167 iff two conditions are satisfied : a. [ war1 ] @xmath168,[0,b , w ] ) = 0 $ ] for every @xmath169,[b , w]\in y'$ ] and b. [ war2 ] @xmath170,[0,b , w ] ) = 0 $ ] for every @xmath171\in y'$ ] . it is seen that is equivalent to @xmath172 , and is equivalent to @xmath173 . in more geometrical terms we can say that @xmath174 iff the set @xmath175 of improper points of @xmath164 is in @xmath176 and @xmath177 for a point @xmath178 and every improper point @xmath179 . interpretation of the isotropic ( singular ) subspaces of higher dimensions in @xmath103 that makes use of the map @xmath180 , analogous to [ lem : anal : izolines ] , remains , clearly , valid . an interested reader can consider this problem as an easy exercise . the construction of the function @xmath180 in falls into the following more general one . [ def : cosik ] let @xmath11 , @xmath181 be vector spaces over a ( commutative ) field @xmath17 with @xmath182 . let @xmath183 be their sets of vectors and @xmath184 be their zero - vectors , respectively . -2pt [ cosik : zal1 ] let @xmath185 be an alternating bilinear map . then @xmath186 and @xmath187 for all vectors @xmath188 . [ cosik : zal2 ] let @xmath189 be a map that satisfies the following conditions 1 . [ delt4 ] @xmath190 , 2 . [ delt3 ] @xmath191 , 3 . [ delt2 ] @xmath192 . for all scalars @xmath193 and @xmath194 . set @xmath195 and @xmath196 . on @xmath39 we define the binary operation @xmath197 by the formula @xmath198,[v_2,u_2 ] ) : = \eta(u_1,u_2 ) - \delta(v_1,v_2).\ ] ] the resulting map @xmath199 is referred to as a _ semiform _ defined on @xmath200 . an alternating bilinear form @xmath8 considered in [ def : cosik ] is _ nondegenerate _ when for each @xmath201 there is @xmath202 such that @xmath203 . the following technical but important formulas are immediate from definition . let @xmath204 $ ] , @xmath205 $ ] . @xmath206 one example is crucial : [ exm : syaps ] let @xmath8 be a null - form defined on @xmath11 . next , let @xmath207 and @xmath208 . definition [ def : cosik ] coincides with the definition of the function @xmath180 in . [ exm : power ] each alternating map @xmath209 is derived from a linear map @xmath210 by the formula @xmath211 ( see any standard textbook , e.g. ( * ? ? ? xix ) ) . we shall write , generally , ( cf . ) @xmath212 for the map defined by @xmath213 . it is a folklore that @xmath214 , where @xmath215 . note that when @xmath216 is fixed then the set @xmath217 is a @xmath45-dimensional vector subspace of @xmath218 . clearly , the operation @xmath219 together with a given @xmath31 determines via a semiform . [ exm : prod ] let @xmath11 be a @xmath220-dimensional vector space . then @xmath221 and we can write @xmath222 , where @xmath223 is a vector product defined on @xmath11 . a standard formula defining @xmath224 is the following : @xmath225\vecprod [ \alpha''_1,\alpha''_2,\alpha''_3 ] = \left [ \varepsilon_1 \left|\begin{array}{ll } \alpha'_2 & \alpha'_3 \\ \alpha''_2 & \alpha''_3 \end{array}\right| , \varepsilon_2 \left|\begin{array}{ll } \alpha'_1 & \alpha'_3 \\ \alpha''_1 & \alpha''_3 \end{array}\right| , \varepsilon_3 \left|\begin{array}{ll } \alpha'_1 & \alpha'_2 \\ \alpha''_1 & \alpha''_2 \end{array}\right| \right]\ ] ] with @xmath226 ( cf . @xcite , @xcite ) . then @xmath199 defined on @xmath227 by the formula @xmath228,[v_2,u_2 ] ) = u_1 \vecprod u_2 - ( v_1 - v_2)\ ] ] is a semiform . in this and the forthcoming subsections [ ssec : synthetic ] and [ ssec : canonical ] most of the proofs consist in direct computations and therefore they are left for the reader . let us give a more explicit representation of a map @xmath31 characterized in [ def : cosik ] . [ lem1:delt ] let @xmath31 meet conditions - of [ def : cosik ] . then the following conditions follow as well : 1 . [ delt0 ] @xmath229 ( by ) ; 2 . [ delt00 ] @xmath230 ( by , ) ; 3 . [ delt1 ] @xmath231 ( by , ) ; 4 . [ delt5 ] @xmath232 ( by , ) ; for all @xmath194 . define @xmath233 by the formula @xmath234 . then @xmath235 is a linear map and @xmath31 is characterized by the formula @xmath236 ad : let @xmath237 be arbitrary . then @xmath238 ad : @xmath239 ad : @xmath240 ad : @xmath241 to prove we compute simply as follows : @xmath242 a map @xmath31 defined by formula is called _ an affine atlas _ , it is _ nondegenerate _ when @xmath235 is an injection ( i.e. if @xmath243 is trivial ) . note that when @xmath244 and @xmath31 is nondegenerate then the representing map @xmath235 is a surjection as well . the following is straightforward [ lem2:delt ] let @xmath233 be a linear map and @xmath31 be defined by . then @xmath31 meets conditions - of [ def : cosik ] . finally , we note that affine atlases can be equivalently characterized by another , less elegant but more convenient for our further characterizations , set of postulates . [ lem3:delt ] let @xmath31 satisfy the postulates , , , of [ def : cosik ] , [ lem1:delt ] . then @xmath31 satisfies as well . by the conditions assumed , the map @xmath245 is linear . then we have @xmath246 from this is easily computed . let @xmath247 , @xmath248 be vector spaces with the common field @xmath17 of scalars . let @xmath249 be a map . consider the following properties : 1 . [ ax1 ] @xmath250 for each @xmath251 . [ ax2 ] if @xmath252 then @xmath253 for each scalar @xmath193 and each vector @xmath254 . [ ax3 ] if @xmath252 then @xmath255 . 4 . [ ax2 - 3 ] if @xmath256 then there is @xmath254 with @xmath257 and @xmath258 . 5 . [ ax4 - 5 ] @xmath259 [ ax4 ] @xmath260 $ ] implies @xmath261 $ ] . [ ax5 ] @xmath262 [ ax7 ] for each @xmath263 there is @xmath264 such that @xmath265 and @xmath266 for all @xmath267 . in view of formulas it is evident that axioms are satisfied by each semiform as defined in . set @xmath268 . with each @xmath264 we associate the map @xmath269 note that if @xmath199 is a semiform defined in [ def : cosik ] then @xmath270 and @xmath271 . recall that in one of the most intensively investigated cases in geometry when we consider a sesquilinear form @xmath199 , @xmath272 , @xmath273 is a linear map , and @xmath274 is semilinear . our axioms lead to a similar situation . [ lem : ax1 ] if @xmath199 satisfies then @xmath275 for each @xmath276 . consequently , @xmath277 . clear : @xmath278 , which gives the claim . [ lem : ax2 + 3 ] assume axiom . let @xmath276 . -2pt [ ax3x ] if @xmath273 is additive then @xmath279 . [ ax2x ] if @xmath273 is multiplicative then @xmath280 . consequently , if axioms and are valid then the map @xmath273 is linear iff @xmath279 . in particular ( cf . [ lem : ax1 ] ) , @xmath281 is a linear map , i.e. the following hold : @xmath282 clearly , @xmath283 and thus @xmath284 is a subspace of @xmath200 . if , moreover , axiom is valid then the assignment @xmath285 is injective . ad : let @xmath273 be additive . we compute @xmath286 therefore , @xmath280 . ad : let @xmath273 be multiplicative . then @xmath287 foregoing consequences of the above are evident . in particular , axiom can be read as ` _ @xmath288 yields @xmath289 _ ' , so the assignment @xmath290 has trivial kernel . [ lem : ax4 + 5 ] -2pt [ ax4x ] set @xmath291 \}$ ] . + then @xmath292 and the set @xmath293 is closed under vector addition . [ ax4a ] assume axiom . then @xmath294 iff the condition @xmath295 holds for all @xmath296 . [ ax5x ] set @xmath297 \}$ ] . + if axiom is valid then the set @xmath298 is closed under scalar multiplication . if axiom is adopted then @xmath299 . consequently , if axioms , , and are valid then @xmath300 is a vector subspace of @xmath200 . moreover , if axioms - are valid then @xmath301 . ad : let @xmath302 and @xmath264 be arbitrary . we have in turn @xmath303 thus @xmath304 . ad : the right - to - left implication is evident , just substitute @xmath305 . the converse implication follows by axiom . ad : let @xmath306 , @xmath264 and a scalar @xmath307 be arbitrary . then @xmath308 therefore , @xmath309 . finally , suppose that @xmath310 , so @xmath311 is a well defined linear map . since @xmath312 we have @xmath313 for each @xmath264 . therefore , @xmath314 and thus @xmath315 . thus @xmath316 . [ lem : rekonstrukcja ] with the axioms - , axiom can be expressed as the following statement : @xmath317 . assume that the axioms - are valid and set @xmath318 , @xmath319 . then @xmath8 is an alternating nondegenerate vector - valued form . the map @xmath31 is a nondegenerate affine atlas ; it is determined by a linear injection @xmath320 by the formula . two first statements are evident in view of [ lem : ax2 + 3 ] . by [ lem : ax4 + 5 ] , @xmath31 meets conditions , , and , since @xmath321 , it satisfies ( note : @xmath322 ) so , by [ lem3:delt ] , @xmath31 satisfies - of [ def : cosik ] . in view of [ lem1:delt ] , @xmath31 is determined by a linear map @xmath320 , @xmath323 , and @xmath324 for @xmath325 . finally , to prove that the map @xmath31 is nondegenerate let @xmath326 for an @xmath327 . by definition , @xmath328 , so @xmath329 . from [ lem : ax4 + 5 ] , @xmath330 and thus @xmath243 is trivial . [ wzorek ] let @xmath331 with @xmath332 , @xmath333 for @xmath334 . then , we have ( cf . ) the following @xmath335 let us compute @xmath336 this is exactly our claim . summing up the above , with not too tedious computation , we close this part by the following representation theorem let @xmath337 be a map . the following conditions are equivalent . -2pt @xmath199 is a semiform defined in accordance with [ def : cosik ] , where @xmath338 are nondegenerate . @xmath199 satisfies axioms - . a nondegenerate semiform @xmath199 is scalar valued ( i.e. @xmath339 ) iff it is associated with a symplectic polar space . let @xmath340 . then the determinant is a symplectic form . therefore the map below ( @xmath341 are elements of the field of scalars @xmath17 of @xmath11 ) @xmath342,[y_1,y_2,y_3 ] ) = \left| \begin{array}{ll } x_2 & x_3 \\ y_2 & y_3 \end{array}\right| - ( x_1 - y_1)\ ] ] is a semiform . the associated aps is determined by the so called line complex in the @xmath220-dimensional projective space over @xmath17 ( cf . 6 ) , ( * ? ? ? 3 ) ) . forthcoming constructions are provided for a fixed nondegenerate semiform @xmath199 defined in [ def : cosik ] . moreover , we assume that @xmath343 . set @xmath344 . let @xmath148 be vectors of @xmath200 , so @xmath345 $ ] , @xmath346 , @xmath347 . by definition , @xmath348 for suitable maps @xmath8 , @xmath235 ( recall , they need to be nondegenerate . as a consequence , @xmath349 ) . there is , generally , a great variety of semiforms . but some of them may lead to isomorphic geometries . write @xmath350 for @xmath199 defined by [ def : cosik ] with @xmath31 defined by . we have evident [ prop : uzmiennic ] [ uzmiennic : war1 ] there is a linear bijection @xmath351 such that for any @xmath352 it holds : @xmath353 [ uzmiennic : war2 ] let @xmath354 , @xmath355 be a non zero scalar . then , clearly , the map @xmath356 defined by @xmath357 is an alternating form . there is a linear bijection @xmath358 such that the following holds for any @xmath352 @xmath359 . we set @xmath360 $ ] in case and @xmath361 ) = [ v , b(u)]$ ] in case . in terms of [ exm : power ] we have @xmath362 . in view of [ prop : uzmiennic ] , till the end of our paper we assume that @xmath199 is defined by the formula of the form @xmath363,[v_2,u_2 ] ) = \eta(u_1,u_2 ) - ( v_1 - v_2).\ ] ] [ [ ssec : afsempol ] ] s ~ imitating [ lem : anal : colin ] , for points @xmath364 of @xmath104 , we put generally @xmath365 from definition it is immediate that @xmath366 . [ lem : lines : adjclosed ] let @xmath148 be two distinct points of and @xmath367 . if @xmath368 then @xmath369 for all @xmath370 . let us write @xmath371 $ ] , @xmath372 , and @xmath373 $ ] . from assumption , @xmath374 . we directly compute that then @xmath375 . this yields @xmath376 for any scalars @xmath377 and closes the proof . in view of [ lem : lines : adjclosed ] , the relation @xmath378 determines the class @xmath105 of lines of by the condition @xmath379 for computation it is convenient to have this criteria ( comp . [ lem : anal : izolines ] ) : @xmath380+\gen{[v , u]}\in\izolines\text { iff } \eta(u_0 , u ) = -v.\ ] ] indeed , if we take @xmath381 $ ] and @xmath382 $ ] , then the line in question is @xmath383 and by we have @xmath384 iff @xmath385 . by the right hand side becomes the equality @xmath386 which trivially gives the required one . it is a straightfoward consequence of and . the class @xmath105 induces the incidence structure @xmath387 that we will take a look into . let us call this structure the _ _ determined by @xmath199 . [ lem : propg ] [ lem : propg : parallel ] the class @xmath105 is unclosed under parallelism , i.e. for every @xmath388 there is an affine line @xmath389 such that @xmath390 . [ lem : propg::alashult ] let @xmath391 , @xmath392 , and @xmath393 . if @xmath87 is an affine line through @xmath90 from the affine plane @xmath394 , then @xmath384 . : let @xmath395+\gen{[v , u]}$ ] , where @xmath396 . suppose that @xmath397 for all @xmath398 . this yields that @xmath399 for all @xmath400 . so take any @xmath401 and note that @xmath402 for some @xmath400 and @xmath403 . this gives @xmath404 . thus @xmath405 , and hence @xmath406 as @xmath8 is nondegenerate . finally , @xmath407 $ ] is the zero of @xmath200 . straightforward computation . : without loss of generality we can assume that @xmath408 $ ] and @xmath409 where @xmath410 , @xmath411 . then @xmath412 for some @xmath413 . applying to @xmath414 and then to @xmath87 we are through . [ thm : gamma ] the determined by a semiform is a @xmath9-space and its every singular subspace carries affine geometry . let @xmath415 be our . the first part follows directly from [ lem : propg ] . the other part is a simple observation that a singular subspace of @xmath387 , in other words , a strong subspace wrt . @xmath378 in @xmath104 , is an affine subspace of @xmath104 . when we deal with a @xmath9-space a question on the form of its triangles may appear important . the following is immediate from and . [ rem : triangle ] a triangle in an @xmath415 has form @xmath416 , \qquad [ v_0 + \eta(u , u_0),u_0 + u ] , \qquad [ v_0 + \eta(y , u_0 ) , u_0 + y],\ ] ] where @xmath417 . [ cor : beztrg ] if @xmath418 for each nonzero vector @xmath419 then the corresponding contains no proper triangle . in that case its maximal singular subspaces are the lines . [ exm : siatki ] in view of [ thm : gamma ] one could expect that s are models of the system considered in @xcite . in case considered in [ exm : power ] and , consequently , in case considered in [ exm : prod ] we have @xmath418 for all @xmath420 and corresponding alternating map @xmath8 . therefore , _ the structure @xmath415 has no triangles_. so , _ s determined by them are not models of the system considered in @xcite_. one can also compute that , e.g. an determined by the exterior power operation _ is not a generalized quadrangle_. in the sequel we shall frequently consider the condition ( with prescribed values @xmath421 ) @xmath422 $ ] ; applying the representation given in [ exm : power ] this can be read as @xmath423 $ ] , which is equivalent to @xmath424 $ ] . this observation allows us to construct quite `` strange '' ( ` locally surjective ' ) alternating maps . as an immediate consequence of [ lem : lines : adjclosed ] and the definition we have [ lem : pomo0 g ] let @xmath373 $ ] be a vector of @xmath200 . the following conditions are equivalent . -2pt there is no line @xmath384 with the direction @xmath254 . the equation @xmath425 is not solvable in @xmath216 . in particular , if @xmath426 and @xmath427 then is not solvable and thus there is no line @xmath384 with the direction @xmath254 . set @xmath428 note , in particular , that when @xmath429 is a surjection for each non zero vector @xmath216 then @xmath430 . exm : prod[cont1 ] let @xmath224 be a vector product in a vector @xmath220-space @xmath11 associated with a nondegenerate bilinear symmetric form @xmath19 and @xmath33 be the orthogonality determined by @xmath19 . then for @xmath431 equation is solvable iff @xmath432 . in that case we have @xmath433\in v\times v\colon u \not\perp v \}.\ ] ] [ lem : pomo1x ] for a fixed @xmath400 , @xmath434 and a scalar @xmath193 the set @xmath435\colon \eta(u_0,u ) = v_0 + \alpha v \big\}\ ] ] is a subspace of . the class of sets of form is invariant under translations of . take @xmath436,[v_2,u_2]\in { \cal z}$ ] and an arbitrary scalar @xmath437 . then we compute @xmath438 which proves that @xmath439\in { \cal z}$ ] and thus @xmath440},{[v_2,u_2 ] } ) \subset { \cal z}$ ] . this proves that @xmath441 is a subspace of . write @xmath442 for the set defined by . let @xmath443\in y$ ] be arbitrary . then @xmath444 ) = [ v+x , u+y ] \in { \cal z}_{u_0,v_0,\alpha } \iff \eta(u_0,u+y ) = v_0 + \alpha ( v+x ) \iff \eta(u_0,u ) = ( v_0 - \eta(u_0,y ) + \alpha x ) + \alpha v \iff [ v , u ] \in { \cal z}_{u_0,v_0 - \eta(u_0,y ) + \alpha x,\alpha}. $ ] thus @xmath445 this closes our proof . [ lem : pomo1y ] let @xmath44 be defined by . then either @xmath44 is an empty set or it is an affine subspace of @xmath104 with the dimension @xmath446 , or with the dimension @xmath447 . if @xmath44 is nonempty then by [ lem : pomo1x ] we can assume that @xmath448\in{\cal z}$ ] . then @xmath44 is characterized by an equation @xmath449 with prescribed values of @xmath450 , @xmath451 and it is the kernel of the linear map @xmath452 , @xmath453\longmapsto \eta(u_0,u ) - \alpha_0 v$ ] . if @xmath454 then , clearly , @xmath455 and thus @xmath456 ) . assume that @xmath457 ; then @xmath44 can be considered as the kernel of the map @xmath458\longmapsto \eta(\frac{1}{\alpha}u_0,u ) - v$ ] . let @xmath459 be a linear basis of @xmath460 and @xmath461 be a basis of @xmath462 . choose one @xmath463 with @xmath464 for each @xmath465 ; then the set @xmath466 is linearly independent . moreover , the subspaces @xmath467 and @xmath462 have only the zero vector in common . a basis of @xmath44 consists of the vectors @xmath468,[d_1,z_1+e_1],\ldots,[d_1,z_1+e_m ] , [ d_2,z_2],\ldots,[d_k , z_k ] \right).\ ] ] consequently , @xmath469 . applying [ lem : pomo1y ] and we get , e.g. a geometrically interesting strengthening of [ lem : propg ] : the set of points that are joinable in an ( determined by a semiform defined on @xmath200 ) with a given point is a subspace of dimension @xmath470 in the surrounding affine space . another corollary of analogous type that will appear important in the sequel is read as follows . [ lem : izoline = przekroj ] assume the following : 1 . if @xmath471 are two vectors of @xmath11 then there is @xmath472 such that @xmath473 and @xmath474 . let @xmath384 pass through @xmath475 $ ] and @xmath476 be a point on @xmath87 . then @xmath477_{\badjac } \colon q \badjac p , p ' \big\}$ ] . by [ thm : gamma ] , @xmath87 is a subset of the considered intersection . by , we can write @xmath478 = [ \btheta , u]$ ] for a nonzero vector @xmath216 . let @xmath479 ; from @xmath480 $ ] for some @xmath114 such that @xmath481 . now , suppose that @xmath482 \badjac [ \btheta , y]$ ] for each @xmath114 with @xmath417 . with we get the following implication @xmath483.\ ] ] let @xmath484 ; then @xmath485 and thus @xmath486 . thus @xmath487 . applying the global assumptions we infer @xmath488 and thus @xmath489 . finally , let us make a few comments that enable us to characterize ( with the help of [ rem : triangle ] ) the geometry of the lines and the planes through a point in an . each alternating map @xmath490 determines the incidence substructure @xmath491 of the projective space @xmath492 with the point set unchanged and with the class @xmath493 of projective lines of the form @xmath494 , where @xmath495 are linearly independent and @xmath496 as its lines . with a fixed basis of @xmath181 one can write @xmath8 as the ( cartesian ) product of @xmath497 bilinear alternating forms @xmath498 : @xmath499,\ ] ] clearly , the @xmath500 need not be nondegenerate . so , each @xmath500 determines a ( possibly degenerate ) null system @xmath501 with the lines @xmath502 . the class @xmath493 is simply @xmath503 . [ lem : geowiazki:0 ] the geometry of the lines and planes of an ( determined by a semiform @xmath199 associated via with an alternating map @xmath8 ) which pass through the point @xmath448 $ ] is isomorphic to @xmath491 . let @xmath475 $ ] . in view of the class of lines through @xmath90 is the set @xmath504 } \colon u \text { is a nonzero vector}\}$ ] so , it can be identified with the point set of @xmath492 under the map @xmath505}\mapsto\gen{u}$ ] . from [ rem : triangle ] we infer that two lines @xmath506}$ ] , @xmath507}$ ] span a plane in the corresponding iff @xmath496 , which closes our reasoning . from the homogeneity of each ( which will be proved later in [ prop : transtiv : gen ] ) one will get that the geometry of the lines and the planes through arbitrary point of an is ( in the above sense ) a generalized null system . to establish the automorphism group of the relation @xmath378 we need some additional assumptions . one of these conditions is read as follows : 1 . [ eq : cond ] the set of directions of @xmath508 can be characterized in terms of the projective geometry of the horizon of @xmath509 with the set of directions of @xmath510 distinguished . clearly , in view of [ lem : pomo0 g ] this condition holds when @xmath199 is scalar valued . let us point out that it is not a unique possibility when this condition is valid . exm : prod[cont2 ] we continue with the notation of [ exm : prod ] . _ let @xmath511 preserve the set of directions @xmath510 . then @xmath512 preserves the vector subspace @xmath513 . _ indeed , the geometric structure of the complement of @xmath510 carries the geometry of the reduct @xmath514 of a hyperbolic polar space of the form considered in [ sssec : hyperbolic ] . our claim follows from [ thm : hippolreduct ] . consequently , the condition is valid here . [ prop : aut : gen ] if @xmath515 is given by the formula @xmath516 ) = [ \psi_1(v ) + \psi_2(u ) + v_0 , \varphi(u ) + u_0]\ ] ] where @xmath517 , @xmath518 , @xmath519 , @xmath520 are linear bijections , @xmath521 , and the following holds : a. [ wrr1 g ] @xmath522 for every vector @xmath216 of @xmath11 , and b. [ wrr2 g ] @xmath523 , for all vectors @xmath121 of @xmath11 , then @xmath515 preserves the relation @xmath378 . in that case the semiform @xmath199 is transformed under the rule @xmath524 for any pair @xmath148 of points of . conversely , under additional assumption that is valid , each linear ( affine ) automorphism of @xmath104 is of the form . if @xmath8 is ` onto ' @xmath525 then given map @xmath526 , condition uniquely determines @xmath527 . similarly , for a given map @xmath526 and vector @xmath450 , condition uniquely determines @xmath528 . assume that @xmath515 is defined by the formula and , hold . let @xmath345 $ ] , @xmath346 , @xmath347 , for @xmath411 . we compute as follows : @xmath529 which proves . this yields , in particular , that @xmath515 preserves @xmath378 . now assume that @xmath515 is an affine automorphism of @xmath104 preserving @xmath378 and that is valid . then , @xmath515 is a composition @xmath530 } \circ f_0 $ ] , where @xmath531 and @xmath532 $ ] is a vector of @xmath200 . the map @xmath533 can be presented in the form @xmath534 ) = [ \psi_1(v ) + \psi_2(u ) , \varphi_1(u ) + \varphi_2(v)]$ ] for suitable linear maps ( @xmath535 ) . by [ lem : pomo0 g ] , the linear part @xmath533 of @xmath515 fixes the subspace @xmath525 and thus @xmath536 . we write @xmath537 . since @xmath515 preserves the relation @xmath378 by definition we obtain the following equivalence : @xmath538 for all vectors @xmath539 , @xmath540 . substituting in @xmath541 and @xmath542 we arrive to the condition . in particular , from we obtain @xmath543 for all @xmath121 in @xmath11 . thus , assuming from we get . finally , @xmath515 has form , as required . [ prop : transtiv : gen ] the group of automorphisms of @xmath378 is transitive . it suffices to compute the orbit @xmath544 of the point @xmath448 $ ] under the group of affine automorphisms of @xmath378 . from [ prop : aut : gen ] , the orbit @xmath544 contains all the vectors @xmath545 = [ v_0,u_0]$ ] with suitable maps @xmath528 , @xmath526 . considering @xmath546 , @xmath547 , @xmath548 we get a class of affine automorphisms of @xmath378 : those defined by the formula @xmath549 ) = [ v+ \eta(u , u_0 ) + v_0 , u + u_0]\ ] ] with arbitrary fixed @xmath550 . so , each point of is in @xmath544 . combining [ prop : transtiv : gen ] and [ lem : izoline = przekroj ] we get a theorem , which is important in the context of foundations of geometry of s. [ thm : izoline = przekroj ] let @xmath551 be the determined by a semiform that meets assumptions of . for each pair @xmath156 of points of @xmath551 such that @xmath552 the set @xmath553 is the line of @xmath551 through @xmath156 . consequently , the class of lines of @xmath551 is definable in terms of the binary collinearity @xmath378 of @xmath551 . in view of [ prop : transtiv : gen ] without loss of generality we can assume that @xmath475 $ ] and then [ lem : izoline = przekroj ] yields the claim directly . the map @xmath512 is given by the formula @xmath556 ) = [ \alpha a + v\circ u + b , \varphi(u ) + w],\quad [ a , u ] \text { is a vector of } { { \mathbb{y}}},\ ] ] where @xmath307 and @xmath557 are scalars , @xmath558 are vectors of @xmath11 , @xmath559 stands for the cartesian " scalar product , and @xmath526 is a linear bijection of @xmath11 such that for a given map @xmath526 , the condition uniquely determines the parameter @xmath193 . similarly , for given @xmath526 and @xmath562 , the map @xmath563 is a functional in @xmath564 and thus there is a vector @xmath565 that satisfies . let @xmath512 have the form such that and hold . let us write @xmath578 ; then assumes form @xmath579 . clearly , the map @xmath580 is a homomorphism of the group of admissible " @xmath526 s and the multiplicative group of @xmath17 . with this notation , the formula can be rewritten in the form @xmath581 ) = \bigl [ |\varphi|a + \eta(\varphi(u),w ) + b , \varphi(u ) + w \bigr].\ ] ] consequently , every @xmath582 can be identified with a triple @xmath583 , where @xmath584 is the group of automorphisms of @xmath585 on @xmath11 , i.e. let @xmath587 be associated with the triple @xmath588 for @xmath589 and let @xmath590 . first , we directly get @xmath591 . from well known formula of ( analytical ) affine geometry @xmath592 , where @xmath593 are linear bijections we obtain @xmath594 and @xmath595 . the obtained rules for transformation of parameters @xmath596 are analogous to the transformation rules of the direct product @xmath597 , but the transformation rule of @xmath557 is more complex . [ lem : aut : cialo ] let @xmath598 , let us fix a natural basis of @xmath11 and for @xmath599 $ ] in @xmath11 let us set @xmath600 $ ] . finally , we set @xmath601 ) = [ \sigma(a),\sigma(u)]$ ] for @xmath125 $ ] in @xmath200 . then @xmath602 . the natural question appears what is the metric " geometry of our symplectic affine spaces i.e. what are characteristic relations defined on the point universe of that characterize our geometry ( except @xmath153 , of course , which _ is sufficient _ , but is more _ affine _ than _ metric _ in spirit ) . it is clear that no relation that is invariant under all the translations can be used here . in particular , no line orthogonality can be used . for pairs @xmath608 , @xmath609 of points of we define @xmath610 clearly , the relation @xmath611 is an equivalence relation . the following relation is crucial . for distinct @xmath148 and arbitrary @xmath90 we have since @xmath153 is expressible in terms of @xmath611 , each automorphism of @xmath611 preserves @xmath153 , so , in view of [ prop : afpol2af ] , it is an automorphism of . from [ prop : aut : af ] , [ prop : afpol2af ] , and the above we can directly compute [ lem : pomo1 ] for fixed @xmath617 the set @xmath618\colon \eta(w , u ) = \beta + \alpha a \bigr\}\ ] ] either is empty ( @xmath571 , @xmath619 , @xmath620 ) , or it is the point set of @xmath104 ( @xmath571 , @xmath621 ) , or it is a hyperplane of . the relation @xmath611 defined by has some properties of an abstract equidistance relation " or a segment congruence " , but note that it is not associated with any norm . from definition we have @xmath624 , which gives another similarity concerns bisector hyperplanes . in the context of a `` metric '' geometry determined by an equidistance relation @xmath611 , with a pair of points @xmath148 one can associate `` bisectors '' of @xmath148 : @xmath628 ( if in a geometry we have @xmath629 for all points @xmath156 , then there is no need to distinguish between these two types of bisectors . in our case the distinction is necessary . ) the following is evident . let @xmath148 be a pair of distinct points of . @xmath636 is empty iff @xmath100 is the direction of @xmath637 . when @xmath100 is not the direction of @xmath87 then @xmath636 , and in any case @xmath638 both are hyperplanes of . comparing the properties of @xmath611 defined here and of the equidistance of a metric affine space ( characterized in @xcite ) we see that the only one axiom of @xcite that is not valid here is this which states that opposite sides of a parallelogram are congruent ( i.e. stating that a translation is an `` isometry '' ) . some other ( we believe : interesting ) properties are valid , instead . let us write @xmath640 for the ( affine ) midpoint operation defined in . simple computation gives given a hyperplane @xmath643 we say that @xmath148 are _ symmetric _ wrt . @xmath643 when @xmath643 is the bisector hyperplane of @xmath148 . in a metric affine geometry this notion can be used to define the axial symmetry . in our geometry we have , formally , two notions of a pair symmetric under a hyperplane . take @xmath656 such that @xmath657 and @xmath658 . from the assumptions we get @xmath659 i.e. @xmath660 . this gives @xmath661 from which @xmath662 follows . thus @xmath663 , so @xmath654 for some @xmath355 . finally , @xmath655 is directly computed . [ aby sie upewnic ] namely , for each fixed @xmath126 we have @xmath680 such that @xmath681 and @xmath682 . assume @xmath683 . then we arrive to @xmath684 is fixed . substituting to the above we obtain @xmath685 . thus @xmath686 . since @xmath126 is arbitrary , in particular , we can take @xmath687 which yields @xmath688 . [ thm : symmetry ] let @xmath643 be a hyperplane of . the relation @xmath697 defined by the condition @xmath698 is not a function ; nevertheless , whenever @xmath699 holds , the translation which maps @xmath700 onto @xmath701 is a subset of @xmath702 . [ prop : bisec = polar ] let @xmath633 $ ] be distinct points of , @xmath706 , and @xmath707 be the direction of the line @xmath708 ( i.e. @xmath709\in { { \mathbb{y}}}$ ] , @xmath710\in { { \mathbb{w}}}$ ] ) . then @xmath711 consequently , @xmath638 is the restriction of the hyperplane @xmath712 of polar of @xmath705 to the point set of . analogously , @xmath636 is the restriction of the hyperplane which has the direction @xmath713_\parallel$ ] as its pole under the underlying null - polarity . authors address : + krzysztof pramowski , mariusz ynel + institute of mathematics , university of biaystok + ul . akademicka 2 , 15 - 267 biaystok , poland + ` krzypraz@math.uwb.edu.pl ` , ` mariusz@math.uwb.edu.pl `

thermal fluctuations in the magnetization of interstellar grains will produce magnetic dipole emission at . we show how to calculate absorption and emission from small particles composed of material with magnetic as well as dielectric properties . the kramers - kronig relations for a dusty medium are generalized to include the possibility of magnetic grains . the magnetic permeability as a function of frequency is discussed for several candidate grain materials . iron grains , or grains containing iron inclusions , are likely to have the magnetic analogue of a frhlich resonance in the vicinity of , resulting in a large magnetic dipole absorption cross section . we calculate the emission spectra for various interstellar grain candidates . while `` ordinary '' paramagnetic grains or even magnetite grains can not account for the observed `` anomalous '' emission from dust in the range , stronger magnetic dipole emission will result if a fraction of the grain material is ferromagnetic , as could be the case given the high fe content of interstellar dust . the observed emission from dust near implies that not more than of interstellar fe is in the form of metallic iron grains or inclusions ( e.g. , in `` gems '' ) . however , we show that if most interstellar fe is in a moderately ferromagnetic material , with the magnetic properties properly adjusted , it could contribute a substantial fraction of the observed emission , perhaps comparable to the contribution from spinning ultrasmall dust grains . the two emission mechanisms can be distinguished by measuring the emission from dark clouds . if ferromagnetic grains consist of a single magnetic domain , and are aligned , the magnetic dipole emission will be linearly polarized , with the polarization depending strongly on frequency . = cmmib10 h * pope-770 * submitted to _ the astrophysical journal _

astro-ph9807009

experiments to map the cosmic background radiation ( cbr ) have stimulated renewed interest in diffuse galactic emission . sensitive observations of variations in the microwave sky brightness have revealed 14 - 90 ghz microwave emission which is correlated with @xmath5 thermal emission from interstellar dust ( kogut et al . 1996 ; de oliveira - costa et al . 1997 ; leitch et al . the origin of this anomalous " emission has been of great interest . while the observed frequency - dependence appears consistent with free - free emission ( kogut et al . 1996 ; leitch et al . 1997 ) , the observed intensities can not be due to free - free emission from interstellar gas : for gas temperature @xmath6 the required spatial variations in emission measure would produce @xmath7 times more h@xmath8 than allowed by observations ( gaustad et al . 1996 ; leitch et al . 1997 ) , while @xmath9 gas would produce powerful x - ray emission which can be ruled out on energetic grounds ( draine & lazarian 1998 ) . it has been proposed that the observed 14 - 90 ghz emission is rotational emission from the population of ultrasmall dust grains believed to be present in the interstellar medium ( draine & lazarian 1998a , b ) . such electric dipole emission is a natural consequence of current models of interstellar dust , and seems likely to be the mechanism responsible for the observed emission . it is nevertheless of interest to investigate whether other emission mechanisms exist which would also be capable of producing strong emission at these frequencies . here we discuss an entirely different emission mechanism : _ thermal fluctuations of the magnetization _ within individual interstellar grains . such thermal fluctuations will result in magnetic dipole radiation . the magnitude of the magnetic fluctuations depends upon the magnetic properties of the grain material , and we consider a number of possibilities , including `` ordinary '' paramagnetism as well as superparamagnetism , ferromagnetism , and ferrimagnetism . if about 30% of the grain mass is carbonaceous , then fe and ni contribute nearly 30% of the mass of the remaining grain material . it would not be surprising , then , if some fraction of the interstellar grain population were strongly magnetic ; metallic iron / nickel , magnetite ( fe@xmath10o@xmath11 ) , and maghemite ( @xmath12fe@xmath13o@xmath10 ) are plausible candidates to consider . at infrared and optical frequencies ( @xmath14 ) it has been customary to neglect the magnetic properties of the grain material when computing absorption or emission of electromagnetic radiation by dust grains ; this is an excellent approximation since most materials have negligible magnetic response to oscillating magnetic fields at frequencies @xmath15 . this is because `` magnetism '' is due to ordering of electron spins , and the maximum frequency for electron spins to reorient is the precession frequency of an electron in the local magnetic field ( due mainly to other electron spins ) in the material ; these gyrofrequencies do not exceed @xmath16 . in contrast , the electron charge distribution in grains can respond strongly to applied electric fields at frequencies as large as @xmath17 . at microwave frequencies materials can respond to both electric and magnetic fields , and both the dielectric constant @xmath18 and the magnetic permeability @xmath19 can play a role in the absorption and emission of electromagnetic radiation . the present study is primarily directed at calculation of thermal emission from magnetic dust grains . the complex permeability @xmath19 obtained here is also relevant to the process of grain alignment via magnetic dissipation ( davis & greenstein 1951 ) , particularly for ultrasmall grains undergoing very rapid rotation ( draine & lazarian 1998a , b ; lazarian & draine 1998 ) and suprathermally rotating graints ( purcell 1979 ) . the optical properties of small particles , including both dielectric and magnetic effects , are reviewd in [ sec : optics ] . in [ sec : kk ] we generalize purcell s ( 1969 ) application of the kramers - kronig relations to interstellar grains to include the case of magnetic grains . the expected form of the frequency - dependence of the complex permeability @xmath19 is discussed in [ sec : magnetic ] . in [ sec : paramag ] we review the properties of ordinary paramagnetic materials , and in [ sec : ferro ] we discuss various classes of iron - rich materials , including : single - domain iron grains ( [ sec : fe_single ] ) ; bulk ( multidomain ) iron ( [ sec : fe_bulk ] ) ; materials with single - domain fe inclusions ( [ sec : fe_incl ] ) , including the superparamagnetic limit where the inclusions contain less than @xmath20 fe atoms each ( [ sec : superpara ] ) ; ferrimagnetic materials such as magnetite and maghemite ( [ sec : ferri ] ) ; and antiferromagnetic materials ( [ sec : antiferro ] ) . in [ sec : prediction ] we consider several ways in which iron can be present in the interstellar grain population , and for each case we calculate the expected thermal emission contributed by magnetic fluctuations . iron grains , if present , would produce strong @xmath21 emission ; this can be used to place an upper limit of @xmath4 on the fraction of the interstellar fe which can be in the form of pure iron grains or inclusions . in [ sec : hypothetical_materials ] we discuss the possibility that the fe could be in a less strongly magnetic form ( e.g. , impure iron ) , and show that there are plausible magnetic properties for which such grains could dominate the 14 - 90 ghz emission . the polarization of the magnetic dipole emission from ferromagnetic grains is discussed in [ sec : polarization ] . we discuss our results in [ sec : discussion ] , and summarize in [ sec : summary ] . for grains which are small compared to the wavelength , the cross section for absorption of electromagnetic waves can be written as the sum of electric dipole and magnetic dipole cross sections ( draine & lee 1984 ) : c_abs c_abs^(ed ) + c_abs^(md ) . [ eq : dipoleapprox ] for nonmagnetic materials ( i.e. , materials with magnetic permeability @xmath22 ) , the magnetic dipole contribution arises from induced eddy currents in the grain , which give it an oscillating magnetic dipole moment which is out of phase with the applied magnetic field ; the absorption cross section for this case has been given previously ( landau & lifshitz 1960 ; draine & lee 1984 ) . for materials with a magnetic as well as electric response , the absorption cross section for small spheres of radius @xmath23 can be shown to be c_abs^(ed)= v c \ { 9_2 |+2|^2 + ( ac)^2 |3 + 2|^2 _ 210 } , [ eq : cabs_ed ] c_abs^(md)= v c \ { 9_2 |+2|^2 + ( a c)^2 |3 + 2|^2 _ 210 } , [ eq : cabs_md ] where @xmath24 , @xmath25 is the dielectric constant , . ] and @xmath26 is the magnetic permeability . the dipole approximation ( [ eq : dipoleapprox ] ) is valid provided ||^1/2 ( ac ) < 1 [ eq : dipolevalidity ] . in eq . ( [ eq : cabs_ed ] ) the first term is due to the polarization in response to the applied electric field , and the second term is due to the oscillating circular magnetization induced by the time - dependent displacement current @xmath27 . similarly , in eq . ( [ eq : cabs_md ] ) the first term is due to the magnetization in response to the applied magnetic field , while the second term is the heating due to the `` eddy currents '' induced by the time - dependent magnetic field @xmath28 . by analogy , we will refer to the second term in eq . ( [ eq : cabs_ed ] ) as being due to `` eddy magnetization '' . under some circumstances the second - order `` eddy '' contributions can be important for example , `` eddy current '' dissipation is important for @xmath29 graphite grains at @xmath30 ( draine & lee 1984 ) . however , for @xmath31 interstellar grains at @xmath32 , @xmath33 and these higher - order terms are generally negligible . the `` eddy '' terms will therefore be neglected in the remainder of this paper . thus , [ eq : cabsmd_over_cabsed ] . ellipsoidal grains with semiaxes @xmath34 are characterized by `` geometrical factors '' or `` depolarization factors '' @xmath35 which depend only on the grain shape ( see bohren & huffman 1983 ) . table [ tab : geom_fact ] contains @xmath36 for selected ellipsoidal shapes . .[tab : geom_fact ] geometrical factors @xmath36 for ellipsoids with semiaxes [ cols="^,^,^,^,^",options="header " , ] solids exhibit a variety of different responses to weak applied oscillating magnetic fields . when a static magnetic field is present , or the substance is spontaneously magnetized , the susceptibility @xmath38 has a tensor nature ( see , e.g. , jones & spitzer 1967 , iiia ) . in the present discussion we will treat @xmath39 as a scalar , but will extend the discussion to spontaneously - magnetized material in [ sec : fe_single ] and appendix [ app : single_domain ] . aside from the diamagnetism of superconductors and materials with no unpaired electron spins , at low frequencies most materials are characterized by @xmath40 , resulting from either `` normal paramagnetism '' or magnetic ordering in the form of superparamagnetism , ferrimagnetism , or ferromagnetism . we consider these cases below , but first discuss the likely behavior of the frequency - dependent susceptibility @xmath41 . suppose that the magnetization @xmath42 obeys the equation of motion = _ 0 ^ 2-/_0 , [ eq : eof_d ] with three parameters : the static response @xmath43 , a resonance frequency @xmath44 , and a characteristic damping time @xmath45 . then we obtain the drude form for the susceptibility,/\tau$ ] . because this lacks an `` inertial '' term , we believe that it overestimates the magnetic response at very high frequencies . ] ( ) = ( 0 ) 1-(/_0)^2-i [ eq : chi_res ] , where we have defined ( _ 0 ^ 2_0)^-1 . [ eq : tau_def ] the real and imaginary parts of @xmath38 are _ 1 = ( 0)1-(/_0)^2 ^2 + ( ) ^2 [ eq : chi1_res ] _ 2 = ( 0)^2 + ( ) ^2 . [ eq : chi2_res ] a susceptibility of the form ( [ eq : chi_res ] ) arises , for example , as a solution to the bloch equations ( c.f . pake 1962 , eq . 6 - 24 ) for a magnetized material . the product @xmath46 determines the shape of @xmath38 . for @xmath47 , @xmath38 has a resonance near @xmath44 . for @xmath48 , the system is `` overdamped '' , and responds to a step function change in the applied @xmath49 with two distinct relaxation times . for @xmath50 , @xmath38 is `` critically - damped '' , with a single relaxation time @xmath51 . we will consider this as a plausible form for the response function @xmath38 for a system which is essentially nonresonant . for this case @xmath39 takes the simple form , shown as the heavy curves in fig . [ fig : chiforms ] : ^(cd)=(0)(1-i/2)^2 , [ eq : chi_cd ] _ 1^(cd ) = ( 0)1-(/2)^2 ^ 2 [ eq : chi_cd1 ] , _ 2^(cd ) = ( 0)^2 [ eq : chi_cd2 ] . the superscript `` ( cd ) '' stands for `` critically - damped '' . we will use @xmath52 to estimate the response associated with the magnetization of a single - domain particle , or of paramagnetic materials . we will identify the characteristic frequency @xmath53 with the gyrofrequency of an electron in the internal field . from eq . ( [ eq : chi_cd]-[eq : chi_cd2 ] ) we see that the response falls off rapidly when the frequency @xmath54 . for ordinary paramagnetism , the zero - frequency susceptibility is ( draine 1996 ) ( 0 ) 410 ^ -2f_p(p5.5)^2 ( 15 t ) , where @xmath55 is the fraction of the atoms which are paramagnetic , with magnetic moments @xmath56 , where the bohr magneton @xmath57 . if essentially all of the interstellar mg , fe , and si are incorporated into material with approximately the composition of mgfesio@xmath11 , then a fraction @xmath58 of the atoms would be fe , presumably in the form of fe@xmath59 or fe@xmath60 ) ions ; these have @xmath61 and 5.9 , respectively ( morrish 1980 ) . fayalite ( fe@xmath13sio@xmath11 ) is antiferromagnetic ( carmichael 1989 ) and it appears that intermediate olivines ( mg@xmath62fe@xmath63sio@xmath11 ) will also be antiferromagnetic ( duff 1968 ) . the magnetic character of amorphous olivine is uncertain . while weak ferrimagnetism seems likely ( see below ) , here we consider paramagnetic behavior . the damping time @xmath64 is the spin - spin relaxation time , which is essentially the time for electron precession in the random magnetic fields within the solid ( see caspers 1967 ) . for amorphous olivine we estimate the r.m.s . gyrofrequency to be ( draine 1996 ) _ draine ( 1996 ) used a different functional form for @xmath65 . ] to estimate the paramagnetic susceptibiity , but we now consider the `` critically - damped '' form @xmath52 [ eq.([eq : chi_cd ] ) ] to provide a better estimate at frequencies @xmath66 . in fig . [ fig : pm ] we show @xmath67 , for @xmath68 . at frequencies @xmath69ghz `` astronomical silicate '' ( draine & lee 1984 ) has @xmath70 , @xmath71 . from eq . ( [ eq : cabsmd_over_cabsed ] ) , it can then be seen that @xmath72 when @xmath73 ( assuming @xmath74 , as is the case for ordinary paramagnetism ) , so for reference we also show @xmath75 in fig . [ fig : pm ] . from fig . [ fig : pm ] it can then be seen that if femgsio@xmath11 grains are merely paramagnetic , the thermal emission from these grains at @xmath76 ghz will evidently be dominated by magnetic dipole emission arising from thermal fluctuations in the magnetization . additional magnetic absorption can arise in paramagnetic material when crystalline electric fields produce stark - effect splitting of the magnetic sublevels , and this splitting corresponds to microwave frequencies ( kittel & luttinger 1948 ) . electric dipole transitions are forbidden , since the sublevels have the same parity , but magnetic dipole transitions are allowed . this type of absorption has been studied in the so - called zero field resonance experiments ( see bramley & strach 1983 and references therein ) . for fe@xmath77 resonances were observed at @xmath78 and @xmath79 ghz in natural amethyst and at @xmath80 and @xmath81 ghz for synthetic brown quartz . similar measurements in sapphire revealed resonances at @xmath82 and @xmath83 ghz . for pure single crystals the absorption is localized in lines , while a continuum of absorption features is expected for impure and noncrystalline interstellar grains . the strength of this absorption can be estimated ( see appendix [ app : stark_effect ] ) . at the temperatures of interstellar grains , it appears that a permeability @xmath84 would result . we see from fig . [ fig : pm ] that this would be a minor contribution to grain absorption and emissions compared even with what we estimate for `` ordinary '' paramagnetism at @xmath85 , and far below what we estimate in [ sec : ferro ] for ferromagnetic or ferrimagnetic materials at microwave frequencies . accordingly , as a continuum emission process this mechanism appears to be unimportant . the absorption due to stark splitting could be concentrated in a few absorption features , rather than being spread into a continuum . even so , the effect will be weak . if , say , 5% of the atoms had stark splitting features within a frequency range @xmath86 , then we see from eq . ( [ eq : mu_stark ] ) that @xmath87 within the narrow feature . at , say , 5 ghz this is only a few times larger than the value of @xmath88 which we estimate for ordinary paramagnetism ( see fig . [ fig : pm ] ) and hence would result in a narrow emission feature with central intensity only a few times what we estimate for the continuum due to ordinary paramagnetism , which we shall see below is itself two orders of magnitude weaker than the observed continuum emission . nevertheless we can not exclude the possibility that future sensitive measurements of the microwave background might disclose weak spectral features arising from the stark effect . as noted in [ sec : materials ] , the substantial fraction of fe in interstellar grains creates the possibility that some fraction of the grains could be magnetic , with the `` exchange interaction '' resulting in ordering of the atomic magnetic moments within a single magnetic domain ( see morrish ( 1980 ) for a review of the physics of magnetism , and dunlop and zdemir ( 1997 ) for an excellent review of magnetic minerals ) . ferromagnetic materials as components of interstellar grains were apparently first considered by spitzer & tukey ( 1951 ) , who discussed the possible formation of ferromagnetic materials in interstellar grains as the result of grain - grain collisions : metallic iron , fe@xmath10o@xmath11 , @xmath12fe@xmath13o@xmath10 , and mgfe@xmath13o@xmath11 were considered likely products . jones & spitzer ( 1967 ) suggested that interstellar grains might contain very small clusters of magnetic materials such as fe@xmath10o@xmath11 or @xmath12fe@xmath13o@xmath10 . shapiro ( 1975 ) proposed that the observed polarization of starlight could be produced by platelets of fe@xmath10o@xmath11 . sorrell ( 1994 ) presented a model for the origin of small fe@xmath10o@xmath11 clusters in h@xmath13o ice mantles irradiated by cosmic rays . ferromagnetic properties of mixed mgo - feo - sio grains were discussed by duley ( 1978 ) . bradley ( 1994 ) argued that certain interplanetary dust particles ( `` glass with embedded metals and sulfides '' , or `` gems '' ) consisted of aggregates of @xmath89 interstellar grains , with nanometer - sized fe - ni metal inclusions . martin ( 1995 ) argued that the properties of these grains were consistent with being interstellar dust , and that the fe - ni inclusions could be superparamagnetic and able to bring about alignment with the galactic magnetic field . goodman & whittet ( 1995 ) noted that the numbers of such inclusions were consistent with the requirements of mathis ( 1986 ) hypothesis to account for the dependence of degree of alignment on grain size . the s in gems appears to be in the form of fes ; if the excess of fe over s is metallic , then the atomic abundances given by bradley ( 1994 ) for 3 bulk gems indicates a volume filling factor @xmath90 for metallic fe - ni ; we will use this value when discussing the properties of silicate grains with metallic inclusions ( [ sec : fe_incl ] ) . there are three classes of magnetically - ordered materials . in _ ferromagnetic _ materials the atomic spins within a domain are parallel ; in _ ferrimagnetic _ and _ antiferromagnetic _ materials the magnetic ions are located on two magnetic sublattices of oppositely directed spins and magnetic moments . in _ antiferromagnetic _ materials the magnetic moments of the two sublattices are exactly opposite in direction and magnitude , so the net magnetization is zero ( for a perfect crystal at zero temperature ) . in _ materials the magnetic sublattices do not exactly compensate , either because the magnetic moments of the two sublattices differ in magnitude ( normal ferrimagnetism ) or are not precisely opposite in direction ( spin - canted ferrimagnetism ) . applied magnetic fields change the magnetization of a bulk sample by two processes : rotation of the magnetization within a domain , and motion of domain walls . grains smaller than a critical radius @xmath91 will contain only a single domain ; for fe @xmath92 ( morrish 1980 ) . such single - domain behavior can take place for clusters of materials which are ferromagnetic ( e.g. , fe ) or ferrimagnetic ( e.g. , fe@xmath10o@xmath11 , or @xmath12fe@xmath13o@xmath10 ) . spontaneous magnetization has been observed for fe clusters as small as @xmath93 atoms ( billas , chtelain & de heer 1994 ) if the magnetic material consists of a single domain , then there is clearly no motion of domain walls involved in the magnetization process , unlike the case for bulk samples , so susceptibilities measured for bulk samples are inapplicable . when a ferromagnetic or ferrimagnetic material consists of a single domain , it will be spontaneously magnetized with magnetization @xmath94 along one of the `` easy '' directions . if a weak oscillating field @xmath95 is now applied , the single - domain sample will exhibit a susceptibility which depends on the direction of @xmath95 . in appendix [ app : single_domain ] we show that at low temperatures only the component of @xmath95 perpendicular to the spontaneous magnetization induces a change in magnetization of the sample : @xmath96 . for fe , we estimate @xmath97 ( see appendix [ app : single_domain ] ) . since changing the magnetization of the sample involves only small reorientations of the magnetic moments , and no motions of domain walls , we assume that the only frequency characterizing the response is the gyrofrequency @xmath98 of a magnetic moment with @xmath99 in the internal field @xmath100 to which each of the dipoles is subject . for fe , with @xmath101 kg , we estimate @xmath102 . we use @xmath52 [ eq.([eq : chi_cd ] ) ] with @xmath103 to estimate @xmath104 ( see fig . [ fig : fe_single ] ) . the energy dissipation rate is proportional to @xmath105 . for isotropic radiation incident on the grain , @xmath106 . assuming the higher order term in eq . ( [ eq : cabs_md ] ) to be negligible , we would therefore take the angle - averaged magnetic dipole absorption cross section for randomly - oriented spheres to be c_abs^(md ) vc 6_,2(_,1 + 2)^2+_,2 ^ 2 [ eq : cabsperp ] whereas for randomly - oriented ellipsoids with semiaxes @xmath107 , magnetized along the long axis , c_abs^(md ) _ j=1 ^ 2 _ , 2l_j^2 [ eq : cabsperp_ell ] with @xmath108 obtained using eq . ( [ eq : chi_cd ] ) with either @xmath109 obtained from eq . ( [ eq : chi_perp_100 ] ) . we now observe that if our estimate for @xmath104 in fig . [ fig : fe_single ] is correct , then fe particles will have the magnetic analogue of a frhlich resonance ( frhlich 1949 ; bohren & huffman 1983 ) a peak in @xmath110 where @xmath111 , thus minimizing the denominator in eq . ( [ eq : cabs_md_ell ] ) or in eq . ( [ eq : cabsperp_ell ] ) . it is apparent that for the functional forms in fig . [ fig : chiforms ] there are either zero or two frequencies where this condition is satisfied ; if there are two , then @xmath110 peaks at the higher of the two frequencies , which we denote @xmath112 . the frhlich resonance condition for spheres requires that @xmath113 for some range of frequencies , or @xmath114 . for @xmath115 [ eq . ( [ eq : chi_cd ] ) ] it is easy to show that the frhlich resonance condition for spheres , @xmath116 , can be satisfied provided @xmath117 , with @xmath118 . however , @xmath112 is associated with a distinct peak in @xmath110 only for @xmath119 . ] whether or not fe particles will have a frhlich resonance will depend on the details of the frequency - dependence of @xmath39 , which at this time is not experimentally - determined . we are therefore forced to rely on model - dependent estimates . with our estimate of @xmath120 ( see fig . [ fig : fe_single ] ) , we estimate that , for spheres , @xmath121 . for ellipsoids with moderate axial ratios , the resonance will be shifted slightly . for example , for a 1:1.5:2 ellipsoid , magnetized along the long axis , the frhlich resonances along axes 1 and 2 occur for @xmath122 and @xmath123 , or @xmath124 and @xmath125 . recalling the kramers - kronig relation ( landau & lifshitz 1960 ) _ 1 ( ) = 2 p_0^ dx [ eq : kk ] ( where @xmath126 denotes principal value integral ) and the fact that @xmath127 , one sees that the integrand is negative for @xmath128 . quite generally , @xmath129 will have @xmath130 , will increase with increasing @xmath131 ( linearly at low frequencies ) , peak , and then eventually decline to zero at high frequencies . if @xmath132 declines more rapidly than @xmath133 at high frequencies , then we expect to have @xmath134 above some frequency . because metallic fe has @xmath135 , it appears likely that there will be a frequency range where @xmath136 , so that the frhlich resonance condition can be satisfied . the functional form @xmath52 used here provides what we consider a reasonable estimate for @xmath120 , but the predicted frhlich frequency @xmath121 is model - dependent and should be regarded as quite uncertain . the static susceptibility of bulk , metallic fe is @xmath137 . the susceptibility of fe as a function of frequency has been measured using fine wires , thin lamina , and powders ( allanson 1945 ; epstein 1954 ) up to @xmath138ghz . the response of bulk material will result from motion of domain walls as well as the deflections of magnetization within single domains as discussed above . we treat these two responses as additive : @xmath139 , where @xmath140 is the contribution from motion of domain walls . the experimental results show considerable scatter , but we can reproduce the general trends if @xmath140 is given by eq.([eq : chi_res ] ) with @xmath141 , @xmath142 , and @xmath143 in fig . [ fig : fe_bulk ] we plot @xmath88 estimated for multidomain metallic fe using these parameters . consider spherical single - domain ferromagnetic inclusions with volume filling factor @xmath144 distributed in a nonmagnetic matrix . suppose each randomly - oriented inclusion to be spontaneously magnetized along one of its `` easy '' directions . we may suppose that , in effect , ( 2/3 ) of these inclusions have their magnetization @xmath145 perpendicular to the direction of the applied @xmath95 , with the remaining 1/3 either parallel or antiparallel to @xmath95 ( and therefore not contributing to the susceptibility ) . the effective susceptibility of the composite material may be estimated using effective medium theory ( bohren & huffman 1984 ) . for the present case of spherical inclusions with small volume filling factor @xmath146 , maxwell - garnett effective medium theory is appropriate , so we estimate _ eff = ( 2/3)_1+(4/3)_(1 - 2/3 ) [ eq : largeinclusions ] where @xmath120 is given by @xmath52 [ eq.([eq : chi_cd ] ) ] . in fig . [ fig : spm ] we show @xmath147 estimated for fe inclusions with a volume filling fraction @xmath148 ( the estimated filling fraction of fe - ni inclusions in gems ; see [ sec : mag_grains ] in a nonmagnetic ( e.g. , silicate ) medium . note the peak in the effective @xmath88 for the composite medium at @xmath149 ; this arises because of the frhlich resonance in the individual single - domain fe inclusions . if the fe atoms are aggregated into single - domain clusters which are sufficiently small , then thermal fluctuations will cause the magnetization to fluctuate in direction , with the most probable magnetization directions being the most energetically favorable ones . if the inclusions are sufficiently small ( @xmath150 for fe at room temperature ) , the energy barrier associated with reorientation of the magnetization is small enough that thermal fluctuations can reorient the magnetization of a single particle on time scales of seconds or less . such clusters are termed _ superparamagnetic_. if a field @xmath49 is applied , the change in magnetization will be greater than would be the case for larger clusters , where only a small deflection in the direction of magnetization takes place , as discussed in [ sec : fe_single ] suppose that the grain consists of a nonmagnetic ( or weakly magnetic ) material containing superparamagnetic clusters of @xmath151 fe atoms , with volume filling factor @xmath152 . a single grain then contains @xmath153 clusters . the susceptibility of the medium may be approximated as + _ sp . the first term is what would be expected if the individual clusters have their magnetization locked into the `` easy '' direction , with a small reorientation of the magnetization resulting from the transverse component of the applied @xmath95 [ see eq . ( [ eq : largeinclusions ] ) ] . the second term is what is expected from alignment of individual clusters assuming them to be free to orient their magnetization in any direction . we note that some previous discussions of the susceptibility of superparamagnetic materials ( e.g. , jones & spitzer 1967 ; lazarian 1995 ; draine 1996 ) have omitted the first term when considering superparamagnetism . in principle , these two estimates should not be treated as simply additive , but in general one or the other will dominate , and factor - of - two accuracy is sufficient at this time . the superparamagnetic contribution at zero frequency is estimated to be _ sp(0 ) .035 _ sp n_cl(15 t ) . in general ( even if spherical ) , a cluster must overcome an energy barrier in order to substantially reorient its magnetization . laboratory experiments show the relaxation process to be be thermally activated , with a characteristic relaxation rate ^-1a ( -n_cl / t ) [ eq : nu0 ] , with @xmath154 and @xmath155 for metallic fe spheres ( bean & livingston 1959 ; jacobs & bean 1963 ) . we will employ ( [ eq : chi_res ] ) with @xmath156 to estimate @xmath157 . [ fig : spm ] shows @xmath88 for silicate grains containing superparamagnetic fe clusters with volume filling factor @xmath158 . such inclusions , if present in all of the silicate grains , would account for @xmath159 of the total fe abundance ( see table [ tab : feminerals ] ) . permeabilities are shown for @xmath160 , and @xmath161 fe atoms per cluster . the superparamagnetic response is large at low frequencies , but is small compared to normal paramagnetism at frequencies @xmath162 ghz , because the estimated relaxation rate is too slow ; at frequencies @xmath162 ghz , the magnetic response is essentially that of the individual domains having their spontaneous magnetizations deflected slightly by the transverse component of the magnetic field . we conclude that while superparamagnetism may lead to enhancements of magnetic dissipation at @xmath163 ghz , at higher frequencies it appears unlikely to make an appreciable contribution to absorption by interstellar grains . therefore the high - frequency response of grains with fe inclusions depends only on the volume fraction @xmath152 of fe inclusions , but is insensitive to the number @xmath151 of fe atoms per inclusion , provided only that @xmath164 so that the inclusions are ferromagnetic . magnetite ( fe@xmath10o@xmath11 ) is a commonly - occuring terrestrial mineral , and is a plausible interstellar grain material . the low temperature spontaneous magnetization @xmath165 , from which we estimate a gyrofrequency @xmath166 . at room temperature magnetite is cubic , and the `` easy '' direction is @xmath167 . at @xmath168 it undergoes a phase transition with a slight distortion of the unit cell from cubic to monoclinic symmetry ; near this temperature the magnetic anisotropy parameter @xmath169 ( see appendix [ app : single_domain ] ) changes sign and the `` easy '' direction changes to @xmath170 . the magnetic anisotropy parameters have been measured down to @xmath171 ( syono & ishikawa 1963 ) . extrapolation to @xmath172 is uncertain ; we take @xmath173 , from which we estimate @xmath174 using eq . ( [ eq : chi_perp_100 ] ) . we have not found data on the high frequency susceptibility of magnetite . taking @xmath175 , we estimate @xmath147 for single - domain fe@xmath10o@xmath11 using @xmath52 [ eq . ( [ eq : chi_cd ] ) ] ; the results are shown in fig . [ fig : fe_3o_4_single ] . the low temperature oxidation product of magnetite is maghemite , @xmath12fe@xmath13o@xmath10 , for which we estimate @xmath176 . from @xmath177 and the crystalline magnetic anisotropy @xmath169 we would estimate @xmath178 ( see appendix [ app : single_domain ] ) . however , measurements by valstyn , hanton , & morrish ( 1962 ) of the permeability as a function of frequency are better reproduced by a slightly larger value , @xmath179 , which we adopt . our estimate for @xmath19 for @xmath12fe@xmath13o@xmath10 is shown in fig . [ fig : fe_3o_4_single ] . valstyn et al . measured @xmath19 for a powder of @xmath180 diameter @xmath12fe@xmath13o@xmath10 spheres suspended in paraffin wax with a volume filling factor 0.15 ; thus in effect there was a volume filling factor of 0.10 for spheres with spontaneous magnetization perpendicular to the applied oscillating field . in fig . [ fig : fe_3o_4_single ] we show their experimental results for @xmath181 divided by 0.10 . the model which we are using appears to be in reasonably good agreement with the measured frequency - dependent susceptibility for this material . in antiferromagnetic materials the exchange interaction leads to magnetic domains with magnetic moments which are ordered , but in such a way that the net magnetization in the domain is zero . the olivine fayalite ( fe@xmath13sio@xmath11 ) is an example of an antiferromagnetic material . these substances have zero frequency susceptibilities @xmath43 which are similar to ( but somewhat smaller than ) those of normal paramagnetic substances with similar concentrations of magnetic ions . our estimate for `` normal paramagnetism '' would therefore be a reasonable guide to antiferromagnetic substances such as fayalite . however , we do not expect interstellar grains to contain perfect crystals ; defects and impurities seem likely to be common , and the materials may be amorphous ( as appears to be the case for interstellar silicates , as indicated by the profile of the interstellar @xmath182 absorption feature due to the si - o stretching mode ) . furthermore , if the magnetic material is in very small inclusions , perfect pairing of spins will not be possible along the boundary of the inclusion . for instance , schuele & deetscreek ( 1962 ) find that small ( @xmath183 ) particles of nio are weakly _ ferro- _ or _ ferri_magnetic although bulk nio is _ antiferromagnetism therefore seems unlikely we instead expect amorphous fayalite , for example , to be at least weakly ferrimagnetic . the high frequency response of such small inclusions would be determined by the single - domain susceptibility ( [ sec : fe_incl ] ) . for a nonrotating grain whose _ internal _ degrees of freedom are in thermal equilibrium at a temperature @xmath184 , the power radiated in frequency interval @xmath185 is simply p_d= 4c_abs ( ) b_(t ) where @xmath186 is the absorption cross section at frequency @xmath187 , and @xmath188 is the planck function . we will use the term `` vibrational emission '' to refer to emission arising from thermal fluctuations in the charge distribution , and therefore the electric polarization , of the grain . this will produce electric dipole emission . we will use the term `` magnetic dipole emission '' to refer to emission arising from thermal fluctuations in the magnetization of the grain material . estimation of this emission is one of the goals of the present paper . we use the term `` rotational emission '' to refer to the emission arising from rotation of the grain . this is primarily due to the electric dipole moment which rotates with the spinning grain ( draine & lazarian 1998a , b ) . in principle , there will be rotational emission arising from the rotating magnetic dipole moment of a spinning magnetized grain . only @xmath189 grains spin at frequencies @xmath190 . the magnetic moment of a spontaneously magnetized grain is just @xmath191 . the largest value of @xmath177 is for metallic fe ( see table [ tab : feminerals ] ) ; an fe grain would have a magnetic moment @xmath192 , somewhat smaller than the electric dipole moment @xmath193 estimated for neutral grains by draine & lazarian ( 1998b ) . other magnetic materials have smaller values of @xmath177 , and hence the rotational emission at @xmath190 is expected to be dominated by electric dipole radiation . we consider 4 possible components of the interstellar grain population : 1 . @xmath194 of the si and fe incorporated into amorphous silicate grains with paramagnetic behavior as in fig . [ fig : pm ] ; 2 . @xmath194 of the fe incorporated into small fe@xmath10o@xmath11 grains ; 3 . 5% of the fe incorporated into bare fe grains ; 4 . 5% of the fe incorporated into small metallic inclusions . for each case we assume a grain temperature @xmath195 . if @xmath196 is the volume per h atom for a grain component , then its _ magnetic dipole _ contribution to the emissivity per h atom is = n_grvn_h 4h^4c^31 9_2(_1 + 2)^2 + _ 2 ^ 2 , where we must remember that there may be additional electric dipole emission [ see eq . ( [ eq : dipoleapprox][eq : cabs_md ] ) ] . the magnetic dipole emissivity for each grain component is shown in figure [ fig : emissivity ] ; also shown is the estimate for the `` vibrational '' or electric dipole emission , sensitive studies of the microwave sky brightness have revealed microwave emission from interstellar matter ; the emissivity per h nucleon has been deduced from the cross - correlation of the microwave sky brightness with far - infrared emission from dust grains , using measurements from the cobe dmr ( kogut et al . 1996 ) , the ground - based saskatoon experiment ( de oliveira - costa et al . 1997 ) , and ovro ( leitch et al . these observational results are shown in fig . [ fig : emissivity ] . we see that a population of fe grains or grains with fe inclusions would be predicted to produce very strong emission near @xmath197 , with the peak occurring near the frhlich resonance for fe spheres . even if the permeability of fe has been estimated correctly , this peak would be broadened if , as expected , the fe grains or inclusions are not all spherical , since the frhlich resonance frequency depends upon the particle shape . the @xmath3 emission reported by kogut et al . ( 1996 ) appears to limit the amount of interstellar fe in metallic form to perhaps @xmath198 of the total fe . in particular , this limits the fraction of interstellar fe which can be present in fe - ni inclusions such as found in `` gems '' ( bradley 1994 ; martin 1995 ) . we have estimated the emission expected from pure iron and magnetite , but the true state of fe in interstellar grains is not known . it is possible that the fe is concentrated in a form which is magnetic not so strongly as pure iron , but more strongly than ferrimagnetic magnetite . this could , for example , be an fe / ni alloy with an appreciable concentration of o , h , si , or other impurities . we now ask whether it is _ possible _ for magnetic dipole emission to account for the observed emission in the 14 - 90 ghz range . we approach this question by seeking to `` customize '' the magnetic properties , within the range of reasonable parameters , to see whether there are possible values which would lead to the observed emission . we make the following assumptions : 1 . the magnetic material is in ( spontaneously - magnetized ) single - domain grains , with a susceptibility @xmath120 perpendicular to the direction of spontaneous magnetization . the frequency dependence of the susceptibility is given by eq . ( [ eq : chi_cd ] ) . the low frequency susceptibility @xmath109 should be somewhat smaller than the value @xmath199 which we have estimated for pure iron . 4 . the spontaneous magnetization should be appreciably smaller than the value @xmath101 kg for pure iron . as a result , the characteristic gyrofrequency @xmath200 should be appreciably smaller than the @xmath201 value estimated for fe . we assume that nearly 100% of the interstellar fe is in the hypothetical material @xmath202 , and assume the fe within the material @xmath202 to contribute a mass density of @xmath203 [ thus the volume per h atom of the hypothetical material is @xmath204 see eq . ( [ eq : v_y ] ) ] . calculations are shown for 4 hypothetical materials , denoted x1 , x2 , x3 , and x4 , with @xmath109 and @xmath44 values as shown in fig . [ fig : custom_emissivity ] . in fig . [ fig : custom_emissivity ] we see that in order to contribute an appreciable fraction of the observed 14 - 90 ghz emission , the hypothetical magnetic material must have @xmath205 and @xmath206 . note that material x1 does not have a frhlich resonance , and materials x2,x3,x4 , with @xmath207 , have only a weak frhlich resonance ( see footnote [ fn : frohlich ] ) . we conclude that _ if _ a large fraction of the interstellar fe were incorporated into a moderately strong magnetic material with properties approximating those of our hypothetical material x4 , then the thermal emission from these grains would approximately reproduce the observed thermal emission in the 14 - 90 ghz range . we stress that we are _ not _ arguing that such material exists indeed , we have previously shown ( draine & lazarian 1998a , b ) that the observed emission seems likely to be due to rotational emission from a population of ultrasmall grains believed to be present for independent reasons . the point of the present discussion is to show that one can not exclude the possibility that magnetic grains make an appreciable contribution to the emission in this frequency range . this issue should be resolved through observations of dark clouds ( see [ sec : discussion ] ) . from fig . [ fig : emissivity ] it is seen that if a small fraction of interstellar fe is in iron grains , these could account for a substantial fraction of the diffuse emission in the @xmath21 region . alternatively , we see in fig . [ fig : custom_emissivity ] that if a large fraction of the fe is in a magnetic material with the properties of our hypothetical material `` x4 '' , then a substantial fraction of the 14 - 90 ghz emission could be magnetic dipole emission from such grains . in recent years there has been significant progress toward understanding the alignment of interstellar grains . analysis of the `` crossover '' phenomenon for grains subject to superthermal rotation now indicates that even ordinary paramagnetic relaxation may suffice to align larger ( @xmath208 ) grains ( lazarian & draine 1997 ) , but radiative torques due to starlight appear to dominate those due to paramagnetic relaxation ( draine & weingartner 1997 ) . if smaller ( @xmath209 ) strongly magnetic grains are present ( e.g. , iron or our hypothetical material x4 ) , it appears likely that ferromagnetic relaxation would effectively align their angular momenta with the galactic magnetic field @xmath210 , so that their `` long '' axes would tend to be perpendicular to @xmath210 . if such grains are present , we then expect frequency - dependent polarization of the microwave emission from interstellar dust . to estimate the likely polarization , we assume the grains to be ellipsoids with semiaxes @xmath107 , spinning with the short axis @xmath211 parallel to the angular momentum @xmath212 ( as expected for suprathermally rotating grains see purcell 1979 ) . we assume `` perfect '' alignment of @xmath212 with the galactic magnetic field @xmath210 , with @xmath210 perpendicular to the line - of - sight ; the polarization of emitted radiation will be taken to be positive when the electric vector is perpendicular to @xmath210 . for imperfect alignment , the polarization for perfect alignment should be multiplied by the `` rayleigh reduction factor '' ( lee & draine 1985 ) r=(3/2)(^2 - 1/3 ) , where @xmath213 is the angle between @xmath210 and @xmath212 . the predicted polarization depends on assumptions concerning the grain structure . we will consider two limiting cases . suppose the grains which dominate the @xmath214 emision each consist of a single magnetic domain . always consist of a single domain ( morrish 1980 ) . ] we will assume the grains to have spontaneously magnetized with @xmath94 along the long axis @xmath215 ( this minimizes the magnetic energy ) . as discussed above in [ sec : fe_single ] , the permeability @xmath147 is anisotropic : @xmath22 for @xmath216 , and @xmath217 for @xmath218 . thus the degree of polarization @xmath219 where i_e= 12 + _ , 2 |1+l_1(_-1)|^2 i_h= _ 2 |1+l_1(-1)|^2 + 12_,2 |1+l_2(_-1)|^2 , [ eq : i_e2 ] the dielectric function of metallic iron can be approximated at low frequencies by a drude model with @xmath220 : ( ) where @xmath221 and @xmath222 . this approximately reproduces the tabulated values of @xmath223 ( palik 1991 ) . with this dielectric constant , the microwave emission is dominated by magnetic dipole radiation ( @xmath224 for @xmath225 ) , so that @xmath226 and the precise value of @xmath223 is not critical . the resulting frequency - dependent polarization @xmath227 is shown in fig . [ fig : polarization ] for different grain shapes . for both fe or x4 grains , at high frequencies ( @xmath228 for fe , @xmath229 for x4 ) the polarization is large ( @xmath230 ) and positive ( i.e. , @xmath231 perpendicular to the `` short axis '' of the grain ) . at these high frequencies @xmath232 so that the values of the `` shape factors '' @xmath233 are unimportant the calculated polarization arises because the thermal fluctuations in the magnetization are perpendicular to @xmath215 ; since @xmath234 the magnetic dipole emission tends to have @xmath235 , and hence @xmath236 , for `` positive '' polarization . at lower frequencies ( @xmath237 for fe , @xmath238 for x4 ) @xmath239 and shape effects now matter . there are still no magnetic fluctuations parallel to @xmath215 , but if @xmath240 is sufficiently small compared to @xmath241 , @xmath242 will be considerably larger than @xmath243 , and the magnetic dipole emission will tend to have @xmath244 , hence @xmath245 , for `` negative '' polarization . this explains the behavior seen in fig . [ fig : polarization ] . suppose the grains to be ellipsoids containing isotropically - oriented magnetic inclusions with filling factor @xmath90 , as found for gems ( see [ sec : mag_grains ] ) . since the grain material is now isotropic , the polarization is given by eq . ( [ eq : pol]-[eq : i_h ] ) . if the inclusions are metallic fe , and contain @xmath246 atoms , we can use the permeability @xmath19 shown in fig . [ fig : spm ] . the resulting polarization for perfectly - aligned spinning grains is shown in fig . [ fig : polarization ] . thermal emission from these grains is predominantly magnetic dipole radiation ; since the grain material is isotropic the emitted radiation tends to have @xmath247 perpendicular to the short axis @xmath211 . however , since @xmath248 ( see fig . [ fig : spm ] ) , the @xmath36-dependent term in eq . ( [ eq : i_e]-[eq : i_h ] ) is only a small correction ; hence @xmath249 and the polarization @xmath219 is small . fe is the 9th most abundant element by number ( after h , he , o , c , n , ne , mg , and si ) , and 5th by mass ( after h , he , o , c ) , and nearly all of the interstellar fe is in dust grains . it is therefore inevitable that some fraction of the interstellar grain material must be quite fe - rich . we do nt know what mineral form the fe mainly resides in ( a number of possibilities are given table [ tab : feminerals ] ) , but the material will be at least paramagnetic , and quite possibly ferrimagnetic or ferromagnetic . it is therefore important to consider the emission of electromagnetic radiation as a result of thermal fluctuations in the magnetization of the grain material . these fluctuations take place at microwave frequencies and below ( the magnetization is constant on timescales short compared to the time for electrons to precess in the internal magnetic fields within the material ; these precession frequencies are at most @xmath16 ( the value for metallic iron ) . as a result , _ thermal _ magnetic dipole emission from interstellar grains @xmath250 may be be stronger than the thermal electric dipole `` vibrational '' emission ( arising from thermal fluctuation is in the charge distribution within the grain ) at frequencies @xmath251 . we show how this magnetic dipole emission can be calculated , but the emission requires knowledge of the magnetic permeability at these frequencies . we have estimated this permeability for various materials of interest , including metallic iron and magnetite ( fe@xmath10o@xmath11 ) , but have been required to extrapolate to the high frequencies of interest . it would be of great value to have direct measurements of the magnetic properties of small fe particles at 50 - 100 ghz . according to our estimates , metallic iron would produce such strong emission near @xmath252 that not more than @xmath4 of the fe can be in the form of pure metallic iron in order not to exceed the observed emission at these frequencies . this limits the fraction of interstellar grains with fe - ni inclusions such as in the gems found by bradley ( 1994 ) . magnetite , on the other hand , would radiate only a fraction of the power emitted by interstellar grains in the 14 - 90 ghz region . however , it is possible that interstellar fe could be present in some fe - rich substance ( e.g. , an fe / ni alloy with an appreciable concentration of mg , si , o , h impurities ) with magnetic properties intermediate between those of metallic iron and magnetite , and we therefore can not rule out the possibility that the bulk of the observed emission in the 14 - 90 ghz range could be thermal magnetic dipole radiation . it would be of great value to have laboratory measurements on various plausible candidate materials ( e.g. , amorphous olivine , or fe - containing alloys ) at these frequencies . rotational emission from spinning dust grains has previously been proposed as the explanation for the observed 15 - 90 ghz emission from dust in diffuse clouds ( dl98a , b ) . there are two ways in which this new emission mechanism magnetic dipole radiation from thermal fluctuations in grain magnetization can be distinguished from the hypothesized rotational emission : * the rotational emission requires ultrasmall grains . since we have reason to believe that ultrasmall grains are depleted in dense regions , we would then expect the rotational emission to be relatively weak in dense gas . on the contrary , thermal magnetic dipole emission from dust should be largely unaffected by coagulation of the dust grains . therefore , observations of ( or upper limits on ) the @xmath253 emission from dense clouds can be distinguish between rotational emission from ultrasmall grains or thermal magnetic dipole emisson from magnetic grain materials . * if produced largely by single - domain grains , the magnetic dipole emission potentially has a complex and strong polarization signature , which could be very different from the polarization expected from spinning dust grains which have been partially - aligned by magnetic dissipation ( lazarian & draine 1998 ) . the principal results of this paper are as follows : 1 . formulae are presented for electric and magnetic dipole absorption cross sections for homogeneous spheres or ellipsoids of material with dielectric function @xmath223 and magnetic permeability @xmath147 . the kramers - kronig relation relating the total grain volume to @xmath254 is generalized to include the case of magnetic grains . the resulting eq . ( [ eq : kkresult ] ) shows that if an appreciable fraction of the grain material is magnetic , then magnetic dipole effects must dominate the extinction cross section at the frequencies @xmath255 where plausible grain materials have a magnetic response . because most of the fe is in solid form in the interstellar medium , it is expected that some fraction of the interstellar grain population must be appreciably magnetic either paramagnetic , superparamagnetic , ferrimagnetic , or ferromagnetic . the frequency - dependent magnetic properties of candidate materials are estimated . the magnetic dipole emission from either paramagnetic grains or magnetite ( fe@xmath10o@xmath11 ) grains is well below the expected electric dipole emission due to vibrational modes at @xmath228 , or rotational emission from very small grains at @xmath0 . fe grains or inclusions will have the magnetic analogue of a frhlich resonance at @xmath256 , where the magnetic dipole absorption cross section peaks because @xmath116 . this will result in strong magnetic dipole absorption from ferromagnetic iron particles near @xmath256 . . if our estimate for the magnetic susceptibility of single - domain fe is correct , not more than @xmath4 of interstellar fe can be in the form of metallic iron ; otherwise the thermal magnetic dipole emission from interstellar dust at @xmath3 would greatly exceed the emission from dust measured by kogut et al . if a substantial fraction of interstellar fe is in a moderately strong magnetic material with the properties resembling our hypothetical material `` x4 '' in fig . [ fig : custom_emissivity ] , then thermal emission from such grains could account for a substantial fraction of the observed 14 - 90 ghz emission . at this time this possibility can not be excluded , even though the 14 - 90 ghz emission seems most likely due to spinning ultrasmall dust grains ( draine & lazarian 1998a , b ) . if nonspherical single - domain ferromagnetic grains are present , the magnetic dipole emission will be polarized . such grains are expected to be spontaneously magnetized along the `` long '' axis . the polarization is expected to depend strongly on frequency , with `` normal '' polarization ( electric vector perpendicular to @xmath210 ) at frequencies @xmath257 , but with a decrease , and perhaps even reversal ( depending upon grain shape ) , of the polarization for @xmath258 . we thank phil myers for helpful discussions , and robert lupton for the availability of the sm package . b.t.d . acknowledges the support of nsf grant ast-9619429 , and a.l . the support of nasa grants nag5 - 2858 and nag5 - 7030 the electric fields within a solid whether crystalline or amorphous can split the magnetic sub - levels of paramagnetic ions in the solid . kittel & luttinger ( 1948 ) noted that while these splittings usually correspond to frequencies @xmath259 , under some circumstances the splittings correspond to microwave frequencies . to estimate the magnitude of the associated absorption , suppose that the solid contain a density @xmath260 of species with energy levels split by frequencies in the interval @xmath261 , and let @xmath262 be the magnetic dipole matrix element between the upper and lower states , giving a spontaneous decay rate a_ul=4 ^ 33c^3|_ul|^2 . if @xmath263 , then @xmath264 . the absorption coefficient is = 2 ^ 23(kt)dnd |_ul|^2c , corresponding to an imaginary component of the magnetic susceptibility @xmath265 where @xmath266 is the bohr magneton . the total atomic density @xmath267 in the solid , so @xmath268 would correspond to 10% of the atoms having resonances within a frequency interval @xmath269 . such a high value would probably only occur if the stark - effect splittings are concentrated near a few frequencies . consider a spherical particle consisting of a single domain . at temperatures @xmath270 , where @xmath271 is the curie temperature , we may take the spontaneous magnetization to be approximately equal to the saturation magnetization @xmath177 . suppose the crystal has cubic symmetry . the free energy is a function of the direction of the magnetization . if @xmath272 , @xmath273 , and @xmath274 are the direction cosines of the magnetization relative to the crystal axes , the `` anisotropy '' free energy is f_k = k_1(_1 ^ 2_2 ^ 2+_2 ^ 2_3 ^ 2+_3 ^ 2_1 ^ 2 ) + k_2_1 ^ 2_2 ^ 2_3 ^ 2 where @xmath169 and @xmath275 are the anisotropy constants ( morrish 1980 ) . if @xmath276 and @xmath278 , then the `` easy '' directions are along the cubic axes ( e.g. , @xmath170 ) . if we now apply a weak magnetic field @xmath279 perpendicular to the magnetization , the magnetization will deflect by an angle @xmath213 . for weak fields , @xmath280 and the anisotropy energy and magnetic energy become f_k = k_1 ^2 + o(^4 ) f_h = -*m*= m_sh_+ o(^2 ) . [ eq : f_h ] minimizing @xmath281 , we find @xmath282 . thus , the transverse susceptibility @xmath283 is _ ( 0 ) = m_s^2 2 k_1 . [ eq : chi_perp_100 ] for fe , @xmath284 , and @xmath285 ( morrish 1980 ) , thus @xmath199 and @xmath286 . this is significantly smaller than the value @xmath287 characterizing bulk ( multidomain ) fe . when @xmath288 , @xmath289 , the `` easy '' directions are along the diagonals ( e.g. , @xmath290 ) . the anisotropy energy is now f_k= k_1 ( 29 - 23 ^ 2 ) + k_2 ( 127 - 29 ^ 2 ) + o(^4 ) , with @xmath291 again still by eq.([eq : f_h ] ) . again minimizing @xmath281 we find @xmath292 $ ] , and _ ( 0 ) [ eq : chi_perp_111 ] maghemite ( @xmath12fe@xmath13o@xmath10 ) is cubic with @xmath293 ( dunlop & zdemir 1997 ) . the magnetic properties do not appear to have been measured at low temperatures ; at room temperature , @xmath294 ( valstyn , hanton , & morrish 1962 ) , from which we estimate @xmath295 . as discussed in [ sec : ferri ] above , a slightly larger value of @xmath179 is in agreement with measurements by valstyn et al . ( 1962 ) at ghz frequencies . the above discussion applies to static fields . for oscillating fields , we estimate @xmath104 using eq . ( [ eq : chi_cd ] ) , with @xmath296 , the precession frequency of an electron in the internal field @xmath100 . allanson , j.t . 1945 , j. inst . ( london ) , 92 , pt . iii , 247 anders , e. , & grevesse , n. 1989 , geochim . acta , 53 , 197 bean , c.p , & livingston , j.d . , j. appl . phys . , 30 , suppl . 120s billas , i.m . , chtelain , a. , & de heer , w.a . 1994 , science , 265 , 1682 bohren , c.f . , & huffman , d.r . 1983 , absorption and scattering of light by small particles ( new york : wiley ) bradley , j.p . 1994 , science , 265 , 925 bramley , r. , & strach , s.j . 1983 , chem . , 83 , 49 carmichael , r.s . 1989 , in crc practical handbook of physical properties of rocks and minerals , ed . r.s . carmichael ( boca raton : crc press ) , p. 301 caspers , w.j . 1964 , theory of spin relaxation ( new york : interscience ) davis , l. , & greenstein , j. 1951 , apj , 114 , 206 de oliveira - costa , a. , kogut , a. , devlin , m.j . , netterfield , c.b . , page , l.a . , & wollack , e.j . 1997 , apj , 482 , l17 draine , b.t . 1996 , in `` polarimetry of the interstellar medium '' , ed . w.g . roberge and d.c.b . whittet , asp conf . 97 , 16 draine , b.t . , & lazarian , a. 1998a , apjl , 494 , l19 ( dl98a ) draine , b.t . , & lazarian , a. 1998b , apj , 508 , 000 ( dl98b ) draine , b.t . , & lee , h.m . 1984 , apj , 285 , 89 draine , b.t . , & weingartner , j.c . 1997 , apj , 480 , 633 duff , e.j . 1968 , j. chem soc . a , 2072 duley , w.w . 1978 , apj , 219 , l129 dunlop , d.j . , & zdemir , . 1997 , rock magnetism ( cambridge : cambridge univ . press ) epstein , d.j . 1954 , in `` dielectric materials and applications '' , ed . a. von hippel , ( new york : wiley ) , 122 . frhlich , h. 1949 , theory of dielectrics , ( london : oxford univ . press ) gaustad , j.e . , mccullough , p.r . , & van buren , d. 1996 , pasp , 108 , 351 goodman , a.a , & whittet , d.c.b . 1995 , apj , 455 , l181 jacobs , i.s . , & bean , c.p . 1963 , in magnetism , iii ( ed . g.t . rado & h. suhl ) , 271 jones , r.v . , & spitzer , l. 1967 , apj , 147 , 943 kittel , c. , & luttinger , j.m . 1948 , phys . , 73 , n2 , 162 kogut , a. , banday , a.j . , bennett , c.l . , gorski , k.m . , hinshaw , g. , & reach , w.t . 1996 , apj , 460 , 1 landau , l.d . , & lifshitz , e.m . 1960 , electrodynamics of continuous media , ( new york : pergamon ) lazarian , a. 1995 , apj , 453 , 229 lazarian , a. , & draine , b.t . 1997 , apj , 487 , 248 lazarian , a. , & draine , b.t . 1998 , in preparation leitch , e.m . , readhead , a.c.s . , pearson , t.j . , & myers , s.t . 1997 , apj , 486 , l23 lee , h .- m . , & draine , b.t . 1985 , apj , 290 , 211 martin , p.g . 1975a , apj , 201 , 373 martin , p.g . 1975b , apj , 202 , 389 martin , p.g . 1995 , apj , 445 , l63 mathis , j.s . 1986 , apj , 308 , 281 morrish , a.h . 1980 , the physical principles of magnetism , ( new york : r.e . krieger ) . pake , g.e . 1962 , paramagnetic resonance ( new york : w.a . benjamin ) . palik , e.d . 1991 , handbook of optical constants of solids ii ( new york : academic press ) . purcell , e.m . 1969 , apj , 158 , 433 purcell , e.m . 1979 , apj , 231 , 404 savage , b.d . , & sembach , k.r . 1996 , araa , 34 , 279 schuele , w.j . , & deetscreek , v.d . 1962 , j.appl.phys . , 33 , 1136s shapiro , p.r . 1975 , apj , 201 , 151 sorrell , w.h . 1994 , mnras , 268 , 40 spitzer , l. , & tukey , j.w . 1951 , apj , 114 , 187 syono , y. , & ishikawa , y. 1963 , j. phys . japan , 18 , 230 valstyn , e.p . , hanton , j.p . , & morrish , a.h . 1962 , phys . , 128 , 2078

we discuss the issue of quasi - particle production by `` analogue black holes '' with particular attention to the possibility of reproducing hawking radiation in a laboratory . by constructing simple geometric acoustic models , we obtain a somewhat unexpected result : we show that in order to obtain a stationary and planckian emission of quasi - particles , it is _ not _ necessary to create an ergoregion in the acoustic spacetime ( corresponding to a supersonic regime in the flow ) . it is sufficient to set up a dynamically changing flow _ either _ eventually generating an arbitrarily small sonic region , but without any ergoregion , _ or _ even just asymptotically , in laboratory time , approaching a sonic regime with sufficient rapidity . pacs : 04.20.gz , 04.62.+v , 04.70.-s , 04.70.dy , 04.80.cc + keywords : analogue models , acoustic spacetime , hawking radiation

gr-qc0604058

it is by now well established that the physics associated with classical and quantum fields in curved spacetimes can be reproduced , within certain approximations , in a variety of different physical systems the so - called `` analogue models of general relativity ( gr ) '' @xcite . the simplest example of such a system is provided by acoustic disturbances propagating in a barotropic , irrotational and viscosity - free fluid . in the context of analogue models it is natural to separate the kinematical aspects of gr from the dynamical ones . in general , within a sufficiently complex analogue model one can reproduce any pre - specified spacetime and the kinematics of fields evolving on it independently of whether or not it satisfies the classical ( or semiclassical ) einstein equations @xcite . indeed , to date there are no analogue models whose effective geometry is determined by einstein equations . in this sense we currently have both analogue spacetimes and analogues of quantum field theory in curved spacetimes , but ( strictly speaking ) no analogue model for gr itself @xcite . in order to reproduce a specific spacetime geometry within an analogue model , one would have to take advantage of the specific equations describing the latter ( for example , for fluid models , the euler and continuity equations , together with an equation of state ) , plus the possibility of manipulating the system by applying appropriate external forces . in the analysis of this paper we will think of the spacetime configuration as `` externally given '' , assuming that it has been set up as desired by external means any back - reaction on the geometry is neglected as in principle we can counter - balance its effects using the external forces . in the context of analogue models this is not merely a hypothesis introduced solely for theoretical simplicity , but rather a realistic situation that is in principle quite achievable . specifically , in this paper we analyze in simple terms the issue of quantum quasi - particle creation by several externally specified @xmath1-dimensional analogue geometries simulating the formation of black hole - like configurations . ( in a previous companion paper @xcite we investigated the causal structure of these , and other , spacetimes . ) in this analysis we have in mind , on the one hand , the possibility of setting up laboratory experiments exhibiting hawking - like radiation @xcite and , on the other hand , the acquisition of new insights into the physics of black hole evaporation in semiclassical gravity . all the discussion holds for a scalar field obeying the dalembert wave equation in a curved spacetime . this means that we are not ( for current purposes ) considering the deviations from the phononic dispersion relations that show up at high energies owing to the atomic structure underlying any condensed matter system . we shall briefly comment on these modifications at the end of the paper . for simplicity , throughout the paper we adopt a terminology based on acoustics in moving fluids ( we will use terms such as acoustic spacetimes , sonic points , fluid velocity , etc . ) , but our results are far more general and apply to many other analogue gravity models not based on acoustics . we summarise the main conclusions below . first of all , we recover the standard hawking result when considering fluid flows that generate a supersonic regime at finite time . ( that is , we recover a stationary creation of quasi - particles with a planckian spectrum . ) we then analyze the quasi - particle creation associated with other types of configurations . in particular , we shall discuss in detail a `` critical black hole '' a flow configuration that presents an acoustic horizon without an associated supersonic region . from this analysis we want to highlight two key results : * the existence of a supersonic regime ( sound velocity @xmath2 strictly smaller than fluid velocity @xmath3 ) is not needed in order to reproduce hawking s stationary particle creation . we demonstrate this fact by calculating the quantity of quasi - particle production in an evolving geometry which generates only an isolated sonic point ( @xmath0 ) , but without a supersonic region , in a finite amount of laboratory time . * moreover , in order to produce a hawking - like effect it is not even necessary to generate a sonic point at finite time . all one needs is that a sonic point develops in the asymptotic future ( that is , for @xmath4 ) _ with sufficient rapidity _ ( we shall explain in due course what we exactly mean by this ) . from the point of view of the reproducibility of a hawking - like effect in a laboratory , the latter result is particularly interesting . in general , the formation of a supersonic regime in a fluid flow normally considered to be the crucial requirement to produce hawking emission is associated with various different types of instability ( landau instability in superfluids , quantized vortex formation in bose einstein condensates , etc . ) that could mask the hawking effect . to reproduce a hawking - like effect without invoking a supersonic regime could alleviate this situation . from the point of view of gr , we believe that our result could also have some relevance , as it suggests a possible alternative scenario for the formation and semiclassical evaporation of black hole - like objects . the plan of the paper is the following : in the next section we introduce the various acoustic spacetimes on which we focus our attention , spacetimes that describe the formation of acoustic black holes of different types . in section [ sec : creation ] we present separately the specific calculations of redshift for sound rays that pass asymptotically close to the event horizon of these black holes . by invoking standard techniques of quantum field theory in curved spacetime , one can then immediately say when particle production with a planckian spectrum takes place . finally , in the last section of the paper we summarise and discuss the results obtained . associated with the flow of a barotropic , viscosity - free fluid along an infinitely long thin pipe , with density and velocity fields constant on any cross section orthogonal to the pipe , there is a ( 1 + 1)-dimensional _ acoustic spacetime _ @xmath5 , where the manifold @xmath6 is diffeomorphic to @xmath7 . using the laboratory time @xmath8 and physical distance @xmath9 along the pipe as coordinates on @xmath6 , the _ acoustic metric _ on @xmath6 can be written as@xmath10 = \omega^2\left[-c^2\,d t^2 + ( \d x+v\,\d t)^2\right ] \;,\label{metric}\]]where @xmath2 is the speed of sound , @xmath3 is the fluid velocity , and @xmath11 is an unspecified non - vanishing function @xcite . in general , all these quantities depend on the laboratory coordinates @xmath12 and @xmath13 . here , we shall assume that @xmath2 is a constant . hence , it is the velocity @xmath14 that contains all the relevant information about the causal structure of the acoustic spacetime @xmath5 . we direct the reader to the companion paper @xcite for a detailed analysis of the causal structure associated with a broad class of @xmath1-dimensional acoustic geometries , both static and dynamic . the sonic points , where @xmath15 , correspond to the so - called acoustic apparent horizons apparent horizons for the lorentzian geometry defined on @xmath6 by the metric ( [ metric ] ) . the fact of having an underlying minkowski structure associated with the laboratory observer makes the definition of apparent horizons in acoustic models less troublesome than in gr ( see reference @xcite , pp . consider a monotonically non - decreasing function @xmath16 such that @xmath17 and @xmath18 for @xmath19 . if one chooses @xmath20 in ( [ metric ] ) , the corresponding acoustic spacetime represents , for observers with @xmath21 , a static black hole with the horizon located at @xmath22 ( in this case apparent and event horizon coincide ) , a black hole region for @xmath23 , and a ( right - sided ) surface gravity@xmath24we can , moreover , distinguish three cases : * @xmath25 and @xmath26 for @xmath23 : a non - extremal black hole ; * @xmath25 and @xmath27 for @xmath23 : a `` critical '' black hole ; * @xmath28 and @xmath27 for @xmath23 : an extremal black hole . now , taking the above @xmath16 , let us consider @xmath13-dependent velocity functions@xmath29with @xmath30 a monotonically decreasing function of @xmath13 , such that @xmath31 and @xmath32 . ( the first condition serves to guarantee that spacetime is flat at early times , whereas we impose the second one only for simplicity . all the analysis in the paper could be performed without adopting this assumption , leaving the physical results unchanged . however , that would require more case - by - case splitting , only to cover new situations without physical interest . ) there are basically two possibilities for @xmath30 , according to whether the value @xmath33 is attained for a finite laboratory time @xmath34 or asymptotically for an infinite future value of laboratory time . in the first case @xmath35 and the corresponding metric ( [ metric ] ) represents the formation of a non - extremal , critical , or extremal black hole , respectively . for small values of @xmath36 we have@xmath37 ^ 2)\;,\label{appxi - bh}\]]where @xmath38 is a positive parameter . hence the function @xmath30 behaves , qualitatively , as shown in figure [ f : xi - bh ] . apart from this feature , the detailed behaviour of @xmath30 is largely irrelevant for our purposes . if instead @xmath39 is attained only at infinite future time , that is @xmath40 , one is describing the asymptotic formation of either a critical black hole ( if @xmath41 ; obviously , in this case choosing the non - extremal or the critical @xmath42 profile is irrelevant ) or an extremal black hole ( if @xmath28 ) . now the function @xmath30 behaves , qualitatively , as shown in figure [ f : xi - cbh ] . the relevant feature of @xmath43 is its asymptotic behaviour as @xmath44 . in the following we shall consider two possibilities for this asymptotics , although others can , of course , be envisaged : exponential : @xmath45 , with @xmath46 a positive constant , in general different from @xmath47 , and @xmath48 ; power law : @xmath49 , with @xmath50 and @xmath51 . for all the situations considered so far , spacetime is minkowskian in the two asymptotic regions corresponding to @xmath52 , and to @xmath4 , @xmath19 ( @xmath53 and @xmath54 , respectively , adopting the notation of reference @xcite ) . starting with a quantum scalar field in its natural minkowskian vacuum at @xmath52 , we want to know the total quantity of quasi - particle production to be detected at the right asymptotic region at late times , @xmath4 , caused by the dynamical evolution of the velocity profile @xmath14 . in the geometric acoustic approximation , a right - going sound ray is an integral curve of the differential equation@xmath55we are interested in sound rays propagating from @xmath56 ( see figure [ f : conf - bh ] ) ; that is , in solutions of ( [ diffeq ] ) that satisfy an initial condition @xmath57 , with @xmath58 in the limit @xmath59 ( so @xmath60 can be thought of as an `` initial '' event corresponding to the emission of the acoustic signal ) . if such a ray ends up on @xmath61 , we can identify `` final '' events @xmath62 on it , with @xmath63 as @xmath64 . for a ray connecting @xmath53 to @xmath54 one can also find an event @xmath65 such that @xmath66 , which corresponds to the crossing of the `` kink '' in @xmath3 , located at @xmath67 according to equation ( [ velocity ] ) , by the sound signal . finally , we can define , for such a ray , two parameters @xmath68 and @xmath69 as follows:@xmath70@xmath71such parameters correspond to null coordinates in spacetime . if an acoustic event horizon @xmath72 is present in the spacetime , the coordinate @xmath68 is regular on it ( , @xmath68 attains some finite value on @xmath72 ) , whereas @xmath69 tends to @xmath73 as @xmath74 is approached . we can express both @xmath68 and @xmath69 in terms of the velocity profile ( shape and dynamics ) and of the crossing time @xmath75 . to this end , we can integrate equation ( [ diffeq ] ) , first between @xmath76 and @xmath77:@xmath78then between @xmath77 and @xmath79:@xmath80on replacing @xmath81 from equation ( [ xi - x ] ) into ( [ u ] ) , we find the value of @xmath68 for a generic right - moving ray that crosses the kink at laboratory time @xmath75:@xmath82similarly , substituting @xmath83 from equation ( [ t - t ] ) into ( [ u ] ) , then adding and subtracting the quantity @xmath84 , we find@xmath85 in the analysis below , our chief goal consists of finding the relation between @xmath68 and @xmath69 for a sound ray that is close to the horizon , , in the asymptotic regime @xmath86 . from such a relation it is then a standard procedure to find the bogoliubov @xmath87 coefficients and hence the total quasi - particle content to be measured , in this case , by an asymptotic observer at @xmath54 . ( see , for example , reference @xcite ) . in the case of an exponential relation between @xmath68 and @xmath69 it is a well established result that a planckian spectrum is observed at late times @xcite , so hawking - like radiation will be recovered . when the apparent horizon forms at a finite laboratory time , say at @xmath88 , an event horizon always exists , generated by the right - moving ray that eventually remains frozen on the apparent horizon , at @xmath22 . for such a ray @xmath89 , and since @xmath35 , the @xmath68 parameter has the finite value@xmath90for a ray with @xmath91 we then obtain , combining equations ( [ uu ] ) and ( [ uh - fin]):@xmath92this _ exact _ equation is now in a form suitable for conveniently extracting _ approximate _ results in the region @xmath93 , corresponding to sound rays that `` skim '' the horizon . on the other hand , when the trapping horizon consists of just one single sonic point located at @xmath94 , it is not obvious that an event horizon exists . loosely speaking , in this case it might happen that the trapping horizon form `` after '' every right - going ray from @xmath53 has managed to cross @xmath22 . since there is a competition between two infinite quantities the time at which the trapping horizon forms , and the time at which the `` last '' right - going signal that connects @xmath95 with @xmath96 crosses @xmath22 a careful case - by - case analysis is in order . this is essentially all that can be said without relying on specific features of @xmath16 . we now consider separately the various situations of interest , focussing first on the issue of the existence of the event horizon . in the case of a non - extremal black hole , the qualitative behaviour of the function @xmath16 is shown , graphically , in figure [ f : v - bh ] . note that , for small values of @xmath97 , one can write@xmath98the function @xmath43 behaves as already shown in figure [ f : xi - bh ] . a sketch of the worldlines of right - moving sound rays is presented in figure [ f : tx - bh ] . note that in the portion of the diagram to the right of the curve @xmath67 ( , to the right of the moving kink in the velocity profile ) , spacetime is static . for @xmath52 , the geometry is minkowskian and the worldlines tend to approach straight lines with slope @xmath99 . the sound ray that generates the event horizon corresponds to a finite ) , given the asymptotic behaviours of @xmath100 and @xmath30 . ] value @xmath101 of the coordinate @xmath68 . hence , in this situation an event horizon always exists . this is also clear from the fact that the vertical half - line @xmath22 , @xmath102 in figure [ f : tx - bh ] is an apparent horizon . the function @xmath16 behaves as shown in figure [ f : v - cbh ] . regarding the right side of the profile , @xmath21 , it is indistinguishable from the profile of a non - extremal black hole ( figure [ f : v - bh ] ) . when the function @xmath43 is of the form ( [ appxi - bh ] ) , that is , when the apparent horizon is formed at a finite amount of laboratory time , the situation is exactly the same as for the non - extremal black hole discussed above . consider now that the sonic point is approached in an infinite amount of time , so the function @xmath43 behaves as in figure [ f : xi - cbh ] . the worldlines of right - moving sound rays are shown in figure [ f : tx - cbh ] . as in the formation of the non - critical black hole , the portion of the diagram to the right of the curve @xmath67 ( , to the right of the moving kink in the velocity profile ) corresponds to a static spacetime , and for @xmath52 the geometry is minkowskian the worldlines tend to approach straight lines with slope @xmath99 . however , now the apparent horizon is just the asymptotic point located at @xmath22 , @xmath4 , and in order to establish whether an event horizon does , or does not , exist one must perform an actual calculation of @xmath101 for the `` last '' ray that crosses the kink . the expression for @xmath101 is again obtained from equation ( [ uu ] ) , noticing that now @xmath103 along the generator of the would - be horizon , so@xmath104the necessary and sufficient condition for the event horizon to exist is that the limit on right hand side of equation ( [ baru - cbh ] ) be finite . the integrand on right hand side of ( [ baru - cbh ] ) can be approximated , for @xmath105 , as @xmath106 , while for @xmath52 it just approaches zero . hence @xmath101 is , up to a finite constant , equal to @xmath107 times the integral of @xmath30 , evaluated at @xmath108 . here we must distinguish between the exponential behaviour and the power law cases ( i ) and ( ii ) . in the former @xmath101 is finite , trivially . for the power law , it turns out that @xmath101 is finite iff @xmath109 . the typical spatial profile function @xmath16 for an extremal black hole is plotted in figure [ f : v - ebh ] . for @xmath12 approaching zero from positive values we can write@xmath110where @xmath111 is a constant . as far as dynamics is concerned , we must distinguish the cases in which the apparent horizon is formed at finite laboratory time @xmath34 , and in an infinite time ( , for @xmath4 ) . the function @xmath43 is of the type shown in figure [ f : xi - bh ] , and the worldlines of right - going sound rays are sketched in figure [ f : rays - ex - fin ] . the event horizon always exists . the function @xmath43 is as shown in figure [ f : xi - cbh ] , and the worldlines of right - going signals are shown in figure [ f : tx - cbh ] . as in the case of the formation of a critical black hole , the apparent horizon forms only asymptotically , for @xmath22 and @xmath4 , so the event horizon exists iff @xmath101 , given by equation ( [ baru - cbh ] ) , has a finite value . using the expansion ( [ appv - ebh ] ) in equation ( [ baru - cbh ] ) , one finds that @xmath112 is always finite when @xmath43 is asymptotically exponential . on the other hand , for a power law , the event horizon exists iff @xmath113 . ( note that the critical value of the exponent , @xmath114 , is now _ not _ the same as for the critical black hole , @xmath115 . ) the configurations we have analyzed until now are the simplest from a purely mathematical point of view . however , having in mind acoustic analogue geometries reproducible in a one - dimensional pipe in the laboratory , it is more sensible to consider double - sided configurations . by this we mean that , after passing ( or approaching ) the sonic / supersonic regime at @xmath22 , and traversing an interval of width @xmath116 , the fluid again goes back to a subsonic regime as @xmath117 . consider for example functions @xmath16 such that @xmath27 for @xmath118,@xmath119and which outside the interval @xmath118 tend monotonically to zero as @xmath97 increases ( see figures [ f : flat - spike ] and [ f : spike ] ) . the corresponding fluid configuration represents what could be called a static `` double - sided critical black hole '' . the formation of such a configuration can be modelled by the velocity function@xmath120with @xmath30 a monotonically decreasing function of @xmath13 , and @xmath100 as above . accordingly , the differential equation for right - going sound rays also splits:@xmath121geometries associated with the formation of non - extremal and extremal black holes can be constructed in the same way ; see figures [ f : v - bh - wh ] and [ f : v - ex - bh ] for plots of the respective @xmath100 functions . the function @xmath43 is of the type illustrated in figure [ f : xi - bh ] . the behaviour of right - going sound rays is shown in figure [ f : rays - spike - fin ] . the apparent horizon is the half - line @xmath22 , @xmath122 , and the event horizon always exists . let us now consider a function @xmath43 of the type illustrated in figure [ f : xi - cbh ] . the behaviour of right - going sound rays is shown in figure [ f : rays - spike - inf ] . for these particular configurations , whether an event horizon does , or does not , actually exist is now a rather tricky issue . the asymptotic behaviour of the function @xmath43 at @xmath123 ensures that all the right - going rays start to the left of the right - moving kink ( , the one in the region @xmath23 ) , then catch up with it , and begin to propagate through the intermediate region at a velocity @xmath124 that depends only on @xmath13 . before they reach the point @xmath22 , such rays might be overtaken by the right - moving kink , but only to start the chase again . after several mutual overtakings ( if the function @xmath43 is sufficiently complicated ) , the rays will always make an ultimate overtaking of the right - moving kink , embarking upon a final encounter with the left - moving kink on the right ( , in the region @xmath21 ) . let us denote by @xmath125 the time of such a last crossing of the right - moving kink , so the corresponding event is @xmath126 . also , let us denote by @xmath127 the time at which the same ray crosses the kink on the right , so that the corresponding event is @xmath128 . from equation ( [ diffeq - dsc ] ) we directly obtain the relation@xmath129\;.\label{ziopollo}\]]between @xmath127 and @xmath125 . ( when the ray crosses the right - moving kink more than once , equation ( [ ziopollo ] ) will be satisfied by more than one value of @xmath125 for any given @xmath127 . in order to avoid cumbersome notation , we shall simply denote by @xmath125 the largest of these roots , corresponding to the last crossing . ) then a necessary and sufficient condition for the existence of an event horizon is that , for @xmath130 , @xmath125 tends to a finite value , say @xmath131 . this guarantees that any right - going ray that last crosses the left kink at a time greater than @xmath131 does not reach the region @xmath21 ( as ray - crossing can not occur under the working hypothesis of this paper ) . applying this condition straightforwardly in order to see whether the event horizon exists is not easy . indeed , that would require us to evaluate the integral in equation ( [ ziopollo ] ) for a generic , finite value of @xmath127 , then solve for @xmath125 as a function of @xmath127 . it is easier to use one of the following two alternative strategies : 1 . instead of asking whether the event horizon exists , one can ask whether the event horizon _ does not _ exist . a necessary and sufficient condition for this is that , for @xmath130 , also @xmath132 . in such a case , we can insert the asymptotic expansions ( [ appv - bh ] ) or ( [ appv - ebh ] ) into equation ( [ ziopollo ] ) to get@xmath133for a non - extremal black hole , and@xmath134for an extremal one . plugging into these expressions the different asymptotic behaviours of the function @xmath43 , one can explicitly solve for @xmath125 as a function of @xmath127 for large values of the latter , and check whether @xmath125 does , or does not , tend to infinity when @xmath130 . 2 . setting @xmath135 into ( [ ziopollo ] ) , one obtains@xmath136\;.\label{xihinf}\]]it is possible to show that the event horizon exists if and only if equation ( [ xihinf ] ) possesses an odd number of finite solutions . is the solution of equation ( [ xihinf ] ) with the largest value . ] in order to establish whether this is the case , it is convenient to define the function of @xmath13@xmath137\;,\label{integral}\]]whose points of crossing with @xmath138 correspond to the solutions of equation ( [ xihinf ] ) . of course , for @xmath139 to be well defined ( and therefore for solutions of ( [ xihinf ] ) to exist at all ) one needs the integral defining it to be convergent . for asymptotically ( @xmath140 ) exponential and power - law behaviours of @xmath43 this happens in the cases already described . now , whenever @xmath139 is well defined , it is clearly a monotonically decreasing function , because the integrand in equation ( [ integral ] ) is always strictly positive . for @xmath123 , @xmath141 is just equal to the integral of the function @xmath124 evaluated at @xmath13 , up to a finite constant . in this limit , @xmath142 so we can write , for @xmath52:@xmath143given the condition @xmath144 , it is clear that the function @xmath141 is always greater than @xmath138 for @xmath52 . on the other hand , for @xmath145 , the asymptotic behaviour of @xmath141 is obtained by expanding @xmath100 in ( [ integral ] ) , which gives@xmath146for critical ( and non - extremal ) black holes and@xmath147for extremal ones . if , for @xmath4 , @xmath141 is smaller ( greater ) than @xmath138 , then equation ( [ xihinf ] ) has an odd ( even ) number of finite solution , and the event horizon does ( does not ) exist . note that , if @xmath148 , one must analyse subdominant terms in the asymptotic behaviour of @xmath43 in order to draw any conclusion . with either method , we find that for @xmath149 the existence of an event horizon in double - sided configurations follows the same rules as in the previously analysed one - sided configurations . when @xmath150 however , it is more difficult to have an event horizon in a double sided configuration , and in general , one has to increase the _ rapidity _ with which one approaches the sonic regime . more specifically , for a critical black hole and an asymptotically exponential @xmath43 , the event horizon exists if @xmath151 , but not when @xmath152 , while for a power law there is no horizon . for an extremal black hole and an exponential @xmath43 the horizon always exists , but in the case of a power law it does not exist if @xmath153 , and it exists for @xmath154 , with the additional condition @xmath155 for the particular value @xmath115 . for a critical black hole with asymptotically exponential @xmath43 and @xmath156 , as well as for an extremal black hole with a power law and @xmath115 , @xmath157 , the asymptotic analysis is not sufficient and one must take into account also subdominant terms in the expansion of @xmath43 for @xmath145 . for those situations in which an event horizon exists , we now find the asymptotic relation between @xmath69 and @xmath68 for rays close to the horizon generator . we also briefly discuss the implications of such a relation for quasi - particle creation in the various cases of interest . consider a sound ray corresponding to a value @xmath91 . for @xmath68 very close to @xmath101 , @xmath75 is very close to @xmath34 , and we can use the approximation ( [ appxi - bh ] ) for @xmath43 . furthermore , we can approximate @xmath100 as in ( [ appv - bh ] ) , so equation ( [ u - fin ] ) gives@xmath158 ^ 2)\;.\label{ut0-bh}\]]this provides us with the link between @xmath68 and @xmath75 . in order to link @xmath75 with @xmath69 , consider the integral on the right hand side of equation ( [ uu ] ) . for @xmath19 , the integrand function vanishes , while near @xmath159 it can be approximated by @xmath160 . then the integral is just given by the difference of the corresponding integrals evaluated at @xmath96 and @xmath161 , respectively , up to a possible finite positive constant . this gives given in reference @xcite . using equation ( 4.2 ) from that paper we have@xmath162expanding , we find again equation ( [ porcoqua ] ) , to the leading order in @xmath163.]@xmath164together with equation ( [ ut0-bh ] ) , this leads to@xmath165this relation between @xmath68 and @xmath69 is exactly the one found by hawking in his famous analysis of particle creation by a collapsing star @xcite . it is by now a standard result that this relation implies the stationary creation of particles with a planckian spectrum at temperature @xmath166 @xcite . for a critical black hole , the results are very different according to whether the sonic regime is attained in a finite or an infinite laboratory time . the calculation of the relation between @xmath68 and @xmath69 is exactly equal to the one presented for the non - extremal black hole case . the two geometries coincide everywhere to the right of the apparent horizon and can not be distinguished by the quasi - particle production observed at @xmath167 . let us suppose that we are in a situation in which the event horizon exists , so @xmath101 is finite . for another right - moving sound ray that corresponds to a value @xmath91 we find , combining equations ( [ uu ] ) and ( [ baru - cbh]),@xmath168\;.\label{deltau - cbh}\]]in the integration interval , @xmath43 is close to zero , so equation ( [ deltau - cbh ] ) can be approximated as@xmath169where the expansion ( [ appv - bh ] ) has been used . equation ( [ deltau-cbh ] ) gives@xmath170for an asymptotically exponential @xmath30 , and@xmath171for a power law with @xmath109 . for the link between @xmath75 and @xmath69 we obtain@xmath172as one can easily check inserting the appropriate asymptotic expansions into equation ( [ uu ] ) . ) which , expanded , gives equation ( [ tu ] ) . the result holds , however , independently of the details of @xmath16 . ] using equation ( [ tu ] ) into equations ( [ deltauexp ] ) and ( [ deltaunu ] ) we find@xmath173for the exponential case , and@xmath174for a power law with @xmath109 . ( remember that for @xmath175 the event horizon does not form . ) it is interesting to compare equations ( [ uuexp ] ) and ( [ uunu ] ) with the corresponding one for the non - critical black hole , equation ( [ uu-bh ] ) . whereas the latter is basically independent of the details of the black hole formation ( which only appear in the multiplicative constant ) , the relation between @xmath68 and @xmath69 in the critical case is not universal , but depends on the dynamical evolution . even for an asymptotically exponential @xmath43 , which leads to an exponential dependence on @xmath69 , the coefficient in the exponent is not universal as in equation ( [ uu-bh ] ) , but depends on dynamics through the parameter @xmath46 . this is not difficult to understand looking back at the way in which equations ( [ uu-bh ] ) and ( [ uuexp ] ) have been derived . for equation ( [ uu-bh ] ) , the exponential dependence was introduced relating @xmath75 with @xmath69 , which only involves sound propagation in the final static region and can not , therefore , be affected by dynamics . on the other hand , when deriving equation ( [ uuexp ] ) it is sound propagation in the initial , dynamical , regime that introduces the exponential ( in the particular case of an asymptotically exponential @xmath30 ) ; hence , it is not surprising that the final result keeps track of the dynamical evolution . however , it is interesting to note that in the limit @xmath176 equations ( [ uu-bh ] ) and ( [ uuexp ] ) coincide . this limit corresponds to a very rapid approach towards the formation of an otherwise - never - formed ( in finite time ) apparent horizon . regarding the creation of quasi - particles , this situation is operationally indistinguishable from the actual formation of the sonic point . however , this `` degeneracy '' might be accidental , given that the origin of the exponential relation is very different in the two cases . as for the case of a critical black hole , we must distinguish between a finite and an infinite time of formation of the event horizon . for a sound ray close to the one that generates the horizon , equation ( [ u - fin ] ) still holds . however , now one must use the expansion ( [ appv - ebh ] ) when approximating the integrand thus obtaining@xmath177 ^ 3)\;.\label{deltaut0-ebh1}\]]using again the approximation ( [ appv - ebh ] ) in the evaluation of the integral on the right hand side of equation ( [ uu ] ) one finds@xmath178finally,@xmath179interestingly , this is the same relation that one finds for the gravitational case @xcite . in particular , this implies that finite time collapse to form an extremal black hole will _ not _ result in a planckian spectrum of quasi - particles @xcite . this is completely compatible with the standard gr analysis , and is one of the reasons why extremal and non - extremal black holes are commonly interpreted as belonging to completely different thermodynamic sectors @xcite . assuming that the event horizon exists , we can again apply equation ( [ deltau - cbh ] ) and use the approximation ( [ appv - ebh ] ) in order to find the relation between @xmath68 and @xmath75 . the results are , for an asymptotically exponential @xmath43:@xmath180for a power law with @xmath181:@xmath182for a power law with @xmath115:@xmath183for a power law with @xmath109:@xmath184using the appropriate expansions in equation ( [ t - t ] ) , one obtains the relation between @xmath75 and @xmath69:@xmath185for an asymptotically exponential @xmath43 this becomes@xmath186for a power law , one must again distinguish between three cases ; for @xmath181:@xmath187for @xmath115:@xmath188for @xmath109:@xmath189putting together equations ( [ utexp - ebh ] ) and ( [ tuexp - ebh ] ) one finds the relationship between @xmath68 and @xmath69 for the exponential case:@xmath190for the power law one finds from equations ( [ utnu<1])([utnu>1 ] ) and ( [ utnu<1])([utnu>1 ] ) , for @xmath181:@xmath191for @xmath115:@xmath192and finally , for @xmath109:@xmath193 in all these cases , quasi - particle production is neither universal , nor planckian . it is not difficult to prove that in the formation , in a finite amount of time , of double - sided non - extremal black holes , double - sided extremal black holes , and double - sided critical black holes , the asymptotic relation between @xmath68 and @xmath69 is identical to that calculated in the corresponding subsections above . the amount and features of quasi - particle creation are then the same . we will demonstrate this in detail for the case of a double - sided critical black hole , and then proceed to consider the situation in which the formation takes place in an infinite amount of time . using the same notation as in section [ paragraph:2cbh2 ] , let us call @xmath125 the largest of the @xmath125 s that satisfy equation ( [ ziopollo ] ) , so @xmath125 is the time at which a right - going ray last crosses the kink on the left . there will be some regular relationship between @xmath68 and @xmath125 , expressed by some differentiable function @xmath139 , so that we can write @xmath194 . for the event horizon to exist , the corresponding @xmath68 must be finite ( equal to some value @xmath101 , say ) , so also @xmath195 must be finite ( as already done in section [ paragraph:2cbh2 ] , we denote by a suffix `` h '' the quantities that correspond to the horizon generator ) . for a ray very close to the horizon generator we have@xmath196where a dot denotes the derivative with respect to @xmath13 . on the horizon , @xmath197 so equation ( [ ziopollo ] ) reduces to@xmath198\;.\label{ziogallo}\]]subtracting ( [ ziogallo ] ) from ( [ ziopollo ] ) we obtain@xmath199 -\int_{t_2}^{t_{\rm h}}\d t\left[c + \bar{v}\left(\xi(t)\right)\right]\;.\label{ziopollogallo}\]]for a ray close to the horizon generator , @xmath127 is close to @xmath200 , and @xmath125 close to @xmath131 , so equation ( [ ziopollogallo ] ) gives , keeping only terms to the leading order:@xmath201together , equations ( [ uf ] ) and ( [ eccola ] ) provide a linear link between @xmath68 and @xmath127 . since the relationship between @xmath127 and @xmath69 is exactly the same as the one between @xmath75 and @xmath69 in equation ( [ porcoqua ] ) , the final result is again the one expressed by ( [ uu-bh ] ) : @xmath202 assuming that we are in a situation for which the event horizon does indeed exist , we can subtract equation ( [ xihinf ] ) with @xmath203 from equation ( [ ziopollo ] ) , finding:@xmath204-\int_{t_2}^{+\infty}\d t\left[c + \bar{v}\left(\xi(t)\right)\right]\;.\label{xi - xi}\]]for a ray close to the horizon generator , @xmath125 is close to @xmath131 and @xmath127 is large , so@xmath205for an asymptotically exponential @xmath43 we find , performing the integral,@xmath206similarly , for a power law with @xmath109:@xmath207 in both cases , the same results as in section [ subsec : cbh ] , equations ( [ uuexp ] ) and ( [ uunu ] ) , follow . in short , the amount and characteristics of the quasi - particle production calculated with the double - sided configurations are exactly the same as those calculated with the simpler profiles in the previous subsections except in two specific situations : the double - sided critical black hole with @xmath150 ( see figure [ f : spike ] ) and the double - sided extremal black hole . in the critical case , only the asymptotically exponential behaviour with @xmath208 produces an event horizon and , therefore , only then we can talk about a stationary and planckian creation of quasi - particles . in the extremal case the results described in section [ subsubsec : ebh2 ] only apply for @xmath209 ( with the further condition @xmath210 in the particular case @xmath115 ) , because otherwise the event horizon itself does not exist . in the present paper we have analyzed different dynamical black hole - like analogue geometries with regard to their properties in terms of quantum quasi - particle production . we have taken several @xmath1-dimensional spacetimes ( considered as externally fixed backgrounds ) , and for each of them we ( i ) have calculated whether it possesses an event horizon or not , and if the answer is `` yes '' , ( ii ) have calculated the asymptotic redshift function that characterizes the amount and properties of the late - time quasi - particle production . in table [ table ] the reader can find a summary of all our results . .this is a summary of the results found for the different configurations analyzed in the paper . to the table we have to add the following comments : for double - sided extremal black hole with @xmath150 , with an infinite time of formation and an asymptotic power law for @xmath43 , the horizon forms in the case @xmath115 if the further condition @xmath155 holds . for the double - sided configurations with @xmath150 and an infinite time of formation , the asymptotic analysis is not sufficient for drawing conclusions when @xmath43 is asymptotically exponential and @xmath156 , and when @xmath43 is asymptotically a power law and @xmath115 , @xmath157 . for those cases , one need also consider subdominant terms in @xmath43 , so the results will depend on the details of formation . [ cols="^,^,^,^,^,^",options="header " , ] the above results are pertinent to a purely mathematical model . their physical relevance has to be assessed with respect to their application to both experimental reproduction of the analogue hawking radiation , and to the lessons they can provide concerning the possible behavior of black hole formation and evaporation in semiclassical gravity . we now turn to separately consider these two issues . the study carried on in this paper has identified several velocity profiles that are potentially interesting for experiments . in particular the critical black hole models seem worth taking into consideration in connection with the realizability of a hawking - like flux in the laboratory . the creation of supersonic configurations in a laboratory is usually associated with the development of instabilities . there are many examples of the latter in the literature ; in reference @xcite it was shown that in an analogue model based on ripplons on the interface between two different sliding superfluids ( for instance , @xmath211he - phase a and @xmath211he - phase b ) , the formation of an ergoregion would make the ripplons acquire an amplification factor that eventually would destroy the configuration . therefore , this analogue system , although very interesting in its own right , will prove to be useless in terms of detecting a hawking - like flux . however , by creating , instead of an ergoregion , a critical configuration one should be able , at least , to have a better control of the incipient instability , while at the same time producing a dynamically controllable hawking - like flux . nevertheless , the actual realization of a critical configuration could also appear as a problematic task for entirely different reasons . the corresponding velocity profiles are characterized by discontinuities in the derivatives , so one might wonder whether they would be amenable to experimental construction , given that the continuum model is only an approximation . let us therefore discuss in some detail the validity of the latter for realistic systems . the main difference between an ideal perfect fluid model and a realistic condensed matter analogue is due to the microscopic structure of the system considered . in particular , it is generic to have a length scale @xmath212 which characterizes the breakdown of the continuum model ( @xmath212 is of the order of the intermolecular distance for an ordinary fluid ; of the coherence length for a superfluid ; and of the healing length for a bose einstein condensate ) . in general , the viability of the analogue model requires one to consider distances @xmath213 of order of at least a few @xmath212 , depending on the accuracy of the experiment performed . in particular , wave propagation is well defined only for wavelengths larger than @xmath212 ( generally with an intermediate regime , for wavelengths between @xmath212 and @xmath213 , where the phenomena exhibit deviations with respect to the predictions based on the continuum model ) . in general , a mathematical description based on the continuum model contains details involving scales smaller than @xmath213 ( for example , in the velocity profile ) . these details should , however , be regarded as unphysical : they are present in the model , but do not correspond to properties of the real physical system . in particular , they can not be detected experimentally , because this would require using wavelengths smaller than @xmath213 , which do not behave according to the predictions of the model ( and for wavelengths smaller than @xmath212 do not even make physical sense ) . for the mathematical models considered in the present paper , all this implies that one will not be able to distinguish , on empirical grounds , between those cases for which the velocity profiles differ from each other only by small - scale details . in particular , double - sided configurations with @xmath150 should be equivalent to configurations with a small , but non - zero , thickness @xmath214 . also , one would not be able to distinguish between two velocity profiles that differ only in a neighborhood @xmath213 of @xmath22 , one of which corresponds to a critical black hole , while the other describes an extremal one . in particular hawking radiation will not distinguish between the models within each of these pairs . this fact would not be troublesome , had our analysis led to identical results for the acoustic black holes of each pair . however , this is not the case ( see table [ table ] ) . but then what shall we see if we realize these models in a laboratory ? in realistic situations , what is relevant for hawking radiation is a coarse - grained profile obtained by averaging over a scale of order @xmath213 , thus neglecting the unphysical small scale details in @xmath16 . this implies that as far as double - sided critical black holes are concerned , the reliable results are those pertinent to the non - zero thickness case ( @xmath215 ) . similarly , since these extremal black holes are never exactly realizable in a laboratory ( as this would require tuning the velocity profile on arbitrary small scales ) , only the predictions based on the critical black hole mathematical model will survive in an experimental setting . indeed , the relevant surface gravity will be defined by averaging the slope of the velocity profile over scales which are of order of @xmath213 . this averaged surface gravity will be non - zero for both the critical and the extremal black hole , but will be approximately equal to the surface gravity at the horizon of the critical black hole , while it will obviously not coincide with the one of the extremal ( which is zero ) . in the body of the paper we have used a terminology particularly suitable to dealing with analogue models based on acoustics . let us now discuss the most relevant features of our findings using a language more natural to gr . when the geometry associated with the formation of a spherically symmetric black hole through classical gravitational collapse ( as , for example , in the oppenheimer - snyder model @xcite ) is described in terms of painlev - gullstrand @xcite coordinates ( whose counterpart , in the context of acoustic geometries , are the natural laboratory coordinates @xmath12 and @xmath13 ) , the apparent horizon forms in a finite amount of coordinate time . in this regard , the painlev - gullstrand time behaves similarly to the proper time measured by a freely - falling observer attached to the surface of the collapsing star . the non - extremal , non - critical @xmath1-dimensional model analysed in this paper , captures the main features of the formation of a ( non - extremal ) black hole . the dynamical collapse is represented by the function @xmath43 in our calculations . in the language of gr , we can think of @xmath43 as the radial distance between the surface of a collapsing star and its schwarzschild radius ; @xmath35 corresponds to the moment in which the surface of the star enters its schwarzschild radius , and this moment corresponds to a finite time ( which we took to be @xmath34 ) . for this model we recovered hawking s result that the formation of ( non - extremal ) black holes causes the quantum emission towards infinity of a stationary stream of radiation with a planckian spectrum , at temperature @xmath166 . the mechanism for particle creation is somewhat `` more than dynamical '' as the characteristics of the stationary stream of particles are `` universal '' and only depend on the properties of the geometry at the horizon , @xmath47 , and not on any detail of the dynamical collapse . indeed , for @xmath43 given by equation ( [ appxi - bh ] ) apparent horizon formation in a finite amount of time we have seen that asymptotic quasi - particle creation does not depend even on the coefficient @xmath38 . that is , particle production does not depend on the velocity with which the surface of the collapsing star enters its schwarzschild radius . this picture leans toward the ( quite standard ) view that hawking s process is not just dynamical , but relies on the actual existence of an apparent horizon and an `` ergoregion '' beyond it , able to absorb the negative energy pairs @xcite . however , by analyzing alternative models , in this paper we have seen two unexpected things : i ) : : one can also produce a truly hawking flux with a temperature @xmath166 through the formation in a finite amount of time of either a single - sided critical black hole , or a double - sided critical black hole of finite thickness " , or even one of zero thickness " ( see figure [ f : spike ] ) . this is an intriguing result , as in none of these cases there is an `` ergoregion '' beyond the apparent horizon , and in the last case there is just a single sonic point . ( in the language of gr , this last configuration corresponds to stopping the collapse of a star at the very moment in which its surface reaches the schwarzschild radius . ) ii ) : : moreover , one can also produce a stationary and planckian emission of quasi - particles by , instead of actually forming the apparent horizon , just approaching its formation asymptotically in time with sufficient rapidity ( @xmath216 ) . in this case the temperature is not @xmath166 but @xmath217 , with@xmath218and ( at any finite time ) there is neither an apparent horizon nor an ergoregion within the configuration . explanations of particle production based on tunneling then seem not viable , and the phenomenon is closer to being interpreted as dynamical in origin . if fact , these configurations interpolate between situations in which the dynamics appears more prominently when @xmath219 we have that the temperature goes as @xmath220 and others in which the characteristics of the approached configuration are the more relevant and `` universality '' is recovered when @xmath221 we have that the temperature goes as @xmath166 , indistinguishable from hawking s result . by looking at our simple critical model , we can say that , in geometrical ( kinematical ) terms , in order to obtain a steady and universal flux of particles from a collapsing ( spherically symmetric ) star there is no need for its surface to actually cross the schwarzschild radius ; it is sufficient that it tend towards it asymptotically ( in proper time ) , with sufficient rapidity . our critical configurations could prove to be relevant also in the overall picture of semiclassical collapse and evaporation of black hole - like objects . our results based on critical configurations suggest an alternative scenario to the standard paradigm . at this stage we are only able to present it in qualitative and somewhat speculative terms . being aware of the various assumptions that could ultimately prove to be untenable , we still think it is worth to present this possible alternative scenario . imagine a dynamically collapsing star . the collapse process starts to create particles dynamically before the surface of the star crosses its schwarzschild radius ( this particle creation is normally associated with a transient regime and has nothing to do with hawking s planckian radiation ) . the energy extracted from the star in this way will make ( due to energy conservation ) its total mass decrease , and so also its schwarzschild radius . by this argument alone , we can see that a process is established in which the surface of the star starts to closely chase its schwarzschild radius while both collapse towards zero ( this situation was already described by boulware in reference @xcite ) . now , the question is : will the surface of the shrinking star capture its shrinking schwarzschild radius in a finite amount of proper time ? let us rephrase this question in the language of this paper . in an evaporating situation our function @xmath43 still represents the distance between the surface of the star and its schwarzschild radius . the standard answer to the previous question is that @xmath43 becomes zero in a finite amount of proper time . to our knowledge , this view ( while certainly plausible ) is not guaranteed by explicit systematic and compelling calculations but still relies on somewhat qualitative arguments . the standard reasoning can be presented as follows : for sufficiently massive collapsing objects , the classical behaviour of the geometry should dominate any quantum back - reaction at any and all stages of the collapse process , as hawking s temperature ( considered as an estimate of the strength of this back - reaction ) is very low ; quantum effects would be expected to become important only at the last stages of the evaporation process . however , in opposition to this standard view , stephens , t hooft , and whiting @xcite have argued for the mutual incompatibility of the existence of external observers measuring a hawking flux and , at the same time , the existence of infalling observers describing magnitudes beyond the apparent horizon . the reason is that the operators describing any feature of the hawking flux do not commute ( and this non - commutation blows up at late times ) with the infalling components of the energy - momentum tensor operator at the horizon . therefore , if we accept this argument , the presence of a hawking flux at infinity would be incompatible with the actual formation of the trapping horizon , which would be destroyed by the back - reaction associated to hawking particles . this fact leads these authors ( seeking for self consistency ) , to look for the existence of a hawking flux ( or at least a flux looking very much like it ) in background geometries in which the collapse process of the star is halted , just before crossing the schwarzschild radius , producing a bounce . ( in our language this could be represented by a function @xmath43 monotonically decreasing from @xmath222 to some @xmath223 , at which it reaches a very small positive value , and then monotonically increasing from @xmath223 to @xmath224 . ) in their analysis they found exactly that : an approximately planckian spectrum of particles present at infinity during a sufficiently long time interval . however , the modified behaviour that deviates the least from the classical collapse picture , and at the same time eliminates the trapping horizon , is that in which @xmath43 does not reach zero , but just `` asymptotically approaches zero '' at infinite proper time , and does that very quickly . this is represented in our critical configurations by the exponential behaviour @xmath225 with a very large @xmath46 . the interesting point is that the analysis in this paper suggests that with quasi - stationary configurations like this , one could expect quasi - stationary planckian radiation at a temperature very close to @xmath166 , just like in the hawking process . standard gr suggests that the surface gravity @xmath47 ( inversely proportional to the total mass of the star ) would increase with time through the back - reaction caused by the quantum dissipation . moreover , it is sensible to think that during the evaporation process @xmath46 would also depend on @xmath13 . as the evaporation temperature increases ( @xmath47 increases ) the back - reaction would become more efficient and therefore we might expect that @xmath46 decreases . then , one could arrive at a situation as the one portrayed in figure [ f : critical - collapse ] . the evolution of the evaporation temperature would interpolate between a starting temperature completely controlled by @xmath47 and a late time temperature completely controlled by @xmath46 , showing a possible semiclassical mechanism for regularizing the end point of the evaporation process . in this scenario the complete semiclassical geometry will have neither an apparent horizon nor an event horizon . in this circumstance there would be no trans - planckian problem , nor information loss associated with the collapse and evaporation of this black hole - like object . whether this scenario is viable or not will be the subject of future work . let us end by making a brief comment concerning modified dispersion relations . everything we said in this paper assumes strict adherence to lorentz symmetry . even if semiclassical gravity contained lorentz - violating traces in the form of modified dispersion relations at high energy , one would still expect that the resulting scenario for the collapse and evaporation of a black hole - like object would keep the quasi - stationary hawking - like flux of particles as a robust prediction @xcite . however , the complete conceptual scenario could be very different . in the presence of dispersion at high energies , the notion of horizon itself shows up only as a low - energy concept . for example , with superluminal modifications of the dispersion relations , high energy signals will be able to escape from the trapped region . the non - analytic behaviour of some sets of modes at the horizon become regularized . therefore , the stephenst hooft whiting obstruction described above forbidding the formation of a ( now approximate ) trapping horizon need no longer apply . we expect that by analyzing different analogue models in which the lorentz violating terms appear at different energy scales one would be able to explore the transition between all these alternative paradigms . the authors would like to thank vctor aldaya and ted jacobson for stimulating discussions . . has been funded by the spanish mec under project fis2005 - 05736-c03 - 01 with a partial feder contribution . c.b . and s.l are also supported by a infn - mec collaboration . the research of m.v . was funded in part by the marsden fund administered by the royal society of new zealand . m.v . also wishes to thank both isas ( trieste ) and iaa ( granada ) for hospitality . novello , m. visser and g. volovik ( eds . ) , _ artificial black holes _ ( singapore , world scientific , 2002).c . barcel , s. liberati and m. visser , `` analogue gravity , '' living rev . relativity * 8 * , 12 ( 2005 ) [ arxiv : gr - qc/0505065 ] . url ( cited on 22 march 2006 ) : http://www.livingreviews.org/lrr-2005-12c . barcel , s. liberati and m. visser , `` analogue gravity from bose - einstein condensates , '' class . quantum grav . * 18 * , 11371156 ( 2001 ) [ arxiv : gr - qc/0011026].m . visser , c. barcel and s. liberati , `` analogue models of and for gravity , '' gen . relativ . gravit . * 34 * , 17191734 ( 2002 ) [ arxiv : gr - qc/0111111].c . barcel , s. liberati , s. sonego and m. visser , `` causal structure of analogue spacetimes , '' new j. phys . * 6 * , 186 ( 2004 ) [ arxiv : gr - qc/0408022].s . w. hawking , `` black hole explosions , '' nature * 248 * , 3031 ( 1974 ) ; + `` particle creation by black holes , '' commun . phys . * 43 * , 199220 ( 1975 ) ; erratum : _ ibid . _ * 46 * , 206 ( 1976 ) . n. d. birrell and p. c. w. davies , _ quantum fields in curved space _ ( cambridge , cambridge university press , 1982).m . visser , `` acoustic black holes : horizons , ergospheres , and hawking radiation , '' class . quantum grav . * 15 * , 17671791 ( 1998 ) [ arxiv : gr - qc/9712010].b . l. hu , `` hawking - unruh thermal radiance as relativistic exponential scaling of quantum noise , '' in _ thermal field theory and applications _ , edited by y. x. gui , f. c. khanna and z. b. su ( singapore , world scientific , 1996 ) , pp . 249260 [ arxiv : gr - qc/9606073].m . visser , `` essential and inessential features of hawking radiation , '' int . j. mod . d * 12 * , 649661 ( 2003 ) [ arxiv : hep - th/0106111].s . w. hawking and g. f. r. ellis , _ the large scale structure of space - time _ ( cambridge , cambridge university press , 1973).r . m. wald , _ general relativity _ ( chicago , university of chicago press , 1984).s . corley and t. jacobson , `` black hole lasers , '' phys . rev . d * 59 * , 124011 ( 1999 ) [ arxiv : hep - th/9806203].r . schtzhold and w. g. unruh , `` gravity wave analogues of black holes , '' phys . d * 66 * , 044019 ( 2002).s . liberati , t. rothman and s. sonego , `` nonthermal nature of incipient extremal black holes , '' phys . d * 62 * , 024005 ( 2000 ) [ arxiv : gr - qc/0002019 ] . s. liberati , t. rothman and s. sonego , `` extremal black holes and the limits of the third law , '' int . j. mod . d * 10 * , 3339 ( 2001 ) [ arxiv : gr - qc/0008018 ] . j. r. oppenheimer and h. snyder , `` on continued gravitational contraction , '' phys . rev . * 56 * , 455459 ( 1939).p . painlev , `` la mcanique classique et la theorie de la relativit , '' c. r. acad . ( paris ) * 173 * , 677680 ( 1921 ) . + a. gullstrand , `` allgemeine lsung des statischen einkrperproblems in der einsteinschen gravitationstheorie , '' ark . mat . * 16 * , 115 ( 1922).l . parker , `` the production of elementary particles by strong gravitational fields , '' in _ asymptotic structure of space - time _ , edited by f. p. esposito and l. witten ( new york , plenum , 1977 ) , pp . 107226.g . e. volovik , `` black - hole horizon and metric singularity at the brane separating two sliding superfluids , '' pisma zh . fiz . * 76 * , 296300 ( 2002 ) [ jetp lett . * 76 * , 240244 ( 2002 ) ] [ arxiv : gr - qc/0208020].d . g. boulware , `` hawking radiation and thin shells , '' phys . d * 13 * , 21692187 ( 1976).a . ashtekar and m. bojowald , `` black hole evaporation : a paradigm , '' class . quantum grav . * 22 * , 33493362 ( 2005 ) [ arxiv : gr - qc/0504029].c . r. stephens , g. t hooft and b. f. whiting , `` black hole evaporation without information loss , '' class . quantum grav . * 11 * , 621647 ( 1994 ) [ arxiv : gr - qc/9310006].w . g. unruh and r. schtzhold , `` on the universality of the hawking effect , '' arxiv : gr - qc/0408009 .

we discuss the excitation spectrum of a disordered , isotropic and translationally invariant spin state in the 2d heisenberg antiferromagnet . the starting point is the nearest - neighbor rvb state which plays the role of the vacuum of the theory , in a similar sense as the nel state is the vacuum for antiferromagnetic spin wave theory . we discuss the elementary excitations of this state and show that these are not fermionic spin-1/2 ` spinons ' but spin-1 excited dimers which must be modeled by bond bosons . we derive an effective hamiltonian describing the excited dimers which is formally analogous to spin wave theory . condensation of the bond - bosons at zero temperature into the state with momentum is shown to be equivalent to antiferromagnetic ordering . the latter is a key ingredient for a microscopic interpretation of zhang s so(5 ) theory of cuprate superconductivity . 2

cond-mat9810216

high - temperature superconductivity occurs in a state which is is frequently referred to as an ` rvb spin liquid ' . this state has no magnetic order , but strong short range antiferromagnetic correlations . undoubtedly the strong repulsion between electrons , which turns the system into a charge - transfer insulator at half - filling persists in the doped case , so that a description in terms of free - electron like slater - determinants is not really adequate . the problem then is , how to describe such a state theoretically . despite its frequently being referred to in the literature , the rvb spin liquid is a rather elusive concept . for example the precise nature of its ground state and low lying elementary excitations is not known to any degree of certainty . in the following we want to address this problem by studying a disordered state for the @xmath1 heisenberg antiferromagnet @xmath2 on an 2d square lattice . here @xmath3 denotes a spin 1/2 operator on site @xmath4 . one might expect that this is a kind of stepping stone also for the doped case , in that the elementary excitations of the undoped spin liquid persist to some degree also for finite doping . + perhaps the best - defined rvb spin liquid is the nearest neighbor rvb state@xcite - at least this wave function can be written down in compact form . we define the singlet generation operator on the bond @xmath5 @xmath6 where @xmath7@xmath8@xmath9 , are the constrained fermion operators , which do not allow the creation of a second electron on an already singly occupied site . introducing the operator @xmath10 where e.g. @xmath11 denotes the nearest neighbor of site @xmath4 in @xmath12-direction , the nearest neighbor rvb state on a 2d square lattice with @xmath13 sites can be written as @xmath14 where @xmath15 is a normalization factor . it corresponds to a superposition of all states which can be obtained by covering the plane compactly with nearest neighbor singlets , all with equal phase . covering the plane with singlets is equivalent to covering it with dimers , a well - known problem from statistical mechanics@xcite . we can therefore rewrite the state as @xmath16 where @xmath17 denotes a dimer covering of the lattice and @xmath18 the state obtained by putting singlets onto the dimers of @xmath17 . + in the following , we want to examine the problem of the possible elementary excitations of such a singlet background , and set up an effective hamiltonian governing their dynamics . the idea of ` expanding ' around a suitably chosen vacuum state is realized in simplest form in linear spin wave theory . the general line of thought here is quite analogous to linear spin wave theory , with the sole exception that the role of the vacuum ( which determines the symmetries of the ground state ) is played by the ` singlet soup ' ( [ rvb ] ) instead of the nel state . a similar approach has previously been applied to dimerized planar heisenberg - type models@xcite to spin ladders@xcite , to strongly coupled heisenberg planes@xcite , and to spin - peierls - like spin chains@xcite . an example where the fluctuations are fermionic rather than bosonic in nature is provided by the kondo lattice@xcite . the main difference as compared to the present work is that in all of these works a rather unique and simple dimer covering of the system is given by the topology or the form of the hamiltonian - the complications that arise from the use of a disordered ` singlet soup ' such as ( [ rvb ] ) then can be avoided . + while the technical complications arising from the use of a dimer basis are considerable , this approach also has some major advantages : in a site basis it is virtually impossible to even write down a disordered spin state , because one has to deal with the spin degeneracy on each single site . the calculation only becomes feasible if this site - degeneracy is lifted , for example by assuming strict nel order as in spin wave theory . on the other hand , the degeneracy is automatically taken care of in the dimer basis , because two interacting sites do have a unique ground state . a further considerable advantage of the dimer basis is , that it is easily enlarged by hole pairs on nearest neighbors , so as to describe a superconducting state . indeed , as will be seen below , the present description of the antiferromagnetic phase most naturally can be generalized to comprise also a superconducting phase , thereby providing a very simple microscopical picture for the so(5 ) rotations which smoothly connect antiferromagnetic and superconducting phase in zhang s theory@xcite of cuprate superconductors . the nearest neighbor - rvb state ( [ rvb1 ] ) has the symmetry properties expected for a homogeneous spin liquid : it is isotropic , translationally invariant , is an exact spin singlet and has no magnetic order . on the other hand , just as the nel state , it is not an eigenstate of @xmath19 . if we take one singlet configuration @xmath18 and act with the exchange term on a bond connecting two different singlets ( see figure [ fig1]a ) we can create a state which no longer can be represented as a superposition of only nearest neighbor singlets . such a state therefore represents a kind of fluctuation and as a first step we need to understand the character of these fluctuations . it might appear@xcite that the energetically most favorable fluctuation is the state shown in figure [ fig1]b : two nearest neighbor singlets are transformed into a configuration with only one nearest neighbor singlet and a second singlet connecting more distant sites . nominally the energy increases by only @xmath20 in this transition , because the only change is the loss of one nearest neighbor singlet . because singlet and triplet are degenerate for sites which are not connected by an exchange bond , we might as well consider the two spins connected by the ` long singlet ' as being unpaired ( see figure [ fig1]c ) . the transition from figure [ fig1 ] a@xmath21b could thus be viewed as pair creation of two unpaired spins . next , by acting with the exchange term on a bond which connects a dangling spin to another singlet ( see figure [ fig1]c ) , we can recouple the spins so as to form a new singlet and leave one of the formerly paired spins unpaired ( see figure [ fig1]d ) . this process corresponds to a propagation = 7.0 cm of the dangling spin . we would thus arrive at the conclusion that the fluctuations out of the nearest neighbor rvb state are unpaired spins , which carry a spin of @xmath22 and consequently must obey fermi statistics . clearly , these excitations should be identified with the ominous ` spinons ' . + further reasoning shows , however , that the line of thought leading to the introduction of the ` spinons ' misses a small but crucial detail . the first reason is that the state in figure [ fig1]b is not orthogonal to the vacuum , = 4.0 cm and thus can not represent a true fluctuation . more precisely , if we introduce ( see figures [ fig1 ] and [ fig2 ] ) ) @xmath23 ) that @xmath24@xmath8@xmath25 , in other words : after the ` transition ' figure [ fig1]a@xmath21b we remain in the original state , figure [ fig1]a , with a probability of @xmath26% . the problem of non - orthogonality is not restricted to the first step in figure [ fig1 ] : the states fig . [ fig1]c and [ fig1]d have an overlap of @xmath22 , and this generalizes to any two states which differ by one hopping process of a ` spinon ' . the non - orthogonality problem thus is omnipresent and severe . + let us therefore return to the first step , figure [ fig1 ] a@xmath21b , and consider how we can remedy the problem . the most natural way to proceed is to form the orthogonal complement @xmath27 so as to see ` what is really new ' in the state @xmath28 . a straightforward computation shows that after normalization to unity the orthogonal complement is @xmath29 here we have introduced the operators@xcite @xmath30 which create the three components of the _ triplet _ on the bond @xmath31 . we arrive at the conclusion that the true fluctuation out of the nearest neighbor singlet background is not the creation of two fermionic ` spinons ' , but rather the creation of two excited dimers , which carry a spin of @xmath32 and consequently should be modeled by bond - bosons@xcite . the further evolution of the created triplets then is quite obvious ( see figure [ fig3 ] ) ( but completely different from that of the ` spinons ' ) : by exchange along bonds connecting the triplets with neighboring singlets the triplets can de - excite while simultaneously the singlet turns into a triplet - this process , which is very much reminiscent of the propagation of a frenkel - type exciton corresponds to the propagation of the excited dimer . note that unlike the ` spinon ' states in figure [ fig1 ] , all different states in figure [ fig3 ] are mutually rigorously orthogonal . as a matter of fact there are problems with non - orthogonalities also for the triplet states - these are ` inherited ' from the original nearest neighbor rvb state . they will be discussed in detail below and be shown to be much less severe than those for the ` spinon ' states . their main effect is to replace the simple excited dimer by a more delocalized object , which resonates between different orientations within a limited spatial region . = 7.0 cm to be more precise , we now discuss the action of the heisenberg exchange on all possible configurations of nearest neighbor singlets and triplets . consider two _ nearest neighbor bonds _ @xmath31 and @xmath33 , and assume that they are connected by a single bond @xmath34 . denoting the heisenberg exchange along the latter bond by @xmath35 we have : @xmath36 these equations show that if we start out from states containing nearest neighbor singlets or triplets on the l.h.s . , the exchange term only produces states which again consist of nearest neighbor singlets or triplets on the r.h.s ( had the two bonds been connected by exchange along another bond , @xmath37 , @xmath38 or @xmath39 , we would have obtained the same equations with the sole difference that in some cases the prefactors change their sign ) . this proves rigorously that acting with an arbitrarily high power of the hamiltonian onto the nearest neighbor rvb state produces only states which can be built up from _ nearest neighbor _ singlets or triplets . + this ` theorem ' in fact holds true in a more general sense : on a single dimer @xmath40 , the @xmath41 states @xmath42 and @xmath43 do form a complete basis@xcite . thus , if we use a fixed dimer covering @xmath17 , the set of states obtained by placing singlets and triplets on the dimers of @xmath17 form a complete basis of the hilbert space . adding up such states obtained from _ all possible _ dimer coverings then clearly produces a highly overcomplete basis of the hilbert space , which therefore must automatically include states with singlets of arbitrary length . it follows that all states with longer - range singlets also can be represented as superpositions of states which are composed exclusively from nearest - neighbor singlets and triplets . these states are therefore redundant , and if we formulate a self - consistent theory in terms of nearest - neighbor singlets and triplets , we have automatically included these longer - ranged singlets . the fact that we are using a nearest - neighbor rvb state as the starting point for constructing our theory therefore means no loss of generality and in particular does by no means imply that we are considering only states with only very short - ranged antiferromagnetic correlations . in fact , it will be shown below that one can construct even states with infinite - range antiferromagnetic order by using exclusively nearest neighbor singlets and triplets . + the preceding considerations suggest that we should model the excitation spectrum of the nearest neighbor singlet vacuum by bosonic excitations , which approximately correspond to excited dimers . assuming that the bonds in the system have been labeled in some way , we denote the triplet operator on bond @xmath4 by @xmath44 . then , we introduce the following basis states @xmath45 where @xmath46 is a normalization factor . they describe a certain number ( @xmath40 ) of triplets which are ` immersed into the singlet soup ' . thereby the singlets fill the space in between the triplets compactly in all possible ways . all states which can be generated by pair creation and propagation of triplet bonds ( such as the ones shown in figure [ fig3 ] ) can be represented in this way . we next consider the triplets as boson - like elementary excitations of the singlet - background , in precisely the same way as misaligned spins are considered as bosonic excitations in a ` nel background ' in antiferromagnetic spin wave theory . re - interpreting the state @xmath47 where the @xmath48 represent boson operators , we may expect to describe the dynamics of these bosons by a hamiltonian of the form@xcite @xmath49 where we have grouped the three triplet components into a 3-vector @xmath50 so as to stress manifest rotation invariance . the first term in ( [ heff ] ) corresponds to the energy of formation of the triplets , the second term describes pair creation processes as in figure [ fig3]a @xmath21 [ fig3]b , and the third term accounts for the propagation of the triplets , see figure [ fig3]c @xmath21 [ fig3]d . the matrix elements @xmath51 and @xmath52 should be obtained by computing matrix elements of the heisenberg hamiltonian @xmath19 between the corresponding states ( [ basis ] ) . of course one thereby has to assume that for example the matrix element for a triplet jumping from bond @xmath40 to bond @xmath15 does not depend significantly on the positions of the other triplets - otherwise a description in terms of a single - particle like hamiltonian would not be feasible . as is the case in spin wave theory , the @xmath53-bosons have to obey a hard - core constraint , and in fact presence of one boson prohibits the presence of another boson not only on the same bond , but also on all bonds which share a site with the original one . + in the following , we will first study the problem of a single excited dimer in the singlet background , in other words we want to compute the ` bare ' boson dispersion @xmath54 in ( [ heff ] ) . as our basis states we consequently choose ( suppressing the @xmath12 @xmath55 or @xmath56 spin - index of the triplet ) : @xmath57 where @xmath58@xmath8@xmath59 denotes the direction of the bond in real space , and @xmath60 a normalization factor . in this state one triplet is put onto the bond @xmath61 and the remainder of the lattice is covered compactly by singlets in all possible ways . next , we introduce the fourier transforms @xmath62 in the hilbert space of bond - bosons , this state would be denoted by @xmath63 . the procedure to be followed then is like this : in a first step we compute the @xmath64 overlap matrix @xmath65@xmath8@xmath66 and diagonalize it . denoting the resulting eigenvectors and eigenvalues by @xmath67 and @xmath68 , the states @xmath69 form an orthonormal basis set and hence can serve as effective single particle orbitals with momentum @xmath70 . since the boson operators which correspond to the original triplets obey@xcite @xmath71 = n_{\alpha,\beta}$ ] , the operators @xmath72 obey the canonical commutation relations for boson operators : @xmath73=\delta_{\nu,\nu'}$ ] . they describe a triplet - like excitation which oscillates between @xmath12 and @xmath55-directed bonds within a certain spatial region whose extent is determined by the range of the real space overlap integrals @xmath74 . + next , we set up the @xmath64 hamilton matrix @xmath75@xmath8@xmath76 . diagonalizing @xmath77 , where @xmath78 denotes the expectation value of @xmath19 in the ` background ' state ( [ rvb1 ] ) , we obtain the desired dispersion relation of a single triplet boson . the pair creation matrix element is obtained in an analogous way . + this procedure in fact is neither new nor unconventional : a completely analogous construction is performed e.g. in the construction of the @xmath79 model@xcite , which describes the dynamics of the ( non - orthogonal ) zhang - rice singlets on the different plaquettes of the cuo@xmath80 plane . the only difference is that here we have two different nonorthogonal objects ( bosons on bonds in @xmath12 and @xmath55-direction ) per unit cell , whereas it was only a single zhang - rice singlet / unit cell in the case of the cuo@xmath80 plane . apart from that the construction is precisely the same . + in the next three sections we will calculate the dispersion relation , the pair creation matrix element , and discuss how these matrix elements depend on the density of triplets . readers which are not interested in these more technical parts are advised to proceed to section vi . to carry out our program we need to compute the real - space matrix elements @xmath81 and @xmath82 . in doing so the concept of a loop covering of the plane@xcite is of crucial importance . for two dimer coverings @xmath17 and @xmath83 the loop covering @xmath84 is obtained by drawing @xmath17 and @xmath83 ` on top of each other ' ( see for example figure 1 in ref . this produces a covering of the plane by closed loops @xmath85 , each of even length @xmath86 ( note that in the following we always measure the length of a loop @xmath87 ` in units of dimers ' ) . let us now consider each loop @xmath85 as an isolated 1d ring with @xmath86 sites . we assume that the sites along the ring are labeled such that the dimer covering @xmath17 corresponds to the state @xmath88 - the dimer covering @xmath83 then must correspond to @xmath89 . expanding the products we get @xmath90 different spin states from each covering , and there are precisely @xmath91 spin states which show up in both @xmath92 and @xmath28 namely the two possible nel states . we thus have @xmath93 . the same holds true for any other loop , whence , using @xmath94 , we find the scalar product of the two singlet distributions@xcite @xmath95 where @xmath96 is the total number of loops in the loop covering @xmath97 . + let us now assume that the singlets on the bond @xmath40 in @xmath18 and on the bond @xmath15 in @xmath98 have been replaced by a @xmath56-like triplet ( due to the explicit rotational invariance of the ` singlet soup ' the result for an @xmath12-like or @xmath55-like boson will be precisely the same - we are choosing the @xmath56-like component because the ambiguous states in this case are again the ones with nel order along the loop ) . then , a necessary condition for the scalar product to be different from zero is that there is a single loop @xmath99 in the resulting loop covering @xmath100 which passes through both bonds @xmath40 and @xmath15 . the reason is that the time - reversal parity of the triplet is negative whereas that of the singlet is positive . a necessary condition for a loop @xmath85 to give a nonvanishing overlap is that the total time - reversal parities ` along the loop ' are equal for both states @xmath18 and @xmath101 . this , however , is only possible if the triplets in @xmath18 and @xmath101 are within the same loop . each loop in @xmath100 therefore must contain either no triplet or both of them . + we can now split up the entire overlap integral @xmath102 into components which differ by the length and topology of the loop @xmath99 which passes through both triplets . the absolute numerical value of the overlap from this particular loop is identical to the case of pure singlet covering . the only change may be an extra minus sign , which originates because the singlets do have an orientation , whereas the triplets do not . we thus can rewrite the overlap as @xmath103 here @xmath104 is @xmath32 if the loop covering @xmath97 contains the loop @xmath85 and zero otherwise . we also note that @xmath105@xmath8@xmath32 , which fixes the normalization factor @xmath60 . with the exception of the @xmath106 all parts in ( [ over - real ] ) can be computed analytically . @xmath106 may be viewed as the norm of a nearest - neighbor rvb state which covers only the exterior of the loop @xmath85 , divided by the norm @xmath60 of the state which covers the exterior of a single bond . if we assume that the norm increases exponentially with the number of sites in the system , @xmath107 , with @xmath108 , one would estimate that @xmath109 . this suggests that @xmath106 is a quite rapidly decreasing function of @xmath87 . in the present work numerical values for @xmath106 were obtained by exact calculations on finite clusters ( see the appendix ) . the @xmath106 thereby indeed turned out to decay rapidly with @xmath87 , so only contributions with @xmath87@xmath110@xmath91 were kept in the present calculation . it should be noted that the computation of the @xmath106 is no fundamental obstacle to the present scheme : it may well be possible to obtain essentially exact values for these parameters by using monte - carlo techniques on large lattices . figure [ fig4 ] then shows the = 8.0 cm pairs of bonds @xmath111 and @xmath112 which can be connected by loops of length @xmath91 and @xmath113 as well as the corresponding signs @xmath114 . in this way we find the overlap matrix @xmath115 with @xmath116@xmath8@xmath32 and @xmath117 we proceed to a calculation of the matrix elements of the hamiltonian . we first recall that for the nearest neighbor rvb state the expectation value of @xmath19 between two dimer coverings @xmath18 and @xmath101 can be decomposed into contributions from each individual loop in the loop covering @xmath100@xcite : @xmath118 \;2^{p(a+b)-n } , \nonumber \\ e(u ) & = & \epsilon_s \ ; [ ( 2 -\delta_{l(u),1 } \ ; ) \;l(u ) + n_b(u)\;],\end{aligned}\ ] ] where @xmath119 is the number of nearest - neighbor bonds in @xmath85 which bridge the loop ( see figure [ fig5 ] ) , and @xmath120@xmath8@xmath121 the exchange energy / singlet . @xmath122 is the number of nearest - neighbor bonds that can be formed from the sites covered by @xmath85@xcite . this formula implies that there is no = 5.0 cm contribution from bonds connecting _ different loops_. the reason is that the exchange along a bond connecting different singlets can only lead to the pair creation of two triplets , see the first equation ( [ st ] ) . in order to maintain time reversal symmetry along each loop , it is then necessary that both triplets belong to the same loop - which is not possible if the bond in question connects different loops . + let us now again assume that the bond @xmath40 in @xmath18 and the bond @xmath15 in @xmath101 are occupied by a triplet , and consider the ` connected matrix element ' of @xmath19 between the two resulting states : @xmath123 . first , let us assume that we act with the exchange along a bond connecting @xmath40 and a neighboring bond @xmath124 ; the triplet can either propagate from @xmath40 to @xmath124 , or decay into two triplets on both @xmath40 and @xmath124 ( see the second equation ( [ st ] ) ) . neglecting the second possibility we obtain a nonvanishing contribution to the matrix element of @xmath19 only if there is a single loop @xmath125 which covers both , @xmath124 and @xmath15 . alternatively , if we act on a bond which connects @xmath15 and a neighboring bond @xmath126 , the triplet jumps from @xmath15 to @xmath126 and we get a nonvanishing contribution only if one single loop @xmath125 covers @xmath126 and @xmath40 . if , on the other hand , we act with the exchange along a bond which does not touch either of the triplet bonds @xmath40 or @xmath15 , both triplets will stay where they are and we get a nonvanishing matrix element only if both , @xmath40 and @xmath15 , are covered by a single loop @xmath125 . the same holds true if we act with the exchange along the bonds @xmath40 and @xmath15 themselves . + we first consider the case that @xmath40 and @xmath15 belong to two different loops , @xmath99 and @xmath127 . in the simplest case both ` loops ' consist only of a single bond , i.e. @xmath99 consists of the single bond @xmath40 and @xmath127 only of @xmath15 . since the two triplets belong to different loops , the overlap @xmath128 is zero . moreover , the matrix elements of the exchange along any bond which does not connect @xmath40 and @xmath15 vanishes - the calculation thus becomes very easy . the matrix element for the triplet hopping from @xmath40 to @xmath15 is @xmath129 , where the signs for different relative positions of the two bonds are shown in figure [ fig6 ] . to ` embed ' the hopping process = 3.0 cm into the singlet background , we need to renormalize this matrix element by @xmath130 which is again estimated from cluster calculations . the first contribution to the hamilton matrix then is : @xmath131 where the elements of the @xmath64 matrix @xmath132 are : @xmath133 to keep things simple we have moreover replaced the different @xmath134 by the average value @xmath135 . + next , we consider the case that one of the loops , e.g. @xmath127 , has a length @xmath136 . in other words , we a considering a process like the one shown in figure [ fig7 ] : the triplet jumps from bond @xmath40 to bond @xmath124 , and the triplet on @xmath124 = 4.0 cm overlaps with the triplet on bond @xmath15 along a loop ( in this case of length @xmath91 ) . there is also an analogous process , where the triplet jumps from @xmath15 to @xmath124 and @xmath124 and @xmath40 overlap by a loop . the respective matrix elements can be factorized into the matrix element for the hopping of the triplet times the overlap along the loop . this means that the matrix element can be written as @xmath137 where @xmath138;\ ] ] here @xmath139 and @xmath140 are the real - space versions of the matrices ( [ kin1 ] ) and ( [ gamma ] ) . fourier transformation gives @xmath141 where it has to be kept in mind that @xmath132 and @xmath142 are _ symmetric _ @xmath64 matrices . the ` embedding factor ' is @xmath143 where @xmath144 are like in figure [ fig7 ] . actually there are two inequivalent relative orientations of a single bond @xmath40 and an @xmath145 loop - the two respective values of @xmath146 do not differ strongly and for simplicity we use the average of the two values for the two configurations . processes involving even longer loops could be treated in an analogous way , but we neglect these . + we proceed to the second case , i.e. we assume that @xmath40 and @xmath15 are covered by a single loop @xmath99 . as was the case for the overlap integrals , the matrix elements then can be split up into contributions differing by the loop @xmath99 which covers both triplet bonds . once this loop is fixed , we can divide all bonds in the plane into three distinct classes . first , bonds belonging to any loop other than @xmath99 are not affected at all by the presence of the triplets and give the same contribution as in the pure singlet state . this becomes @xmath147 2^{p(a+b)-1 } \delta_{a+b , u_0}. \label{bulk1}\end{aligned}\ ] ] this is an energy of order of the system size @xmath148 ; in the end it must be canceled , up to terms of order @xmath149 , by a corresponding contribution in @xmath150 . this cancellation is the analogue of the familiar linked - cluster theorem of many - body physics . + second , we consider those bonds in @xmath99 which are not bridging bonds . they will be covered by either a singlet or a triplet in either @xmath17 or @xmath83 , whence these bonds together give the contribution @xmath151 \nonumber \\ & \;&\;\;\;\;\;\;\;\ ; 2^{1-l(u_0 ) } ( -1)^{\sigma(u_0 ) } \chi(u_0).\end{aligned}\ ] ] this is a ` connected ' contribution of order @xmath149 . + this leaves us with the bridging bonds ( see figure [ fig5 ] ) , which may give a nontrivial contribution . however , since the bridging bonds occur only for @xmath152 we neglect their contribution altogether . + the subtracted contribution , @xmath150 , may be rewritten as @xmath153 where we have used the expanded form ( [ over - real ] ) of @xmath154 . this is again an energy of order @xmath148 , which cancels the bulk term ( [ bulk1 ] ) up to terms of order @xmath149 . after some reshuffling ( using @xmath155@xmath8@xmath148 ) we can rewrite the contribution to the matrix element as @xmath156 with @xmath157\ ; \chi(u ) , % \nonumber \\ \epsilon_j(u ) & = & j\;(1 -\delta_{l(u_0),1 } \;)\ ; 2^{1-l(u ) } ( -1)^{\sigma(u ) } \chi(u ) \nonumber \\ \epsilon_{0}(u ) & = & \frac { ( -1)^{\sigma(u)}}{n_1 } \sum_{a , b } [ \;\sum_{u\in a+b}\bar{e}(u)\ ; ] 2^{p(a+b)-n } \delta_{a+b , u } , \nonumber \\ \bar{e}(u ) & = & e(u ) - l(u ) \frac{e_0}{n}. \label{mats}\end{aligned}\ ] ] the first term on the rhs of ( [ hmat ] ) is the ` bare ' on - site energy of the triplet . it is always proportional to the overlap integral , so upon switching to the effective bosons this term becomes a @xmath70-independent constant shift . the quantity @xmath158 may be thought of as describing a ` loss of resonance energy ' . it is the difference of the two contributions ( [ bulk1 ] ) and ( [ bulk2 ] ) and occurs because the loop @xmath99 is fixed , whence the area covered by this loop is not available for resonating between different singlet coverings . for example we have @xmath159 i.e. this quantity is an additive renormalization of the energy of formation for a single triplet due to its being ` embedded into the singlet soup ' . numerical evaluation in a cluster shows that this additive correction is quite large - for one triplet in a pure singlet background we find @xmath160 . while this is surprising at first sight it should be noted that a similar large value ( @xmath161 ) was previously found in spin ladders@xcite . a fixed triplet obviously leads to a quite substantial loss of resonance energy . the numerical values of @xmath158 and @xmath78 were again obtained by cluster calculations ( appendix ) . introducing @xmath162@xmath8@xmath163 we can write down the third part of the hamilton matrix : @xmath164 we can now add up the three contributions , ( [ eps1 ] ) , ( [ eps2 ] ) , and ( [ eps3 ] ) to obtain the total ` connected ' hamilton matrix @xmath165 . this is still expressed in terms of the non - orthogonal orbitals @xmath166 . what remains to be done therefore is to transform the hamilton matrix to the orthogonal orbitals ( [ ortho ] ) . to that end we take matrix elements of the form @xmath167 . introducing the @xmath64 transformation matrix @xmath168 the transformed hamiltonian then can be expressed as @xmath169 in this section we proceed to a discussion of the pair creation amplitude @xmath170 in equation ( [ heff ] ) . as discussed in section ii , by starting from an arbitrary singlet covering of the plane and acting with the hamiltonian along a bond connecting two different singlets , we _ only _ create a state where both singlets are replaced by triplets , see the first of eqs . ( [ st ] ) . the situation where the two singlets in question are ` parallel ' to each other , i.e. that the @xmath41 sites belonging to the @xmath91 singlets form a square with edge @xmath32 , requires special attention . as discussed above we have the identity @xmath171 i.e. the state with two parallel triplets can be expressed as a linear combination of the two perpendicular combinations of parallel singlet states . in other words : this state is already exhausted by the singlet background , and consequently must be omitted from our reduced hilbert space . the triplets thus have to obey the additional constraint of never being parallel to each other - one implication is that we must set the respective pair creation matrix element to zero . + to compute the numerical value of the matrix element , let us consider the action of @xmath19 on the rvb state ( [ rvb ] ) . we have @xmath172 the first term on the rhs originates from processes where the hamiltonian has ` hit ' a bond covered by a singlet , the second one originates from processes where the exchange has acted along bonds connecting two singlets on bonds @xmath40 and @xmath15 ( the prime on the sum indicates that only pairs of bonds connected by a bond are summed over ) . the modulus of the respective matrix element is @xmath173 ( see equation ( [ st ] ) ) and there is an extra sign which depends on the relative orientation of the bonds @xmath40 and @xmath15 . the dependence of this sign on the orientation is shown in figure [ fig8 ] . also , we have approximated the normalization factor which is included in the definition of @xmath174 by @xmath175 the factor of @xmath176 remains from the definition of ( [ rvb ] ) . to obtain the pair creation matrix element we now form the overlap between ( [ hrvb ] ) and the state @xmath177 . = 8.0 cm defining the @xmath64 matrix @xmath178 and bearing in mind that @xmath179 , we find for the pair creation matrix : @xmath180 this completes the derivation of the single - particle terms of the triplet hamiltonian . before we proceed , let us briefly return to the problem of non - orthogonality . strictly speaking , the state with two triplets is not orthogonal to the singlet background either . the reason is that if one draws a loop passing through both triplets , the time reversal parities of the two triplets cancel , and the state has a nonvanishing overlap with a state where the loop is covered only by singlets ( see figure [ fig9 ] ) . = 4.0 cm however , this overlap is rather small : in the case shown in figure ( [ fig9 ] ) it is for example @xmath181 , and obviously this is the most unfavorable case . for other relative orientations of the two triplets the overlap can be only achieved by a loop of length @xmath182 , whence these overlaps are @xmath183 . the non - orthogonalities thus are much more benign that those for the spinon - states , whence neglecting them altogether ( as we will do henceforth ) is probably quite justified . in the preceding sections we have computed the various overlap integrals and matrix elements for triplet propagation , pair creation and interaction . in all of these cases the matrix element could be factorized into a contribution from a ` local ' transition between different singlet / triplet coverings along a single or two neighboring loops , and a factor which describes the embedding of these active loops into the singlet background . thereby we have always given expressions for these ` embedding factors ' which are valid in the limit of vanishing triplet concentration , i.e. we have computed them as they would be for a pure singlet covering of the system . clearly , this is inappropriate for the real system , where quantum fluctuations have admixed a finite density of triplet bosons . in the following we want to discuss how we have to modify our theory to take the effect of a finite triplet concentration into account . it should be stressed from the very beginning that this is quite obviously a very complex problem and we will be forced to apply some relatively crude approximations . + if we want to derive single - particle like matrix elements for finite triplet concentration we should consider overlap integrals or matrix elements of @xmath19 of the type @xmath184 triplets stay unchanged , and only a single one ( which without loss of generality can be taken to be the first one ) changes its position ( but maintains its spin ) . in other words , we should calculate the embedding factors for singlet coverings containing a certain number of ` inert ' triplets . thereby we actually have to make the major _ assumption _ that the matrix element does not depend significantly on the positions of these inert triplets - otherwise , the very idea of a single particle - like propagation of the triplets would be invalid . one might then expect that approximate values can be obtained by distributing the @xmath40 inert triplets in all possible ways ( we call the number of possible distributions d(m ) ) over the system , and taking the average of the respective matrix elements computed for all d(m ) possible distributions . in this way , the embedding factors acquire a dependence on the density of triplets . + the numerical calculation in a finite cluster then proceeds as follows : we choose @xmath40 bonds @xmath185 , which obey the various constraints on the relative positions of triplets , and evaluate the ground state norm and energy according to @xmath186\;2^{p(a+b)-n}\prod_{\nu=1}^{m } \delta_{a+b , i_\nu}. \label{hugo}\end{aligned}\ ] ] the calculation of the various embedding factors then proceeds in an entirely analogous fashion i.e. in ( [ over - real ] ) , ( [ etadef ] ) , and ( [ mats ] ) we replace @xmath187 in this way we obtain all embedding factors for fixed distribution of inert triplets , and the value for triplet concentration @xmath188 is obtained by averaging over all allowed distributions of the bonds @xmath185 . in practice this calculation requires quite a substantial numerical effort so this was performed only for the @xmath189 cluster . = 4.0 cm one might expect , however , that the finite size effects are in fact smaller for the more relevant higher triplet concentrations : the main effect of the fixed triplets obviously is to reduce the importance of long loops . since the bonds occupied by the static triplets must be identical in the bra and ket state , loops of length @xmath190 which pass through these bonds are impossible . when the density of triplets gets appreciable , the probability to find enough space for forming longer loops becomes smaller and smaller . since longer loops may cause problems in the small clusters , a suppression of these may therefore even reduce finite - size effects . moreover , the suppression of longer loops is actually beneficial for our entire approach : it tends to eliminate the problems with the nonorthogonalities in the singlet soup and makes the truncation of the length @xmath191 a better approximation . + as an illustration figure [ fig10 ] shows the dependence of @xmath192 , i.e. the additive renormalization of the energy of formation of a triplet ( see ( [ renorm ] ) , evaluated in the @xmath189 cluster . the concentration dependence is as expected : for low triplet concentration @xmath192 is large because an added triplets blocks many long loops along which resonance could have occurred . as the triplet concentration gets higher , these longer loops a blocked anyway , so adding a further triplet does not increase the energy too much any more . in the high - density limit the additive renormalization approaches zero , as expected . the figure shows , however , that the concentration dependence is quite significant , i.e. this effect should not be neglected . + in addition to reducing the importance of longer loops , for a finite density of triplets we have to take care of the excluded volume constraint which the bond bosons have to obey . placing a triplet on one given bond @xmath40 blocks a total of @xmath193 other bonds , on which no more triplets can be placed ( see figure [ fig11 ] ) . = 4.0 cm of these , @xmath194 bonds are blocked because they share a site with @xmath40 , the remaining @xmath91 are blocked because they are ` parallel ' to @xmath40 and the state with two parallel triplets is actually a linear combination of singlets ( see ( [ combo ] ) ) . in order to take care of this _ blocking effect _ , we resort to a gutzwiller - type approximation . it has been shown by ogawa _ et al._@xcite that the essence of the gutzwiller approximation is the neglect of the difference in phase between states where the particles in question are distributed in different ways over the lattice . with this approximation , any real space distribution of @xmath40 bosons contributes the same number ( which without loss of generality can be taken to be @xmath32 ) to the norm of any state with @xmath40 bosons . the total norm of any such state then becomes simply the number of possible ways to distribute @xmath40 bosons over the plane , i.e. @xmath195 . in the present case , the situation is somewhat more complicated , because we still have to take into account the fact that due to the loop problem the norm of a state with @xmath40 fixed triplets is not @xmath32 ( unlike the case if the triplets were just ordinary bosons ) . we thus approximate the normalization factor of any state with @xmath40 triplets by @xmath196 where @xmath197 is the value obtained by averaging ( [ hugo ] ) over all allowed triplet configurations @xmath185 . clearly , this replacement will lead to a renormalization of the various embedding factors . to evaluate these , we usually have to consider a certain area @xmath198 , covered by one or two loops , in which the actual overlap , hopping or pair creation process takes place . we then determine ( again by simulation on the @xmath189 cluster ) the number of ways to distribute @xmath40 triplets over the exterior of the area @xmath198 . thereby we request that putting a triplet anywhere within @xmath198 always gives an allowed configuration of @xmath199 triplets . we call the number of allowed triplet configurations @xmath200 . defining the various areas @xmath201 as = 4.0 cm in figure [ fig12 ] , we then have to renormalize the various parameters as follows : @xmath202 for the pair creation amplitude we replace @xmath203 to summarize this section , figure [ fig10 ] shows the values of the different parameters as estimated by the procedure outlined above , for all possible triplet concentrations in the @xmath189 cluster . let us stress again that the approximations leading to the parameters in figure [ fig10 ] are probably rather crude - one may expect , however that we get roughly correct orders of magnitude and that the ratios of the different parameters come out approximately correct . let us first briefly summarize the discussion so far : we have shown that the elementary excitations of the singlet soup correspond to excited dimers , which must be modeled by bond - bosons . combining the matrix elements computed in the two preceding sections we obtain a hamiltonian of the form @xmath204 thereby the matrix elements @xmath205 and @xmath54 are functions of the triplet density . for given triplet density @xmath206 these parameters can be computed and the hamiltonian , which after the gutzwiller - type renormalization of the matrix elements we take to be a free - boson hamiltonian , is solved by bogoliubov transformation . combining the two different @xmath53-operators into a 2-vector @xmath207 the ansatz reads @xmath208 where the real @xmath209 matrices @xmath85 and @xmath210 have to fulfill @xmath211 the inverse transformation of ( [ bogo ] ) is therefore @xmath212 the hamiltonian can be transformed to free particle form : @xmath213,\end{aligned}\ ] ] provided the transformation matrices obey @xmath214 here we have introduced the @xmath64 matrix @xmath215 . the triplet density is @xmath216 , \label{den}\ ] ] where @xmath217 is the bose function . the requirement @xmath218 then provides a self - consistency condition for the density . + numerical evaluation shows that there is a minimum triplet concentration @xmath219 . for @xmath220 , the parameters are such that in a certain area around @xmath221 the eigenvalue problem ( [ eigenvalue ] ) does not have real eigenvalues . for @xmath219 the minimum of the triplet dispersion , which occurs at @xmath0 is precisely zero . with the coarse mesh of concentrations possible in the @xmath222-site cluster , we find @xmath223 . we therefore linearize all matrix elements @xmath12 , i.e. @xmath224 , using the values at @xmath225 and @xmath226 to determine @xmath227 and @xmath228 . this gives @xmath229 . + with increasing temperature , the self - consistent value of @xmath230 increases slowly , approaching @xmath219 for @xmath231 . figure [ fig13 ] shows the dispersion relation for the respective self - consistently determined triplet density at different temperatures . since we have two degrees of freedom / site ( the triplet in @xmath12 and @xmath55 direction ) we obtain two bands . while one of these bands is practically dispersionless , the dispersive one resembles results obtained for other spin = 5.0 cm liquids , such as the two - leg ladder@xcite or the bi - layer heisenberg antiferromagnet@xcite : the dispersion starts at relatively high energy at @xmath232 , takes a shallow maximum near the antiferromagnetic zone boundary and , as stated above , takes a more or less pronounced minimum at @xmath233 . for low temperatures the bandwidth is @xmath234 , which is quite close to the value for antiferromagnetic spin waves . the case of zero temperature requires special attention : it is obvious from figure [ fig13 ] that the gap at @xmath235 approaches zero for @xmath236 . assuming that the gap vanishes at zero temperature , we may then assume that this momentum becomes macroscopically occupied by a triplet density @xmath237 . to determine this condensed fraction we note that the limiting density @xmath238 is defined such that the gap at @xmath235 is exactly zero with the parameters calculated for this density . then , for @xmath239 we evaluate the density @xmath240 of uncondensed triplets ( i.e. we exclude @xmath241 from the sum in ( [ den ] ) ) and finally determine the condensate density from @xmath242 this is entirely analogous to the treatment of hirsch and tang@xcite of the condensation of schwinger - bosons in the mean - field theory of arovas and auerbach@xcite . as will be seen in a moment , just as for the schwinger bosons the condensation of the triplets corresponds to antiferromagnetic ordering . the ground state energy then is @xmath243\ ] ] where @xmath244 is the energy of the singlet background for triplet density @xmath230 . this is calculated from ( [ hugo ] ) . we obtain a value of @xmath245/bond , which is quite close to the true ground state energy of the 2d heisenberg antiferromagnet@xcite . an interesting figure is the lowering of the energy as compared to the original rvb - vacuum ( [ rvb ] ) . in the @xmath189 cluster ( which has been used to compute all matrix elements ) the expectation value of the pure nearest - neighbor rvb state is @xmath246/bond@xcite , so that the admixture of the triplets lowers the energy only by a tiny @xmath247/bond . in view of the strong approximations we were forced to make this result probably has little quantitative significance - it shows quite clearly , however , that the energy of the rvb - vacuum is lowered only by a very small energy by the triplet fluctuations . this is what must come out because in the thermodynamic limit the energy of the rvb state is @xmath248/bond@xcite , compared to the exact ground state energy of approximately @xmath249/bond@xcite . + we proceed to a discussion of the spin correlation function . since the bond - bosons are actually objects which extend over more than one unit - cell , they have something like a structure factor which depends on the type of operator by which they are probed . let us consider a single bond , @xmath31 , and introduce even and odd combinations of spin operators on this bond : @xmath250 due to their opposite parity under exchange of @xmath4 and @xmath251 these two operators have an entirely different effect on the @xmath41 possible spin states of the bond : @xmath252 annihilates the singlet and @xmath253 but converts @xmath254 and @xmath255 , whence @xmath252@xmath8@xmath256 . @xmath257 on the other hand converts the singlet into @xmath253 and vice versa , but annihilates both @xmath258 and @xmath252 , whence @xmath252@xmath8@xmath259 . if we still restrict ourselves to only a single bond @xmath15 we have to ` translate ' @xmath260 , \label{corr1}\end{aligned}\ ] ] where @xmath261 is the unit vector along the bond @xmath15 and @xmath262 denotes its center of gravity . let us next discuss how these matrix elements are modified due to the embedding into the singlet background . we act with the operator @xmath263 on some given state . the first term simply changes the @xmath56-spin and momentum of a triplet . it does not affect the number of triplets , so we need no further renormalizations . for the second term , let us start out from a state with one triplet , @xmath264 , and assume that the triplet is annihilated , i.e. converted into a singlet . this leaves us with the state @xmath265 the matrix element with the pure rvb state then is @xmath266 because it is easy to see that a state with one fixed singlet has an overlap of @xmath267 with the rvb state . if we want to extrapolate this to finite triplet density we again have to use the values of @xmath15 and @xmath60 computed with @xmath40 fixed triplets , and renormalize the matrix element by @xmath268 we thus find @xmath269.\end{aligned}\ ] ] by using the inverse transformation ( [ bogo_inv ] ) this can now be converted to the eigenvectors @xmath270 . the dynamical spin correlation function of the spin liquid thus consists of two components : first , there is a two - particle continuum , which dominates for momenta around @xmath271 . second , there is a single - particle like contribution , which dominates the cross section for momentum transfers near @xmath0 , and hence should be identified with the excitations seen in neutron scattering around this momentum . the situation is the same for ladders , where the single - particle spectrum is observable in the channel with momentum transfer perpendicular to the ladder , @xmath272 , and the two - particle continuum for @xmath273@xcite . + next , we discuss the relationship between condensation of triplets and antiferromagnetic ordering . the operator of staggered magnetization ( which is a _ vector _ ) can be written as @xmath274 here we have used the fact that for the high - symmetry momentum @xmath235 the overlap and hamilton matrix are trivially diagonalized by the symmetric and antisymmetric combinations of the bonds in @xmath12 and @xmath55-direction , @xmath275 . then , introducing the 3d unit vector @xmath276 , we can construct the coherent state @xmath277 this state corresponds to a condensate of the @xmath278 bosons , whose density is given by @xmath279 . the staggered magnetization per site is @xmath280 inserting the value for the condensate density @xmath237 obtained from ( [ conden ] ) we obtain the value @xmath281 . most current estimates for the 2d heisenberg antiferromagnet are around @xmath282@xcite . + as mentioned above , the identification of antiferromagnetic ordering and condensation of some kind of effective bosons is quite reminiscent of the treatment of hirsch and tang@xcite in the framework of schwinger - boson mean - field theory@xcite . in the present theory the direction of the staggered magnetization is given by the unit vector @xmath276 , which can be chosen arbitrarily . condensation of the triplet bosons determines the total density of triplets , but not their distribution over the three spin species . the ` ground state ' thus is actually an entire manifold of states which can be transformed into one another via so(3 ) rotations of the vector @xmath276 . one might thus conjecture the existence of low - energy states where the direction of @xmath276 changes slowly in real space . these states , which corresponds to a slow fluctuation of the antiferromagnetic order parameter may be describable in terms of a nonlinear @xmath283-model - we defer a detailed discussion of this issue to a separate paper . + to conclude this section , we discuss the relationship with zhang s so(5 ) symmetric theory of cuprate superconductivity@xcite . the preceding discussion has shown that an antiferromagnetic state can be viewed as a spin liquid where the triplet - like bond bosons have condensed into the state which has momentum @xmath0 and @xmath284-like symmetry under point group operations . the antiferromagnetic order parameter , a real 3-vector , then is the vector of condensation amplitudes of the three bond - boson species . this interpretation of the antiferromagnetic state in fact is a key ingredient for a microscopic interpretation of zhang s so(5 ) theory of cuprate superconductor . namely the @xmath285-operator , which acts as the ` ladder operator in charge direction ' in zhang s representation of the so(5 ) angular momentum algebra@xcite , precisely converts an @xmath284-like combination of bond triplets with momentum @xmath0 into a nearest neighbor @xmath286-wave hole pair . this can be seen by writing the @xmath285-operator in real space : @xmath287.\end{aligned}\ ] ] introducing a further bond boson @xmath288 , which stands for the hole pair , this could be written as @xmath289 where the momentum @xmath290 . acting with @xmath291 onto an undoped state with the triplet condensate densities @xmath292 will therefore convert this state into a condensate of @xmath286-like nearest neighbor hole - pairs with momentum @xmath271 . viewed in this way , the connection between the antiferromagnetic and the superconducting state appears quite natural . the only drawback would be that the hole pairs are strictly nearest - neighbor pairs , i.e. the present version of the @xmath285-operator neglects charge fluctuations . in summary , we have discussed the excitation spectrum of a completely disordered and homogeneous spin system . thereby we used an approach which might be viewed as a generalization of spin wave theory : whereas spin wave theory describes fluctuations around a nel ordered ` vacuum ' , we have instead used the nearest neighbor rvb state as basis for self - consistently constructing the excitation spectrum . this general idea of treating a completely disordered state is probably more widely applicable , for example to treat the ` orbital liquids ' proposed recently@xcite for manganates . + as a first key result , we then found that the elementary excitations of the singlet soup - like vacuum are not fermionic ` spinons ' , but rather bosonic excitations @xmath293 , which can be viewed as excited dimers propagating through the system while constantly resonating between @xmath12 and @xmath55 direction of the dimer . following a similar procedure as in the zhang - rice derivation of the t - j model , we could write down a hamiltonian for these bosonic excitations . the hamiltonian contains an energy of formation for the triplets , a term describing their propagation and a term describing pair creation and annihilation . all quantities in the derivation are quite well - defined , and can in principle be obtained by numerical techniques . in the present work we have actually attempted an ` ab initio calculation ' of the various parameters in the effective hamiltonian - a promising alternative to this somewhat clumsy approach might be a semi - empirical approach , where the parameters in the hamiltonian are adjusted to match e.g. experimental data . + the obtained ground state is an exact spin singlet , translationally invariant and isotropic - in short , it has precisely the symmetry properties expected for a spin liquid . the elementary excitations is a triplet mode , which reaches its lowest energy at @xmath235@xmath8@xmath0 . such a triplet mode is the most natural generalization of an antiferromagnetic spin - wave to the spin - liquid state . condensation of these bond bosons into the state with momentum @xmath0 then corresponds to antiferromagnetic ordering of the system . the breaking of rotational symmetry in spin space is due to fixing ( different ) condensation amplitudes for the three components of the triplet mode . the latter correspond to the three possible components of the antiferromagnetic order parameter . + in the present work we have chosen the completely isotropic nearest neighbor rvb - state as the starting point for constructing the triplet - hamiltonian . this is appropriate for an truly isotropic system , which probably is realized at finite doping and/or sufficiently high temperature . using a translationally invariant state , however , is not mandatory . for example by redefining the operator @xmath294 as @xmath295 with @xmath296 we can obviously generate a ` singlet soup ' with a preference for singlet orientation in @xmath12-direction . it is tempting to speculate that this may be appropriate for orthorhombic la@xmath297sr@xmath298cuo@xmath299 , where at @xmath300 columnar order is known to exist . analogous calculation as above , but with a finite @xmath301 may thus be appropriate to discuss the excitation spectrum of this material . + finally we note that the present scenario for the excitation spectrum of the spin - liquid provides a corner stone for a microscopic interpretation of zhang s so(5 ) symmetric theory of cuprate superconductors@xcite . as discussed above , an antiferromagnetic state may be thought of as being generated by condensing triplet - like spin excitations of the rvb spin - liquid into the state with momentum @xmath0 . the antiferromagnetic order parameter corresponds to the vector of condensation amplitudes of the three triplet species . then , so(5 ) symmetry states that such a triplet - excitation with momentum @xmath0 is ` dynamically equivalent ' to a @xmath302 hole pair@xcite . in fact the @xmath285-operator@xcite , which plays the role of a ladder operator for ` rotations in charge - direction ' in the so(5 ) theory@xcite , precisely replaces a bond - triplet with momentum @xmath0 by a hole pair with momentum @xmath271 . the fact that the @xmath285-operator is an approximate eigenoperator of the hamiltonian , @xmath303=\omega_0 \pi$ ] , then implies that the triplet and the hole pair are dynamically indistinguishable and differ only by their energy of formation . the @xmath285-operator thus would convert a condensate of triplets into a condensate of hole pairs , and , by the dynamical equivalence of these two objects , thus converts the antiferromagnetic ground state at half - filling into a superconducting state at finite - doping one . the somewhat ` toy - model - type ' discussion of ref.@xcite thus may very well be transferable almost literally to the fully planar t - j model . + in the present work we have restricted ourselves to a pure spin system . one my expect , however , that all considerations go through also for the doped system , with the sole difference that we get additional renormalizations of the matrix elements in the effective hamiltonian due to the fact that the doped holes reduce the volume available for pair generation and propagation of triplets . moreover , we will need terms which describe the coupling of the triplet branch to the mobile holes . in the simpler case of hole motion in a spin ladder this program has in fact already been carried out@xcite , leading to quite satisfactory agreement with numerical results . for the planar case we defer this to future work . + i would like to thank w. hanke , o. p. sushkov and shou - cheng zhang for instructive discussions and helpful comments . in this appendix we discuss the numerical work needed to obtain the values of @xmath106 , @xmath158 and the @xmath135 . to that end all possible dimer coverings of an @xmath304 cluster with periodic boundary conditions are generated on the computer the sums in ( [ over - real ] ) are evaluated numerically . restrictions on computer memory and cpu time do not allow to use @xmath305 , so that in practice only @xmath306 are possible . tu study the size dependence , we consider the quantities @xmath307 ; these give the ratio of the norm of an rvb state covering the exterior of a loop of length @xmath308 to the norm of an rvb state covering the entire cluster . similarly , we define @xmath309 , which gives the gain or loss in energy due to a fixed loop of length @xmath308 . the values for the different @xmath310 and @xmath311 are given in table [ tab1 ] and actually show already a quite remarkable independence of systems size . moreover , the expectation value of the energy / bond , @xmath312 , is @xmath313 for @xmath314 , @xmath315 for @xmath316 ; the estimate for @xmath317 is @xmath318@xcite . all in all the data suggest that already the @xmath189 cluster gives reasonably accurate estimates for the different parameters . + moreover , the data show that the @xmath319 and @xmath320 decrease quite rapidly with @xmath308 . truncation of the series after @xmath145 thus seems to be quite a reasonable approximation as well . s. kivelson , d. rokhsar , and j. p. sethna , phys . rev . b * 35 * , 865 ( 1987 ) . b. sutherland , phys . b * 37 * , 3786 ( 1988 ) . m. kohmoto , phys . b * 37 * , 3812 ( 1988 ) . e. fradkin , _ field theories of condensed matter systems _ , addison - wesley publishing company , 1991 . kasteleyn , physica * 27 * ( 1961 ) 1209 . s. sachdev and r. n. bhatt , phys . b * 41 * , 9323 ( 1990 ) . r. r. p. singh , m. p. gelfand , and d. a. huse , phys . * 61 * , 2484 ( 1988 ) . s. gopalan , t. m. rice , and m. sigrist , phys . b * 49 * , 8901 ( 1994 ) . r. eder , phys . b * 57 * , 12823 ( 1998 ) . a. v. chubukov and d. k. morr , phys . rev . b * 52 * , 3521 ( 1995 ) . v. n. kotov , o. sushkov , zheng weihong , and j. oitmaa , phys . lett . * 80 * , 5790 ( 1998 ) . han - ting wang , jue - lian shen , and zhao - bin su , phys . b. * 56 * , 14435 , ( 1997 ) . w. brenig , phys . b. * 56 * , 14441 , ( 1997 ) . r. eder , o. stoica , and g. a. sawatzky , phys . b. * 55 * , 6109 , ( 1997 ) ; r. eder , o. rogojanu , and g. a. sawatzky , phys . b. * 58 * , 7599 , ( 1998 ) . zhang , science * 275 * , 1089 ( 1997 ) . s. liang , b. doucot , and p. w. anderson , phys . lett . * 61 * , 365 ( 1988 ) . a. v. chubukov , jetp lett . * 49 * , 129 ( 1989 ) . it should be noted that by defining the states @xmath322 one could also construct a basis of ` spinon ' states and re - interpreting @xmath323 where the @xmath324 now correspond to spin-1/2 fermions one might be tempted to try and derive a _ fermionic _ hamiltonian for the ` spinons ' : @xmath325 readers uncomfortable with the triplet excitation picture put forward in this work are invited to try their hands on this calculation - the present author trusts that this experience will convince anybody that the triplets are the one and only sensible way of discussing the excitation spectrum of the spin liquid . g. stollhoff and p. fulde , j. chem . phys . * 73 * , 4548 ( 1980 ) . f. c. zhang and t. m. rice , phys . b * 37 * , 3759 ( 1988 ) . t. ogawa , k. kanda , and t. matsubara , progr . . phys . * 53 * , 614 ( 1975 ) ; see also d. vollhardt , rev . phys . * 56 * 99 ( 1984 ) . j. e. hirsch and s. tang , phys . b * 39 * , 2850 ( 1989 ) . d. p. arovas and a. auerbach , phys . b * 38 * , 316 ( 1988 ) . e. manouskais , rev . phys . * 63 * , 1 ( 1991 ) . r. eder , a. dorneich , m. g. zacher , w. hanke , and shou - cheng zhang , preprint cond - mat/9805120 . l. f. feiner , a. m. oles , and j. zaanen , phys . lett . * 78 * , 2799 ( 1997 ) . o. p. sushkov , preprint cond - mat/9808302 .

second rpa ( srpa ) calculations of nuclear response are performed and analyzed . unlike in most other srpa applications , the ground state , approximated by the hartree - fock ( hf ) ground state , and the residual couplings are described by the same hamiltonian and no arbitrary truncations are imposed on the model space . finite - range interactions are used and thus divergence problems are not present . we employ a realistic interaction , derived from the argonne v18 potenial using the unitary correlation operator method ( ucom ) , as well as the simple brink - boeker interaction . representative results are discussed , mainly on giant resonances and low - lying collective states . the focus of the present work is not on the comparison with data , but rather on technical and physical aspects of the method . we present how the large - scale eigenvalue problem that srpa entails can be treated , and demonstrate how the method operates in producing self - energy corrections and fragmentation . the so - called diagonal approximation is conditionally validated . stability problems are traced back to missing ground - state correlations .

0910.1674

extended rpa theories such as second rpa ( srpa ) @xcite , which go beyond first - order rpa , are often used to describe the strength , decay width and fine structure of nuclear giant resonances ( grs ) and other collective low - lying states . they also appear particularly useful in the case of unitarily transformed realistic interactions , which have not been calibrated for first - order rpa methods , but rather rely on extended model spaces to provide converged results . such interactions also make self - consistent " extended - rpa applications possible , in the sense that the ground state and the residual couplings can be described by the same hamiltonian . first applications using a renormalized argonne v18 potential , derived with the unitary correlation operator method ( ucom ) @xcite , were presented in ref . @xcite . self - consistent extended - rpa applications in large spaces without arbitrary truncations can be envisioned , in principle , with any properly constructed finite - range effective interaction . however , they are hardly ever performed for various technical and conceptual reasons . phenomenological effective interactions are fitted to sets of experimental data using mostly hartree - fock(-bogoliubov ) and selected ( quasi - particle ) rpa results . their range of applicability is then restricted to the selected observables and many - body methods . zero - range effective interactions , which greatly simplify numerical applications , can not be employed in second - order rpa methods , because they do not provide a natural cutoff in momentum space . consequently , no effective interactions have been fitted to calculations beyond first - order rpa and consistency in the treatment of the ground and excited states is ordinarily abandoned in practical applications of such methods . it is the purpose of the present work to perform and analyze large - scale ( i.e. , without arbitrary truncations ) , self - consistent " ( i.e. , with a single interaction as the sole input ) srpa calculations . we employ mostly the @xmath0 interaction used in ref . it is derived from the argonne v18 potential by means of a unitary transformation , which renormalizes it , while preserving the phase shifts and retaining the complex structure of the realistic interaction . we also use the brink - boeker interaction , @xmath1 @xcite , which is a simple , central , phenomenological effective interaction . no explicit three - body force is used at this point . we will not focus on producing realistic results and comparing them with data , as was done in ref . @xcite , but rather on technical and physical aspects of the method . we present how the large - scale eigenvalue problems that srpa involves can be treated , demonstrate how the method operates in producing self - energy corrections and fragmentation , and discuss consistency and stability problems . in the next section we present the srpa formalism and in sec . [ s : solve ] the methods we have used to solve the srpa eigenvalue problem . in sec . [ s : res ] we discuss our results with the help of illustrative examples . we conclude in sec . [ s : concl ] . in the following we assume a nuclear hamiltonian consisting , in general , of a one - body part and a two - body part , @xmath2 three - body terms are not included . if the total hamiltonian is considered , then @xmath3 is the total kinetic energy , while @xmath4 contains the interactions of particle pairs . if the intrinsic hamiltonian is considered , then @xmath5 and @xmath6 includes the intrinsic kinetic energy of the system , @xmath7 we will employ the srpa as it was formulated in ref . @xcite in analogy to rpa . the derivation is based on the equations - of - motion method and relies on a quasi - boson approximation . we will consider closed-(sub)shell spherical nuclei and their excited states of definite angular momentum and parity @xmath8 . excited states are expanded in the space of particle - hole ( @xmath9 ) and two - particle - two - hole ( @xmath10 ) configurations . the symbol @xmath11 ( or @xmath12 ) will represent all the quantum numbers of a particle ( hole ) state except the magnetic quantum number @xmath13 @xmath14 , i.e. , the set of quantum numbers @xmath15 of the nodes ( @xmath16 ) , orbital angular momentum , total angular momentum , and isospin . the combined label @xmath17 is used to label the @xmath18 combination uniquely ; then @xmath19 $ ] and @xmath20 + 1/2 $ ] ( where @xmath21 $ ] is the integer part of @xmath22 ) . the greek letters @xmath23 will be used to denote single particle states of either kind ( @xmath11 or @xmath12 ) . the operator @xmath24 that creates an excited state @xmath25 of energy @xmath26 with respect to the @xmath27 ground state @xmath28 and of angular momentum @xmath29 , @xmath30 is written as @xmath31 where @xmath32 creates a @xmath9 state and @xmath33 creates a @xmath10 state , coupled to the given quantum numbers . in particular , we have @xmath34 @xmath35 henceforth the indices @xmath29 will be omitted , but implied throughout . the @xmath10 state ( and the corresponding creation operator ) is characterized , besides @xmath29 , by @xmath36 and @xmath37 , the angular momenta to which the two particle states and the two hole states , respectively , are coupled . the same holds for the amplitudes @xmath38 , @xmath39 . moreover , an ordering of the single - particle states is introduced and only the operators with @xmath40 , @xmath41 are included in the expansion ( [ e : qop ] ) to avoid multiple counting of configurations . for example , the present convention is that @xmath42 if @xmath43 , or , for states of the same isospin , if @xmath44 , or , when all other quantum numbers are the same , if @xmath45 . the srpa ground state , which formally is the vacuum of the annihilation operators @xmath46 , is approximated with the hartree - fock ( hf ) ground state . the latter is the slater determinant that minimizes the expectation value of the given hamiltonian , @xmath47 . the forward ( @xmath48 , @xmath38 ) and backward ( @xmath49 , @xmath39 ) amplitudes are the solutions of the srpa equations in @xmath50 space @xmath51 the vanishing blocks are due to the choice of ground state . @xmath52 and @xmath53 are the usual @xmath54 rpa matrices ( @xmath55 the number of @xmath9 configurations ) , whose angular momentum - coupled forms are given by @xmath56_{ph;p'h ' } & = & ( e_p - e_h)\delta_{pp'}\delta_{hh ' } \nonumber \\ & & + \langle ph^{-1};j | h_2 | p'{h'}^{-1};j\rangle \nonumber \\ & = & ( e_p - e_h)\delta_{pp'}\delta_{hh ' } \nonumber \\ & & + \sum_{j_1}(-1)^{j_{h}+j_{p'}-j_1 } ( 2j_1 + 1 ) \nonumber \\ & & \times \left\ { \begin{array}{ccc}j_{p } & j_{h ' } & j_1\\ j_{p ' } & j_{h } & j \end{array } \right\ } \langle ph';j_1 | h_2 | h p';j_1\rangle \label{e : amat } \end{aligned}\ ] ] @xmath57_{ph;p'h ' } & = & \langle ( ph^{-1};j ) ( p'{h'}^{-1};j ) | h_2 | 0\rangle \nonumber \\ & = & \sum_{j_1}(-1)^{j_{h}+j_{p'}+j - j_1 } ( 2j_1 + 1 ) \nonumber \\ & & \times \left\ { \begin{array}{ccc}j_{p } & j_{p ' } & j_1\\ j_{h } & j_{h ' } & j \end{array } \right\ } ( 1+\delta_{pp'})^{1/2 } ( 1+\delta_{hh'})^{1/2 } \nonumber \\ & & \times \langle pp';j_1 | h_2 | h h';j_1\rangle . \label{e : bmat } \end{aligned}\ ] ] the @xmath58 submatrix @xmath59 ( @xmath60 the number of @xmath10 configurations ) describes the coupling between @xmath9 and @xmath10 states , @xmath61_{ph;p_1p_2h_1h_2j_pj_h } = } \nonumber \\ & & \langle ph^{-1};j | h_2 | ( p_1p_2;j_p)(h_1h_2;j_h)^{-1};j \rangle \nonumber \\ & = & [ 1-(-1)^{j_{h_1}+j_{h_2}-j_h } p(h_1,h_2 ) ] \delta_{h_1h } \nonumber \\ & & \times ( -1)^{j_{p}+j_{h_2}+j+j_h}(1+\delta_{h_1h_2})^{-1/2 } \hat{j}_p\hat{j}_h \nonumber \\ & & \times \left\ { \begin{array}{ccc}j_p & j & j_h\\ j_{h_1 } & j_{h_2 } & j_p \end{array } \right\ } \langle p_1p_2;j_p | h_2 | ph_2;j_p\rangle \nonumber \\ & - & [ 1-(-1)^{j_{p_1}+j_{p_2}-j_p } p(p_1,p_2 ) ] \delta_{p_1p } \nonumber \\ & & \times ( -1)^{j_{p_1}+j_{p_2}+j+j_h}(1+\delta_{p_1p_2})^{-1/2 } \hat{j}_p\hat{j}_h \nonumber \\ & & \times \left\ { \begin{array}{ccc}j_h & j & j_p\\ j_{p_1 } & j_{p_2 } & j_h \end{array } \right\ } \langle hp_2;j_h | h_2 | h_1h_2;j_h\rangle , \label{e : a12mat } \end{aligned}\ ] ] while the @xmath62 matrix @xmath63 contains the @xmath10 states and their interactions , @xmath64_{p_1p_2h_1h_2j_pj_h;p_1'p_2'h_1'h_2'j_p'j_h ' } = } \nonumber \\ & & \delta_{p_1p_1'}\delta_{h_1h_1'}\delta_{p_2p_2'}\delta_{h_2h_2 ' } \delta_{j_pj_p'}\delta_{j_hj_h'}(e_{p_1}+e_{p_2}-e_{h_1}-e_{h_2 } ) \nonumber \\ & + & \langle ( p_1p_2;j_p)(h_1h_2;j_h)^{-1};j | h_2 | ( p_1'p_2';j_p')({h_1'}{h_2'};j_h')^{-1};j \rangle \nonumber \\ & = & \delta_{p_1p_1'}\delta_{h_1h_1'}\delta_{p_2p_2'}\delta_{h_2h_2 ' } ( e_{p_1}+e_{p_2}-e_{h_1}-e_{h_2 } ) \nonumber \\ & + & \delta_{p_1p_1'}\delta_{p_2p_2 ' } \delta_{j_pj_p'}\delta_{j_hj_h ' } [ 1+(-1)^{j_p}\delta_{p_1p_2 } ] ( 1+\delta_{p_1p_2})^{-1 } \nonumber \\ & & \times \langle h_1'h_2';j_h | h_2 | h_1h_2;j_h\rangle \nonumber \\ & + & \delta_{h_1h_1'}\delta_{h_2h_2 ' } \delta_{j_pj_p'}\delta_{j_hj_h ' } [ 1+(-1)^{j_h}\delta_{h_1h_2 } ] ( 1+\delta_{h_1h_2})^{-1 } \nonumber \\ & & \times \langle p_1p_2;j_p | h_2 | p_1'p_2';j_p\rangle \nonumber \\ & + & [ 1-(-1)^{j_{p_1}+j_{p_2}-j_p } p(p_1,p_2 ) ] ( 1+\delta_{p_1p_2})^{-1/2 } \nonumber \\ & & \times [ 1-(-1)^{j_{h_1}+j_{h_2}-j_h } p(h_1,h_2 ) ] ( 1+\delta_{h_1h_2})^{-1/2 } \nonumber \\ & & \times [ 1-(-1)^{j_{p_1'}+j_{p_2'}-j_p } p(p_1',p_2 ' ) ] ( 1+\delta_{p_1'p_2'})^{-1/2 } \nonumber \\ & & \times [ 1-(-1)^{j_{h_1'}+j_{h_2'}-j_h } p(h_1',h_2 ' ) ] ( 1+\delta_{h_1'h_2'})^{-1/2 } \nonumber \\ & & \times \delta_{h_2h_2'}\delta_{p_2p_2 ' } ( -1)^{1+j_{p_1}+j_{p_2}+j_{h_1}+j_{h_2 } } \hat{j}_p \hat{j}_p ' \hat{j}_h \hat{j}_h ' \nonumber \\ & & \times \sum_{l}(-1)^{j_h - j_{h'}+j - l } ( 2l+1 ) \left\ { \begin{array}{ccc}j_p & j_p ' & l\\ j_h ' & j_h & j \end{array } \right\ } \nonumber \\ & & \times \left\ { \begin{array}{ccc}j_p & j_p ' & l\\ j_{p_1 ' } & j_{p_1 } & j_{p_2 } \end{array } \right\ } \left\ { \begin{array}{ccc}j_h & j_h ' & l\\ j_{h_1 ' } & j_{h_1 } & j_{h_2 } \end{array } \right\ } \nonumber \\ & & \times \sum_{j_1}(-1)^{j_{h_1}+j_{p_1'}-j_1 } ( 2j_1 + 1 ) \nonumber \\ & & \times \left\ { \begin{array}{ccc}j_{p_1 } & j_{h_1 ' } & j_1\\ j_{h_1 } & j_{p_1 ' } & l \end{array } \right\ } \langle p_1h_1';j_1 | h_2 | p_1'h_1;j_1\rangle . \label{e : a22mat } \end{aligned}\ ] ] in the above , @xmath65 are the hf single - particle energies and the operator @xmath66 exchanges the indices @xmath67 and @xmath68 . the two terms in @xmath59 describe the free propagation of the hole @xmath12 while the particle @xmath11 interacts with an intermediate @xmath9 state ( terms with @xmath69 or @xmath70 , i.e. , containing @xmath71 ) , and the free propagation of a particle while the hole @xmath12 interacts with an intermediate @xmath9 state ( terms with @xmath72 ) . thus self - energy corrections via bubble diagrams are introduced either to a particle or a hole state , or a @xmath9 excitation is exchanged between a particle and a hole state . the four terms in @xmath63 describe : the free propagation of a @xmath10 state ; the free propagation of the two particles while the two holes interact ; the free propagation of the two holes while the two particles interact ; and the free propagation of a @xmath9 pair while the other particle and hole interact . in the following we will omit the @xmath36 , @xmath37 indices for simplicity . if we neglect the coupling amongst the @xmath10 states , @xmath63 becomes diagonal and its elements are determined by the unperturbed @xmath10 energies ( diagonal approximation ) , @xmath73_{p_1h_1p_2h_2,p_1'h_1'p_2'h_2 ' } = } \nonumber \\ & & \delta_{p_1p_1'}\delta_{h_1h_1'}\delta_{p_2p_2'}\delta_{h_2h_2 ' } ( e_{p_1}+e_{p_2}-e_{h_1}-e_{h_2 } ) \label{ea22 } . \end{aligned}\ ] ] the validity of the diagonal approximation is examined in sec . [ s : diag ] . as long as we are interested only in the single - particle response , determined by the @xmath9 amplitudes , @xmath74 and @xmath75 , we may eliminate the @xmath76 , @xmath77 amplitudes from eq . ( [ esrpa ] ) and reduce the srpa problem to an energy - dependent eigenvalue problem of the dimension of the rpa matrix @xcite , @xmath78 in general , the expression for @xmath79 will involve the inverse of @xmath80 $ ] , where @xmath81 the @xmath62 identity matrix . within the diagonal approximation we have simply @xmath82_{ph;p_1p_2h_1h_2}[\mathcal{a}^t_{12}]_{p_1p_2h_1h_2;p'h'}}{e - ( e_{p_1 } + e_{p_2 } - e_{h_1}-e_{h_2 } ) + i\eta_2 } . \label{e : srpared2 } \end{aligned}\ ] ] a finite constant @xmath83 is used in applications to smoothen the poles of the function and to introduce a width to the @xmath10 states . the quantities of interest are transition strength distributions , or strength functions , @xmath84 of transition operators @xmath85 @xmath86 and their energy moments @xmath87 determined , in general , by the amplitudes @xmath48 , @xmath49 , @xmath38 , and @xmath39 through @xmath88 } \nonumber \\ & & + \!\!\ ! \sum_{p_1p_1h_1h_2 } \!\!\ ! [ f_{p_1p_2h_1h_2}{\mathcal{x}_{p_1p_2h_1h_2}^{\lambda}}^{\ast } - \tilde{f}_{p_1p_2h_1h_2}{\mathcal{y}_{p_1p_2h_1h_2}^{\lambda}}^{\ast } ] , \label{e : fl0 } \end{aligned}\ ] ] where the coefficients @xmath89 , @xmath90 depend on the operator and ground state . centroid energies can be defined as @xmath91 where the sums @xmath92 may be evaluated over the whole spectrum ( unrestricted summations in eq . ( [ e : moments ] ) ) , or only in the energy region of a resonance . the width of a distribution or a resonance can be expressed as @xmath93 smoothened strength functions can be produced , for analysis purposes , by folding the discrete strength functions with a lorentzian of width @xmath94 , which yields @xmath95 in the limit @xmath96 , eq . ( [ e : rdiscr ] ) is recovered . we will consider is and iv transitions of definite spin and parity @xmath8 , described by standard single - particle transition operators @xcite . then , for the hf ground state , we have @xmath97 \langle p || f || h \rangle |^2 .\ ] ] the two - body @xmath38 , @xmath39 amplitudes do not contribute . the energy moments @xmath98 and @xmath99 will be the same in srpa as in rpa @xcite . a @xmath100 multiplicity will be included in our final results . two - body operators can also be considered . then all amplitudes contribute to the transition matrix element . for example , let us consider the double dipole resonance , excited by the two - body operator @xmath101_{j^+}$ ] , where @xmath102 the usual ( single - particle ) isovector @xmath103 operator and @xmath104 . we then have @xmath105 f_{ph}^{\mathrm{ddr};j } } \nonumber \\ & & + \!\!\ ! \sum_{p_1p_2h_1h_2 } \!\!\ ! [ { x_{p_1p_2h_1h_2}^{\lambda \ , \ , \ast}}+{y_{p_1p_2h_1h_2}^{\lambda \ , \ , \ast } } ] f_{p_1p_2h_1h_2}^{\mathrm{ddr};j } , \end{aligned}\ ] ] where @xcite @xmath106 and @xmath107 .\end{aligned}\ ] ] finally , we may define the strength distribution of a @xmath108 configuration , coupled to a given angular momentum state , via the quantity @xmath109 where the summation is over all eigenstates with @xmath110 . ( in the unperturbed case the centroid of @xmath111 is trivially identical to the hf @xmath9 energy @xmath112 and its width is zero . ) similarly , the strength distribution of a @xmath10 state can be defined using the @xmath38 and @xmath39 amplitudes . energy moments and centroids , as well as smoothened distributions , can be defined as usual . we note that the total @xmath98 and @xmath99 ( and centroid ) of @xmath113 will be the same in rpa and srpa , since @xmath113 is the sum of the strength functions related to the operator @xmath114 and its adjoint . @xmath98 , in particular , should always amount to one . by setting the coupling matrices @xmath59 and @xmath63 and the @xmath10 amplitudes @xmath38 , @xmath39 equal to zero in eq . ( [ esrpa ] ) , we recover the usual rpa problem . if , in addition , we neglect the @xmath9 residual interaction ( i.e. , @xmath115 and @xmath116 ) , we obtain a trivial , unperturbed problem , where the eigenstates @xmath117 are the @xmath9 configurations at the hf level and the @xmath49 amplitudes vanish . by setting only @xmath118 in eq . ( [ esrpa ] ) , we obtain a second - order tamm - dancoff approximation ( stda ) , which amounts to solving the eigenvalue problem of the @xmath52 block of the srpa matrix . the backward amplitudes @xmath49 and @xmath39 vanish and ground - state correlations implicitly taken care of by those are ignored . setting also the coupling matrices @xmath59 and @xmath63 equal to zero , one gets the usual , first - order tamm - dancoff approximation ( tda ) . in a manner analogous to tda , stda is equivalent to a diagonalization of the hamiltonian in the @xmath50 space . it should be noted , though , that for @xmath119 the hf ground state does not decouple from the stda space ( unlike tda ) . a diagonalization in the model space that includes in addition the hf state would produce a new ground state of lower energy . the @xmath27 excitation spectrum would also be affected . in sec . [ s : cons ] we will investigate the possible influence of ignored ground - state correlations ( gsc ) on our results . a rigorous way to do that would be the use of an extended srpa method built on a correlated ground state as self - consistently as possible ( or other extended methods such as a self - consistent green s function method @xcite ) . since , however , that is a demanding project going beyond the scope of the present work , we resort instead to simpler approaches . one way to assess the role of gsc is to ignore them completely . this is accomplished within the ( s)tda . as a second approach , we have devised a simple and rather heuristic renormalized version of srpa ( rsrpa ) , which takes into account to some extent the depletion of the fermi sea . we start with a renormalized rpa approach ( rrpa ) . following the simplified rrpa method of refs . @xcite see also ref . @xcite we assume partially occupied single - particle states and renormalize the residual couplings and the transition matrix elements accordingly . in particular , when calculating the matrix elements of @xmath52 , the @xmath120 terms in eq . ( [ e : amat ] ) are multiplied by a factor @xmath121 } = d_{ph}^{1/2}d_{p'h'}^{1/2 } \ , , \label{e : dfact1}\ ] ] where @xmath122 and @xmath123 is the occupation probability of the orbital @xmath67 . the matrix elements of @xmath53 , eq . ( [ e : bmat ] ) , are multiplied by the same factor . finally , the single - particle transition matrix elements @xmath124 , @xmath125 , are renormalized by a factor @xmath126 . in rsrpa we have to renormalize the matrix elements of @xmath127 and @xmath128 as well . the former , eq . ( [ e : a12mat ] ) , will be multiplied by the expression ( [ e : dfact1 ] ) , taking account of the occupation probabilities of the @xmath11 , @xmath12 states , as well as by a factor @xmath129 } = \frac{1}{2 } [ d_{p_1h_1,p_2h_2 } + d_{p_1h_2,p_2h_1 } ] , \label{e : dfact2}\ ] ] to take into account the occupation probabilities of the @xmath130 , @xmath131 states . similarly , the @xmath120 terms in the @xmath128 matrix elements , eq . ( [ e : a22mat ] ) , will be multiplied by @xmath132 } d_{p_1'h_1'p_2'h_2'}^{[2 ] } $ ] . in practice , we will not solve these equations iteratively . the single - particle energies and eigenstates needed to evaluate the above matrix elements will be the hf ones , whereas the occupation probabilities are calculated using the shell model . obviously , this method is neither rigorous nor consistent , but it should help us get an idea regarding the sensitivity of our results to gsc . it is well known , that in self - consistent rpa ( meaning that the same hamiltonian is used to calculate the hf ground state and the residual interaction ) the spurious state related to the cm momentum operator @xmath133 ( single spurious state ) will appear at zero energy and be exactly separated from the physical spectrum , provided that all @xmath9 and @xmath134 configurations available in the ( sufficiently large ) single - particle space are taken into account . the same does not hold , however , when extensions of rpa are considered , whether that means considering a correlated ground state or higher - order configurations . as was formally shown in ref . @xcite , once @xmath10 configurations are included in the model space , a necessary condition for the single spurious state to appear at zero energy is that all single - particle amplitudes be taken into account : this means not only @xmath9 and @xmath134 amplitudes , but also @xmath135 and @xmath136 . it was in fact shown that @xmath135 and @xmath136 amplitudes affect the spurious state even with a hf ground state , because at energies equal to @xmath137 or @xmath138 they do not vanish . obviously our approach does not include them . this problem will be examined quantitatively in sec . [ s : cons ] . a more severe problem in hf - based srpa is the onset of instabilities . contrary to rpa , the self - consistent " use of the hf ground state does not guarantee that the srpa stability matrix will be positive - definite : thouless s theorem was proven specifically for @xmath9 excitations @xcite . thus rpa is related to the stability conditions for the hf solution , but srpa seems related to an extended variational problem @xcite . in practice , we will find that low - lying states appear at imaginary or negative energies . is normalized to + 1 , i.e. , the norm of the positive - energy counterpart is negative . ] it is not difficult to demostrate how negative eigenvalues can occur , already at the stda level ( @xmath118 ) , if we consider eq . ( [ e : srpared2 ] ) : for energies @xmath139 below typical @xmath10 energies , strong ( or many ) @xmath59 elements , regardless of sign , can cause even the diagonal elements of @xmath140 to become negative and destroy the positive - definitness of the matrix . additional ground - state correlations could cure this problem , through a renormalization of the @xmath59 couplings , as illustrated above , or by filling the vanishing @xmath141submatrices of eq . ( [ esrpa ] ) with finite elements ( see also ref . @xcite for a related discussion within an extended dressed rpa ) . the question whether srpa gives meaningful results for giant resonances despite the problematic solutions at low energies is also tackled in sec . [ s : cons ] . in practice we proceed as follows . we first choose a single - particle space , consisting of harmonic - oscillator eigenstates . the larger the space , the better the convergence of our results to their final values . all angular - momentum coupled two - body matrix elements of the given interaction within the harmonic - oscillator basis are calculated in advance as described in ref . @xcite and stored . we solve the hf equations within the given space and then we use all @xmath9 and @xmath10 configurations , which are available within the space and can couple to the desired angular momentum and parity , to construct the srpa matrix . the single - particle space is characterized by either * @xmath142 , the number of energy quanta in the highest oscillator shell considered . this means that all single - particle states within the lowest @xmath143 shells are used . an additional cutoff @xmath144 in @xmath145 may also be imposed . or * @xmath146 and @xmath144 . a cutoff @xmath144 in @xmath145 is considered and each possible @xmath147 state is expanded in the lowest @xmath148 harmonic - oscillator @xmath145 states . since angular momentum is preserved , for an rpa calculation of the @xmath8 response , one has to include @xmath149 s up to the maximal occupied one plus @xmath150 , so that all @xmath151 s with @xmath152 are included in the space , and may omit higher @xmath151 s . in srpa there is no @xmath149-cutoff provided by angular momentum conservation . in principle , particle states with infinitely large @xmath149 s can contribute to allowed @xmath10 configurations and such configurations can couple with @xmath9 ones through finite matrix elements of @xmath59 . in particular , infinitely large @xmath153 and @xmath154 can couple to any given @xmath36 , thus resulting in finite values of @xmath155 see r.h.s . of eq . ( [ e : a12mat ] ) . for restricted @xmath146 , however , upper bounds are provided by the transformation ( moshinsky ) brackets needed to evaluate the above matrix element in the ho basis @xcite . natural cutoffs on @xmath146 or @xmath156 should be provided by the properties of the interaction . perhaps such considerations can be used , along with the convergence behavior of the srpa solutions with respect to the single - particle basis , for the optimization of the latter , but so far this has not been done . the next task is to solve the srpa eigenvalue problem . first we note that all the submatrices comprising the srpa matrix , eq . ( [ esrpa ] ) , are real in the cases studied here . the submatrices @xmath52 , @xmath53 , and @xmath63 are also symmetric . the same holds for the @xmath157 blocks indicated in eq . ( [ esrpa ] ) ( separated by lines ) . the dimension @xmath158 of the srpa matrix is given by twice the number of @xmath9 and of @xmath10 configurations available in the model space ( @xmath159 ) . the latter , @xmath60 , can be rather large : for the purposes of the present work we encountered problems with @xmath160 up to @xmath161 . in the following we present technical information on how we deal with the srpa eigenvalue problem in practice . we shall distinguish three cases : * all eigenvalues and eigenvectors are calculated . this is feasible for relatively small problems , e.g. , @xmath162 . * for large problems the complete solution becomes impossible . then only a small portion of consecutive eigenvalues and the respective eigenvectors are calculated . in most cases , all excitations of interest lie at the lower end of the spectrum , so it suffices to evaluate only the lowest ( a few tens or hundreds ) positive eigenvalues . * no eigenvalue problem is solved . the response function is evaluated directly . the strategy is well known and very practical when one is interested in the single - particle response , and only in the final result , namely a smoothened strength function . standard library routines can be used to solve small enough eigenvalue problems . the solution can be sped up considerably if the srpa problem is reduced to half the dimension , @xmath157 . a method relying on a cholesky decomposition of the matrix @xmath163 or @xmath164 is the most common way to do it , with the additional bonus that it produces a _ symmetric _ @xmath157 eigenvalue problem @xcite . however , the method will not work if neither @xmath163 nor @xmath164 is positive - definite , a problem that can occur when there are instabilities , or just spurious states at imaginary energies . we employ a modified method instead , relying on a generalized cholesky decomposition @xcite , which is equivalent to the original one when the matrix under decomposition is positive definite ( thus still producing a symmetric eigenvalue problem ) and still works when it is not ( producing a non - symmetric eigenvalue problem of half the dimension ) , with minimal additional computational effort . for details , see ref . @xcite . when the dimension of the matrix is very large , the first problem that occurs is the storage of the matrix elements . fortunately , most of the srpa matrix matrix elements vanish , especially when the diagonal approximation is used , eq . ( [ ea22 ] ) , so it becomes possible to store all the finite ones in memory . obviously , one needs to store only one @xmath157 matrix ( e.g. , @xmath52 or @xmath165 ) and one small , @xmath54 matrix ( @xmath53 ) . moreover , only the upper ( or lower ) triangles need be stored , since the matrices are symmetric . the dimension and degree of sparseness of the srpa matrix depends on the nucleus , type of response , and model space . as an example , let us mention that in the case of the @xmath166 response of @xmath167ca with the @xmath0 , and for spaces large enough for reasonable convergence , there are about a million configurations . @xmath168 gb of storage are needed in the diagonal approximation , but @xmath169 gb could be required for the full problem ( @xmath170 elements in double precision ) . again the dimension of the srpa problem can be reduced by half , to speed up the numerical solution . in order to save matrix operations we chose not to perform a decomposition , but resort to a more straightforward reduction method @xcite : we solve the non - symmetric @xmath157 problem @xmath171 where the eigenvectors @xmath172 obey the normalization condition @xmath173 ( real and positive @xmath174 ) . the properly normalized @xmath48 and @xmath49 arrays ( @xmath175 ) are then given by @xmath176 r^{\lambda } \\ y^{\lambda } & = & \frac{1}{2 } [ \sqrt{e_{\lambda } } \mathcal{i}_n - \frac{1}{\sqrt{e_{\lambda } } } ( a+b ) ] r^{\lambda } , \end{aligned}\ ] ] where @xmath177 is the @xmath157 identity matrix . finally , an arnoldi iteration procedure from the arpack package @xcite is employed to solve the problem ( [ e : ab ] ) only for the @xmath178 lowest positive eigenvalues @xmath179 , where @xmath180 . in principle , it is possible to solve for the first @xmath178 eigenvalues lying above a given energy value @xmath181 , not necessarily equal to zero . in practice , however , such a strategy can prove problematic : already at moderate energies the density of eigenstates can be so large ( see sec . [ s : res ] ) , that the arnoldi iteration will likely fail to converge . as long as one - body transition operators are considered , we may just solve the reduced srpa problem , eq . ( [ e : srpared1 ] ) . in general , the reduction procedure involves the inversion of a large matrix ( of the dimension of the @xmath10 space ) , but when @xmath63 is diagonal , that is reduced to a trivial number inversion , see eq . ( [ e : srpared2 ] ) . there are ways to solve such an energy - dependent eigenvalue problem @xcite . an efficient alternative is to employ the response - function formalism . then , instead of explicitly solving the eigenvalue problem , one can obtain directly the strength function of interest @xcite . first , the @xmath9 green s function , a @xmath54 matrix , is evaluated , @xmath182 ( @xmath183 ) by numerical matrix inversion . the response function for a given single - particle field is given by @xmath184 where the elements of the @xmath185dimensional ( assumed real ) array @xmath186 ( @xmath187 ) , in @xmath9 space , are the matrix elements of the transition operator , @xmath124 ( @xmath188 ) see eq . ( [ e : fl0 ] ) . finally , the strength function of interest is given by @xmath189 in practice , the green s function , response function , and finally the strength function are evaluated over the energy range of interest , which is represented by mesh points @xmath190 . the constants @xmath191 , which provide the @xmath9 and @xmath10 states , respectively , with a finite width should , in principle , be small enough for the structure of the strength function to be resolved as desired . the mesh size @xmath192 should be smaller than both of them ( a factor 3 - 4 suffices ) . the choice @xmath193 produces a smoothened strength function that is practically the same as the discretized strength function ( obtained by explicit diagonalization of the srpa matrix ) folded with a lorentzian of width @xmath194 , cf . ( [ e : lor ] ) . in the following , we will discuss the features of the srpa solutions with the help of illustrative examples . we note that we will use the acronym srpa0 when referring specifically to the diagonal approximation , eq . ( [ ea22 ] ) , and srpa(0 ) when referring explicitly to both solutions , with and without the diagonal approximation . the intrinsic nuclear hamiltonian is employed . in all cases we set the oscillator length parameter of the single - particle basis equal to @xmath195 fm ( for @xmath0 ) or @xmath196 fm ( for @xmath1 ) . it will prove instructive to consider , among others , a relatively small srpa problem , for which we can calculate all eigenstates with and without the diagonal approximation . the purpose of such an exercise is not to perform a realistic calculation , but to illustrate and discuss selected features of the method : in particular , how the large amount of @xmath10 configurations influences the response function and the distribution of the srpa eigenstates . as such a toy model " we choose the monopole ( @xmath27 ) response of @xmath197o in a rather small single - particle space , consisting of seven oscillator shells , and using the @xmath0 interaction . in total , there are 4148 positive - energy eigenstates to evaluate ( @xmath198 , @xmath199 ) . the single - particle , isoscalar monopole strength distribution is shown in fig . [ f : toy ] in linear ( fig . [ f : toy]a ) and logarithmic ( fig . [ f : toy]b ) scale . further examples will be introduced in the following as needed . when looking at fig . [ f : toy]a , we observe that the results of the diagonal approximation , srpa0 , are very close to the exact srpa results . it looks as though the effect of the @xmath10 space on the single particle response is approximately the same , whether or not the @xmath10 states are considered unperturbed . we have verified that the approximation is equally good in larger spaces and for different types of response always of single - particle operators . see , e.g. , the ivd response of @xmath197o calculated within a space of 13 oscillator shells , as shown in fig . 1 of ref . @xcite . the approximation remains quite good for heavier nuclei , for example @xmath200ca , whose isq response is shown in fig . [ f : full]a , calculated within a space of 9 shells . we expect the diagonal approximation to be reliable for soft , perturbative interactions in general . the additional couplings within the @xmath10 space , ignored in the diagonal approximation , constitute higher - order corrections to the excitation propagator with respect to the interaction , as can be demonstrated diagrammatically @xcite . results with the brink - boeker potential , @xmath1 , corroborate this speculation . @xmath1 is even softer than the @xmath0 . for example , it produces much smaller second - order corrections to the nuclear binding energies in the perturbation expansion beyond hartree - fock @xcite . in fig . [ f : full]b we show the ivd response of @xmath197o calculated using @xmath1 . the diagonal approximation is very good in this case . to be on the safe side , it is always advisable to verify the quality of the approximation within some tractable space . in ref . @xcite , for example , it was found that it is quite bad in the case of metallic clusters when the bare coulomb interaction is used . from fig . [ f : toy]a we deduced that the srpa and srpa0 solutions yield almost the same results for the giant resonance . however , when we look at the strength distribution on a logarithmic scale , fig . [ f : toy]b , it becomes obvious that the two solutions give different results for the majority of eigenstates , which are mostly of @xmath10 character . one implication is that the diagonal approximation can not be relied upon when examining , e.g. , double giant resonances and the response function of two - particle operators in general . as an example , in fig . [ f : ddr ] we plot the strength function of the @xmath27 component of the double dipole resonance . we notice that the srpa0 strength function is very close to the unperturbed one ( hf ) and differs significantly from the srpa result . in fig . [ f : eigen05 ] we show the density of eigenstates ( number of states per 5 mev excitation energy ) when solving the srpa problem with and without the diagonal approximation , as well as the density of unperturbed @xmath9 and @xmath10 states ( hf ) . one immediately notices that for the most part the srpa0 density of states is very similar to the unperturbed one . finally , below about 40 mev , where the giant resonance lies , the srpa and srpa0 densities practically coincide . this is not surprising , given the very small number of eigenstates present there , but it is not always the case . we note , for example , that in the case of the @xmath201 response of @xmath200ca , where there are several unperturbed @xmath10 states at low energies , the srpa0 density of states follows mostly the hf one in that energy region . the single - particle strength functions , however , are almost identical ( fig . [ f : full]a ) and the diagonal approximation remains well justified . in figs . [ f : toy ] and [ f : full ] we notice that the gr lies lower in srpa(0 ) than in rpa . this is a rather general result see also ref . it is a general feature of the coupling to @xmath10 excitations @xcite ( and not particular to nuclei @xcite ) , and shows why traditional effective interactions can not be used in srpa . the self - energy corrections responsible for the modification of the @xmath9 energies and the lowering of the resonance energies ( see also sec . [ s : fragm ] ) are already parameterized in the interaction , so that realistic results are obtained already at the rpa level . employing such an interaction in srpa will result in double counting of those effects . ths problem is circumvented in practical applications in the literature through the use of realistic single - particle energies and subtracting procedures , which remove the real part of the self energy , responsible for the shift . as a microscopic theory of collisional damping , srpa has been used extensively to describe the spreading width and strength fragmentation and quenching of collective excitations . let us examine our results in this context . no spreading width is observed in fig . [ f : toy]a , as there are no configurations available in the vicinity of the resonance . but when we look at fig . [ f : toy]b , we realize that there is practically a continuum of weak srpa or srpa0 eigenstates ( @xmath202 ) extending to high energies . most are predominantly of @xmath10 nature and their contribution to the single - particle strength is rather small . nevertheless , they carry a non - negligible percentage of the total strength and provide a mechanism of strength quenching for the giant resonance ( by contrast , there are only @xmath198 rpa eigenstates ) . we note , in particular , that : ( i ) as expected , the total energy - weighted strength @xmath99 is found to be practically the same in all three cases ( rpa , srpa0 , srpa ) , and the total strength @xmath98 is found just about @xmath203 larger in srpa or srpa0 than in rpa , but ( ii ) the strongest ism peak appears at 21.36 mev in the case of rpa , 16.50 mev in srpa0 , and 16.25 mev in srpa and ( iii ) the six lowest eigenstates carry a total strength of 313.4 @xmath204 in rpa , which is almost all the rpa strength ( they lie at 21 to 47 mev excitation energy ) , 304.5 @xmath204 in srpa0 ( 16 to 34 mev ) and 304.0 @xmath204 in srpa ( 16 to 34 mev ) , i.e. , about @xmath203 less in srpa(0 ) than in rpa ; finally , ( iv ) the width of the ism strength functions below 40 mev is 3.86 mev in rpa and 3.73 mev in srpa , while the widths over the whole spectrum are 4.65 and 18.94 mev respectively . the coupling with @xmath10 configurations affects not only the collective excitations , but at the same time the single - particle states . within the present formalism one can demonstrate this through the strength of the @xmath9 configurations . in fig . [ f : phfraglin ] , for example , we plot the hf , rpa and srpa strength distribution @xmath111 , eq . ( [ esph ] ) , of the @xmath9 configurations @xmath205 , contributing to the @xmath27 strength of fig . [ f : toy ] . in hf , these are well - defined transitions at energies equal to @xmath112 with strength 1 . in rpa the @xmath9 transitions appear fragmented . there is a cluster of configurations at energies around 30 mev , which can be attributed to the @xmath206 transition @xmath207 . a second cluster of states is visible around 60 mev roughly corresponding to @xmath208 configurations , and so on . in srpa the strength distribution appears even more fragmented , as well as shifted to lower energies . the shift reflects an effective compression of the single - particle spectrum ( a modification of the nucleon effective mass ) , with respect to hf , and leads to the downward shift of the collective states , discussed above . although we can visually identify the different shells contributing to the strength distribution , and observe an energetic shift , the total @xmath98 and @xmath99 and thus the centroid of any given @xmath9 configuration is the same in rpa and srpa . nevertheless , in rpa we find only 1% of the @xmath207 @xmath98 strength above 40 mev . in srpa the strength in the same region is 14% . when we look at the distributions in logarithmic scale , fig . [ f : phfraglog ] , we realize that the @xmath9 strength distributions span the whole space available , reflecting the extended spectroscopic functions of the particle and hole states . as is obvious from figs . [ f : toy ] and [ f : eigen05 ] , the majority of basis states ( unperturbed @xmath9 and @xmath10 states ) lies at high energies relative to the giant resonance . figure [ f : eph2max ] ( see caption for details ) illustrates what would happen if we truncated the model space by setting an upper cutoff @xmath209 to the @xmath10 energies taken into account . had we set , e.g. , @xmath210 mev , we would have excluded 75% of the configurations available in this particular space and the eigenenergies would have been at least 2 mev higher than their converged ( with respect to @xmath209 ) values . inclusion of the energetically lowest 50% of configurations would still not warrant good convergence . we found that the convergence of the eigenenergies and the saturation of the model space follow the same pattern as demonstrated in fig . [ f : eph2max ] , regardless of model - space size and type of response . finally , we should note that most finite @xmath59 elements are small . the distribution of their values is always cusp shaped around zero and a vast majority of them have amplitutudes no larger than 5% the value of the strongest element . it would be computationally economical to neglect those elements . we found , however , that such a procedure influences the position of the resonances as well as the shape of the strength distribution and can not be blindly trusted . the self - consistent rpa that we have used produces a spurious state at practically zero energy , and leaves the rest of the spectrum uncontaminated @xcite , as expected . the srpa dipole spectrum , however , may contain spurious admixture , as discussed in sec . [ s : cons1 ] . in order to quantify this problem , we have examined the is dipole response . we found that relatively strong spurious states appear mostly at about 5 to 8 mev . using a transition operator of the usual radial form ( @xmath211 ) and its uncorrected form ( @xmath212 ) , we found that only the lowest part of the dipole spectrum is strongly affected by the choice of operator , while there are no significant contaminations in the spectrum around and beyond the iv gdr peak . an example is shown in fig . [ f : spurious ] . let us note , that including a small percentage of @xmath10 states would not shift the spurious state far away from zero , but for large spaces such as those used here the effect is noticeable . as we couple the @xmath9 states to more and more @xmath10 configurations , the spurious rpa state moves away from zero and may occur at imaginary , or even negative energies ( or , equivalently , positive energies , but assuming negative norm this is the case in fig . [ f : spurious ] ) . it may also appear fragmented , as spurious admixtures enter the rest of the spectrum , or as additional eigenstates occur in its proximity . as anticipated in sec . [ s : cons1 ] , another problem with the hf - based srpa is the onset of instabilities . we find , in particular , that low - lying states ( @xmath213 @xmath201 states and the collective @xmath214 excitation ) appear at imaginary or negative energies . this is the case , for example , in fig . [ f : full]a : the norm of the collective quadrupole state at @xmath215 mev is , in fact , negative ; the positive - norm counterpart appears at @xmath216 mev . as discussed previously , the problem seems to be the inadequate treatment of ground - state correlations . obviously , srpa is not appropriate for describing low - lying states . the question is then whether it still gives meaningful results for giant resonances . we have tested the sensitivity of our results to gsc with the help of the renormalized srpa devised in sec . [ s : rsrpa ] and of stda ( srpa setting @xmath118 ) and , in first order , the corresponding rrpa and tda . examples are shown in fig . [ f : gsc ] for the octupole response of @xmath197o and table [ t : gsc ] for the quadrupole response of @xmath200ca . in all cases the diagonal approximation has been employed . .isoscalar @xmath201 spectrum of @xmath200ca calculated in a basis of 9 oscillator shells : energy @xmath139 ( in mev ) and strength @xmath53 ( in @xmath204 ) of the two main low - lying states and of the giant resonance peak ( gqr ) within rpa and srpa , renormalized " rpa and srpa ( rrpa , rsrpa see text ) , and tda and stda ( srpa with @xmath118 ) . ( whenever the gqr is split into two or three major peaks , the centroid and total strength is given . ) [ t : gsc ] [ cols="<,^,^,^,^,^,^",options="header " , ] the general trend can be described as follows : we begin with minimal gsc , in stda . inclusion of the @xmath53 matrix and the backward amplitudes ( srpa ) pushes the solutions to somewhat lower energies and in general affects their strength . the downward shift , which can lead to imaginary solutions , was interpreted recently in the framework of random - matrix theory @xcite : the matrix @xmath53 couples the positive - and negative - energy branches of the ( s)rpa solutions and causes an attraction between them . strong enough coupling leads to a merging of the two branches at zero energy and eventually to imaginary or complex solutions . finally , we find that subsequent renormalization of the matrix elements ( rsrpa ) shifts the solutions back to higher energies , i.e. , slightly closer to the rpa solutions , plausibly because of the weaker @xmath217 couplings . the effect of renormalization appears rather small in the case of giant resonances and higher - lying solutions in general , but this is not the case for the low - lying quadrupole and octupole states : their energy shifts in rsrpa , relative to rpa , are noticably more moderate than in srpa . the stda results confirm in general that the lower - lying states are more sensitive to gsc than the higher ones . we note that , since stda is a hermitean problem , it will eliminate the imaginary solutions of srpa by construction see , e.g. , fig . [ f : gsc]a . as we observe in table [ t : gsc ] , however , the same does not hold for the negative solutions . it is thus confirmed that the strong resonance shifts ( with respect to rpa ) in this case are induced by the @xmath59 couplings , as discussed in sec . [ s : cons1 ] . in most cases the rsrpa solutions lie higher than the stda , i.e. , renormalization has a stronger effect on the energies , with the notable exception of the giant monopole resonance ( not shown ) , where the backward amplitudes seem more relevant . finally , regarding our first - order results , renormalization affects the rpa results very weakly . the same conclusion was reached in ref . @xcite , where a more consistent rrpa was applied . the backward amplitudes , though , missing in tda , are found important for the description of low - lying states . the present results suggest that giant resonances ( at least for @xmath218 ) in srpa are only moderately sensitive to the treatment of gsc . of course , a comprehensive inclusion of gsc in the srpa formalism would involve filling up the @xmath141sections of the srpa matrix with many more finite elements . it can not be predicted how strongly those could affect the results . in the applications shown in ref . @xcite , the so - called extended srpa including correlations did not produce strong corrections to the srpa strength functions , especially after renormalizing the ground state @xcite . the present work constitutes a feasibility and justification study of large - scale srpa calculations . the motivation was the prospect of studying nuclear collective excitations using srpa and unitarily transformed nuclear hamiltonians , which have not been fitted to first - order rpa results . the discussion , however , has been kept general . we showed how the large model spaces of srpa can be treated and discussed the salient features of the solutions , including the energetic shift relative to the rpa solutions and the fragmentation of strength . the diagonal approximation was found reliable for soft interactions and for single - particle strength distributions . we found that low - lying states become unstable in srpa , due to an inadequate treatment of ground - state correlations . nevertheless , giant resonances and higher - lying solutions in general do not appear sensitive to ground - state correlations . we have thus concluded that srpa can be applied in the giant - resonance region with reasonable confidence . srpa is primarily the theory of collisional damping . a more comprehensive method to study nuclear collective states should consider also coupling to collective low - lying phonons @xcite . it is not easy to tell at this point how strongly those could enhance or ( partly ) cancel the effect of coupling to the large amount of @xmath10 states considered in srpa . in ref . @xcite it was shown that the resonance shifts due to the @xmath10 states are stronger than those due to collective phonons , although both mechanisms are expected to contribute to the damping width . notwithstanding some particular shortcomings of the present srpa formalism , this work paves the way for systematic studies of giant resonances using finite - range interactions and extended rpa theories without adjustable parameters and arbitrary truncations of the model space . discussions with vladimir ponomarev and jochen wambach are gratefully acknowledged . this work was supported by the deutsche forschungsgemeinschaft through sfb 634 , by the helmholtz international center for fair within the framework of the loewe program launched by the state of hesse , and by the bmbf verbundforschung ( contract 06da90401 ) .

let be an irreducible crystallographic root system with weyl group and coroot lattice , spanning a euclidean space . let be a positive integer and be the arrangement of hyperplanes in of the form for and . it is known that the number of bounded dominant regions of is equal to the number of facets of the positive part of the generalized cluster complex associated to the pair by s. fomin and n. reading . we define a statistic on the set of bounded dominant regions of and conjecture that the corresponding refinement of coincides with the-vector of . we compute these refined numbers for the classical root systems as well as for all root systems when and verify the conjecture when has type , or and when . we give several combinatorial interpretations to these numbers in terms of chains of order ideals in the root poset of , orbits of the action of on the quotient and coroot lattice points inside a certain simplex , analogous to the ones given by the first author in the case of the set of all dominant regions of . we also provide a dual interpretation in terms of order filters in the root poset of in the special case . [ section ] [ theorem]proposition [ theorem]corollary [ theorem]definition [ theorem]remark [ theorem]lemma [ theorem]example [ theorem]examples [ theorem]conjecture [ theorem]question

math0605685

let @xmath3 be an @xmath19-dimensional euclidean space , with inner product @xmath20 . let @xmath0 be a ( finite ) irreducible crystallographic root system spanning @xmath3 and @xmath4 be a fixed nonnegative integer . we denote by @xmath21 the collection of hyperplanes in @xmath3 defined by the affine equations @xmath6 for @xmath22 and @xmath8 , known as the @xmath4th _ extended catalan arrangement _ associated to @xmath0 . thus @xmath21 is invariant under the action of the weyl group @xmath1 associated to @xmath0 and reduces to the coxeter arrangement @xmath23 for @xmath24 . let @xmath25 denote the generalized cluster complex associated to the pair @xmath11 by s. fomin and n. reading @xcite . this is a simplicial complex which reduces to the cluster complex @xmath26 of s. fomin and a. zelevinsky @xcite when @xmath14 . it contains a natural subcomplex , called the positive part of @xmath25 and denoted by @xmath10 , as an induced subcomplex . the complex @xmath25 was also studied independently by the second author @xcite when @xmath0 is of type @xmath15 or @xmath16 ; see section [ pre ] for further information and references . the weyl group @xmath1 acts on the coroot lattice @xmath2 of @xmath0 and its dilate @xmath27 , where @xmath13 denotes the coxeter number of @xmath0 . hence @xmath1 acts also on the quotient @xmath28 . for a fixed choice of a positive system @xmath29 , consider the partial order on @xmath30 defined by letting @xmath31 if @xmath32 is a nonnegative linear combination of positive roots , known as the _ root poset _ of @xmath0 . an _ order filter _ or _ dual order ideal _ in @xmath30 is a subset @xmath33 of @xmath30 such that @xmath34 and @xmath31 in @xmath30 imply @xmath35 . the filter @xmath33 is called _ positive _ if it does not contain any simple root . the following theorem connects the objects just discussed . parts ( i ) , ( ii ) and ( iii ) appear in ( * ? ? ? * corollary 1.3 ) , ( * ? ? ? * proposition 2.13 ) and ( * ? ? ? * theorem 7.4.2 ) , respectively . the last statement was found independently in @xcite . let @xmath0 be an irreducible crystallographic root system of rank @xmath19 with weyl group @xmath1 , coxeter number @xmath13 and exponents @xmath36 . let @xmath4 be a positive integer and let @xmath37 the following are equal to @xmath12 : 1 . the number of bounded regions of @xmath5 which lie in the fundamental chamber of @xmath38 , 2 . the number of facets of @xmath10 and 3 . the number of orbits of the action of @xmath1 on @xmath39 . moreover , for @xmath14 this number is equal to the number of positive filters in the root poset of @xmath0 . [ thm0 ] the purpose of this paper is to define and study a refinement of the number @xmath12 and prove that it has similar properties with the one defined by the first author @xcite for the total number @xmath40 of regions of @xmath5 in the fundamental chamber of @xmath38 . to be more precise let @xmath41 be the affine hyperplane in @xmath3 defined by the equation @xmath6 and @xmath42 be the fundamental alcove of the affine weyl arrangement corresponding to @xmath0 . wall _ of a region @xmath43 of @xmath5 is a hyperplane in @xmath3 which supports a facet of @xmath43 . for @xmath44 we denote by @xmath45 the number of regions @xmath43 of @xmath21 in the fundamental chamber of @xmath38 for which exactly @xmath46 walls of @xmath43 of the form @xmath47 separate @xmath43 from @xmath42 , meaning that @xmath48 holds for @xmath49 . the numbers @xmath45 were introduced and studied in @xcite . let @xmath50 and @xmath51 be the @xmath52th entries of the @xmath13-vector of the simplicial complexes @xmath25 and @xmath10 , respectively . it follows from case by case computations in @xcite that @xmath53 for all @xmath52 when @xmath0 is of classical type in the cartan - killing classification . we define @xmath54 as the number of bounded regions @xmath43 of @xmath21 in the fundamental chamber of @xmath38 for which exactly @xmath46 walls of @xmath43 of the form @xmath47 do not separate @xmath43 from the fundamental alcove @xmath42 . theorem [ thm0 ] implies that the sum of the numbers @xmath54 , as well as that of @xmath55 , for @xmath44 is equal to @xmath12 . the significance of the numbers @xmath54 comes from the following conjecture , which can be viewed as the positive analogue of ( * ? ? ? * conjecture 3.1 ) . for any irreducible crystallographic root system @xmath0 and all @xmath56 and @xmath44 we have @xmath57 . [ conj0 ] our first main result ( corollary [ cor : conj ] ) establishes the previous conjecture when @xmath14 and when @xmath0 has type @xmath15 , @xmath16 or @xmath17 and @xmath4 is arbitrary . our second main result provides combinatorial interpretations to the numbers @xmath54 similar to the ones given in @xcite for @xmath45 . to state this result we need to recall ( or modify ) some definitions and notation from @xcite . for @xmath58 consider the stabilizer of @xmath59 with respect to the @xmath1-action on @xmath60 . this is a subgroup of @xmath1 generated by reflections . the minimum number of reflections needed to generate this subgroup is its _ rank _ and is denoted by @xmath61 . we may use the notation @xmath62 for a @xmath1-orbit @xmath63 in @xmath60 since stabilizers of elements of @xmath60 in the same @xmath1-orbit are conjugate subgroups of @xmath1 and hence have the same rank . a subset @xmath64 of @xmath30 is an _ order ideal _ if @xmath65 is a filter . an increasing chain @xmath66 of ideals in @xmath30 is a _ geometric chain of ideals _ of length @xmath4 if @xmath67 holds for all indices @xmath68 with @xmath69 and @xmath70 holds for all indices @xmath68 , where @xmath71 for @xmath72 and @xmath73 for @xmath74 . such a chain is called _ positive _ if @xmath75 contains the set of simple roots or , equivalently , if @xmath76 is a positive filter . a positive root @xmath77 is _ indecomposable _ of _ rank _ @xmath4 with respect to this increasing chain of ideals if @xmath77 is a maximal element of @xmath78 and it is not possible to write @xmath79 with @xmath80 and @xmath81 for indices @xmath82 with @xmath83 . the following theorem refines part of theorem [ thm0 ] . let @xmath0 be an irreducible crystallographic root system of rank @xmath19 with weyl group @xmath1 , @xmath4 be a positive integer and @xmath84 be the set of orbits of the action of @xmath1 on @xmath85 . for any @xmath44 the following are equal : 1 . the number @xmath86 , 2 . the number of positive geometric chains of ideals in the root poset @xmath30 of length @xmath4 having @xmath52 indecomposable elements of rank @xmath4 , 3 . the number of orbits @xmath87 with @xmath88 and 4 . the number of points in @xmath89 which lie in @xmath52 walls of @xmath90 . in particular , the number of positive geometric chains of ideals in @xmath30 of length @xmath4 is equal to @xmath12 . [ thm1 ] the equivalence of ( iii ) and ( iv ) follows essentially from the results of ( * ? ? ? * section 7.4 ) . in the special case @xmath14 the arrangement @xmath21 consists of the hyperplanes @xmath91 and @xmath92 for all @xmath7 and is known as the _ catalan arrangement _ associated to @xmath0 , denoted @xmath93 . moreover a geometric chain of ideals consists of a single ideal @xmath64 in @xmath0 . this chain is positive if @xmath64 contains the set of simple roots or , equivalently , if @xmath94 is a positive filter and in that case the set of rank one indecomposable elements is the set of maximal elements of @xmath64 . we write @xmath95 instead of @xmath54 when @xmath14 . part of the next corollary is implicit in the work of e. sommers ( * ? ? ? * section 6 ) . let @xmath0 be an irreducible crystallographic root system of rank @xmath19 with weyl group @xmath1 and @xmath96 be the set of orbits of the action of @xmath1 on @xmath97 . for any @xmath98 the following are equal : 1 . the number of ideals in the root poset @xmath30 which contain the set of simple roots and have @xmath52 maximal elements , 2 . the number @xmath99 of bounded regions @xmath43 of @xmath93 in the fundamental chamber of @xmath38 such that @xmath52 walls of @xmath43 of the form @xmath92 do not separate @xmath43 from @xmath42 , 3 . the number of orbits @xmath100 with @xmath88 , 4 . the number of points in @xmath101 which lie in @xmath52 walls of @xmath102 and 5 . the entry @xmath103 of the @xmath13-vector of the positive part of @xmath104 . [ cor1 ] our last theorem provides a different interpretation to the numbers @xmath105 in terms of order filters in @xmath30 . for any irreducible crystallographic root system @xmath0 and any nonnegative integer @xmath52 the number @xmath95 is equal to the number of positive filters in @xmath30 having @xmath52 minimal elements . [ thm2 ] theorem [ thm1 ] is proved in sections [ som ] and [ proof ] by means of two bijections . the first is the restriction of a bijection of ( * ? ? ? * section 3 ) and maps the set of positive geometric chains of ideals in @xmath30 of length @xmath4 to the set of bounded regions of @xmath21 in the fundamental chamber ( theorem [ cor : bij1 ] ) while the second maps this set of regions to the set of @xmath1-orbits of @xmath60 . in the case @xmath14 the composite of these two bijections gives essentially a bijection of sommers @xcite from the set of positive filters in @xmath30 to @xmath106 . the proof of theorem [ thm1 ] in these two sections parallels the one of theorem 1.2 in @xcite and for this reason most of the details are omitted . the main difference is that the unique alcove in a fixed bounded region of @xmath21 which is furthest away from @xmath42 plays the role played in @xcite by the unique alcove in a region of @xmath21 closest to @xmath42 . the existence of these maximal alcoves was first established and exploited in the special case @xmath14 by sommers @xcite . in section [ f ] we prove conjecture [ conj0 ] when @xmath14 and when @xmath0 has type @xmath15 , @xmath16 or @xmath17 and @xmath4 is arbitrary ( corollary [ cor : conj ] ) using the fact that @xmath107 holds for all @xmath52 in these cases . a key ingredient in the proof is a new combinatorial interpretation ( see part ( iii ) of theorems [ thm : eleni1 ] and [ thm : eleni2 ] ) to the @xmath108-numbers defined from the @xmath45 and @xmath54 via the usual identity relating @xmath108-vectors and @xmath13-vectors of simplicial complexes . in section [ class ] we compute the numbers which appear in theorem [ thm1 ] for root systems of classical type and those in corollary [ cor1 ] for root systems of exceptional type . we also prove theorem [ thm2 ] by exploiting the symmetry of the distribution of the set of all filters in @xmath30 by the number of minimal elements , observed by d. panyushev @xcite . some useful background material is summarized in section [ pre ] . we conclude with some remarks in section [ remarks ] . apart from @xcite , our motivation for this work comes to a great extent from the papers by fomin and reading @xcite , fomin and zelevinsky @xcite and sommers @xcite . in this section we introduce notation and terminology and recall a few useful facts related to root systems , affine weyl groups , generalized cluster complexes and the combinatorics of @xmath5 . we refer to @xcite and @xcite for further background and references and warn the reader that , throughout the paper , some of our notation and terminology differs from that employed in @xcite ( this is done in part to ease the co - existence of order filters and order ideals in this paper , typically denoted by the letters @xmath33 and @xmath64 , respectively , and in part to match some of the notation of @xcite ) . * root systems and weyl groups . * let @xmath3 be an @xmath19-dimensional euclidean space with inner product @xmath20 . given a hyperplane arrangement @xmath109 in @xmath3 , meaning a discrete set of affine subspaces of @xmath3 of codimension one , the _ regions _ of @xmath109 are the connected components of the space obtained from @xmath3 by removing the hyperplanes in @xmath109 . let @xmath0 be a crystallographic root system spanning @xmath3 . for any real @xmath110 and @xmath22 we denote by @xmath41 the hyperplane in @xmath3 defined by the equation @xmath111 and set @xmath112 . we fix a positive system @xmath29 and the corresponding ( ordered ) set of simple roots @xmath113 . for @xmath114 we denote by @xmath115 the orthogonal reflection in the hyperpane @xmath116 , called a _ simple reflection_. we will often write @xmath117 instead of @xmath0 , where @xmath118 is an index set in bijection with @xmath119 , and denote by @xmath120 the parabolic root system corresponding to @xmath121 . if @xmath0 is irreducible we denote by @xmath122 the highest positive root , by @xmath123 the exponents and by @xmath13 the coxeter number of @xmath0 and set @xmath124 , where @xmath4 is a fixed positive integer . the following well known lemmas will be used , as in @xcite . ( ( * ? ? ? * lemma 2.1 ) ) ( i ) if @xmath125 with @xmath126 and @xmath127 then there exists @xmath52 with @xmath128 such that @xmath129 . \(ii ) ( cf . @xcite ) if @xmath130 and @xmath131 then @xmath132 or there exists @xmath52 with @xmath133 such that @xmath134 . @xmath135[lem : ro ] ( ( * ? ? ? * ch . 6 , 1.11 , proposition 31 ) @xcite ) if @xmath0 is irreducible and @xmath136 then @xmath137 . [ lem : h ] we denote by @xmath23 the _ coxeter arrangement _ associated to @xmath0 , i.e. the collection of linear hyperplanes @xmath91 in @xmath3 with @xmath22 , and by @xmath1 the corresponding _ weyl group _ , generated by the reflections in these hyperplanes . thus @xmath1 is finite and minimally generated by the set of simple reflections , it leaves @xmath0 invariant and acts simply transitively on the set of regions of @xmath38 , called _ chambers_. the _ fundamental chamber _ is the region defined by the inequalities @xmath138 for @xmath139 . a subset of @xmath3 is called _ dominant _ if it is contained in the fundamental chamber . the _ coroot lattice _ @xmath2 of @xmath0 is the @xmath140-span of the set of coroots @xmath141 > from now on we assume for simplicity that @xmath0 is irreducible . the group @xmath1 acts on the lattice @xmath2 and on its sublattice @xmath142 , hence it also acts on the quotient @xmath143 . we denote by @xmath84 the set of orbits of the @xmath1-action on @xmath144 and use the notation @xmath145 and @xmath146 when @xmath14 . we denote by @xmath147 the _ affine coxeter arrangement _ , which is the infinite hyperplane arrangement in @xmath3 consisting of the hyperplanes @xmath41 for @xmath22 and @xmath148 , and by @xmath149 the _ affine weyl group _ , generated by the reflections in the hyperplanes of @xmath147 . the group @xmath149 is the semidirect product of @xmath1 and the translation group in @xmath3 corresponding to the coroot lattice @xmath2 and is minimally generated by the set @xmath150 of _ simple affine reflections _ , where @xmath151 is the reflection in the hyperplane @xmath152 . for @xmath153 and @xmath44 , the reflection @xmath115 is a _ right ascent _ of @xmath154 if @xmath155 , where @xmath156 is the length of the shortest expression of @xmath154 as a product of simple affine reflections . the group @xmath149 acts simply transitively on the set of regions of @xmath157 , called _ alcoves_. the _ fundamental alcove _ of @xmath157 can be defined as @xmath158 note that every alcove can be written as @xmath159 for a unique @xmath160 . moreover , given @xmath139 , there exists a unique integer @xmath161 , denoted @xmath162 , such that @xmath163 holds for all @xmath164 . the next lemma is a reformulation of the main result of @xcite . * theorem 5.2 ) ) . let @xmath165 be an integer for each @xmath139 . there exists @xmath153 such that @xmath166 for each @xmath167 if and only if @xmath168 for all @xmath169 with @xmath170 . [ lem : alc ] we say that two open regions in @xmath3 are _ separated _ by a hyperplane @xmath171 if they lie in different half - spaces relative to @xmath172 . if @xmath43 is a region of a subarrangement of @xmath157 or the closure of such a region ( in particular , if @xmath43 is a chamber or an alcove ) , we refer to the hyperplanes of @xmath147 which support facets of the closure of @xmath43 as the _ walls _ of @xmath43 . * generalized cluster complexes . * let @xmath0 be crystallographic ( possibly reducible ) of rank @xmath19 . the generalized cluster complex @xmath25 is an abstract simplicial complex on the vertex set @xmath173 consisting of the negative simple roots and @xmath4 copies of each positive root ; we refer to ( * ? ? ? * section 1.2 ) for the definition . it is a pure complex of dimension @xmath174 ( * ? ? ? * proposition 1.7 ) . if @xmath0 is a direct product @xmath175 then @xmath25 is the simplicial join of @xmath176 and @xmath177 . we denote by @xmath10 the induced subcomplex of @xmath25 on the set of vertices obtained from @xmath173 by removing the negative simple roots and call this simplicial complex the _ positive part _ of @xmath25 . for @xmath178 we denote by @xmath179 and @xmath180 the number of @xmath181-dimensional faces of the complex @xmath25 and @xmath10 , respectively . these numbers are related to the @xmath50 and @xmath55 by the equations @xmath182 and @xmath183 respectively . following @xcite we give explicit combinatorial descriptions of the complexes @xmath25 and @xmath10 when @xmath0 has type @xmath15 , @xmath16 or @xmath17 . for @xmath184 let @xmath185 be a convex polygon with @xmath186 vertices . a diagonal of @xmath185 is called _ @xmath4-allowable _ if it divides @xmath185 into two polygons each with number of vertices congruent to 2 mod @xmath4 . vertices of @xmath25 are the @xmath4-allowable diagonals of @xmath185 and faces are the sets of pairwise noncrossing diagonals of this kind . for @xmath187 or @xmath188 let @xmath189 be a centrally symmetric convex polygon with @xmath190 vertices . a vertex of @xmath25 is either a diameter of @xmath189 , i.e. a diagonal connecting antipodal vertices , or a pair of @xmath4-allowable diagonals related by a half - turn about the center of @xmath189 . a set of vertices of @xmath25 forms a face if the diagonals of @xmath189 defining these vertices are pairwise noncrossing . in all cases the explicit bijection of @xmath173 with the set of allowable diagonals of @xmath185 or @xmath189 just described is analogous to the one given in ( * ? ? ? * section 3.5 ) for the usual cluster complex @xmath104 , so that the negative simple roots form an @xmath4-snake of allowable diagonals in @xmath185 or @xmath189 and @xmath10 is the subcomplex of @xmath191 obtained by removing the vertices in the @xmath4-snake . the _ negative part _ of a face @xmath192 of @xmath25 is the set of indices @xmath193 , where @xmath194 , which correspond to the negative simple roots contained in @xmath192 . the next lemma appears as ( * ? ? ? * proposition 3.6 ) in the case @xmath14 and follows from the explicit description of the relevant complexes in the remaining cases . assume that either @xmath14 or @xmath117 has type @xmath195 or @xmath17 . for any @xmath121 the map @xmath196 is a bijection from the set of faces of @xmath197 with negative part @xmath198 to the set of faces of @xmath199 . in particular @xmath200 where @xmath201 if @xmath202 for any complex @xmath203 . @xmath135[lem : fr ] for @xmath14 essentially the same equation as ( [ ff+fr ] ) has appeared in the context of quiver representations in ( * ? ? ? * section 6 ) . * regions of @xmath5 and chains of filters . * let @xmath0 be irreducible and crystallographic of rank @xmath19 . for @xmath44 let @xmath45 be the number of dominant regions @xmath43 of @xmath21 for which exactly @xmath46 walls of @xmath43 of the form @xmath47 separate @xmath43 from @xmath42 , as in section [ intro ] . we recall another combinatorial interpretation of @xmath45 from @xcite using slightly different terminology . we call a decreasing chain @xmath204 of filters in @xmath30 a _ geometric chain of filters _ of length @xmath4 if ( [ bi2 ] ) and ( [ bi1 ] ) hold under the same conventions as in section [ intro ] ( the term _ co - filtered chain of dual order ideals _ was used in @xcite instead ) . a positive root @xmath77 is _ indecomposable _ of _ rank _ @xmath4 with respect to this chain if @xmath205 and it is not possible to write @xmath79 with @xmath206 and @xmath207 for indices @xmath208 with @xmath83 . let @xmath209 be the set of points @xmath210 which satisfy @xmath211 for @xmath212 , where @xmath213 . the following statement combines parts of theorems 3.6 and 3.11 in @xcite . the map @xmath214 is a bijection from the set of geometric chains of filters of length @xmath4 in @xmath30 to the set of dominant regions of @xmath5 . moreover a positive root @xmath77 is indecomposable of rank @xmath4 with respect to @xmath33 if and only if @xmath47 is a wall of @xmath209 which separates @xmath209 from @xmath42 . in particular @xmath45 is equal to the number of geometric chains of filters in @xmath30 of length @xmath4 having @xmath215 indecomposable elements of rank @xmath4 . @xmath135[thm : ath2 ] by modifying the definition given earlier or using the interpretation in the last statement of the previous theorem we can define the numbers @xmath216 when @xmath0 is reducible as well . clearly @xmath217 for any crystallographic root systems @xmath218 . in this section we generalize some of the results of sommers @xcite on bounded dominant regions of @xmath93 and positive filters in @xmath30 to bounded dominant regions of @xmath21 and positive geometric chains of ideals and establish the equality of the numbers appearing in ( i ) and ( ii ) in the statement of theorem [ thm1 ] . the results of this and the following section are analogues of the results of sections 3 and 4 of @xcite on the set of all dominant regions of @xmath21 . their proofs are obtained by minor adjustments from those of @xcite , suggested by the modifications of the relevant definitions , and thus are only sketched or omitted . let @xmath0 be irreducible and crystallographic of rank @xmath19 and let @xmath64 be a positive geometric chain of ideals @xmath219 in @xmath30 of length @xmath4 , so that ( [ bi2 ] ) and ( [ bi1 ] ) hold , where @xmath220 , and @xmath221 . we define @xmath222 for any @xmath139 . observe that @xmath223 is well defined since @xmath221 and that @xmath224 for @xmath225 , with @xmath226 if and only if @xmath227 . if @xmath228 and @xmath229 for all @xmath52 then @xmath230 [ lem0 ] _ proof . _ this is clear from the definition . @xmath135 let @xmath139 and @xmath231 . 1 . if @xmath232 then @xmath225 . if @xmath233 then there exist @xmath234 with @xmath79 and @xmath235 . moreover we may choose @xmath236 so that @xmath237 . [ lem1 ] _ proof . _ analogous to the proof of ( * ? ? ? * lemma 3.2 ) . @xmath135 if @xmath238 and @xmath239 are integers such that @xmath240 then @xmath241 or @xmath242 . [ lem : cp ] _ proof . _ by induction on @xmath243 , as in the proof of ( * ? ? * lemma 3.3 ) . @xmath135 we have @xmath244 whenever @xmath238 . [ cor : k ] _ proof . _ the second inequality is a special case of lemma [ lem0 ] and the first follows from lemma [ lem : cp ] letting @xmath245 and @xmath246 . @xmath135 we denote by @xmath247 the set of points @xmath210 which satisfy the inequalities in ( [ map ] ) ( thus we allow a slight abuse of notation since the same set was denoted by @xmath209 in section [ pre ] , where @xmath33 is the chain of complementary filters @xmath248 ) . since @xmath249 we have @xmath250 for all @xmath114 and @xmath251 and therefore @xmath247 is bounded . there exists a unique @xmath153 such that @xmath252 for @xmath139 . moreover , @xmath253 . in particular , @xmath247 is nonempty . [ prop : min ] _ proof . _ the existence in the first statement follows from lemma [ lem : alc ] and corollary [ cor : k ] while uniqueness is obvious . for the second statement let @xmath139 and @xmath254 . part ( i ) of lemma [ lem1 ] implies that @xmath224 if and only if @xmath255 . hence from the inequalities @xmath256 which hold for @xmath164 , we conclude that @xmath257 . @xmath135 let @xmath258 be the map which assigns the set @xmath247 to a positive geometric chain of ideals @xmath64 in @xmath30 of length @xmath4 . conversely , given a bounded dominant region @xmath43 of @xmath21 let @xmath259 be the sequence @xmath260 where @xmath261 is the set of @xmath167 for which @xmath262 holds in @xmath43 . clearly each @xmath261 is an ideal in @xmath30 . the map @xmath258 is a bijection from the set of positive geometric chains of ideals in @xmath30 of length @xmath4 to the set of bounded dominant regions of @xmath21 , and the map @xmath263 is its inverse . [ cor : bij1 ] _ proof . _ that @xmath258 is well defined follows from proposition [ prop : min ] , which guarantees that @xmath247 is nonempty ( and bounded ) . to check that @xmath263 is well defined observe that if @xmath43 is a bounded dominant region of @xmath21 and if @xmath264 and @xmath265 hold for @xmath49 then @xmath266 must hold for @xmath49 , so that @xmath267 satisfies ( [ bi2 ] ) similarly , @xmath267 satisfies ( [ bi1 ] ) . that @xmath221 follows from ( * ? ? ? * lemma 4.1 ) . it is clear that @xmath258 and @xmath263 are inverses of each other . @xmath135 let @xmath268 be a bounded dominant region of @xmath21 , where @xmath269 . let @xmath270 denote the element of the affine weyl group @xmath149 which is assigned to @xmath64 in proposition [ prop : min ] . the following proposition implies that @xmath271 is the alcove in @xmath43 which is the furthest away from @xmath42 . in the special case @xmath14 the existence of such an alcove was established by sommers ( * ? ? ? * proposition 5.4 ) . let @xmath43 be a bounded dominant region of @xmath21 . the element @xmath270 is the unique @xmath153 such that @xmath272 and whenever @xmath139 , @xmath273 and @xmath274 holds for some @xmath49 we have @xmath275 for all @xmath164 . [ prop : wr ] _ proof . _ analogous to the proof of ( * ? ? ? * proposition 3.7 ) . @xmath135 we now introduce the notion of an indecomposable element with respect to the increasing chain of ideals @xmath64 . given @xmath254 , a root @xmath139 is indecomposable of rank @xmath161 with respect to @xmath64 if @xmath225 and 1 . @xmath231 , 2 . it is not possible to write @xmath79 with @xmath80 and @xmath81 for indices @xmath82 with @xmath276 and 3 . if @xmath277 for some @xmath278 then @xmath279 . [ def ] observe that , by part ( i ) of lemma [ lem1 ] , the assumption @xmath280 in this definition is actually implied by condition ( i ) . for @xmath281 the definition is equivalent to the one proposed in section [ intro ] , as the following lemma shows . a positive root @xmath77 is indecomposable of rank @xmath4 with respect to @xmath64 if and only if @xmath77 is a maximal element of @xmath282 and it is not possible to write @xmath79 with @xmath283 and @xmath81 for indices @xmath82 with @xmath83 . [ lem : m ] _ proof . _ suppose that @xmath284 is indecomposable of rank @xmath4 . since @xmath285 we must have @xmath286 . hence to show that @xmath77 satisfies the condition in the statement of the lemma it suffices to show that @xmath77 is maximal in @xmath75 . if not then by lemma [ lem : ro ] ( i ) there exists @xmath287 such that @xmath288 . then clearly @xmath289 and @xmath290 by corollary [ cor : k ] . hence @xmath291 and condition ( iii ) of definition [ def ] leads to a contradiction . for the converse , suppose that @xmath284 satisfies the condition in the statement of the lemma in view of part ( i ) of lemma [ lem1 ] , condition ( iii ) in definition [ def ] is satisfied since @xmath77 is assumed to be maximal in @xmath75 . hence to show that @xmath77 is indecomposable of rank @xmath4 it suffices to show that @xmath285 . this is implied by the assumption that @xmath292 and part ( i ) of lemma [ lem1 ] . @xmath135 suppose that @xmath77 is indecomposable with respect to @xmath64 . 1 . we have @xmath293 whenever @xmath79 with @xmath234 . we have @xmath294 whenever @xmath295 . [ lem : ind ] _ proof . _ analogous to the proof of ( * ? ? ? * lemma 3.10 ) . for part ( ii ) , letting @xmath231 and @xmath296 , we prove instead that @xmath297 . this implies the result by corollary [ cor : k ] . @xmath135 the following theorem explains the connection between indecomposable elements of @xmath64 and walls of @xmath247 . if @xmath64 is a positive geometric chain of ideals in @xmath30 of length @xmath4 with corresponding region @xmath268 and @xmath254 then the following sets are equal : 1 . the set of indecomposable roots @xmath139 with respect to @xmath64 of rank @xmath161 , 2 . the set of @xmath139 such that @xmath298 is a wall of @xmath43 which does not separate @xmath43 from @xmath42 and 3 . the set of @xmath139 such that @xmath298 is a wall of @xmath271 which does not separate @xmath271 from @xmath42 . [ mylemma ] _ proof . _ we prove that @xmath299 for the three sets defined in the statement of the theorem as in the proof of ( * ? ? ? * theorem 3.11 ) , replacing the inequalities @xmath300 which appear there by @xmath301 and recalling from the proof of proposition [ prop : wr ] that @xmath302 holds for all @xmath139 and @xmath251 . @xmath135 we denote by @xmath303 the subset of @xmath149 consisting of the elements @xmath270 for the bounded dominant regions @xmath43 of @xmath21 ; see figure [ figure1 ] for the case @xmath304 and @xmath305 . we abbreviate this set as @xmath306 in the case @xmath14 . the elements of @xmath306 are called _ maximal _ in @xcite . for any nonnegative integers @xmath307 the following are equal : 1 . the number of positive geometric chains of ideals in @xmath30 of length @xmath4 having @xmath308 indecomposable elements of rank @xmath161 for each @xmath254 , 2 . the number of bounded dominant regions @xmath43 of @xmath21 such that @xmath308 walls of @xmath43 of the form @xmath298 do not separate @xmath43 from @xmath42 for each @xmath254 and 3 . the number of @xmath309 such that @xmath308 walls of @xmath310 of the form @xmath298 do not separate @xmath310 from @xmath42 for each @xmath254 . [ cor : ik ] _ proof . _ combine theorems [ cor : bij1 ] and [ mylemma ] . @xmath135 the following corollary is immediate . for any nonnegative integer @xmath52 the numbers which appear in ( i ) and ( ii ) in the statement of theorem [ thm1 ] are both equal to the number of @xmath309 such that @xmath52 walls of @xmath310 of the form @xmath47 do not separate @xmath310 from @xmath42 . @xmath135[cor : part1 ] as was the case with @xmath45 , the interpretation in part ( ii ) of theorem [ thm1 ] mentioned in the previous corollary or the original definition can be used to define @xmath54 when @xmath0 is reducible . equivalently we define @xmath311 for any crystallographic root systems @xmath218 . we now consider the special case @xmath14 . a positive geometric chain of ideals @xmath64 of length @xmath4 in this case is simply a single ideal @xmath64 in @xmath30 such that @xmath312 , meaning that @xmath313 is a positive filter . by lemma [ lem : m ] the rank one indecomposable elements of @xmath64 are exactly the maximal elements of @xmath64 . for any nonnegative integer @xmath52 the following are equal to @xmath314 : 1 . the number of ideals in the root poset @xmath30 which contain all simple roots and have @xmath52 maximal elements , 2 . the number of bounded dominant regions @xmath43 of @xmath93 such that @xmath52 walls of @xmath43 of the form @xmath92 do not separate @xmath43 from @xmath42 , 3 . the number of @xmath315 such that @xmath52 walls of @xmath310 of the form @xmath92 do not separate @xmath310 from @xmath42 and 4 . the number of elements @xmath315 having @xmath52 right ascents . [ cor : shi ] _ proof . _ this follows from the case @xmath14 of corollary [ cor : ik ] and ( * ? ? ? * lemma 2.5 ) . @xmath135 in this section we complete the proof of theorem [ thm1 ] ( see corollary [ cor : proof ] ) . we assume that @xmath0 is irreducible and crystallographic of rank @xmath19 . as in ( * section 4 ) , by the reflection in @xmath1 corresponding to a hyperplane @xmath41 we mean the reflection in the linear hyperplane @xmath91 . we let @xmath316 , as in section [ pre ] , and @xmath317 . the following elementary lemma , for which a detailed proof can be found in ( * ? ? ? * section 7.4 ) , implies that @xmath318 is a set of representatives for the orbits of the @xmath1-action on @xmath144 . the natural inclusion map from @xmath318 to the set @xmath84 of orbits of the @xmath1-action on @xmath144 is a bijection . moreover , if @xmath319 then the stabilizer of @xmath59 with respect to the @xmath1-action on @xmath144 is the subgroup of @xmath1 generated by the reflections corresponding to the walls of @xmath320 which contain @xmath59 . in particular , @xmath61 is equal to the number of walls of @xmath320 which contain @xmath59 . @xmath135[lem : hai ] we will define a bijection @xmath321 such that for @xmath309 , the number of walls of @xmath310 of the form @xmath47 which do not separate @xmath310 from @xmath42 is equal to the number of walls of @xmath320 which contain @xmath322 . let @xmath323 be the region of @xmath5 defined by the inequalities @xmath324 for @xmath114 . let @xmath325 be the unique element @xmath154 of @xmath303 such that @xmath326 . we define the map @xmath327 by @xmath328 for @xmath309 . observe that , by lemma [ lem : h ] , the alcove @xmath329 can be described explicitly as the open simplex in @xmath3 defined by the linear inequalities @xmath330 for @xmath331 and @xmath332 . for any @xmath254 we define the simplex @xmath333 so that @xmath334 . for any @xmath19-dimensional simplex @xmath335 in @xmath3 bounded by hyperplanes @xmath41 in @xmath147 with @xmath336 we denote by @xmath337 the wall of @xmath335 orthogonal to @xmath122 or @xmath338 , if @xmath339 or @xmath340 , respectively . we write @xmath341 instead of @xmath342 for @xmath153 . the reader is invited to test the results that follow in the case pictured in figure [ figure1 ] . the map @xmath343 is a bijection from @xmath303 to @xmath318 . moreover for any @xmath309 , @xmath254 and @xmath44 , the point @xmath344 lies on the wall @xmath345 if and only if the wall @xmath346 of @xmath310 is of the form @xmath298 and does not separate @xmath310 from @xmath42 . [ bij ] _ proof . _ analogous to the proof of ( * theorem 4.2 ) . @xmath135 = 3.0 in for any nonnegative integers @xmath307 each of the quantities which appear in the statement of corollary [ cor : ik ] is equal to the number of points in @xmath318 which lie in @xmath308 walls of @xmath347 for all @xmath254 . [ ik : new ] _ proof . _ this follows from theorem [ bij ] . @xmath135 the next corollary completes the proof of theorem [ thm1 ] . for any @xmath44 the following are equal to @xmath348 : 1 . the number of points in @xmath318 which lie in @xmath52 walls of @xmath320 and 2 . the number of @xmath309 such that @xmath52 walls of @xmath310 of the form @xmath47 do not separate @xmath310 from @xmath349 . [ cor : proof ] _ proof . _ by lemma [ lem : hai ] , the number of orbits @xmath87 with rank @xmath88 is equal to the number of points in @xmath318 which lie in @xmath52 walls of @xmath320 . the statement now follows by specializing corollary [ ik : new ] and recalling that @xmath334 . @xmath135 part ( ii ) of corollary [ cor : proof ] implies that @xmath350 is equal to the cardinality of @xmath351 . an argument similar to the one employed in ( * ? ? ? * remark 4.5 ) shows that @xmath352 is equinumerous to @xmath353 . therefore from theorem [ thm0 ] ( iii ) we conclude that @xmath354 . since the reduced euler characteristic @xmath355 of @xmath356 is equal to @xmath357 , it follows from the results of the next section that @xmath358 if @xmath0 has type @xmath15 , @xmath16 or @xmath17 ( see ( * ? ? ? * ( 3.1 ) ) for the corresponding property of @xmath191 ) . @xmath135[rem ] the following conjecture is the positive analogue of ( * ? ? ? * conjecture 3.7 ) . for any crystallographic root system @xmath0 and all @xmath56 the complex @xmath10 is pure @xmath359-dimensional and shellable and has reduced euler characteristic equal to @xmath360 . in particular it is cohen - macaulay and has the homotopy type of a wedge of @xmath361 spheres of dimension @xmath174 . let @xmath0 be a crystallographic root system of rank @xmath19 spanning the euclidean space @xmath3 . we define numbers @xmath362 and @xmath364 by the relations @xmath365 and @xmath366 respectively . comparing to equation ( [ f+fr ] ) we see that conjecture [ conj0 ] for the pair @xmath11 is equivalent to the statement that @xmath367 for all @xmath52 , where @xmath180 is as in section [ pre ] . we will give a combinatorial interpretation to the numbers @xmath368 and @xmath369 as follows . for @xmath370 we denote by @xmath371 the collection of @xmath110-dimensional ( nonempty ) sets of the form @xmath372 where @xmath373 can be @xmath374 and @xmath375 and @xmath376 denote the two open half - spaces in @xmath3 defined by the inequalities @xmath262 and @xmath274 , respectively . observe that each element of @xmath371 is dominant . we also denote by @xmath377 the elements of @xmath371 which are bounded subsets of @xmath3 or , equivalently , the sets of the form ( [ cells ] ) with @xmath378 or @xmath379 for @xmath114 . in the special case @xmath14 part ( iii ) of the following theorem is the content of remark 5.10 ( v ) in @xcite . for any irreducible crystallographic root system @xmath0 and all @xmath56 and @xmath370 the number @xmath380 counts 1 . pairs @xmath381 where @xmath43 is a dominant region of @xmath21 and @xmath382 is a set of @xmath383 walls of @xmath43 of the form @xmath47 which separate @xmath43 from @xmath42 , 2 . pairs @xmath384 where @xmath33 is a geometric chain of filters in @xmath30 of length @xmath4 and @xmath385 is a set of @xmath383 indecomposable roots of rank @xmath4 with respect to @xmath33 and 3 . the elements of @xmath371 . [ thm : eleni1 ] _ proof . _ > from ( [ f_i ] ) we have @xmath386 which cleary implies ( i ) and ( ii ) ( see theorem [ thm : ath2 ] ) . to complete the proof it suffices to give a bijection from the set @xmath387 of pairs @xmath381 which appear in ( i ) to @xmath371 . given such a pair @xmath388 let @xmath389 be the intersection ( [ cells ] ) , where @xmath373 is chosen so that @xmath390 unless @xmath281 and @xmath391 , in which case @xmath392 . let @xmath393 and let @xmath394 be the intersection of the hyperplanes in @xmath382 . it follows from ( * ? ? ? * corollary 3.14 ) that @xmath382 is a proper subset of the set of walls of an alcove of @xmath147 and hence that @xmath394 is nonempty and @xmath110-dimensional . to show that @xmath389 is nonempty and @xmath110-dimensional , so that @xmath395 is well defined , we need to show that @xmath394 is not contained in any hyperplane @xmath298 with @xmath139 and @xmath212 other than those in @xmath382 . so suppose that @xmath396 with @xmath139 and @xmath397 . then there are real numbers @xmath398 such that @xmath399 and @xmath400 . observe that the @xmath401 are minimal elements of the last filter in the geometric chain of filters in @xmath30 corresponding to @xmath43 and hence that they form an antichain in @xmath30 , meaning a set of pairwise incomparable elements . it follows from the first main result of @xcite ( see the proof of ( * ? ? ? * corollary 6.2 ) ) that the coefficients @xmath402 in ( [ lambdas ] ) are nonnegative integers . hence either @xmath403 or @xmath404 and @xmath281 for some @xmath52 , so that @xmath405 . to show that @xmath406 is a bijection we will show that given @xmath407 there exists a unique @xmath408 with @xmath409 . let @xmath410 be the linear basis of @xmath3 which is dual to @xmath119 , in the sense that @xmath411 observe that if @xmath412 with @xmath388 , @xmath63 is a point in @xmath413 and @xmath414 are sufficiently small positive numbers then @xmath415 since regions of @xmath5 are pairwise disjoint this implies uniqueness of @xmath43 , and hence of @xmath416 . to prove the existence let @xmath43 be the unique region of @xmath5 defined by ( [ epsilon ] ) . equivalently @xmath43 can be obtained by replacing all hyperplanes of the form @xmath47 in the intersection ( [ cells ] ) defining @xmath413 by @xmath417 . it suffices to show that any such hyperplane @xmath47 is a wall of @xmath43 since then , if @xmath382 is the set of hyperplanes of the form @xmath47 which contain @xmath413 then @xmath418 and @xmath419 . suppose on the contrary that @xmath420 is not a wall of @xmath43 . it follows from theorem [ thm : ath2 ] that @xmath77 is not indecomposable of rank @xmath4 with respect to the geometric chain @xmath33 of filters @xmath421 corresponding to @xmath43 and hence that one can write @xmath422 for some @xmath206 , @xmath207 with @xmath423 . since @xmath424 for @xmath425 and @xmath426 and @xmath427 hold for @xmath49 , so that @xmath428 and @xmath429 hold for @xmath425 , we must have @xmath430 and @xmath431 for @xmath425 . however one of @xmath432 must be less than @xmath4 and this contradicts the fact that @xmath413 can be contained in @xmath298 for @xmath139 only if @xmath281 . @xmath135 the proof of the next theorem is entirely similar to that of theorem [ thm : eleni1 ] and is omitted . for any irreducible crystallographic root system @xmath0 and all @xmath56 and @xmath370 the number @xmath433 counts 1 . pairs @xmath381 where @xmath43 is a dominant bounded region of @xmath21 and @xmath382 is a set of @xmath383 walls of @xmath43 of the form @xmath434 which do not separate @xmath43 from @xmath42 , 2 . pairs @xmath435 where @xmath64 is a positive geometric chain of ideals in @xmath30 of length @xmath4 and @xmath385 is a set of @xmath383 indecomposable roots of rank @xmath4 with respect to @xmath64 and 3 . the elements of @xmath377 . @xmath135 [ thm : eleni2 ] the reader is invited to use part ( iii ) of theorems [ thm : eleni1 ] and [ thm : eleni2 ] as well as figure [ figure1 ] to verify that @xmath436 , @xmath437 , @xmath438 , @xmath439 , @xmath440 and @xmath441 when @xmath304 and @xmath305 . it should be clear that apart from the necessary modifications in the statements of part ( i ) , theorems [ thm : eleni1 ] and [ thm : eleni2 ] are also valid when @xmath0 is reducible . for any crystallographic root system @xmath117 and all @xmath56 and @xmath442 we have @xmath443 [ lem : ff+ ] _ proof . _ for @xmath193 let @xmath444 be the linear span of the simple roots indexed by the elements of @xmath198 and let @xmath445 be the orthogonal projection onto @xmath446 . we define the _ simple part _ of @xmath447 as the set of indices @xmath448 such that @xmath449 . observe that if @xmath198 is the simple part of @xmath413 then for @xmath139 and @xmath425 we have @xmath450 it follows that @xmath451 induces a bijection from the set of elements of @xmath452 with simple part @xmath198 to @xmath453 . hence counting the elements of @xmath454 according to their simple part proves the lemma . @xmath135 the same type of argument as that in the next corollary appears in the proof of ( * ? ? ? * proposition 6.1 ) . if for some pair @xmath11 every parabolic subsystem of @xmath0 satisfies ( [ ff+fr ] ) and we have @xmath107 for all @xmath52 then we have @xmath57 for all @xmath52 as well . [ cor : if ] _ proof . _ > from the assumption we have @xmath455 for all @xmath110 . equation ( [ ff+fr ] ) and lemma [ lem : ff+ ] imply that @xmath456 for all @xmath110 via mbius inversion on the set of pairs @xmath457 partially ordered by letting @xmath458 this is equivalent to the conclusion of the corollary . @xmath135 conjecture [ conj0 ] holds for root systems of type @xmath15 , @xmath16 or @xmath17 and any @xmath459 and for all root systems when @xmath14 . [ cor : conj ] _ proof . _ this follows from lemma [ lem : fr ] , the previous corollary and the fact that the equality @xmath107 can be checked case by case from the explicit formulas given in @xcite in the cases under consideration . we conclude this section with a combinatorial interpretation to @xmath460 similar to those provided in parts ( i ) and ( ii ) of theorem [ thm : eleni2 ] . for any irreducible crystallographic root system @xmath0 and all @xmath56 and @xmath370 the number @xmath433 counts 1 . pairs @xmath381 where @xmath43 is a dominant region of @xmath21 and @xmath382 is a set of @xmath383 walls of @xmath43 of the form @xmath434 which separate @xmath43 from @xmath42 such that @xmath382 contains all such walls of @xmath43 with @xmath461 and 2 . pairs @xmath384 where @xmath33 is a geometric chain of filters in @xmath30 of length @xmath4 and @xmath385 is a set of @xmath383 indecomposable roots of rank @xmath4 with respect to @xmath33 which contains all simple indecomposable roots of rank @xmath4 with respect to @xmath33 . [ thm : eleni3 ] _ proof . _ the sets in ( i ) and ( ii ) are equinumerous by theorem [ thm : ath2 ] . to complete the proof one can argue that the map @xmath406 in the proof of theorem [ thm : eleni1 ] restricts to a bijection from the set in ( i ) to @xmath377 . alternatively , arguing as in the proof of corollary [ cor : if ] , it suffices to show that @xmath462 where @xmath194 and @xmath463 denotes the cardinality of the set of pairs , say @xmath464 , which appears in ( ii ) . let @xmath465 denote the set of pairs defined in ( ii ) of the statement of theorem [ thm : eleni1 ] . for @xmath466 call the set of simple roots which are indecomposable of rank @xmath4 with respect to @xmath33 and are not contained in @xmath385 the _ simple part _ of @xmath384 and for any @xmath193 denote by @xmath467 the order filter of roots @xmath468 for which @xmath469 for some @xmath470 . it is straightforward to check from the definitions that the map which sends a pair @xmath384 to @xmath471 , where @xmath472 denotes the chain obtained from @xmath33 by removing @xmath467 from each filter of @xmath33 , induces a bijection from the set of elements of @xmath465 with simple part @xmath198 to @xmath473 . therefore counting the elements of @xmath474 according to their simple part gives the desired equality . @xmath135 in this section we compute the numbers @xmath54 and @xmath475 in the cases of the classical root systems . the number @xmath54 is equal to lr @xmath476 , & if @xmath184 , + + @xmath477 , & if @xmath187 or @xmath188 , + + @xmath478 , & if @xmath479 . [ thm : abcd ] _ proof . _ the proof can be obtained from that of ( * ? ? ? * proposition 5.1 ) by replacing the quantity @xmath480 by @xmath481 and using theorem 1.3 ( iii ) instead of ( * ? ? ? * theorem 1.2 ( ii ) ) . @xmath135 the following corollary is a straightforward consequence of proposition [ thm : abcd ] and equation ( [ f_i+ ] ) . the number @xmath433 is equal to lr @xmath482 , & if @xmath184 , + + @xmath483 , & if @xmath484 or @xmath188 , + + @xmath485 , & if @xmath479 . @xmath135[cor : abcd ] in the case @xmath14 , the following corollary for @xmath184 and @xmath486 is a special case of ( * ? ? ? * ( 34 ) ) and ( * ? ? ? * ( 46 ) ) , respectively . the number @xmath487 is equal to lr @xmath482 , & if @xmath184 , + + @xmath483 , & if @xmath187 or @xmath188 . moreover @xmath488 [ cor : ab ] _ proof . _ combine corollaries [ cor : conj ] and [ cor : abcd ] . @xmath135 the number of positive filters in @xmath30 with @xmath52 minimal elements has been computed for the exceptional root systems by victor reiner as shown in the following table . .the numbers @xmath95 for the exceptional root systems . [ cols="<,<,<,<,<,<,<,<,<,<",options="header " , ] _ proof of theorem [ thm2 ] . _ we will prove the statement of the theorem without the assumption that @xmath0 is irreducible . let @xmath19 be the rank of @xmath194 . we write @xmath489 instead of @xmath490 , so that @xmath491 counts the filters in @xmath492 with @xmath110 minimal elements as well as the ideals in @xmath492 with @xmath110 maximal elements . let @xmath493 denote the number of positive filters in @xmath30 with @xmath110 minimal elements . counting filters in @xmath492 by the set of simple roots they contain gives @xmath494 similarly , counting ideals in @xmath492 by the set of simple roots they do not contain gives @xmath495 since it is known @xcite that @xmath496 the previous relation can also be written as @xmath497 comparing ( [ dual1 ] ) and ( [ dual2 ] ) and using mbius inversion as in section [ f ] gives @xmath498 for all @xmath193 , which is the statement of the theorem . the following reciprocity relation @xmath499 was observed by fomin and reading ( * ? ? ? * ( 2.12 ) ) . we will show that , as suggested by s. fomin ( private communication ) , this relation is in fact an instance of ehrhart reciprocity . let @xmath500 be the cardinality of @xmath501 for @xmath502 . it is clear that the vertices of the simplex @xmath503 have rational coordinates in the basis @xmath119 of @xmath3 of simple roots , hence also in the basis @xmath504 of @xmath3 . therefore the function @xmath500 is the ehrhart quasi - polynomial of @xmath503 with respect to the lattice @xmath2 ( see ( * ? ? ? * section 4.6 ) for an introduction to the theory of ehrhart quasi - polynomials ) . ehrhart reciprocity ( * ? ? ? * theorem 4.6.26 ) implies that @xmath505 for @xmath502 and hence , setting @xmath506 and consulting ( * ? ? ? * theorem 1.1 ) , that @xmath507 remark [ rem ] asserts that @xmath508 and hence ( [ rec ] ) holds . \3 . after this paper was completed the following came to our attention . ( i ) the numbers @xmath509 of positive filters in @xmath30 with @xmath52 minimal elements are also discussed and partially computed in ( * ? ? ? * section 3 ) . ( ii ) theorem 2.7 in @xcite implies that lemma [ lem : fr ] is valid for all pairs @xmath11 . in view of ( ii ) and the equality @xmath107 ( see @xcite ) when @xmath479 , it follows from corollary [ cor : if ] that conjecture [ conj0 ] is also valid for root systems of type @xmath510 and arbitrary @xmath4 . _ acknowledgements_. we are grateful to sergey fomin and nathan reading for making their work @xcite available to us . we also thank sergey fomin for useful discussions and victor reiner for providing the data of table [ except ] as well as the software to confirm it . athanasiadis , _ generalized catalan numbers , weyl groups and arrangements of hyperplanes _ london math . * 36 * ( 2004 ) , 294302 . athanasiadis , _ on a refinement of the generalized catalan numbers for weyl groups _ , * 357 * ( 2005 ) , 179196 . athanasiadis and v. reiner , _ noncrossing partitions for the group @xmath511 _ , siam j. discrete math . * 18 * ( 2004 ) , 397417 . n. bourbaki , lie groups and lie algebras , chapters 46 , springer - verlag , berlin , 2002 . f. chapoton , _ enumerative properties of generalized associahedra _ , smin . loth . de combinatoire * 51 * ( 2004 ) , article b51b , 16pp ( electronic ) . s. fomin and n. reading , _ root systems and generalized associahedra _ , in _ geometric combinatorics _ , ias / park city mathematics series ( to appear ) . s. fomin and n. reading , _ generalized cluster complexes and coxeter combinatorics _ , incomplete preliminary draft of january 6 , 2005 , 30pp . s. fomin and n. reading , _ generalized cluster complexes and coxeter combinatorics _ , int . math res . not . * 44 * ( 2005 ) , 27092757 . s. fomin and a.v . zelevinsky , _ @xmath512-systems and generalized associahedra _ , ann . of math . * 158 * ( 2003 ) , 9771018 . haiman , _ conjectures on the quotient ring by diagonal invariants _ , j. algebraic combin . * 3 * ( 1994 ) , 1776 . humphreys , reflection groups and coxeter groups , cambridge studies in advanced mathematics * 29 * , cambridge university press , cambridge , england , 1990 . r. marsh , m. reineke and a. zelevinsky , _ generalized associahedra via quiver representations _ , trans . * 355 * ( 2003 ) , 41714186 . panyushev , _ ad - nilpotent ideals of a borel subalgebra : generators and duality _ , j. algebra * 274 * ( 2004 ) , 822846 . d.i . panyushev , _ two covering polynomials of a finite poset , with applications to root systems and ad - nilpotent ideals _ , preprint , 2005 , 20pp ( arxiv preprint math.co/0502386 ) . shi , _ alcoves corresponding to an affine weyl group _ , j. london math . * 35 * ( 1987 ) , 4255 . e. sommers , _ @xmath16-stable ideals in the nilradical of a borel subalgebra _ , canad . math . bull . * 48 * ( 2005 ) , 460472 . stanley , enumerative combinatorics , vol . 1 , wadsworth & brooks / cole , pacific grove , ca , 1986 ; second printing , cambridge university press , cambridge , 1997 . e. tzanaki , _ polygon dissections and some generalizations of cluster complexes _ , j. combin . theory series a ( to appear ) .

we present a numerical work aiming at the computation of excised initial data for black hole spacetimes in full general relativity , using dirac gauge in the context of a constrained formalism for the einstein equations . introducing the isolated horizon formalism for black hole excision , we especially solve the conformal metric part of the equations , and assess the boundary condition problem for it . in the stationary single black hole case , we present and justify a no - boundary treatment on the black hole horizon . we compare the data obtained with the well - known analytic kerr solution in kerr - schild coordinates , and assess the widely used conformally flat approximation for simulating axisymmetric black hole spacetimes . our method shows good concordance on physical and geometrical issues , with the particular application of the isolated horizon multipolar analysis to confirm that the solution obtained is indeed the kerr spacetime . finally , we discuss a previous suggestion in the literature for the boundary conditions for the conformal geometry on the horizon .

0901.2052

trying to accurately describe black holes solutions as evolving physical objects in numerical simulations is of direct interest in astrophysics . numerical simulations have made a great leap forward in the past few years , mainly with the first stable simulations of black hole mergers in full general relativity by pretorius @xcite , campanelli et al . @xcite , baker et al . @xcite , and a few other groups ( see @xcite for a review ) . several of these simulations model black holes in their equations by punctures . these punctures basically change the topology of spacetime to handle evolution of singular objects ( see @xcite for the first proposal of this method ) . notable exceptions to this are references @xcite , that use an excision approach for black hole evolution : the simulations only evolve the spacetime outside two - spheres that are supposed to encircle the black hole singularities . in the same spirit , we are here trying to describe black holes as physical objects characterized by their horizons . defining the physical laws for event horizons of black holes has been notably done by @xcite , @xcite and @xcite in a so - called _ membrane paradigm_. the aim was to describe them as fluid - like two - membranes with physical properties . however , applying evolution laws to event horizons is problematic due to their teleological ( non causal ) behavior ( see for example @xcite for a description of the phenomenon ) . being defined as a global property of spacetime , a local notion of causality actually does not apply to event horizons . alternative local characterizations have been formulated in the past 15 years by @xcite , @xcite and @xcite . they are based on the concept of trapped surfaces , dating back to penrose s singularity theorem @xcite . defined locally , those objects behave in a causal way in a general dynamical context , with local evolution laws following from the projection of einstein equations , e.g the navier - stokes @xcite `` fluid bubble '' analogy or the area evolution law in @xcite . for this work we shall use local characterizations for isolated horizons , prescribing the physics of non - evolving black hole horizons . following the prescriptions of @xcite , @xcite , @xcite and pursuing the numerical explorations of @xcite and @xcite among others , we try to numerically implement those objects as boundary conditions imposed on the ( 3 + 1 ) form of einstein equations , in a three - slice excised by a two - surface ( single black hole case ) . this is done here using the fully constrained formalism ( fcf ) of @xcite ( see also @xcite ) , with maximal slicing and dirac gauge , based on the ( 3 + 1 ) formalism , and with spectral methods - based numerical resolution using the lorene library @xcite . an important point here is that we drop out the usual conformal flatness hypothesis and solve for the conformal geometry , so that we can exactly recover a slice of a stationary rotating vacuum spacetime . contrary to free evolution schemes , which are the most used prescription for ( 3 + 1 ) simulations in numerical relativity , an important feature of constrained schemes is the necessity to solve constraints on each three - slice , in the form of elliptic equations . these equations generally need additional conditions to be imposed on grid boundaries , following reasonable geometrical and physical prescriptions . our approach particularly requires a specific handling of boundary conditions for the two dynamical gravitational degrees of freedom . this is a crucial point of our calculation ; we will justify and apply here a no - boundary treatment for these quantities . the paper is organized as follows : we first review in sec . [ s : isol_hor ] fundamental geometrical properties associated with isolated horizons in general relativity . in sec . [ s : fcf ] , we quickly give the basic features of the fully constrained formalism and the methods we use to treat the conformal part . sec . [ s : bcs ] discusses implementation of boundary conditions for the system of equations , and specifically discusses the conformal metric part . [ s : resus ] gives the numerical results obtained , and confronts them to a battery of tests characterizing the physics of the solutions . a discussion follows in sec . [ s : discus ] , in regard of previous works concerning the computation of the conformal part in the black hole initial data problem . we also raise the question of applicability of this scheme to other more general astrophysical cases . throughout all this paper , greek letters will denote indices spanning from 0 to 3 , latin indices from @xmath0 to @xmath1 shall denote indices from @xmath2 , and indices from @xmath3 to @xmath4 have the range of @xmath5 . all formulae and values are given in geometrical units ( @xmath6 ) . we also use the einstein summation convention . the concept of a trapped surface in a lorentzian frame has been first defined by penrose in 1965 @xcite in connection with the singularity theorems . it relies on the notion of expansion of the light rays emitted from a surface , that we explain here . we start by a closed spacelike two - surface @xmath7 embedded in spacetime , topologically related to a two - sphere . we assign to it a two - metric @xmath8 induced by the ambient four - metric @xmath9 , and its associated area form @xmath10 . two future null directions orthogonal to @xmath7 are associated with this surface . representative vector fields are denoted @xmath11 and @xmath12 , being respectively oriented outwards and inwards . we can for example assume that our spacetime is asymptotically flat , so the orientation can be defined without ambiguity . the expansion @xmath13 of @xmath7 along @xmath11 is the area rate of change along this vector : @xmath14 . @xmath15 is the lie derivative , here along the vector @xmath11 . same definition goes for the vector field @xmath12 . for a two - sphere embedded in a minkowski spacetime ( flat metric ) , we have typically @xmath16 and @xmath17 . @xmath7 is said to be a trapped surface if both expansions are negative or zero : @xmath18 , @xmath19 . this clearly characterizes strong local curvature . a marginally ( outer ) trapped surface will be characterized by @xmath20 and @xmath21 . two theorems make the connection between those objects and black holes : provided the weak energy condition holds , the singularity theorem of penrose @xcite ensures that a spacetime containing a trapped surface necessarily contains a singularity in its future . following this result , provided the cosmic censorship holds , another result by hawking and ellis @xcite conveys that a spacetime containing a trapped surface necessarily contains a black hole region enclosing this surface . marginally outer trapped surfaces ( mots ) are intended as models for the black hole boundary ( see @xcite for a discussion of its relation with the boundary of the black hole trapped region ) . local horizons ( trapping horizons for @xcite , isolated and dynamical horizons in @xcite , @xcite ) are defined as three - dimensional tubes `` sliced '' by mots , with additional geometrical properties . the isolated horizon case is detailed below ; for a review , see @xcite . in vacuum stationary spacetimes , all horizons are at the same location , which is also the location of the event and apparent horizons ( constructed with outermost motss ) . in the more general case , and assuming cosmic censorship , local horizons are always situated inside the event horizon in general relativity . the notion of isolated horizon is aimed at describing stationary black holes . it is based on the notion of non - expanding horizons , and defined as a three - dimensional tube @xmath22 foliated by mots , and with a null vector field @xmath11 as generator . the three - metric induced on the tube has then a signature @xmath23 . we also define a ( 3 + 1 ) spatial slicing for our spacetime , and @xmath7 a two - slice of our isolated horizon at a certain value of the time parameter @xmath24 . the spacelike two - metric on @xmath7 is denoted @xmath8 . the shear tensor @xmath25 on the two - surface is defined along @xmath11 as @xmath26 . \label{e : shear}\ ] ] using the fact that @xmath27 and the dominant energy condition , the raychaudhuri equation for null tubes @xcite ensures that both the shear along @xmath11 and the energy - momentum tensor projected on @xmath11 , and evaluated on the surface must vanish : @xmath28 and @xmath29 . these are additional properties constraining the geometry of the horizon . an isolated horizon is also required to be such that the extrinsic geometry of the tube is not evolving along the null generators : @xmath30 = 0 $ ] , where @xmath31 is the connection on the tube induced from the ambient spacetime connection . if this last condition is dropped , we only retrieve a non - expanding horizon ( linked to the notion of `` perfect horizon '' @xcite ) . the isolated horizon formalism has already been studied extensively in numerics , as a diagnosis for simulations involving black holes , where marginally trapped surfaces are found _ a posteriori _ with numerical tools called apparent horizon finders ( see for example @xcite , @xcite , @xcite , @xcite ) . let us note that very often , apparent horizon finders actually locate mots on the three - slice considered ( not necessarily outermost ones ) . a thorough study of geometrical properties of isolated horizons located _ a posteriori _ can be found in @xcite . in the present paper we employ isolated horizons as an _ a priori _ ingredient in the numerical construction of cauchy initial data for black hole spacetimes . more precisely , we impose conditions on the excised surface characterizing it as the slice of a non - expanding horizon ( see below ) . this approach to the modeling a black hole horizon in instantaneous equilibrium has been investigated in @xcite , where a prescription for the conformal metric is assumed . the main feature of this work is the inclusion of the conformal metric in the discussion , not through analytical prescriptions , but indeed by numerical calculation . this problem has also been recently addressed in @xcite . in sec . [ s : bcs ] we tackle the description of our special treatment of the conformal metric on the excised boundary and compare with previous results , in particular through the numerical recovery of excised kerr initial data . all the following is a summary of the physical and technical assumptions set in @xcite . for a review of ( 3 + 1 ) formalism in numerical relativity , the reader is referred to @xcite , or more recent reviews like @xcite and @xcite . we consider an asymptotically flat , globally hyperbolic four - dimensional manifold @xmath32 , associated with a metric @xmath33 of lorentzian signature @xmath34 . we define on @xmath32 a slicing by spacelike hypersurfaces @xmath35 , labeled by a timelike scalar field @xmath24 ; in this way , the four - metric can be written in its usual ( 3 + 1 ) form : @xmath36 here @xmath37 and @xmath38 are the usual lapse scalar field and shift vector field . @xmath39 is the spacelike three - metric induced on @xmath35 . we also define the second fundamental form of @xmath35 , or extrinsic curvature tensor , as : @xmath40 with @xmath41 the future - directed vector field normal to @xmath35 . writing the vacuum einstein equations with this formalism , one comes up with the classical ( 3 + 1 ) vacuum einstein equations system(see for example @xcite ) : @xmath42 @xmath43 and @xmath44 being , respectively , the connection and the ricci tensor associated with the three - metric @xmath45 . quantities without indices represent tensorial traces . these equations are referred to respectively as the hamiltonian constraint , momentum constraint and evolution equations . now we must choose a set of variables and a gauge , to get a partial differential equations system that we solve numerically . the first ingredient in the formalism presented in @xcite is the conformal decomposition of the three - metric @xcite . we define on each slice @xmath46 an extra metric noted @xmath47 , that will have a vanishing riemann tensor ( flat metric ) and will be time independent . the existence of such a metric in a neighborhood of spatial infinity is ensured by our sub - manifold being asymptotically flat . the associated flat connection is noted @xmath48 . we introduce in @xmath49 a conformal metric such that its determinant coincides with that of @xmath47 , as : @xmath50 the tensor field @xmath51 we use to encode the conformal degrees of freedom is the deviation of the conformal metric from the flat one : @xmath52 we also define in our equation sources the following conformal traceless extrinsic curvature : @xmath53 we choose for a gauge the generalized dirac gauge for the conformal metric : @xmath54 and we add to this prescription the maximal slicing condition , i.e the vanishing of the trace in the extrinsic curvature : @xmath55 . therefore , @xmath56 contains all the information about extrinsic geometry . under those conditions , we can rewrite the ( 3 + 1 ) einstein equations in what we shall call the fcf system : @xmath57 @xmath58 is the usual scalar flat laplacian ( which expression from a spectral point of view is recalled in the appendix [ s : appendix ] ) . the actual sources @xmath59 , in general non - linear in the variables and time - dependent , can be retrieved by the reader from @xcite . we must supplement this system with the kinematical relation between the three - metric and extrinsic curvature of the slice , deduced from ( [ e : kij ] ) and ( [ e : aij ] ) ( see equation ( 92 ) of @xcite ) . this fully constrained scheme is strictly equivalent to the one presented in @xcite . a slightly different version has been presented recently in @xcite , focusing on non - uniqueness issues . although the scheme in @xcite would probably pose no additional difficulty in the present setting ( except maybe some more boundary conditions to prescribe to additional variables ) , there has been no significant indication of problems involving non - uniqueness of solutions in our study , that suggested modifications of the original formalism . here we choose as variables the quantities @xmath60 , @xmath61 , @xmath38 and @xmath62 . we especially come up with three elliptic equations , two scalar and one vectorial . those are derived directly from the hamiltonian and momentum constraints of the ( 3 + 1 ) system , together with the trace part of the dynamical equations . in an evolution scheme , these will be the conditions enforced at each value of time @xmath24 . we do this for one particular slice . we are then left in general with a second - order tensorial hyperbolic equation ( [ e : fcf ] ) dealing with the variable @xmath62 , that is obtained by the geometrical relation between @xmath45 and @xmath63 , and the dynamical part of einstein equations . the goal here is to simulate as accurately as possible stationary spacetimes containing one black hole , represented by an isolated horizon . in this respect , we shall assume a coordinate system that is adapted to stationarity . this will mean that a stationary timelike killing vector field will be identified with our time evolution vector field @xmath64 . using this prescription , all the time derivatives in our equations vanish , so that our sources and operators simplify somewhat . in particular , the tensorial equation is written using an ellipticlike operator acting on @xmath62 : @xmath65 our problem is then totally equivalent to an actual initial data problem , where quantities have to be determined on a three - slice by elliptic equations , before evolving them . the main difference with classical initial data schemes like the conformal transverse traceless ( ctt ) , the extended conformal thin sandwich ( xcts ) scheme or the conformal flat curvature ( cfc ) system , is an additional elliptic equation for the conformal geometry of the three - slice . up to now , a vast majority of initial data computations have been done using an _ ad hoc _ prescription for the conformal geometry . the most common one is the conformally flat approach , where @xmath66 is simply approximated to be the 3d flat metric . this has been done in numerous computations , and this type of initial data is the most frequently used for black hole evolution simulations . however , though this conformally flat approximation turns out to be well - behaved in most cases , we know that it is a strong limitation when trying to compute stationary black hole spacetimes : it has been proven that the rotating kerr - newman spacetime does not admit any conformally flat slice ( see @xcite ) . other prescriptions for the conformal geometry include data suggested by the post - newtonian formalism @xcite , or superposition of additional gravitational wave content ( see @xcite ) . let us mention the work of @xcite for neutron - star binary initial data , which also computes the conformal geometry using a prescription in @xcite , that considers as well the dynamical einstein equations for the conformal variables . finally , the exact scheme we have explicated above has been applied by one of the authors in the case of a single rotating neutron star in equilibrium @xcite . it has led to the computation of strictly stationary initial data , that can be directly extended into future and past time directions . this is exactly what we are trying to do here in the black hole case . apart from the boundary condition problem ( that we discuss in sec . [ s : bcs ] ) , our approach for the resolution of the tensorial equation presents some peculiarities that we explain here . the system of equations is composed of equation ( [ e : dh1 ] ) and the gauge condition : @xmath67 that we supplement with a condition on the determinant of @xmath68 , following from our definition of the conformal factor : @xmath69 we are left with a tensorial equation for a symmetric tensor with four constraints : the system has two degrees of freedom . we now try to make them explicit and solve for the related variables . any second - rank symmetric tensor @xmath62 can be decomposed in the following way into a divergence free part , and a symmetrized gradient part : @xmath70 with @xmath71 . we shall use here variables associated only with the divergence - free part @xmath72 , meaning that the gauge component ( gradient part ) of the tensor considered has no influence on them . we choose to encode the information in @xmath72 in the two scalar spectral potentials @xmath73 and @xmath74 presented in @xcite , and whose definitions are quickly recalled in the appendix [ s : appendix ] . ( a more extensive study shall be performed in @xcite ) . what is remarkable about quantities @xmath73 and @xmath74 is that they can actually be decomposed into scalar spherical harmonics , and that the tensorial poisson equation @xmath75 decouples into scalar elliptic equations @xmath73 and @xmath74 ( see the appendix [ s : appendix ] ) . this is not exactly the case for the nonlinear modified elliptic operator ( [ e : probncf ] ) ; however , in our numerical scheme , we just slightly modify the sources of the equation at each iteration so that we can write : @xmath76 the elliptic operator @xmath77 being defined in the appendix [ s : appendix ] . we keep the lie derivative notation for scalar fields , to show that this is directly related to the operator in eq . ( [ e : dh1 ] ) ; of course , in the scalar case , this operator simply reduces to @xmath78 . during the iteration of the resolution algorithm , the sources of the equations are updated so that they stay coherent with the original equation in @xmath62 . equations ( [ e : eqa1 ] ) and ( [ e : eqb1 ] ) are the two elliptic equations that we solve at each iteration . specifically , at each step , we proceed as follows : once the scalars @xmath73 and @xmath74 are determined by the resolution of ( [ e : eqa1 ] ) and ( [ e : eqb1 ] ) , the dirac gauge and unit determinant conditions allow us to totally reconstruct a divergence - free tensor , as the expected solution of our tensorial equation . this is done by inverting two differential systems ( ( [ e : diracmag ] ) and ( [ e : diracel ] ) of the appendix [ s : appendixb ] ) , that express the dirac gauge conditions and definitions of the scalars @xmath73 and @xmath74 in function of the tensor components . those differential systems involve scalars , which are components of @xmath62 in a tensor spherical harmonics basis ( see the appendix [ s : appendix ] , and sec . v of @xcite ) . the differential systems require three boundary conditions on the excised surface , to be inverted ( see @xcite ) ; we discuss them in sec . [ s : bcs ] , in a detailed description of the scheme . in addition , the trace of our tensor with respect to the flat metric is kept fixed , so that the calculated determinant at this step is one . this reduces to an algebraic nonlinear condition for the tensor components . finally , we update our sources for the next step . we note here that the resolution for the variable @xmath79 introduced in the appendix is not necessary in this scheme : for a divergence - free tensor , @xmath79 is unambiguously determined by the knowledge of @xmath74 and the trace . with this tensorial scheme , the gauge is necessarily enforced by construction , so no gauge - violating mode can occur . this is in the same spirit as the global fully constrained formalism ( equations ( [ e : fcf1]-[e : fcf2 ] ) ) for our equation system , that forbids _ a priori _ all constraint - violating modes . we also emphasize the fact that , in the general case and with an arbitrary source for ( [ e : dh1 ] ) , we do not recover an actual solution of the equation by reconstructing our tensor this way . this is only true if the elliptic equation admits a solution that actually satisfies the dirac gauge and the determinant condition . we can see it as an integrability condition for our equation , that is for example not generically true during an iteration . however , since in our case we are looking for stationary axisymmetric data for a single black hole , we know that our entire system does admit a solution : it is the kerr - newman spacetime in dirac gauge . as a consequence , if our scheme converges , we know that the tensor field @xmath62 we obtain shall satisfy the dynamical einstein equations , thus equation ( [ e : probncf ] ) . the missing ingredient for solving all our system of equations in an excised spacetime is the setting of boundary conditions for our partial differential equations , following part of the geometrical prescriptions of the isolated horizon formalism , namely non - expanding horizon boundary conditions . besides the prescription of asymptotic flatness at infinity and the bulk stationarity prescription , all the physics of our system will be contained in the boundary conditions we shall put on our excised surface . this section follows largely the prescriptions of @xcite . we consider our excised two - surface to be a slice of an non - expanding horizon , i.e. a mots with vanishing outgoing shear . following sec . [ s : isol_hor ] , this translates into several geometrical prescriptions , namely the vanishing of the outgoing expansion and the shear two - tensor : @xmath80 and @xmath81 . being an instantaneous non - expanding horizon , the evolution of the excision surface will be a null tube . since we are adapting our coordinates to stationarity , another important condition on the excised boundary consists in prescribing the time evolution vector field of our coordinates to be tangent to the null tube . thus , we are ensured that our horizon location stays instantaneously fixed during an evolution . those prescriptions on the horizon will suffice to give four boundary conditions for the constraint equations ( one is scalar and the other vectorial ) , as we see below . we certainly have freedom to prescribe the coordinate location of our excision surface in our coordinate system . for simplicity , we choose the surface to be a coordinate sphere , fixed at a radius @xmath82 . we shall denote by @xmath83 the unit outer spacelike normal to the surface , that will be tangent to the three - slice @xmath35 . the shift vector is then decomposed into two orthogonal parts adapted to the geometry of the horizon : @xmath84 . the vanishing of the expansion can be expressed as a condition for the conformal factor on the horizon : @xmath85 where we have used the conformal rescaling @xmath86 , and the notation @xmath87 for the connection associated with the conformal three - metric . multiplying ( [ e : thetaeq0 ] ) by @xmath61 , it can be seen as a non - linear robin condition for the quantity @xmath61 . the requirement for the time evolution vector field on the horizon to be tangent to the null tube provides the equality @xmath88 . this is a natural way to fix the component of the shift normal to the two - sphere . we must also fix @xmath89 , the part of the shift tangent to the two - surface . for this we make use of the vanishing of the symmetric shear tensor @xmath25 . it can be shown ( @xcite and @xcite ) that the vanishing of the shear is equivalent to the following equation for @xmath89 : @xmath90 here @xmath91 is the connection associated with @xmath8 on the surface . this means that @xmath89 is a conformal killing symmetry for the two - sphere ( in particular , quantities in ( [ e : sheareq0 ] ) can be substituted by _ tilded _ conformal ones ) . defining coordinates @xmath92 on our two - sphere , we prescribe @xmath93 as : @xmath94 and we shall verify _ a posteriori _ that this is a ( conformal ) axial symmetry . the constant @xmath95 will be called the rotation rate of the horizon , @xmath96 being the azimuthal coordinate . in the case of the kerr spacetime , there is an analytical relation between the areal radius of the apparent horizon , the ( reduced ) angular momentum parameter @xmath97 , and @xmath95 . from a more general point of view , different values for @xmath95 will likely affect directly the angular momentum . in the general case , we define a parameter @xmath3 for the angular momentum associated with the entire spacetime , from the dimensionless relation : @xmath98 with @xmath99 the adm mass of the 3-slice , and @xmath100 the komar angular momentum of the 3-slice at infinity ; the latter is tentatively defined with the ( presumably ) killing vector @xmath101 ( see equations ( 7.14 ) and ( 7.104 ) in @xcite for explicit expressions for @xmath99 and @xmath100 ) . note that we do not impose any killing symmetry , except on the horizon : we know however , by the black hole rigidity theorem @xcite , that an accurate resolution of einstein equations would impose this vector to be so . we discuss the dependence between all those quantities in sec . [ s : resus ] . once we have set boundary conditions for the conformal factor and the three components of the shift vector , we must still fix the lapse function on the horizon . different prescriptions have been considered in the literature ( e.g. @xcite ) . in the spirit of the effective approach in @xcite , we arbitrarily impose the value of the lapse to be a constant @xmath102 on the excised sphere . as mentioned before , previous boundary conditions define with no ambiguity our excised surface to be a slice of a non - expanding horizon . moreover , our choice for the lapse , the horizon location and the conformal killing symmetry on the horizon fixes coordinates on the two - surface . only the conformal two - geometry of the excised sphere remains to be fixed . this is done in relation with the resolution scheme for the @xmath62 equation . we recall that , with the approach developed in sec . [ s : fcf ] , the resolution of our tensorial problem in dirac gauge reduces to two elliptic - like scalar equations , to be solved on a three - slice excised by a two - sphere : we should normally provide two additional boundary conditions for those equations . a result by @xcite shows that in the full evolution case for this tensorial equation ( equation ( [ e : fcf ] ) ) and in a dirac - like gauge , the characteristics of the equation are not entering the resolution domain when the spacetime is excised by a null or spacelike marginally trapped tube . this means that in the evolution case , once the initial data are set , there is no boundary condition whatsoever to prescribe to the hyperbolic equation . the problem is of course different here , where we are left with an elliptic equation instead of a hyperbolic one . however , a simple analysis will hint that in our particular single horizon case , there will not be any inner boundary condition to be prescribed on our data . let us examine the case of the elliptic equation in @xmath73 , that we recall here : @xmath103 we will try and exhibit a simplified linear operator acting on the variable @xmath73 , that will contain the most relevant terms . the double lie derivative operator acting on @xmath73 can be separated in : @xmath104 the second term contains all the remaining components of the double lie derivative . at this point , and with a fixed system of spherical coordinates , we are allowed to make a decomposition into spherical harmonics for all the scalar variables . we write in this respect : @xmath105 where @xmath106 are the spherical harmonics of order @xmath107 , defined as eigenfunctions of the angular laplace operator : @xmath108 . we now point out the fact that , due to our coordinate choice , we have @xmath109 on the horizon . we can use a second - order taylor expansion to write the @xmath110 part of the factor in front of the first term of ( [ e : seplie ] ) , close to our surface coordinate radius @xmath82 : @xmath111_{(l=0)}\partial_{r}^{2}a = [ 1 + \alpha(r -r_{\mathcal{h } } ) + \nonumber \\ \delta(r - r_{\mathcal{h}})^{2 } + \mathcal{o}(r - r_{\mathcal{h}})^{3}]\partial_{r}^{2}a,\end{aligned}\ ] ] where @xmath112 and @xmath113 are two real numbers that can be directly computed during one iteration , from the values of @xmath37 , @xmath61 and @xmath114 at the excised surface . our global equation can be rewritten for each spherical harmonic @xmath115 as : @xmath116 \frac{\partial^{2}}{\partial r^{2}}a_{l m } + \frac{2}{r } \frac{\partial}{\partial r}a_{l m } \nonumber \\ -\frac{l(l+1)}{r^{2}}a_{l m } = a_{\mathcal{s}}+ \frac { \psi^{4}}{n^{2 } } ( \mathcal{l}_{\beta } \mathcal{l}_{\beta}a)^{**}_{l m},\end{aligned}\ ] ] where we only keep on the left - hand side the terms given in ( [ e : opmoda ] ) , and put the rest ( denoted with @xmath117 ) with the source . the latter contains the remaining components of the double lie derivative , and involves either terms that are not second - order in the radial derivative , or that are multiplied by the higher harmonics of @xmath118 ( supposedly smaller than the main term , explicitly developed in ( [ e : opmodalm ] ) ) . thus , we have isolated a linear operator @xmath119 , depending on two real numbers @xmath112 and @xmath113 : @xmath120 \frac{\partial^{2}}{\partial r^{2 } } \nonumber \\ & + & \frac{2}{r } \frac{\partial}{\partial r } -\frac{l ( l + 1)}{r^{2}}i,\end{aligned}\ ] ] other contributions being taken as source terms . this operator is different from the ordinary laplace operator by the factor in front of the second order differential term , which vanishes on the excision boundary . it can be shown that the space of analytic solutions on @xmath121 minus the excised horizon belonging to the kernel of @xmath122 is generally of dimension one . this is in contrast with the case of the laplace equation , where it is of dimension two . in practice , this will mean that for a numerical resolution of an equation @xmath123 , there is only one boundary condition to fix for the unknown , mainly the behavior at infinity . no additional information is needed at the excised boundary for the _ effective _ operators ( [ e : opq ] ) . operators of this kind are known in the mathematical literature as elliptic operators with weak singularities @xcite . the very same scheme can be applied to the equation ( [ e : eqb1 ] ) for @xmath74 , the only difference being that the original laplace operator is replaced by a slightly modified one ( see the appendix [ s : appendix ] ) . as explained in sec.[ss : non_cf ] , after solving for the two main equations ( [ e : opmodalm ] ) and its equivalent in @xmath74 , the inversion of the gauge differential systems ( [ e : diracmag]-[e : diracel ] ) ( explicited in the appendix [ s : appendixb ] ) for the reconstruction of @xmath62 requires three extra boundary conditions , in addition to the vanishing of all quantities at infinity . we obtain them as compatibility conditions based on the original elliptic tensorial equation ( [ e : dh1 ] ) : we express three decoupled elliptic scalar equations for three components of @xmath62 in the spin - weighted tensor spherical harmonics basis , denoted @xmath124 , @xmath125 and @xmath126 , which are directly related to the usual tensorial components of @xmath62 and defined in appendix [ s : appendix ] . from the tensor equation ( [ e : dh1 ] ) , we deduce : @xmath127 where @xmath128 superscripts indicate the corresponding components of @xmath62 in the tensor spherical harmonics basis ( see appendix [ s : appendix ] ) . as for the equation involving @xmath73 , we can rewrite the above equations by extracting the weakly singular operator @xmath122 acting on the principal variable , the other contributions being put on the right - hand side of the equations . for example , the equation in @xmath126 can be rewritten the following way : @xmath129 with the lie derivative term containing all the left - hand - side contributions of equation ( [ e : ellmu ] ) not taken into account . we do not need to invert this equation : however , as the leading order term in @xmath122 vanishes at the horizon @xmath130 , we can accordingly write a robin - like boundary condition for the @xmath126 quantity : @xmath131 this will be used as a boundary condition for the gauge system ( [ e : diracmag ] ) , the source terms being computed with quantities from the previous iteration . using the same method , we can write very similar expressions ( that we do not explicitly give here ) for the fields @xmath124 and @xmath125 , to be used as robin boundary conditions applied to the gauge differential system ( [ e : diracel ] ) . the three boundary conditions are sufficient to invert the two gauge systems ( [ e : diracmag]-[e : diracel ] ) @xcite , and reconstruct the whole @xmath62 tensor from the tensor spherical harmonics components ( see the appendix [ s : appendixb ] and @xcite for details ) . to summarize , the method employed for the resolution of the whole @xmath62 system is iterative and can be decomposed for each step in the following way ( more technical details are provided in the appendices ) : 1 . after calculating the source @xmath132 from equation ( [ e : dh1 ] ) , we deduce the right hand side of the equation ( [ e : opmodalm ] ) for @xmath73 , using values from the previous iteration . the same is done for the quantity @xmath74 and its corresponding source terms . we invert equations ( [ e : opmodalm ] ) and its equivalent for @xmath74 , only by imposing that the fields are vanishing at infinity . we compute the value of the trace from equation ( [ e : det1 ] ) ( a more explicit expression can be found in equation ( 169 ) of @xcite ) . this allows us to write the two differential systems ( dirac gauge systems ) mentioned in sec . [ ss : non_cf ] and expressed in appendix [ s : appendixb ] , involving the spherical harmonics components of @xmath62 ( scalar quantities ) . 4 . we invert these two gauge differential systems using three boundary conditions similar to eq . ( [ e : boundmu ] ) , for the three scalar spherical harmonics components @xmath124 , @xmath125 and @xmath126 . as those are compatibility conditions expressing information already contained in eq . ( [ e : dh1 ] ) , we provide in this way no additional physical information . this gives us the spherical harmonics components of @xmath62 . we reconstruct the whole tensor @xmath62 from the spherical harmonics components . we have not proven here that no boundary condition has to be put generically for the resolution of the two scalar equations involving @xmath73 and @xmath74 in the tensorial system . however , if we implement numerically the resolution by the inversion of the operator @xmath119 at each iteration , we will not have to impose any boundary condition , but only informations coming from the einstein equations . moreover , a convergence of the entire @xmath62 system would support the coherence of the reasoning , and hint that there is , in our case , a deeper physical motive preventing the prescription of additional information on the horizon . the results in sec . [ s : resus ] show this is the case . all the numerical and mathematical tools we use here are available in the open numerical relativity library lorene @xcite . our simulation is made on a 3d spherical grid , using spherical harmonics decomposition for the angular part and multidomain tau spectral methods ( see @xcite for a review ) . the mapping consists in four shells and an outer compactified domain , so that infinity is part of our grid and we have no outer boundary condition to put at a finite radius . our grid size is typically @xmath133 . we also have checked our code by setting @xmath134 , to verify that no deviation from axisymmetry occurred . our innermost shell has a boundary at the radius @xmath82 , which will be the imposed location of a mots , and will be used as the unit of length in all the results presented here . we impose the values of all the fields to be equivalent at infinity to those of a flat three - space . finally , trying to get stationary data , we prescribe our coordinates to be adapted to this stationarity , so that all the time derivatives in the einstein ( 3 + 1 ) system are set to zero . however , even if we expect to get axisymmetric data ( the only vacuum stationary solution for a black hole being the kerr solution ) , we are always able to solve our equations in three dimensions . we proceed with our scheme in the following way : during one iteration , all the variables are updated immediately after they have been calculated , so that the sources for the next equations are modified . the tensorial equation for @xmath62 is the last solved in a particular iteration , and we obtain at each step a local convergence for the whole tensorial system ( including the determinant condition ) , before we proceed to the update of all quantities , and to the next iteration . we impose on the sphere of radius @xmath82 the conditions of zero expansion ( [ e : thetaeq0 ] ) and shear ( [ e : sheareq0 ] ) , via respectively a robin condition on the quantity @xmath61 and a dirichlet condition on the partial shift @xmath89 . we also impose the horizon - tracking coordinate condition on the radial shift component @xmath135 . having set the shape and the location of the surface in our coordinates , we are only left with two free parameters , which are the boundary value of the lapse function and the rotation rate @xmath136 . as we said , the lapse function , which is merely a slicing gauge choice , is fixed to a constant value @xmath137 on the horizon . we generate two sets of data on our three - slice , spanning the rotation rate from zero ( schwarzschild solution ) to a value of about @xmath138 , where our code no longer converges . one set will give the solution for the whole differential system ( the non - conformally flat ( ncf ) data , supposed to converge to the rotating kerr solution ) , while the other will compute conformally flat ( cf ) data , by putting @xmath139 . from a spacetime point of view , the cf data can also be seen as a computation of black hole spacetime using the so - called isenberg - wilson - mathews approximation to general relativity @xcite , @xcite . figure [ f : precision ] presents , on the one hand , the absolute accuracy obtained for the einstein constraints ( in the form expressed in @xcite ) in the ncf case . regarding fulfillment of the einstein dynamical equation , figure [ f : precision ] also shows the accuracy of the ncf fully stationary solution , as well as its violation in the conformally flat case . we see the expected improvement for precision of resolution of dynamical equations in the full ncf case . let us note that a verification of the gauge conditions is not even necessary , as it is fulfilled by construction ( we only solve for variables satisfying the gauge ) . this is one of the strengths of our algorithm . a non - trivial issue of our computation is the link between the two physical characteristics of the system ( the mass and angular momentum of the data ) and the two input quantities supposed to fix them , namely the boundary value for the lapse and the rotation rate @xmath140 on the horizon . we choose here in our sequence to fix the value of the horizon coordinate radius , removing it from the list of variables . results are shown in figure [ f : omega_asm ] . the value of the lapse being also fixed , we observe that an increase in @xmath140 not only affects the angular momentum , but also the adm mass of the spacetime . moreover , fixing the rotation rate does not amount to the prescription of the angular momentum to an _ a priori _ given value . a decrease in the value of @xmath102 on the horizon results also in an increase in @xmath97 ( defined in section [ ss : bccons ] ) . this stems from the fact that our choice for the slicing directly influences in this approach the physical parameters ( e.g the areal radius ) of the solution obtained . we note also that for a fixed value of @xmath102 , the correspondence between @xmath95 and @xmath141 is slightly different in the conformally flat case and in the ncf case . with our algorithm , a larger value of the lapse gives a slightly better convergence of the code for high rotation rates of the black hole ( until @xmath142 approximatively ) . for each lapse the code stops converging at a certain value of the rotation rate . we do not yet know whether this is a problem of our algorithm to be improved , or if this has deeper physical reasons : constant values for the lapse and the rotation rate might not be `` good '' variables for the kerr black hole in dirac gauge , once we reach high rotation rates . the only conclusion we can draw from this is that there is a non trivial correspondence between our `` effective parameters '' @xmath102 and @xmath95 , and the physical ones , namely the adm mass and komar angular momentum . this correspondence is likely to be one to one for values of @xmath143 below a certain threshold of about @xmath144 . reaching higher values for @xmath141 is left to future numerical investigations . let us mention again the remark made by @xcite about the boundary condition for the lapse in the xcts scheme . although it is necessary to fix the slicing of the spacetime by an arbitrary boundary condition on the lapse , we have the freedom to decide what kind of condition to impose . the authors in @xcite suggest that an arbitrary condition of neumann or robin type would be preferable , because it is more flexible in view of a numerical algorithm . in particular , not fixing a value for the lapse on the horizon , but rather giving a first order prescription , allows the data to `` adapt '' to potentially high tidal distortions . however , having also tried to impose neumann conditions for the lapse in our configurations , we do not see any clear improvement in the robustness of the algorithm . this is why we still keep a dirichlet boundary condition as the simplest prescription . one of the tests of stationarity to be made can be the comparison between the adm mass and the komar mass at infinity , defined with the ( presumably ) killing vector @xmath145 ( equation ( 7.91 ) of @xcite ) . the results of this test are displayed in figure [ f : error_jm ] . the comparison between the adm mass and the komar mass is actually directly linked to the virial theorem of general relativity put forth by @xcite . the concordance between those masses is equivalent to the vanishing of the virial integral , and has been also used as a stationarity marker by @xcite . we have also computed in both ncf and cf cases an estimate of the amount of gravitational radiation contained outside the black hole in the 3-slice . following the prescriptions of @xcite , we calculate the difference between the adm mass and what could be called the isolated horizon mass , defined in geometrical units by : @xmath146 where @xmath147 is the areal radius of @xmath148 . @xmath149 is nothing but the formula for the christodoulou mass @xcite calculated from the komar angular momentum @xmath150 on the horizon ( defined with the same supposed killing symmetry as @xmath100 ) . if we have an isolated horizon extending to future infinity , the difference between @xmath151 and @xmath149 gives exactly the radiation energy emitted at future null infinity for the data @xcite . in non - stationary cases ( for example binary systems ) , this is an appropriate estimate of the radiation content at an initial given time . results for comparison between the two cases studied here are shown in figure [ f : error_jm ] . although the gravitational energy available for ncf spacetimes is contained under @xmath152 whatever the rotation rate might be , in the cf case , the increase in energy with @xmath153 is patent . this measure of energy available with respect to @xmath97 gives us a way of approximating _ a priori _ the amount of what is usually called `` junk '' gravitational radiation , that could be emitted on a spacetime evolution with conformally flat initial data . in the same spirit , we have also computed the accuracy in the verification of a penrose - like inequality for axisymmetric data , that can be written as : @xmath154 where @xmath155 is the minimal area of a surface containing the horizon , @xmath100 is the komar angular momentum at infinity and @xmath99 is the adm mass at infinity . being a little more stringent that the actual penrose inequality , it has been first proposed by @xcite for axisymmetric spacetimes . this inequality is supposed to be verified for all axisymmetric data containing an apparent horizon , and to be an equality only for actual kerr data ( this is referred in @xcite as _ dain s rigidity conjecture _ @xcite ) . the results are presented in figure [ f : epsilon ] . we observe that , if the equality is very well verified in the actual kerr case , this is definitely not true for cf data , even for reasonable values of @xmath97 . in @xcite ( cf . @xcite for a general context ) , it has been proposed that this quantity @xmath156 ( _ dain s number _ ) should be understood as a strong diagnosis tool for distinguishing between kerr horizons and other isolated or dynamical horizons . this numerical observation shows strong support in favor of this claim , pulling apart actual kerr data and reasonable approximations of these data . let us also point out the virtual costlessness of this tool , as we only have to rely on a single real value . we also note that , when computing the rescaled difference of komar angular momentum between the horizon and infinity @xmath157 , we come up in all cases with a difference at the level of numerical precision for resolution ( see figure [ f : error_jm ] ) . this is of course coherent with the fact that gravitational waves can not carry any angular momentum in axisymmetric spacetimes . this result ensures us the equivalence in practice between the estimation of radiation exterior to the horizon and the verification of penrose inequality via dain s number . to be much more complete about the geometry of the constructed horizons , one could rely on the source multipole decomposition of the two - surface lying on our three - slice . this feature has first been presented by @xcite , based on an analogy with electromagnetism , and first studied in @xcite in the case of dynamical horizons . we here implement the computation of multipole moments in the isolated horizon case , which is the strict situation where they have been defined in @xcite . a prerequisite is the existence of a preferred divergence - free vector field @xmath158 on the sphere , from which the angular momentum of the horizon is defined ( the divergence - free condition on @xmath158 ensures that all definitions will be gauge - independent ) . as mentioned above , our chosen vector field will be the one associated with the azimuthal coordinate , namely @xmath159 . another important feature is the construction of a preferred coordinate system , so that the legendre polynomials associated with spherical harmonics will possess the right orthonormality properties ; as expressed in the implementation of @xcite , this reduces to finding a set of coordinates @xmath160 where the metric on the two - surface can be written as : @xmath161 with @xmath147 the areal radius of the sphere and @xmath162 determined in terms of the two - dimensional ricci scalar and the norm of @xmath158 @xcite . in the axisymmetric case studied here for the horizon , the integral curves for the coordinate @xmath96 are already defined by the orbits of the vector field @xmath163 . the coordinate @xmath164 is defined by @xmath165 an appropriate normalization should be added , that ensures that @xmath166 . in the kerr case , those coordinates turn out to correspond with the boyer - lindquist coordinates , with @xmath167 in spherical coordinates @xcite . the mass and angular momentum multipoles of order @xmath168 are then defined , by analogy with electromagnetism @xcite : @xmath169 with this definition and using the gauss - bonnet theorem it is trivial to see that @xmath170 and @xmath171 , the komar angular momentum on the horizon . we should emphasize that these multipoles , except for @xmath172 and @xmath173 , are in general different from the field gravitational multipoles that can be defined at infinity . however , the authors in @xcite have pointed out that the knowledge of all the multipoles of an isolated horizon allows to reconstruct the whole horizon , and also the spacetime in a vicinity of this horizon . the multipoles then discriminate exactly every isolated horizon , and the spacetime at its vicinity . figure [ f : multipoles ] shows the capacity of telling apart the horizon of a cf axisymmetric slice and the one of a ncf slice , in dirac gauge . data are also compared with an analytic kerr solution in kerr - schild coordinates . apart from the accuracy obtained for our ncf data ( and a further confirmation that we indeed have obtained the actual kerr spacetime ) , we see the clear distinction made by this computation between the kerr horizon and a conformal approximation of it . together with dain s number , this study has proven that those two tools are very well - suited to study isolated horizon properties , and the distance between data obtained from , say an evolution scheme , and the eventual equilibrium black hole data it is supposed to reach . ultimate tests on the characterization of the obtained data as slices of kerr could be achieved by implementing the schemes proposed in @xcite . the data we get with our simulations are interesting at several levels . they allow to make a direct comparison between the conformally flat approximation and the exact solution for axisymmetric spacetimes containing a black hole , that are both calculated _ a priori _ and in the same gauge . as we have seen in sec . [ s : resus ] , this gives us insight about the geometric features of the exact solution ; we can single important issues concerning for example the intrinsic geometry of the horizon , via multipoles and the penrose inequalities . numerical tools are in this respect implemented and their efficiency tested . at a more theoretical level , the method we used to get those data is a little bit heterodox : providing standard non - expanding horizon conditions for ( 3 + 1 ) variables such as @xmath38 , @xmath61 and @xmath174 , we choose in addition not to prescribe any further geometrical information for the conformal part , symbolized here by the tensorial field @xmath62 . this has been motivated by the fact that , given the tensorial equation corresponding to @xmath62 in our formalism , it appears that we most likely _ can not _ prescribe anything else in the studied setting . by numerical transformation of the operator acting on @xmath62 we ensured that at every iteration step no boundary condition was required . the fact that our system of einstein equations written this way converges to the required kerr solution shows that indeed , no additional information was needed in the single horizon case . to be more precise , our method suggests that in this case , the conformal geometry of the mots is directly encoded in the einstein equations , when written with an equilibrium ansatz : we directly use these equations to justify a no - boundary treatment . in the light of this numerical study , we can make a parallel with the proposition made in @xcite . in that paper , the authors suggested , after a gauge dependence analysis for @xmath68 on the horizon , that a prescription could be made for the conformal geometry on the horizon . in this respect , they justify , for the projection of the three - metric on the two - surface , the following choice : @xmath175 with @xmath176 the usual diagonal round metric for a two - sphere in spherical coordinates adapted to the horizon . this choice should not affect the physics of the three - slice , and suffices to recover the solution for einstein dynamical equation with a slice of a spacetime containing an isolated horizon . their study is made in a differential gauge generalizing the dirac gauge we are using , namely @xmath177 , with @xmath89 a regular vector field on the three - slice . our case corresponds to @xmath178 , which is precisely the one treated in detail in @xcite . when comparing to our data , we find that the projection of our three - metric on the two - surface is not conformally related to the flat metric in adapted spherical coordinates . this means that in our particular case , and in regard of the particular no - boundary argument we use , a boundary condition of the same type as ( [ e : propcook ] ) is probably inconsistent with our data , and is likely a choice that we do not have the freedom to make ( note however that geometric conditions in @xcite making full use of the isolated horizon structure are indeed compatible with the present results , i.e. they are identically satisfied in the present kerr case , whose horizon is indeed an isolated horizon ) . unfortunately , the authors in @xcite did not present any numerical results to support their claims , that we could have compared with ours . we insist here on an important caveat for our argumentation : assumptions can only be justified in the very particular case we are studying here , which is the axisymmetric vacuum spacetime . this spacetime has very specific and non - trivial properties , all related to the uniqueness theorem of carter @xcite . although the reasoning we have made on sec . [ s : fcf ] for the operator could apply in other isolated horizon studies , we are not certain that our algorithm would globally converge when applied to a more general case ( e.g. a black - hole binary system ) ; a failure of this behavior would probably mean that an additional information about the conformal two - geometry has to be given to the system . geometric fully isolated horizon boundary conditions proposed in @xcite could then be enforced ( note that geometric inner boundary conditions in @xcite are not necessarily tied to the particular analytic setting here discussed and , more generally , they would also apply in schemes not enforcing the coordinate adaptation to stationarity at the horizon , @xmath179 , crucial for the singular nature of operators ( [ e : opq ] ) ) . finally , let us point out the fact that we made those simulations by prescribing only the geometry of the horizon , and the geometry of spacetime at infinity . no assumption has been made for axisymmetry in the three - slice ( computations can easily be made in the full 3d case and give the same results ) . prescribing a vanishing expansion and a conformal killing symmetry on a horizon , together with asymptotically flat hypothesis , our code converges to the only solution of the kerr spacetime . without claiming any rigorous demonstration here , this numerical result is most likely a support to the well known black hole rigidity theorem @xcite , where the same hypotheses lead to a uniqueness theorem involving the kerr solution as the only one with no electromagnetic field . we have used the prescription for a fully constrained scheme of ( 3 + 1 ) einstein equations in generalized dirac gauge @xcite to retrieve stationary axisymmetric black hole spacetime , and compared it with the analytical solution of kerr type . an advanced handling of the conformal geometry of our three - slice allowed us to reach actual stationarity with good resolution precision for our scheme . although we used standard quasi - equilibrium conditions concerning boundary values for other metric fields in the excised horizon , we found that the conformal geometry on the horizon required no prescription whatsoever in the single horizon case . this is in contrast with suggestions available in the literature @xcite , and probably suggests an underlying physical feature of the horizon geometry ( maybe related to uniqueness of the kerr solution ) . to our knowledge , it is the first time the conformal part is numerically computed in a black hole spacetime using only a prescription on the stationarity of spacetime ( and without resorting to additional symmetries ) . the application of this feature to the more general initial data problem is evident : in the same spirit as the work done in @xcite for neutron - star binaries , using it for the black - hole binary system could lead to significant improvement in the available initial data for evolution codes . further numerical work will clarify this issue . we have implemented and used in our study numerical tools aimed at characterizing the geometry and physical properties related to horizons embedded in spacetime ; those tools , among which a complete multipole analysis for two - surfaces as gravitational sources , have proven very accurate for diagnostics involving the horizon geometry and physical features . they will be more thoroughly presented , and tested in more general cases , in an upcoming work . the authors warmly thank ric gourgoulhon and philippe grandclment for many valuable discussions , and a careful reading of the manuscript . we also thank the referee for numerous useful comments . nv and jn were supported by the a.n.r . grant blan07 - 1_201699 entitled `` lisa science '' . jlj has been supported by the spanish micinn under project fis2008 - 06078-c03 - 01/fis and junta de andaluca under projects fqm2288 , fqm219 . we here give the definition of the three spectral quantities introduced in sec . [ ss : non_cf ] , that describe the divergence - free degrees of freedom ( with respect to the dirac gauge ) associated with a rank two symmetric tensor . the reader is also invited to go to @xcite or @xcite where more detailed calculations are provided . we first define a set of spin - weighted tensor spherical harmonics components for a symmetric rank-2 tensor , directly linked to the tensor spherical harmonics as introduced by mathews and zerilli @xcite . we shall give the expression for these components of the tensor @xmath62 using the classical spherical coordinate basis , which is used in practice in our computations . with the notation @xmath180 , the six pure spherical harmonics components of @xmath62 are defined as : @xmath181 the fifth and sixth scalar fields being simply the tensor trace @xmath182 with respect to the flat metric and the @xmath124 spherical component . let us note that these relations are more tractable when using a scalar spherical harmonics decomposition ( introduced in sec . [ ss : bchij ] ) for all fields . indeed , an angular laplace operator acting on a field reduces then to a simple algebraic operation on every spherical harmonics component . inverse relations can also be computed to retrieve the classical components of @xmath62 from spherical harmonics quantities . we now derive the main variables related to our study : with the divergence - free decomposition @xmath183 , and @xmath71 , a choice for three quantities defined from @xmath62 and verifying : @xmath184 can be expressed as the following scalar fields ( see @xcite ) : @xmath185 these quantities can also be decomposed onto a scalar spherical harmonics basis . the equivalence in ( [ e : def2abc ] ) is achieved up to boundary conditions . to show how the quantities @xmath73 , @xmath186 and @xmath187 behave with respect to the laplace operator , we shall assume in the following that the tensor @xmath62 is the solution of a poisson equation of the type @xmath75 . we can deduce a scalar elliptic system verified by @xmath73 , @xmath186 and @xmath187 as : @xmath188 where @xmath189 , @xmath190 and @xmath191 are the corresponding quantities associated with the source @xmath192 . a simple way of decoupling the last two elliptic equations is to define the variables @xmath193 and @xmath194 with : @xmath195 thus , we can write an equivalent system for ( [ e : deltaa],[e : deltab],[e : deltac ] ) as : @xmath196 with the following elliptic operators defined for each spherical harmonic index l : @xmath197 @xmath198 is the identity operator . @xmath73 , @xmath74 and @xmath79 are then defined as three scalar fields characterizing only the divergence - free part of a symmetric rank two tensor @xmath62 , and giving a system of three decoupled scalar elliptic equations in the poisson problem for this tensor . hence they are very well suited to the study of a tensorial elliptic problem in dirac gauge . let us finally note that the quantities @xmath74 and @xmath79 are directly related to each other by the trace of the considered tensor @xcite : if we know _ a priori _ the value of the trace for @xmath62 , then the knowledge of @xmath74 suffices to recover @xmath79 with no additional information ( the converse being equally true ) . in this section , we come up with technical details for resolution of the gauge differential system introduced in sec . [ ss : non_cf ] , to reconstruct the tensor @xmath62 from the quantities @xmath73 and @xmath74 . we begin by expressing components of the vector field representing the divergence of @xmath62 : @xmath199 with the dirac gauge for @xmath200 . adopting the vector spherical harmonics decomposition suggested in @xcite , the three spherical harmonics components of @xmath201 are expressed , in function of the spherical harmonics components of @xmath62 ( see appendix [ s : appendix ] ) , as : @xmath202 , \label{e : heta}\\ h^\mu & = & { \delta_{\theta\varphi}}\left [ \frac{\partial \mu}{\partial r } + \frac{3\mu}{r } + \frac{1}{r } \left ( { \delta_{\theta\varphi}}+ 2 \right ) { \mathcal{x}}\right ] . \label{e : hmu}\end{aligned}\ ] ] those three expressions , alongside with definitions of the quantities @xmath73 and @xmath74 , will allow to express two decoupled differential systems . the first one , involving the spherical harmonics components @xmath126 and @xmath203 , combines the expression for the scalar field @xmath73 , as well as the fact that @xmath204 vanishes under the dirac gauge : @xmath205 the second system is composed of the definition of @xmath74 for each of its spherical harmonic component @xmath206 , as well as the vanishing of @xmath207 and @xmath208(again , due to dirac gauge ) : @xmath209 with the expressions ( [ e : def_b ] , [ e : def_c ] ) of @xmath186 and @xmath187 as functions of the spherical harmonics components of @xmath62 . in this system , the trace is given _ a priori _ , so that only three spherical harmonics components are considered as unknowns . we refer to the analysis of @xcite to affirm that , when solving our equations in @xmath121 minus an excised inner sphere , one boundary condition has to be provided at the surface for the system ( [ e : diracmag ] ) , and two for the system ( [ e : diracel ] ) . as pointed out in sec.[ss : bchij ] , these conditions are retrieved as compatibility conditions based on the original elliptic tensorial equation . overall , we are able to invert the two dirac differential systems , and retrieve all the spherical harmonics components of @xmath62 from the sole knowledge of @xmath73 , @xmath74 and the trace @xmath182 . j. l. jaramillo , j. a. valiente kroon and e. gourgoulhon , class . quant . grav . * 25 * , 093001 ( 2008 ) a. ashtekar , j. engle , t. pawlowski and c. van den broecke , class . quantum grav . * 21 * , 25492570 ( 2004 ) .

the theory of real relativistic fluids is in the rather unique situation that there is a natural relativistic extension of the nonrelativistic theory , but it is physically untenable . on the other hand , mounting evidence that matter created in relativistic heavy ion collisions behaves as a relativistic fluid with small but finite viscosity has given the quest for an alternative a definite goal . we shall review different approaches to relativistic real fluids , their link to relativistic kinetic theory , and their application to the analysis of heavy ion collisions

1310.0841

the goal of these lectures is to discuss the theory of relativistic real fluids under the excuse of its application to the description of relativistic heavy ion collisions @xcite . more concretely , our goal is to discuss , in the simplest possible terms , why the `` einstein elevator '' concept is not useful in the development of this theory . namely , we can not build a theory of relativistic real fluids by just asking that it reduces to its nonrelativistic counterpart , namely the navier - stokes equations , in an inertial frame where the fluid is at rest . for this reason we shall begin with a `` derivation '' of the nonrelativistic theory of ideal fluids ( those which flow with no entropy production ) from thermodynamics @xcite . for ideal fluids , the relativistic theory may be found as the covariant extension of the nonrelativistic theory enforced in the rest frame . we shall then show that already in this simple framework we may obtain a working picture of how a relativistic heavy ion collision works , yielding remarkably accurate predictions regarding several concrete observables . among these observables there is the so - called elliptic flow , namely the anisotropy of the particle yield in the transverse plane to the beam direction . if the matter deposited in the collision region behaves as a nearly ideal fluid , then elliptic flow arises as the simple transduction of pressure gradients in the original configuration into velocities in the asymptotic expanding state . this is an obvious prediction of the euler equations which is surprisingly hard to reproduce in non - hydrodynamic models . the success in predicting that there must be an elliptic flow is one of the most compelling reasons to believe that matter in the collision region , from some very early time after the actual collision up to the break up in the individual hadrons which are eventually detected , behaves as a fluid . however , ideal hydrodynamics overestimates the amount of elliptic flow @xcite . this suggests there must be a mechanism that tends to isotropize flow , and indeed viscosity acts in precisely that direction . therefore , the next natural step is to consider hydrodynamic models based on real fluid hydrodynamics . at this point we return to the formal theory , and generalize our previous `` thermodynamic '' derivation to the case of real fluids . the resulting non relativistic theory is given by the navier - stokes equations , which is satisfactory , but the relativistic theory emerging from the `` einstein elevator '' is essentially flawed - if covariance is enforced , then the theory has no stable solutions @xcite . how to get out of this cul de sac is the subject of the rest of the lectures . since the `` thermodynamic '' approach led us nowhere , we shall begin anew from a more fundamental point of view , that provided by relativistic kinetic theory . in this case , there is an agreed upon relativistic formulation to build on @xcite . the problem is how to reduce it to the hydrodynamic level . in the nonrelativistic case , the method of choice is the so - called chapman - enskog expansion , which leads to the navier - stokes equations @xcite . in the relativistic case , it works likewise , and so it must be rejected . however , a second well known technique , the grad expansion , is successful in building a relativistic hydrodynamics where linear perturbations of an equilibrium state evolve causally . however , the application to relativistic heavy ion collisions forces us to face the relationship between kinetic and hydro descriptions a second time , because what is actually measured are individual hadrons . this means that there must be some space time `` break up '' or `` freeze out '' surface ( or region ) where the fluid nucleates into individual particles that fly away to the detectors ( a process that is by no means simple , but does not belong to these lectures ) . we must be able to read the one particle distribution function of these particles out of the hydrodynamic state of the fluid just before break up . if we simple invert the grad expansion , we are led to a one particle distribution function which is not nonnegative throughout phase space , a defect that contaminates the predictions of the theory regarding observables which are sensitive to high momenta . this drawback of the grad approach , and also the need to include nonlinear corrections to the linearized equations of motion without spoiling stability and covariance , prompted us to suggest that the grad approach may be just the linear approximation to a more comprehensive view of the kinetic - hydro relationship @xcite . the basic insight is that any realistic kinetic theory displays a whole hierarchy of relaxation times , from the relaxation times characteristic of large scale inhomogeneities to the much shorter relaxation times of hard modes . fluctuations in the hydrodynamic modes relax on the longer time scales , and are perceived by the harder modes as externally imposed thermodynamic forces which prevent their relaxation to true equilibrium . in other words , the truly kinetic modes relax not to equilibrium but to a nonequilibrium steady state constrained by the instantaneous configuration of the hydrodynamic modes . it has been known from long ago that such steady states are the solutions to variational problems @xcite . prigogine among others has proposed that they are the extrema of the entropy _ production _ , as opposed to the extrema of the entropy itself , which are the true equilibria @xcite . known proofs of the so - called `` prigogine theorem '' are restricted to linear irreversible thermodynamics @xcite ; we shall appeal to it on a heuristic rather than formal basis - the idea is that most kinetic modes are in the linear regime most of the time anyway , so a theory which is good in the linear regime is good enough to compute global observables such as stress tensor components , but we shall not attempt to formulate this insight in any rigorous way @xcite . we shall therefore conclude these lectures with a brief presentation of this `` entropy production variational method '' , its relationship to positivity of the one particle distribution function , and its nonlinear generalization . the lectures are organized as follows . in next section we present the theory of relativistic ideal fluids , as derived from thermodynamics , and its application to the description of relativistic heavy ion collisions ( rhics ) . the main success of the theory , namely the prediction of elliptic flow , is also its fatal drawback , because experimental data show that ideal hydrodynamics overestimates the flow anisotropy @xcite . we are therefore motivated to extend the theory to include viscosity , whereby we hit on the stability problem @xcite . we present the theory of real fluids in section iii ; since the thermodynamic derivation fails , we base our discussion on relativistic kinetic theory . we must confront the _ closure _ problem , that is , to relate the one particle distribution function in the kinetic description to the hydrodynamic degrees of freedom . we present the chapman - enskog solution ( which reproduces the results from the thermodynamic analysis , and therefore is unsuitable ) and the grad solution , and finally identify the grad solution of the closure problem as the linearized approximation to an approach based on the entropy production variational method ( epvm ) . we conclude with some brief final remarks . if relativistic hydrodynamics may be founded on kinetic theory , kinetic theory itself comes from quantum field theory , the main subject matter of this school . we show the main lines of the derivation of kinetic theory from quantum field theory in the appendix , which relies heavily on ref . let us begin by reviewing non relativistic hydrodynamics , as `` derived '' from thermodynamics . the basic tenets of thermodynamics we need to keep in mind are the following : we have a ( simple ) system described by some intensive parameters ( temperature @xmath0 , chemical potential @xmath1 , pressure @xmath2 , etc . ) whose meaning we take for granted ( e.g. we already know everything about the zeroth law ) and extensive parameters , such as energy @xmath3 , entropy @xmath4 , volume @xmath5 , and particle number @xmath6 ( we use particle number for concreteness , but any , or several , conserved charge(s ) would serve just as well ) . in equilibrium , all these quantities are position independent . their first deviations from equilibrium are related by the first law tds = du + pdv -dn [ ( 12.1 ) ] extensive quantities are homogeneous functions of each other , so we must have ts = u + pv - n [ ( 12.2 ) ] from the differential of this second identity we obtain the gibbs - duhem relation dp = sdt + nd [ ( 12.3 ) ] where @xmath7 and @xmath8 are the entropy and particle number densities , respectively . this means .| _ & = & s = t.|_t & = & n [ ( 12.4 ) ] where @xmath9 is the energy density . let us now consider a non equilibrium , inhomogeneous configuration . the above relations still hold for the densities of the conserved quantities energy , momentum ( which we must include ) and particle number . the dynamics is given by the conservation laws & = & -_j j^ j_u&=&-_j t^ij&=&-_j j^j here @xmath10 , @xmath11 and @xmath12 are the energy , momentum and charge densities , @xmath13 , @xmath14 and @xmath15 the respective currents . @xmath14 must be symmetric @xcite . the entropy density s=1 t obeys the first law tds = d - v_idp^i - dn which implies the gibbs - duhem relation dp = sdt+p^idv_i+nd in these formulae we have introduced the velocity @xmath16 and the chemical potential @xmath17 . we therefore have & = & 1t - v_i-&= & t_j j^ j_u - v_i_j t^ij -_j j^j we then have = -_j1t+_j(1 t ) -1tt^ij_jv_i - tj^j_j to proceed we write @xmath18 . in the terms involving the velocity we eliminate the gradient of the chemical potential by using the gibbs - duhem relation & = & -_j1t+_j(1 t ) -1tt^ij_jv_i&-&1tv^j- t^j_j so far & = & -_j1t&+&_j(1 t ) & -&1t_jv_i- t^j_jusing again the second law , we may write j^ j_u+pv^j - v_it^ij - j^j - tsv^j = j^ j_u-(+p)v^j - v_i(t^ij - p^iv^j - p^ij)-^j and so & = & -_j1t&+&_j(1 t ) & -&1t_jv_i-^j_j[dsdt ] where @xmath19 . since we have allowed for the possibility of the energy and particle number flux not being collinear , we must adopt a convention about what is the fluid velocity . we shall adopt the so - called landau - lifshitz prescription , namely , that @xmath20 and @xmath21 vanish when @xmath22 @xcite . we define an ideal fluid as one that flows with no entropy production . thus for an ideal fluid t^ij&=&p^iv^j+p^ijj^ j_u&=&(+p)v^j^j&=&0 the symmetry of @xmath14 requires @xmath11 to be proportional to @xmath23 . the equations for an ideal fluid are thus + _ j nv^j&=&0+_j ( + p)v^j&=&0+_j p^iv^j+^ij_jp&=&0 we now generalize the above framework of thermodynamics to a relativistic fluid evolving in a spacetime with an arbitrary metric @xmath24 . all derivatives shall be covariant derivatives with respect to the levi - civita connection , so that @xmath25 . we write the minkowsky metric as @xmath26 , @xmath27 and @xmath28 , where a @xmath29 refers to coordinate @xmath30 to simplify matters we will describe the construction of a covariant theory in terms of a set of rules @xcite : \(a ) intensive quantities ( @xmath0 , @xmath2 , @xmath1 ) are associated with scalars , which represent the value of the quantity at a given event , as measured by an observer at rest with respect to the fluid . \(b ) extensive quantities ( @xmath4 , @xmath5 , @xmath6 ) are associated with vector currents @xmath31 , @xmath32 , @xmath33 . if a given observer measures a density @xmath34 and a flux @xmath35 for the quantity @xmath36 , then @xmath37 the quantity @xmath32 associated with volume is the fluid 4-velocity ; in minkowsky space u^=1(1,c ) @xmath32 obeys the additional constraint @xmath38 . we call density tout court the density measured by an observer comoving with the fluid , namely @xmath39 . if the quantity @xmath36 is conserved , then @xmath40 . if we consider the variation of the entropy content within some spatial region as a function of time , the second law demands that the increase in entropy should be higher than the entropy flow through the boundary . thus the covariant statement of the second law is that entropy production must be positive , i.e. @xmath41 . \(c ) energy and momentum are combined into a single extensive quantity described by an energy - momentum tensor @xmath42 which is symmetric . therefore any given observer will identify the energy density as @xmath43 , the energy flux as @xmath44 , the momentum density as @xmath45 and the momentum flux as @xmath14 . the symmetry of the energy momentum tensor implies that the momentum density is @xmath46 times the energy flux @xcite . we define the rest frame of the fluid as the frame where the energy flux vanishes . as before , the energy density @xmath10 is defined as the energy density in the rest frame . both the fluid velocity and energy density may be found from the eigenvalue equation @xmath47 let us now describe a relativistic ideal fluid , namely one that flows with no entropy production . we shall proceed by correspondence between the relativistic and non relativistic theories . to this end , we assume that in the rest frame we have the decomposition n^=(n,0 ) t^= ( cc & 0 + 0&p^ij ) so in an arbitrary frame we must have n^ & = & nu^t^&= & u^u^+ p^ [ ( 12.11 ) ] where @xmath48 . the entropy current @xmath31 in the rest frame is given by @xmath49 , where @xmath50 . in an arbitrary frame @xmath51 or s^= -t^ _ + pu^ - n^ where the inverse temperature vector @xmath52 and @xmath19 . observe that @xmath53 . then s^ _ ; = -_t^_;- n^ _ ; [ ( 12.15 ) ] this means that entropy production vanishes for an ideal fluid , provided the conservation laws of energy - momentum and particle number hold . the theory of relativistic real fluids has been brought to the limelight by the mounting evidence that one such system has been seen in actual experiments , namely , relativistic heavy ion collisions ( rhics ) . indeed , one can get a working understanding of rhics from the theory of ideal fluids we have just described . nevertheless , precisely the demands of a better fit between theory and experiment will prompt us to develop a theory or real fluids , namely , to add viscosity to the flow equations . in this section we shall briefly describe rhics , and discuss what can and can not be reproduced by models based on ideal hydrodynamics . rhics have been produced in several experiments , namely the super proton synchrotron ( sps ) at cern , the relativistic heavy - ion collider ( rhic ) at brookhaven and the large hadron collider ( lhc ) once again at cern . the rhic experiments in particular are described in detail in the so - called ` white papers ' , which are possibly the most reliable source on the subject @xcite . one of the goals of the rhic program is to probe into possible new phases of nuclear matter at high energies . in such a high energy phase , matter is expected to form a plasma of gluons and ( massless ) quarks ( quark - gluon plasma , qgp ) . virtually all theoretical analysis of rhics assume a space - time picture of the collision provided by the bjorken model @xcite . the colliding nuclei are seen as slabs of quark and gluon matter . in the center of mass frame , both slabs approach each other at near light speed . upon collision , the two slabs of matter will mostly go through each other , leaving behind a wake of hot plasma . we may then distinguish three different regions : the two fragmentation regions corresponding to the receding slabs , and the central region corresponding to the plasma in between . we are interested in phenomena in the central region . at the time of crossing a number of hard scattering processes will occur , whose products will reach directly the detectors . these hard processes are unrelated to the nonequilibrium dynamics of the plasma ; and may presumably be predicted on perturbative qcd grounds . in what follows , we will assume this hard component has been isolated despite great difficulty to achieve this in reality . the hot plasma will expand and cool , and eventually fragment into ordinary particles in flight intercepted by the detectors . we wish to predict the number of particles of each species to be detected , as a function of the angle @xmath54 between the direction of flight and the direction @xmath55 of the beam . it is remarkable that with this simple picture we can state a first observable prediction already . indeed , because of lorentz contraction , we may think of the approaching slabs as infinitely thin in the direction of motion @xmath55 , and in a first approach to the problem , as infinite and homogeneous in the transverse directions @xmath34 and @xmath56 . this picture is invariant under boost in the @xmath55 direction , and so is the final distribution of particles . so if we parametrize the momentum of an out - going particle as @xmath57 , @xmath58 and @xmath59 , then the distribution of particles may depend only upon the transverse momentum and @xmath60 . in particular , it must be independent of @xmath61 since @xmath62 is not invariant . it is conventional to plot the yield of the collision in terms of the _ rapidity _ @xmath63 , defined by @xmath64 , or rather the pseudorapidity @xmath65 , $ ] @xmath66 rapidity and pseudo - rapidity agree at momenta which are large compared to the mass of the particle . then the prediction in this picture is that there is a _ plateau _ in the ( pseudo ) rapidity distribution , at least for small rapidity ( @xmath67 corresponds to the fragmentation rather than the central region ) . although this prediction is not quantitatively borne out by the rhic data @xcite , the experimental curve flattens enough at low rapidity that it may be accepted as a working rough approximation . we may elaborate the bjorken picture further . let us assume that the plasma is formed on the plane @xmath68 at the time @xmath69 of the collision , and then expands along the @xmath55 direction . a given plasma element will cool according to its own proper time @xmath70 . eventually , at some given constant @xmath70 surface , the plasma will be cold enough ( and/or dilute enough ) to break up into hadrons . assuming that the product hadrons are thermally distributed , massless and at zero chemical potential , the bose - einstein distribution predicts that the energy per particle is @xmath71 @xmath0 . since temperature is constant on the break up surface , this means that in all collisions particles should have the same average energy . indeed , it is observed that the energy per particle is about @xmath72 , regardless of center - of - mass energy and impact parameter . to obtain a more quantitative description of the process , we describe the plasma as a relativistic ideal fluid . to close the hydrodynamic system of equations we must provide the equation of state . the central feature of this is the `` softening '' near the critical point , meaning that the speed of sound @xmath73 as we approach the transition point . the softening of the equation of state affects the evolution of the fireball , which then becomes a signal of whether the transition point has been reached or not . since perfect fluids conserve entropy , the total entropy within the fireball remains constant , and @xmath0 scales as @xmath74 so , if the expansion is one - dimensional , and we consider the volume enclosed between two fixed rapidities , then @xmath75 where @xmath76 is the proper time . in particular , the energy density scales as @xmath77 . we now consider more closely the phenomenon of break - up @xcite . assume this occurs on a 3-dimensional surface @xmath78 defined by some equation @xmath79 if @xmath80 is a solution , then the normal vector at @xmath80 is @xmath81 , @xmath82 we shall assume that @xmath83 is timelike . the invariant measure on @xmath84 is given by @xmath85 let us assume that right up to break - up we can describe matter as a perfect relativistic fluid , and as a noninteracting relativistic gas thereafter . let @xmath86 be the four killing vectors of minkowski space . then gauss theorem shows that the quantities @xmath87 and @xmath88 are continuous across the break - up surface ( we shall consider only one conserved current , corresponding to , say , the baryon number ) . these conditions plus the equation of state of the hadronic phase define the energy density , pressure , baryon number density ( or equivalently , the temperature and chemical potential ) and the four - velocity of the hadrons at break - up . the detailed spectrum is found by assuming that the hadrons are thermally distributed . the total number of emitted particles is @xmath89 where the integral is over some @xmath90@xmath91 surface well to the future of the collision . because of gauss theorem , we may replace the integral by an integral over the break - up surface ( we may have to complete this surface to get a cauchy surface , but the particle density flux will vanish on these additions anyway ) . but then we may use the matching conditions to express this integral in terms of the particle current before break - up . we obtain the total number of emitted particles as @xmath92 in practice , we may wish to smear a little the position of the break - up surface , thus writing the total number of emitted particles as @xmath93 \sigma _ { , \mu } n_{hydro}^{\mu } \ ] ] the total number of particles of species @xmath94 with momentum @xmath95 is @xmath96 \delta \left ( p_{i}^{2}-m_{i}^{2}\right ) \frac{\sigma _ { , \mu } p_{i}^{\mu } } { \left [ \exp \left ( -\beta _ { \nu } p_{i}^{\nu } -\mu_{i}\right ) -\varepsilon _ { i}% \right ] } \label{spec1}\ ] ] where the temperature four vector and chemical potentials are read out from the ( ideal ) energy momentum tensor and charged currents of the fluid just before break up . the two basic observables are the total number of particles with transverse ( respect to the beam axis ) momentum @xmath97 , which is usually given in terms of the transverse mass @xmath98 and the elliptic flow coefficient @xmath99 which results from fitting the particle spectrum in the transverse plane to a second harmonic @xmath100 where @xmath101 is the angle measured from the reaction plane . this is equivalent to considering an elliptic fireball , in which case @xmath102 measures the eccentricity of the ellipse . the first harmonic is called directed flow , and would represent a shifted spherical fireball in the transverse plane . the agreement of predictions from hydrodynamical simulations with experimental data is good , provided the simulation is started very early ( earlier than @xmath103 /@xmath104 after the collision ) . if one believes that the validity of hydrodynamics demands ( local ) equilibration , this very short time is somewhat of a puzzle . of the several predictions of hydrodynamic models , elliptic flow is one of the most compelling , because it is hard to match by alternative models . let us show how elliptic flow arises in hydro models . let the velocity of the fluid ( with respect to laboratory time ) in the longitudinal direction be @xmath105 . we may define a longitudinal rapidity y_l=12 let us take a boost invariant flow in the beam direction as background solution . then @xmath106 and the longitudinal rapidity becomes identical with the space time rapidity y_l==12 we introduce milne time @xmath107 and milne coordinates @xmath108 . define @xmath109 . we find -_=1-v_^2 introducing the transverse rapidity y_=12 then @xmath110 and the four velocity becomes u^=y _ ( , , _ ) or else in milne coordinates u^=y _ ( 1,0 , _ ) the interval element in milne coordinates is @xmath111 , and so there are nontrivial christoffel symbols ^_= and ^_=1 in the strictly boost invariant case @xmath112 we have @xmath113 and @xmath114 , although @xmath115 . the covariant euler equation is identically satisfied , @xmath116 , while the energy density @xmath117 for conformal matter . obviously boost invariant expansion can not create elliptic flow . the next step is to consider a nontrivial dynamics in the transverse plane . it is convenient to parametrize the transverse plane in terms of lagrangian coordinates @xmath118 . then the full coordinates are @xmath119 ; the transformation back to euler coordinates is given by the system _ q^ a= v^i ( , x^i ) with initial condition @xmath120 . now write ds^2&=&-d^2+^2d^2+_ijdx^idx^j&=&-d^2+^2d^2+_ij ( dq^a+v^id ) ( dq^b+v^jd ) in the new coordinates the metric reads =( ccc - ( 1-g^abn_an_b ) & 0 & n_b + 0 & ^2 & 0 + n_a & 0 & g_ab ) where g_ab=_ij n_a = v_i observe that @xmath121 , so @xmath122 . the lapse function is then @xmath123 and the four metric determinant is @xmath124 , where @xmath125 . in milne - lagrange coordinates the four velocity is @xmath126 , so u^_;=1y _ and the ideal hydrodynamic equations t+13u^_,&=&0^+^t&=&0 give t= and ^2y_^a_00 + 1t=0 we shall consider only linear deviations from homogeneity , where x^a = q^a+^a ( , ) @xmath127 n^a = v^a=^a ( , ) = 1+^a_,a ^a_00=^a ( , ) so , writing @xmath128 -13 - 13^b_,ba = -_,a given our initial conditions , there are no transverse velocity components . writing @xmath129 we get -13 - 13*=-write * = e^ikq_k then = e^ikq_k -13+_k=-_k write @xmath130 -13 + 13 _ k=0 the solution which is regular at @xmath131 is @xmath132 . as @xmath133 we get @xmath134 , @xmath135 , so c_k=(k)^-1/3_k we may now find the velocities v_a=_,a=_,a=()e^ikqk()^2/3_kj_1/3(k/ ) at early times @xmath136 , and thus the flow transforms an initial density ( or pressure ) gradient into a flow anisotropy . this simple effect is remarkably hard to reproduce in parton models , under any realistic initial condition . the important conclusion is that even in this simple model we get a qualitative picture of a relativistic heavy ion collision that serves as a framework for further analysis . in reality the flow is not really boost invariant , and it is certainly not homogeneous in the transverse plane . the hydrodynamic models offer a natural way to couple flow to pressure gradients in the transverse plane , and in this respect are superior to all other known alternatives . on the other hand , ideal hydrodynamic models overestimate the amount of anisotropy in the transverse flow @xcite ; this underlines the need for a relativistic theory of real fluids to really model relativistic heavy ion collisions . let us return to our analysis of non relativistic fluids . our starting point is eq . ( [ dsdt ] ) , where now we allow for entropy production , which must be positive . this suggests modifying the constitutive relations for an ideal fluid . we must take into account that @xmath14must be symmetric , and the relativistic constraint @xmath137 this suggests writing t_ij=12(p^jv^i+v^jp^i)+p^ij+_ij where _ ij=-_ij-_ij_kv^k _ ij=_iv_j+_jv_i-23_ij_kv^k and j^ j_u=(+p)v^j+12(p^jv^2-v^j(v_ip^i))+v_i^ij _ j=-_j in what follows we assume @xmath138 . with these constitutive relations we find = -_js^j+(_j)^2 + t^ij_ij+t(_jv^j)^2 + 1 t ( ) the last term vanishes in the rest frame and will not affect our argument . to find the relativistic generalization of this theory , we observe that in the rest frame t^= ( cc & 0 + 0&p^ij+^ij ) thus in any frame we may write t^= t^_0+^ where @xmath139 and ^=-^ where @xmath140 is the shear tensor _ = ^_^_this is the so - called chapman - enskog prescription , and although it leads to a covariant generalization of the navier - stokes equations , it is untenable . the conservation equations read t^_0,+^_,=0 now t^_0,=_,u^u^+(+p)u^_,u^+(+p)u^u^_,+^p_,=0 therefore + ( + p)u^_,-u_^_,&=&0^+^&=&0 moreover , since @xmath141 is both traceless and transverse , we may write the first equation as + ( + p)u^_,+12^_=0 since in the chapman - enskog approach the viscous energy - momentum tensor ( vemt ) is slaved to the four velocity , the only dynamic degrees of freedom are the components of the four velocity themselves . let us consider the linear perturbations around a solution with @xmath142 constant . let us write @xmath143 , @xmath144 and @xmath145 . we seek solutions that depend on a single variable @xmath146 . we further decompose @xmath147 , @xmath148 . since the theory is covariant , we may always perform a lorentz transformation to the rest frame , namely we introduce new time and space variables t&=&u^0t - uxx&=&u^ 0x - ut and new frequency and wave number &=&u^0-ukk&=&u^ 0k - uwith @xmath149 and @xmath150 . we seek solutions where @xmath151 is real and @xmath152\le 0 $ ] . in this frame @xmath153 , @xmath154 and @xmath155 . the linearized perturbations must have @xmath156 . for the other variables we seek a solution t=t e^-i t+ik_jx^j u^i=^ i e^-i t+ik_jx^j the equations of motion become -i+ic^2k_j^j&=&0-i^j+ik^j+ ( ( k^2+k_^2)^j+13k^jk_k^k ) & = & 0 where @xmath157 . for a transverse perturbation @xmath158 , @xmath159 we must have @xmath160 . we observe that because @xmath161 , the imaginary parts of @xmath162 and @xmath163 have the same sign , so we can check stability by checking the imaginary part of @xmath163 . under a galilean transformation @xmath164 and @xmath165 , so the theory is stable in all frames . however , under a relativistic transformation @xmath166 . if @xmath167 we get a quadratic equation for @xmath163 . for @xmath168 this is ^2 - 2(ku+ ) + = 0 with solution =ku when @xmath169 , one solution is the expected one @xmath170 but the second solution becomes @xmath171 and is unstable . for a longitudinal perturbation @xmath172 and the dispersion relation is ^2-c^2 ( k^2+k_^2)+43i ( k^2+k_^2)=0 the same analysis as before leads to a cubic equation . when @xmath169 two roots correspond to damped sound waves ; the third root @xmath173 is unstable . in the kinetic theory description @xcite the transport equation reads p^_f=(p^0)i_col @xmath174 is the so - called relaxation time . the current is j^=edpp^f and the emt is t^=dpp^p^f where dp = = ( ( p^0-p ) + ( p^0+p ) ) for simplicity we assume massless particles . we do not assume particle number conservation . to enforce energy - momentum conservation we require dpp^(p^0)i_col=0 assuming for simplicity maxwell - boltzmann statistics , the equilibria are of the form f_0=\{-|_p^| } for a given @xmath42 we can always find a local equilibrium distribution @xmath175 such that the ideal fluid energy momentum tensor built from it t_0^=dpp^p^f_0=u^u^+p^ [ emt0 ] obeys t_0^u_=-u^ [ ll0 ] with the same energy density and four velocity as from the landau - lifshitz prescription . we then define a temperature from = t^4 where @xmath176 . using the maxwell - juttner distribution rather than the bose - einstein one is equivalent to the approximation _ n=1n^-4=1 whereby @xmath140 becomes @xmath177 . we make this approximation from now on . it follows that the viscous energy momentum tensor ^=t^-t_0^ [ vemt ] is traceless and transverse ^u_=0 [ ttc ] we parametrize f = f_0 with @xmath178 at equilibrium . the transversality condition becomes dpp^(-u_p^)f_0z=0 [ trans ] we assume a simple boltzmann type entropy flux @xcite s^=-dp((p^0 ) ) p^fwe get the entropy production s^_,=dpi_col the chapman - enskog procedure seeks a formal expansion for @xmath179 in powers of @xmath174 . to this end , one parametrizes z = z_1+^2z_2+ the `` spatial '' derivatives @xmath180 and @xmath181 are regarded as zeroth order quantities , while the `` time '' derivatives @xmath182 and @xmath183 are derived from energy - momentum conservation , which is also a necessary consistency condition . to find @xmath184 we only need the linearized collision integral . this must be a symmetric operator , it must obey the energy momentum conservation constraint , must lead to non negative entropy production and must admit thermal distributions as the only homogeneous distributions . to satisfy these requirements , we write i_col ( p ) = ff_0 ( p ) the second term is there to enforce the constraints , namely dp(p^0)p^f f_0z = dpdp(p^0)p^ff_0kf f_0z since the kernel @xmath185 is symmetric and these hold for any @xmath179 we must have dpk(p^0)p^f f_0=p^(p^0 ) it also makes @xmath186 to vanish when @xmath179 is just a variation in @xmath187 . we adopt the anderson - witting prescription @xmath188 @xcite . write k = k_p^(p^0)p^(p^0 ) then k_i^=^ _ where i^=u_dp(p^0)p^p^p^ f_0=a_3 a_3=dp|-u_p^|^3 f_0=t^5 this means that k_=a_3 ^ -1finally the collision integral is i_col ( p ) =| -u_p^| f_0 ( p ) if @xmath179 satisfies the constraint eq . ( [ trans ] ) , then the second term is zero and the entropy production is positive , provided @xmath189 . to first order z_1=- _ , [ foc ] now in general _ , & = & 1t&=&1tso , using the zeroth order time derivatives and @xmath190 z_1=_p^p^ [ trans2 ] the viscous energy momentum tensor is _ 1^=dpf_0_=-^ [ vemtce1 ] where = dpf_0|-p^u_|^3=t^4 the chapman - enskog procedure can not generate a true dynamical equation for the viscous emt , and thus does not solve the causality / stability problems . the grad approach takes the different strategy of keeping the form eq . ( [ trans2 ] ) , but replacing @xmath191 by a new tensor @xmath192 regarded as a new independent variable . z_g = c_p^p^ [ gradansatz ] @xmath192 is defined up to a multiple of @xmath24 , so we may assume it is traceless , and because of the constraint eq . ( [ trans ] ) it must be transverse . the viscous energy momentum tensor becomes _ g^=dpf_0c_=-c^ [ vemtg ] the next step is to substitute this into the boltzmann equation . the nonlocal term vanishes and we get ( p^u_)f_0z=[be2 ] we wish to extract from there an equation for @xmath193 . the time - honored procedure is to consider the moments of this equation . since @xmath179 is even the zeroth moment vanishes , and the first moment gives back energy - momentum conservation . dpp^p^f_0(1+z ) = 0 therefore the first non trivial choice is to use the second moments dp(p^0)p^p^p^f_0(1+z ) = dpp^p^|-p^u_|f_0z to evaluate these equations , we use the following identities c^=c^ where c^= \{3u^u^u^ + ^u^+^u^+^u^-u^c^- u^c^-u^c^ } it is not really possible to kill all second moments simultaneously . we shall be happy to kill the transverse traceless contribution , namely t^5c^=c _ which reduces to u^c_&-&_c^ -_c^&+&13_c^+13_c^=c _ [ ge ] keeping only linear terms , this is an equation of maxwell - cattaneo type for @xmath192 @xcite _ + c_-_+=0 actually , we may get an expression for the time derivative of @xmath179 directly from the transport equation and then use it to find an equation for the viscous energy momentum tensor @xcite . however , beyond the leading order this equation involves integrals which can not be simple expressed in terms of the known moments of the distribution function . therefore to obtain a definite equation we must provide a _ closure _ , i. e. , an expression for @xmath179 allowing us to compute the integrals . for example , if we use the grad ansatz @xmath194 we get an equation with the same structure as eq . ( [ ge ] ) , though the coefficients may be different . we wish to check that the grad approach is consistent with causality and stability @xcite . introducing a new perturbation for @xmath195 we get the equations -i+ic^2k_j^j&=&0-i^i+ik^i -ik_jc^ij&=&0-ic^ij+c^ij - i ( k^i^j+k^j^i-23^ijk_k^k ) & = & 0 we revert to the chapman - enskog equations with the replacement @xmath196 . for transverse perturbations we now get a quadratic equation ( 1-i)+ik^2=0 it is easy to see that both roots are stable when @xmath197 . for general @xmath198 stability obtains if @xmath199 . it is interesting to observe that with the expressions above @xmath200 . we now consider longitudinal waves with @xmath168 . the dispersion relation is ( 1-i)+43u^2_0i ( k - u)^2=0 we rearrange this as +=0 where c^2=(1 + ) this is (-)(+ ) + = 0 in the formal limit @xmath201 we find three real roots @xmath202 , where @xmath203 . at large but finite @xmath174 , the solution that goes to zero behaves as _ 0= and it is stable . the solutions which converge to @xmath204 behave as 2ck_(-_)+ ( ku_)^2=0 so they are stable too . of course , we also have solutions with @xmath205 and @xmath206 , which are obviously stable . the grad approach as we have presented it still has the problems that the grad ansatz eq . ( [ gradansatz ] ) does not lead to a non negative one particle distribution function ( quite the opposite , since @xmath207 must be negative in some direction in momentum space ) and that it is unclear how to introduce nonlinear terms . to overcome these problems we need a better motivated closure for @xmath179 . we shall try to find it by seeking the value of @xmath179 which minimizes entropy production for a given vemt , and satisfies the constraint eq . ( [ ttc ] ) @xcite . adding lagrange multipliers @xmath208 we get the equation \{i_col+}= - _ p^p^f_0 [ epvm ] the point is that this equation has bounded solutions for any value of the right hand side . we may also regard it as a means to obtain a formal series solution for @xmath179 in powers of @xmath174 , where , if we keep only the first order term , we recover the grad ansatz . it is convenient to introduce a new unknown @xmath209 $ ] , with inverse transformation @xmath210 . in terms of the new unknown , the equation reads i_col= -2 _ p^p^f_0 - 12[epvm3 ] to lowest order we may neglect the second term in the right hand side . observe that in any case @xmath208 is defined up to a multiple of @xmath24 . we use this freedom to require @xmath211 . transversality requires 0=dp(p^0)p^f_0 _ p^p^=_a_3therefore @xmath208 must be traceless and transverse . to lowest order = 2 let us further expand the exponential . in the rest frame the energy momentum tensor reads t^00&=&t^4t^0i&=&0t^ij&=&13 t^4 ^ij-^ij [ tfromz ] the kinetic equation to lowest order in @xmath174 reads p^_f_0=2 ( -u_p^ ) f_0 the meaning of this equation is as a generating equation for its moments . the first order moments give energy momentum conservation . to find an equation for @xmath208 we go to the rest frame and multiply both sides by @xmath212 . integrating and discarding the trace part , we get _ ij-_ij-= we now turn to consider nonlinear corrections to this equations . the strategy we are following consists on improving the stability of the theory by adding new variables obeying dynamical equations of their own . in many approaches , such as the so - called israel - stewart theory @xcite or extended thermodynamics @xcite the extra variables are the components of the viscous energy - momentum tensor @xmath213 itself . the dynamical equations for @xmath213 have been derived in a number of ways , such as carefully taking moments of the kinetic equation @xcite , a systematic gradient expansion of the kinetic theory @xcite , from ads - cft correspondence @xcite or simply writing down all terms consistent with the symmetries of the theory up to a certain order @xcite . we shall call these theories `` second order fluid dynamics '' ( sofd ) for short , and take the presentation in @xcite as a suitable representative . in our approach ( epvm , or entropy production variational method ) , on the other hand , the new variables are the lagrange multipliers enforcing the constraints on entropy production ; @xmath213 itself may depend nonlinearly on the lagrange multipliers . our approach is therefore closer to the geroch - linblom divergence type theories @xcite , although we shall not demand that the resulting theory conforms to the dissipative type framework . to obtain further insight on the meaning of these theories , we shall apply them to the case of boost invariant flow . adopting milne coordinates @xmath214 is diagonal ; we write @xmath215 , @xmath216 , @xmath217 . energy - momentum conservation yields + 1 ( 43- ) = 0 we also write @xmath218 , @xmath219 . recall that @xmath220 , @xmath221 . we get , discarding a derivative of the temperature , _ r+=43t - a_1 _ r [ foeq ] where we have written @xmath222 for the relaxation time to avoid confusion with milne time , and a_1= the linearized collision term we are using is too simplistic to allow for a reliable derivation of nonlinear terms ; however , in this case we may exploit the symmetries of the problem , which indicate that there is a single true degree of freedom @xmath223 underlying the viscous energy momentum tensor . therefore we generalize eq . ( [ tfromz ] ) to = + h_1 ( ) ^2 and eq . ( [ foeq ] ) to _ r+=43t - a_1_r - h_2^ 2 [ soeq ] we determine the new transport coefficients @xmath224 by asking that , in a stationary situation , the relation between @xmath225 and the shear @xmath226 , to second order in @xmath222 , agrees with a systematic expansion for a boltzmann gas @xcite . indeed , to second order we may write = and so , neglecting time derivatives = 43-a_1_r - ( h_2-h_1 ) ^2 matching against sofd yields @xcite _ r & = & & = & a_1&=&43h_2-h_1&=&14 t where @xmath227 is the shear viscosity and @xmath228 the entropy density , both extracted from a realistic equation of state . an analysis of the possible origins of the @xmath224 coefficients shows that both are of the same order of magnitude , @xmath229 . when this holds , the behavior of the model is relatively insensitive to the actual values of @xmath230 and @xmath231 . to break the degeneracy between @xmath230 and @xmath231 , let us consider a situation where the shear @xmath232 terms are negligible and the transport coefficients are constant . the sofd equation admits a stationary solution with @xmath233 ; @xmath225 is otherwise unbounded . the epvm equations on the other hand imply a lower bound @xmath234 , which is realized when @xmath235 . there is a steady solution when @xmath236 , which implies @xmath237 $ ] . in both cases , the steady solutions are unstable . for sofd , this implies the existence of runaway solutions , namely , solutions which begin below the fixed point ( if it is negative ) run away to minus infinity . in the epvm , on the other hand , we may eliminate the runaway solutions by demanding that the steady solution coincides with the lower bound , and adopting , for each allowed value of @xmath225 , the value of @xmath223 above the fixed point . this requires @xmath238 , and therefore @xmath239 . introducing the transport coefficient = the sofd equation reads @xcite _ r+=-43_r - ^2 [ sofd ] while the extended epvm yields the system @xcite _ r+&=&43t-43_r - & = & t one way to visualize the difference between these models is to consider the free decay of @xmath225 . suppose @xmath225 is observed to have the value @xmath240 at some time @xmath241 late enough that the @xmath242 terms in the equations may be neglected , and the transport coefficients regarded as constant . then , according to sofd , the further decay of the vemt is given by = where @xmath243 and @xmath244 , while from epvm we get = where = -1 besides the absence of runaway solutions already noted , the epvm provides for a faster decay of large fluctuations ( see fig . ( [ decay ] ) ) . the theory of relativistic real fluids has a curious history because while the stability problems we have discussed have been known for a long time @xcite , yet they do not seem to have elicited any strong response until fairly recently . there were known ways to improve the theory ( foremost the israel - stewart @xcite and extended thermodynamics theories @xcite ) , and also a family of theories which were known to be free of such problems on a rigorous basis ( the geroch - lindblom dissipative type theories @xcite ) . however , the former were presented as successive approximations to a yet unknown theory , and the physical foundations of the latter remained elusive @xcite . it was only the realization that relativistic real fluids might be produced in rhics that triggered an all - out attack on the problem , to the extent that it would be impossible to describe all this activity in a short review such as this . we have therefore aimed to present just the fundamental ideas behind the theory , what the main problems are , and which lines of thought seem to us likely to be fruitful , and this almost entirely from the formal side , leaving phenomenology to more knowledgeable authors . concerning the present state of the theory , it seems fair to say that we have a reliable understanding of evolution during the hydrodynamic era , and the beginnings of a theory of the freeze out transition . the early times of the collision are relatively much more poorly understood . in particular , we do not know how hydrodynamic behavior may arise on such short time scales as demanded by theory , though the work on non - abelian instabilities @xcite and qgp turbulence @xcite , on one hand , and on ads - cft correspondence on the other @xcite , makes those scales look not so unrealistic as they used to . we can only conclude that we are only witnessing the early childhood of the relativistic real fluids - rhics connection , and this is what makes this such an exciting field to work on . this work has been developed in collaboration with jernimo peralta ramos . it is supported in part by universidad de buenos aires , conicet and anpcyt ( argentina ) in this appendix we shall discuss the derivation of kinetic theory from quantum field theory . the presentation follows @xcite . @xmath256 is causal and @xmath257 is even , and both are real . a good deal of our discussion will revolve around the different properties of the propagators of the theory , that is , the expectation values of binary products of field operators with respect to the initial state . since field operators at different locations do not generally commute , we have several different propagators according to the ordering of the field operators within the expectation value . we can now begin the discussion of our subject matter . our goal is to recast the equations for the propagators in a way suitable to discuss the last stages of the equilibration process . we shall reduce the equations to the form of a kinetic equation , the so - called kadanoff - baym equation . further approximations reduce this to the boltzmann equation . to succeed , we need a way to identify the important terms and discard the irrelevant ones . this will be provided by the so - called adiabatic expansion , under the assumption that the propagators are almost translation invariant close enough to equilibrium . we say that a @xmath269 is almost translation - invariant if , when partially fourier transformed with respect to @xmath270 , the fourier transform is weakly dependent on the centroid " variable @xmath271 , i.e. , expressions involving @xmath273 may be classified according to their _ adiabatic order _ , namely , the number of @xmath36 derivatives appearing in the expression . we call this the _ adiabatic expansion_. when almost translation - invariance is verified , we may further reject all terms above a given adiabatic order . we call such a truncation of an adiabatic expansion an _ adiabatic approximation_. in other words , the adiabatic order is used as a tag to bunch together certain terms in the equations of motion in accordance to their derivative orders and the adiabatic approximation determines how many of those terms are kept . g_1eq ( x , k)=()g_eq ( x , k)where @xmath278 is the bose - einstein distribution . we generalize this to a nonequilibrium situation by defining the _ density of states _ @xmath279 out of the fourier transform of the jordan propagator to obtain the dynamics of the distribution function @xmath284 , we make use of the equation involving the noise kernel . let us call @xmath285 , @xmath286 , @xmath287 and @xmath288 . then to first adiabatic order we get @xmath289 where @xmath290 @xmath291 and @xmath186 is the _ collision integral _ for weakly coupled theories , a series of approximations allow us to reduce the off - shell kinetic equation to the more familiar boltzmann kinetic equation . we observe that in terms of the coupling constant @xmath293 we have , for a generic momentum @xmath2 , @xmath294 while @xmath295 . a second observation is that in general @xmath296 , which involves the coupling constants , will be much smaller than @xmath297 for a generic choice of @xmath2 . when the coupling constants go to zero @xmath298 , but the retarded propagator has a well - defined asymptotic value , and the density of states becomes @xmath299 in this limit the propagators are insensitive to the behavior of the distribution function off shell " ( i. e. , when @xmath300 ) , because the distribution function is always multiplied by the density of states , and this is very small there . therefore , only `` on shell '' modes ( i. e. , those for which @xmath301 ) really contribute to the field correlation functions . if our only concern is to follow the evolution of the distribution function on shell , we are allowed to replace the @xmath302 and @xmath303 coefficients in ( [ ke1 ] ) by their `` on shell '' values , namely @xmath304 and @xmath305 we thus obtain the kadanoff - baym equations @xcite the nontrivial content of the kadanoff - baym equations is given by the form of the collision integral , namely , which feynman graphs contribute to the self energies . we recognize the structure of the collision term as the difference between a gain and a loss term for particles moving in or out of a phase space cell around the point @xmath307 per unit time . taking @xmath308 for simplicity , we see that @xmath309 is the gain term , with @xmath310 accounting for stimulated emission of particles into the cell , while the other term is the loss term , which is proportional to the number of particles @xmath311 already there . 99 w. hiscock and l. lindblom , ann . 151 , 466 ( 1983 ) ; phys . d 31 , 725 ( 1985 ) ; contemporary mathematics 71 , 181 ( 1988 ) . d. rischke , proceedings of the 11th chris engelbrecht summer school in theoretical physics , cape town , feb . 4 - 13 , 1998 e. calzetta and b - l hu , nonequilibrium quantum field theory ( cambridge university press , cambridge ( england ) , 2008 ) p. romatschke , arxiv 0912.3663 p. v. ruuskanen , acta phys . b18 , 551 ( 1986 ) p. m. chaikin and t. c. lubensky , principles of condensed matter principles ( cambridge university press , cambridge ( england ) , 1995 ) w. israel , in l. oraifeartaigh ( ed . ) general relativity : papers in honour of j. l. synge ( clarendon press , oxford , 1972 ) , p. 201 . w. israel , in a. anile and y. choquet - bruhat ( eds . ) , relativistic fluid dynamics ( springer , new york , 1988 ) . s. chapman and t. cowling , the mathematical theory of non - uniform gases ( cambridge university press , cambridge , england , 1939 ) ( reissued 1990 ) ) . j. peralta ramos and e. calzetta , phys . d 82 , 106003 ( 2010 ) ; phys . d86 , 125024 ( 2012 ) ; phys . rev . d 87 , 034003 ( 2013 ) s. ono , in i. prigogine , ed . , advances in chemical physics , vol iii ( john wiley , new york , 1961 ) e. t. jaynes , ann . 31 , 579 ( 1980 ) . i. prigogine , introduction to thermodynamics of irreversible processes ( john wiley , new york , 1955 ) . r. landauer , phys . a12 , 636 ( 1975 ) . s. bruers , arxiv:0604482 ; s. bruers , c. maes and k. netocny , arxiv:0701035 yu . l. klimontovich , turbulent motion and the structure of chaos ( springer , berlin , 1991 ) yu . l. klimontovich , statistical theory of open systems ( springer , berlin , 1995 ) t. olson , ann . phys . 199 , 18 ( 1990 ) . t. olson and w. hiscock , ann . 204 , 331 ( 1990 ) g. k. batchelor , an introduction to fluid dynamics ( cambridge up , cambridge ( england ) , 1967 ) l. d. landau and e. m. lifshitz , fluid mechanics ( pergamon press , oxford , 1959 ) brahms collaboration , nucl . a757 , 1 ( 2005 ) . phobos collaboration , nucl . a757 , 28 ( 2005 ) . star collaboration , nucl . a757 , 102 ( 2005 ) . phenix collaboration , nucl . a757 , 184 ( 2005 ) . bjorken , phys . d * 27 * , 140 ( 1983 ) . t. sagawa , in m. nakahara and s. tanaka ( eds . ) , lectures on quantum computing , thermodynamics ans statistical physics ( world scientific , singapore , 2013 ) . j. l. anderson and h. r. witting , physica 74 , 466 ( 1974 ) ; ibid . 74 , 489 ( 1974 ) . m. takamoto and s .- inutsuka , physica a 389 , 4580 ( 2010 ) . j. c. maxwell , philos . trans . soc . london 157 , 49 ( 1867 ) . d. d. joseph and l. preziosi , rev . 61 , 41 ( 1989 ) ; addendum , rev . 62 , 375 ( 1990 ) . g. s. denicol , t. koide , and d. h. rischke , phys . 105 , 162501 ( 2010 ) , g. s. denicol , e. molnr , h. niemi and d. h. rischke , eur . j. a , 48 11 ( 2012 ) 170 w. israel , ann . ( ny ) * 100 * , 310 ( 1976 ) . w. israel and j. m. stewart , ann . ( ny ) * 118 * , 341 ( 1979 ) . d. jou , j. casas - vazquez and g. lebon , extended irreversible thermodynamics ( springer , berlin , 2001 ) . r. baier , p. romatschke and u. a. wiedemann , phys . c 73 , 064903 ( 2006 ) s. bhattacharyya , v. e. hubeny , s. minwalla , and m. rangamani , j. high energy phys . * 02 * , 045 ( 2008 ) ; m. natsuume and t. okamura , phys . d * 77 * , 066014 ( 2008 ) ; erratum - ibid . d * 78 * , 089902(e ) ( 2008 ) . r. baier , p. romatschke , d. t. son , a. o. starinets and m. a. stephanov , jhep 04 , 100 ( 2008 ) ; m. luzum and p. romatschke , phys . c 78 , 034915 ( 2008 ) ( ( e ) phys c 79 , 039903(e ) ( 2009 ) ) a. el , z. xu , and c. greiner , phys . c81 , 041901 ( 2010 ) . r. geroch and l. lindblom , phys . rev . d 41 , 1855 ( 1990 ) ; r. geroch and l. lindblom , ann . ( ny ) 207 , 394 ( 1991 ) ; e. calzetta , class . 15 , 653 ( 1998 ) ; j. peralta - ramos and e. calzetta , phys . c 82 , 054905 ( 2010 ) . g. b. nagy and o. a. reula , j. phys . a 28 , 6943 ( 1995 ) ; e. calzetta and m. thibeault , phys . rev . d 63 , 103507 ( 2001 ) ; j. peralta - ramos and e. calzetta , phys . d 80 , 126002 ( 2009 ) . s. mrowczynski , phys . b 314 , 118 ( 1993 ) ; s. mrowczynski and m. h. thoma , annu . rev . nucl . part . 57 , 61 ( 2007 ) ; b. schenke , m. strickland , c. greiner , and m. h. thoma , phys . rev . d 73 , 125004 ( 2006 ) ; c. manuel and s. mrowczynski , phys . d 74 , 105003 ( 2006 ) ; m. attems , a. rebhan , and m. strickland , phys . rev . d 87 , 025010 ( 2013 ) ; j. peralta - ramos and e. calzetta , arxiv:1309.5412 . s. floerchinger and u. a. wiedemann , jhep 11 , 100 ( 2011 ) . r. a. janik and r. peschanski , phys . rev . d 73 , 045013 ( 2006 ) ; y. v. kovchegov , nuclear physics a 830 , 395c ( 2009 ) ; v. balasubramanian , a. bernamonti , j. de boer , n. copland , b. craps , e. keski - vakkuri , b. mller , a. schfer , m. shigemori , and w. staessens phys . d 84 , 026010 ( 2011 ) ; d. garfinkle and l. a. pando zayas , phys . rev . d 84 , 066006 ( 2011 ) . l. kadanoff and g. baym , quantum statistical mechanics ( benjamin , new york , 1962 ) . e. calzetta and b - l . hu , phys . d37 , 2878 ( 1988 )

we discuss integral field spectra of the compact star - forming complex that is the brightest near - infrared ( nir ) source in the central regions of the starburst galaxy ngc 253 . the spectra cover the and passbands and were recorded with the gemini nir spectrograph during sub - arcsec seeing conditions . absorption features in the spectrum of the star - forming complex are weaker than in the surroundings . an absorption feature is found near m that coincides with the location of a c bandhead . if this feature is due to c then the star - forming complex has been in place for at least a few hundred myr . emission lines of br , [ feii ] , and hei m do not track the nir continuum light . pockets of star - forming activity that do not have associated concentrations of red supergiants , and so likely have ages myr , are found along the western edge of the complex , and there is evidence that one such pocket contains a rich population of wolf - rayet stars . unless the star - forming complex is significantly more metal - poor than the surroundings , then a significant fraction of its total mass is in stars with ages myr . if the present - day star formation rate is maintained then the time scale to double its stellar mass ranges from a few myr to a few tens of myr , depending on the contribution made by stars older than myr . if as suggested by some studies the star - forming complex is centered on the galaxy nucleus , which presumably contains a large population of old and intermediate - age stars , then the nucleus of ngc 253 is currently experiencing a phase of rapid growth in its stellar mass . = 0.8 cm = 1.0 cm

1602.01400

interactions and mergers can spur the large - scale re - distribution of material and angular momentum within a galaxy , thereby altering basic observational properties such as morphology . some of the material that is displaced by such events will be channeled into the central regions of galaxies , where elevated levels of star - forming activity will be triggered in concert with those at large radii . observations suggest that while elevated episodes of star - forming activity in a galaxy may die out first at large radii , it may continue for some time in the central regions ( e.g. soto & martin 2010 ) . given that the typical damping time for star burst activity in moderately low mass galaxies is many hundreds of myr ( mcquinn et al . 2010 ) , then we might expect to find evidence for extensive star - forming activity in the central regions of late stage starbursts that is spread over a broad time period . the barred spiral galaxy ngc 253 is the dominant member of one of the three concentrations that collectively make up the sculptor group ( karachentsev et al . 2003 ) . ngc 253 is experiencing a starburst ( e.g. rieke et al . 1980 ) , and is one of the nearest such examples of this phenomenon . ngc 253 does not have an entourage of bright companions , and it is not clear what event ( or events ) triggered the starburst . based on a survey of bright evolved stars in the ngc 253 disk , davidge ( 2010 ) suggests that ngc 253 interacted with a now defunct companion within the past 0.2 gyr . such an encounter could have spurred the formation of the bar , and caused the re - distribution of gas and dust throughout the galaxy ( e.g. hopkins et al . 2009 ) . alternatively , a flyby encounter with another large galaxy could also have triggered bar formation and starburst activity ( lang et al . 2014 ) . if a flyby happened more than @xmath8 myr in the past then identifying the perturbing galaxy may be problematic . the central regions of ngc 253 contain clues about the age and starburst triggering mechanism . the center of ngc 253 is heavily obscured at visible wavelengths , and so most previous studies have been conducted in the near - infrared ( nir ) . engelbracht et al . ( 1998 ) used long - slit spectra and narrow - band images to investigate the ionization properties , stellar content and structural characteristics in a 7.5 arcsec radius aperture that contains most of the [ neii ] , br@xmath4 , h@xmath3 , and mid - infrared ( mir ) flux . they conclude that the interstellar medium ( ism ) is excited by a radiation field that is powered by objects with t@xmath9 k , indicating that a large population of young stars is present . this young component formed within the past @xmath10 myr and contributes @xmath11 of the stellar mass in the region studied . there is a compact bright structure that defines the photometric center of ngc 253 in the nir . this object , which has a sub - arcsec characteristic width , and its immediate surroundings are the subject of this paper . following gunthardt et al . ( 2015 ) , we refer to this object as the ir core ( hereafter irc ) . modelling of br@xmath4 emission by kornei & mccrady ( 2009 ) suggests that the irc contains a large number of ionizing stars that have an age of 5.7 myr , and they estimate that its total stellar mass is @xmath12m@xmath13 . kornei & mccrady ( 2009 ) further note that the irc has been the site of an extended period of star - forming activity during recent epochs . this conclusion is based on the deep co absorption bands that are seen in the integrated spectrum at @xmath14 m . deep co bands in the integrated spectra of young clusters originate in red supergiants ( rsgs ) , and so are signatures of systems with ages older than @xmath15 myr . the irc is offset by a few arcsec from what has been identified as the kinematic center of the galaxy ( muller - sanchez et al . 2010 ; rosenberg et al . however , rosenberg et al . ( 2013 ) also find differences between the distributions of nir continuum light , br@xmath4 emission , and h@xmath3 emission , indicating that the stellar content near the center of ngc 253 has complex spatial and kinematic properties . additional evidence of complexity is presented by gunthardt et al . ( 2015 ) , who find sub - structuring and a complicated projected distribution of stellar types in and around the irc . they find that the irc coincides with the rotational center of h@xmath3 emission , and argue that the irc is the true nucleus of ngc 253 . the complex projected distribution of stars and gas , coupled with the evidence for sub - structuring , demonstrates the need to map the distribution of spectroscopic features in and around the irc . this task is well - suited to spectrographs that are equipped with an integral field unit ( ifu ) . in the present study , deep ifu spectra of the central regions of ngc 253 that cover the @xmath16 m wavelength interval are discussed . this wavelength interval contains absorption and emission features that can be used to trace the projected distribution of stars that span a wide range of ages , and identify areas of recent star - forming activity . the nucleus of the late - type disk galaxy ngc 7793 , which is the dominant member of another sculptor group concentration ( karachensheva et al . 2003 ) , was also observed to represent the nucleus of a non - bursting , isolated spiral galaxy . to the best of our knowledge , these are the first nir spectra of the central regions of ngc 7793 . bibby & crowther ( 2010 ) estimate from the h@xmath17 flux that the sfr in ngc 7793 is 0.45 m@xmath13 year@xmath18 , and this is @xmath19 lower than the sfr of ngc 253 measured by ott et al . carson et al . ( 2015 ) find that the photometric size of the ngc 7793 nucleus grows with decreasing wavelength , as expected if there is a spatially extended young population . unlike ngc 253 , there are indications that ngc 7793 has evolved in isolation throughout much of its life . the disk of ngc 7793 spans many scale lengths ( vlajic et al . 2011 ) , and radburn - smith et al . ( 2012 ) suggest that the outermost regions of its disk were populated by radial migration from smaller radii , rather than by the disruption of an earlier structure . davidge ( 1998 ) found that the ngc 7793 disk contains stars that span a broad range of ages , and the oldest of these have an age @xmath20 gyr . the age and radial extent of the ngc 7793 disk are thus consistent with it having been in place for a long period of time and not having been disrupted by a merger or interaction . details of the observations and the reduction of the data can be found in section 2 . archival nir images that provide supplemental information for interpreting the spectra are discussed in section 3 . the strengths and on - sky distributions of spectroscopic features that are used to probe stellar content are discussed in sections 4 and 5 . comparisons are made with model spectra in section 6 . a summary and discussion of the results are presented in section 7 . the spectra were recorded at gemini south ( gs ) on the night of ut november 21 , 2004 ut with the gemini nir spectrograph ( gnirs ; elias et al . 2006 ) as part of program gs2004b - sv-22 ( pi : davidge ) . the gnirs ifu is described in detail by dubbeldam et al . the ifu was an image slicer design with 21 slitlets . it was used in low resolution mode for these observations , and so a @xmath21 arcsec field was covered on the sky . each slitlet is 0.15 arcsec wide and subtends 32 pixels on the @xmath22 insb array in gnirs , so that the sampling along the slit is 0.15 arcsec pixel@xmath18 . each spaxal thus subtends @xmath23 arcsec . the irc was positioned at the center of the ifu science field . while there are no point sources in the ifu observations from which the image quality can be measured directly , the irc has a fwhm @xmath24 arcsec in the @xmath25band continuum wavelength interval defined in section 5.1 . given that the irc is an extended object , then this sets an upper limit to the size of the seeing disk . a complete observation consisted of ( 1 ) five exposures of the target galaxy , with each exposure recorded at slightly different dither positions to assist with the identification and suppression of bad pixels and cosmic rays , followed by ( 2 ) five observations of a background sky field recorded with the same exposure times as the galaxy observations . this galaxy@xmath26sky sequence was repeated three times in each filter . individual exposure times were 20 seconds for ngc 253 and 40 seconds for ngc 7793 . the spectra presented here thus have total on - source integration times per filter of @xmath27 seconds @xmath28 seconds for ngc 253 , and @xmath29 seconds @xmath30 seconds for ngc 7793 . ancillary data were recorded to monitor instrumental and atmospheric characteristics . an ar lamp spectrum was recorded after each galaxy@xmath26sky pair , while a series of flat - field frames were recorded at various times during the night , with the slit illuminated by a light source in the gemini facility calibration unit . telluric absorption features were monitored with spectra of the f4v star hr8999 , which was observed at different times throughout the night . hydrogen absorption lines in the hr8999 spectrum were removed by subtracting voigt profiles that were fitted to those lines . the first step in the data reduction sequence was the removal of sky emission . a sky emission template was constructed for each galaxy spectrum by interpolating in time between the sky observations , and the result was subtracted from the spectrum . the sky - subtracted frames in a given filter at each dither position were then co - added , and the results were divided by a flat field frame to correct for pixel - to - pixel sensitivity variations and non - uniformities in optical throughput . the mean flat - fielded exposures were aligned by centering on the peak signal in each slitlet , and the results were combined by averaging the signal at each shifted pixel location . this was followed by wavelength calibration , which was done using a dispersion solution obtained from the arc spectra . the wavelength - calibrated galaxy spectra were then divided by the telluric star spectrum . the central regions of ngc 253 have a complex morphology , and even a single pointing with the gnirs ifu samples a number of different structures . moderately deep archival images of the center of ngc 253 were thus examined in an effort to better understand the region near the irc . f128n and f164n wfc3 images that were recorded by the hubble space telescope for program 12206 ( pi : westmoquette ) and that had been processed with the wfc3 pipeline were downloaded from the canadian astronomical data center . while a combination of broad and narrow - band images were recorded for this program , the central regions of ngc 253 are saturated in the broad - band images , and so only the narrow - band data are considered here . three images were recorded in each filter , and these were aligned and median - combined on a filter - by - filter basis . the left hand panel of figure 1 shows the central @xmath31 arcsec of the stacked f164n image , with the ifu footprint indicated in black . the stacked f128n image was smoothed to match the angular resolution of the f164n image , and f128n - f164n colors are displayed in the right hand panel of figure 1 . the f128n f164n color varies throughout the area sampled with the ifu . a filament with moderately red color runs e / w through the irc , and is flanked by two pockets with blue f128n f164n colors . the detection of areas with very blue colors in the irc is perhaps not surprising given the spectroscopic evidence for very young stars there ( section 4 ) . the red filament may be a dust lane , or a region between the blue pockets that is dominated by red stars . the filament has a color that is comparable to the area to the south of the irc and is not as red as some of the suspected dust features , and this is consistent with the latter possibility . that the f128n f164n color varies in the irc foreshadows one of the general results of this paper , which is that the age distribution of stars in and around the irc varies with location the irc is not a simple stellar population , but is instead a compact star - forming complex . additional evidence to support this claim is presented in section 5 . aside from the irc , other less prominent structures are also sampled by the ifu , with most of these to the east and north of the irc . kornei & mccrady ( 2009 ) note the presence of dust lanes within a few arcsec of the irc in their nir images , and it is likely that some of the structures in figure 1 are not physically distinct entities , but may instead be artifacts of dust absorption . the regions between the irc and the knots that are @xmath32 arcsec to the north east of the irc are separated by a filament that owing to its very red color is a candidate dust lane . the north east corner of the ifu field contains an area with a blue f128n f164n color , and in section 5 it is demonstrated that this area shows spectroscopic signatures of very recent star formation . finally , the north west corner of the ifu samples a region with red f128n f164n colors , and a filamentary structure that appears to be a dust lane is also seen there in the f164n image . using the f164n image in figure 1 as a guide , spectra that sample three different regions were extracted for detailed investigation : ( 1 ) the irc , ( 2 ) the eastern portion of the ifu , where faint sub - structures are evident , and ( 3 ) the western portion of the ifu , where the light distribution is more uniform than elsewhere in the area sampled . the irc spectrum was constructed by summing signal within a 0.4 arcsec radius circular gather centered on the irc , which is where the signal was at least one half of the peak value . the other two spectra were obtained by combining signal in the @xmath33 arcsec areas at the eastern and western edges of the ifu . these regions thus contain signal from sources that are at least 0.8 arcsec from the center of the irc , and so contamination from the irc should not be a concern . the extraction regions are indicated in figure 1 . the spatial distributions of prominent spectroscopic features at finer angular resolutions are examined in section 5 , while the extracted spectra are compared with model spectra in section 6 . the extracted spectra of the irc are shown in figures 2 ( @xmath0 ) and 3 ( @xmath1 ) , where prominent emission and absorption features are identified . the continuum has been divided out using a low - order function that was fit to each spectrum . a high - order continuum function was not required as high frequency components in the system response were removed during the division by the telluric standard . spectra of the ngc 7793 nucleus are shown at the bottom of the figures . the ngc 7793 spectra were extracted by summing the signal in a 0.4 arcsec radius aperture , which is where the signal is at least one half of the peak value . there are obvious differences between the irc and ngc 7793 spectra . whereas there is a rich mix of absorption and emission features in the irc spectra , there is no obvious line emission in the ngc 7793 spectra . this suggests that the nucleus of ngc 7793 does not contain a large population of stars with ages @xmath34 myr . the first and second overtone bands of co , which are the most prominent absorption features in this wavelength region , are deeper in ngc 7793 than in the irc . however , a direct comparison of absorption features is affected by nebular continuum emission , which causes absorption features in the irc spectrum to appear weaker than they would be in the absence of this emission . the contribution made by continuum emission to the spectra can be estimated in a purely empirical manner using the line and continuum measurements made by armand et al . ( 1996 ) , who observed five compact galactic hii regions . the compact morphology of these hii regions makes them areas of concentrated line emission , and so at least some of the conditions near the center of ngc 253 may be replicated , albeit at much smaller levels of activity . the relative strengths of emission lines and the continuum depends on environmental properties such as the temperatures of the ionizing stars and the density of gas and dust . one way to identify a reference hii region for comparison with the irc would be to consider the ultraviolet spectral - energy distribution ( sed ) , as this is the wavelength range where the photons that power the emission originate . however , such a comparison is not possible for ngc 253 given the heavy obscuration towards the center of that galaxy . photometric measurements made in the mir , where the signal is dominated by thermal emission from dust that is heated by hot stars , offer an alternative . both the hardness of the radiation field and the properties of the dust affect the mir emission , and a template hii region was selected based on the w1 w2 color obtained from wise ( wright et al . 2010 ) all sky survey images . the w1w2 color of the hii region iras 12073 - 6233 matches the w1w2 color of the center of ngc 253 to within 0.06 magnitudes . in contrast , iras 17200 - 3550 and 17455 - 2800 have w1w2 colors that are 0.3 0.4 magnitudes redder than the center of ngc 253 . thus , iras 120736233 was selected to serve as the template for estimating the contribution made by nebular continuum emission to the nir light from the irc . the br@xmath4 line width in the extracted irc spectrum is @xmath35 m , and so using the entries for iras 12073 - 6233 in table 1 and 2 of armand et al . ( 1996 ) the nebular continuum is @xmath36 of the mean strength of the br@xmath4 line flux . the nebular continuum then accounts for @xmath37 of the signal from the irc in @xmath1 . the starburst99 models summarized in figure 8 of reines et al . ( 2010 ) predict that this level of continuum emission near @xmath38 m is expected for a system with an age 6 8 myr . rather than identifying a single hii region to serve as a reference object , an alternate approach might be to combine the results for all of the hii regions in the armand et al . ( 1996 ) sample to obtain a mean line - to - continuum emission ratio . in fact , the ratio of the br@xmath4 to continuum strength in iras 12073 - 6233 is close to the mean value of all 5 hii regions . hence , adopting mean values from all hii regions would yield similar results to those obtained using iras 12073 - 6233 alone . a 10% contribution from the nebular continuum has a non - trivial effect on the absorption spectrum . this is demonstrated in figures 4 and 5 , where a featureless continuum that accounts for 10% of the total light has been added to the ngc 7793 spectra . not surprisingly , the addition of this continuum weakens the absorption features in ngc 7793 , producing a better match with the irc spectra . the agreement between the absorption features in the irc and the ngc 7793 spectra with a 10% contribution from nebular continuum is examined in figures 4 and 5 , where the differences between the irc and the ngc 7793 nucleus @xmath26 continuum spectra are shown . subtracting the ngc7793 @xmath26 continuum spectrum from the irc spectrum results in a reasonable suppression of absorption features in the @xmath39band . weak emission features in the ngc 253 spectrum can also be more cleanly identified when the absorption spectrum is suppressed . the absorption features in the @xmath25band are not as well matched as those in the @xmath39band . in particular , the residuals near the 2.32@xmath40 m ( 31 ) co band indicate that this feature is stronger in the ngc 7793 spectrum than in the irc spectrum , although the residuals near the 2.295@xmath40 m ( 20 ) band indicate better agreement . a wavelength - dependent trend in the residuals could result if there is a significant thermal component in the irc continuum . the experiments considered here assume a nebular component that contributes a fixed percentage of the light . in fact , there is significant thermal emission associated with star - forming knots in the central regions of ngc 253 ( e.g. fernandez - ontiveros et al . 2009 ) , and this emission may contribute significantly near the long wavelength end of the @xmath25band . the residuals near 2.32@xmath40 m in figure 5 might then be the consequence of failing to account for thermal emission . the comparisons in figures 4 and 5 ostensibly suggest that the brightest red stars in the irc have spectroscopic properties that more - or - less match those in the nucleus of ngc 7793 . this agreement is somewhat surprising , as it suggests similarity in the red stellar content . more specifically , if the nir light in both systems is dominated by red stars then the agreement in line strengths could indicate comparable metallicities . however , the total @xmath1 magnitudes of these galaxies differ by @xmath41 magnitudes , in the sense that ngc 253 is the brighter of the two ( jarrett et al . 2003 ) , and so the irc might be expected to have a higher metallicity than the nucleus of ngc 7793 . perhaps the gas from which stars in the irc are forming is moderately metal - poor , possibly having originated in an accreted companion . however , a problem with this speculation is that there is no evidence for metal - poor gas outside of the irc . indeed , the co bands in the area around the irc are deeper than those in the nucleus of ngc 7793 ( section 4.2 ) , and have depths that are consistent with a super - solar metallicity ( section 6 ) . a basic assumption in the preceeding paragraph is that the nir continuum in the irc is dominated by red stars . however , if the contribution from hot stars to the total nir light is sufficiently large then absorption features from red stars will be veiled by the nearly featureless stellar continuum associated with ob stars . this can occur if there are large numbers of stars with ages @xmath6 myr , as rsgs do not form in this age range the composite sed of such a young population is then dominated by hot stars . in fact , it is evident from the right hand panel of figure 1 that the irc has relatively blue nir colors . the effect of a contribution from a large population of hot stars to the nir spectrum of the irc is investigated further in section 6 . the ballick - ramsey c@xmath3 bandhead near @xmath2 m is a prominent absorption feature in c star spectra , and so might be expected in the spectrum of a system that experienced large - scale star formation during intermediate epochs . in fact , an absorption trough with a depth of a few percent is seen in the irc spectrum near 1.78@xmath40 m . absorption at this wavelength is also seen in the east and west spectra ( section 4.2 ) . the absorption trough near 1.78@xmath40 m is not an artifact of continuum removal . this is demonstrated in the figure 2 inset , where the irc spectrum between 1.76 and 1.79@xmath40 m without continuum removal is shown . the absorption trough near @xmath2 m is clearly seen in the spectrum with the continuum in place . absorption near 1.78@xmath40 m is also evident in the @xmath39band irc spectrum in figure 2 of kornei & mccrady ( 2009 ) . the spectra of k and m giants do not show obvious absorption features at this wavelength ( rayner et al . 2009 ) . however , it is still not obvious that c@xmath3 has been detected in the irc , given the very high levels of extinction towards the center of ngc 253 . engelbrecht et al . ( 1998 ) estimate that a@xmath42 within 7.5 arcsec of the nucleus based on the brackett decrement , and a@xmath43 from nir colors . kornei & mccrady ( 2009 ) estimate that a@xmath44 from [ feii ] lines , and a@xmath45 from nir colors . because of the substantial obscuration , there will likely be significant absorption by diffuse interstellar bands ( dibs ) . geballe et al . ( 2011 ) discuss the detection of dibs between 1.77 and @xmath46 m in the spectra of galactic stars with a@xmath47 magnitudes , and an absorption feature at @xmath48 m is the deepest dib discovered in their spectrum of gcs3 - 2 . given that the extinction towards the center of ngc 253 is comparable to that towards the stars examined by geballe et al . ( 2011 ) then the c@xmath3 wavelength region likely also contains dib signatures . however , these signatures are not expected to be present in the nuclear spectrum of ngc 7793 , where the line - of - sight extinction is much lower than towards the irc . the residuals in figure 4 suggest that the irc and the ngc 7793 nucleus spectra have similar levels of absorption in the wavelength interval that hosts the c@xmath3 band . the spectra of ngc 253 east and west are shown in figures 6 ( @xmath39band ) and 7 ( @xmath25band ) . the difference between the two spectra is shown at the bottom of each figure . the largest differences between the spectra occur at wavelengths where there are strong emission lines , with many ( but not all ) of these lines being stronger in ngc 253 east , which is an area where bluer f128n f164n colors are seen ( figure 1 ) . as for absorption features , the co bands in the east and west spectra have similar depths . the scatter in the @xmath49 m interval , which covers part of the wavelength range where c@xmath3 absorption might be expected , is similar to that seen at other wavelengths . however , there is much larger scatter in the differenced spectra at wavelengths @xmath50 m . this likely is due to the amplification of the noise that occurs near @xmath51 m when correcting for telluric water absorption . many absorption and emission features in the @xmath0 and @xmath1 spectra of ngc 253 east and west have markedly different strengths when compared with their counterparts in the irc spectra . emission lines of h , hei , and [ feii ] in ngc 253 east and west are weaker than in the irc . this suggests that while very young stars are present outside of the irc , their density when measured with respect to the stars that dominate the nir continuum is lower than in the irc . h@xmath3 emission is an exception . the strongest h@xmath3 features in ngc 253 east and west are comparable to or even stronger than those in the irc , with no difference between ngc 253 east and west . this suggests that the source of the h@xmath3 excitation across the ifu field is well - mixed with the stars that dominate the nir light when averaged over arcsec angular scales . as noted by rosenberg et al . ( 2013 ) , this is consistent with the stars that power h@xmath3 emission having lower masses than those that power br@xmath4 and hei emission , as these objects would then have had more time to mix throughout the central regions of ngc 253 . the co bands in ngc 253 east and west are deeper than in the irc , and this is most noticeable for the first - overtone transitions . it is unlikely that this difference is a consequence of veiling of the irc spectrum by non - stellar continua , which was discussed in section 4.1 . in order for nebular continuum emission alone to veil the co bands in ngc 253 east and west so that they match what is seen in the irc then the contribution from this emission would have to be @xmath52 higher than estimated for the irc in section 4.1 . thermal emission also can not explain the differences in the depths of the @xmath53 m features , as the second overtone bands in the @xmath39band circumnuclear spectra are deeper than in the irc . the impact of a thermal contribution at these shorter wavelengths is negligible . rather than a non - stellar continuum , the differences between the irc and east and west spectra could be due to large numbers of hot stars in the irc , and this is considered in section 6 . the spectra that were investigated in section 4 were constructed by combining data over large portions of the area sampled by the ifu . this binning was done at the expense of angular resolution to boost the s / n ratio . however , there are some features ( most of them emission lines ) that are sufficiently strong that their projected distribution can be examined without binning . the on - sky behaviour of these features is examined in this section . there are wavelength intervals that are free of strong emission features , and the signal in these can be used to map the distribution of light from the stars that dominate the nir continuum free of contamination from prominent emission lines . two such wavelength intervals are between @xmath54 m and @xmath55 m ( hereafter referred to as hcont ) and @xmath56 m and @xmath57 m ( hereafter kcont ) . the distribution of light in kcont and the hcont kcont color are shown in figure 8 . the kcont and hcont kcont images show features that match those in figure 1 . the hcont kcont color of the irc is @xmath58 magnitudes bluer than that of the surroundings . mapping the distribution of absorption features provides important information about stellar content and the mix of stellar types . however , obtaining maps for all but the deepest absorption features is problematic given the modest s / n ratio in spaxals that are external to the irc . the first overtone co bands , which are the deepest absorption features in these spectra , are an exception . the depth of the ( 2,0 ) co band was measured in the wavelength interval @xmath59 m @xmath60 m , and a co index was constructed by dividing the mean signal in this wavelength interval by the kcont flux . the spatial behaviour of the co index near the center of ngc 253 is shown in the lower panel of figure 8 . the spectroscopic properties of the irc differ from those in the surrounding area . the co index in the irc is 10 15% weaker than in the surroundings , in agreement with the extracted spectra discussed in sections 4.1 and 4.2 . it is perhaps surprising that there are only minor variations in the co index in the area external to the irc given the large range of f128n f164n colors that are seen there in figure 1 . an exception is the pocket of deep co absorption near the northern edge of the ifu , which has a slightly redder hcont kcont color than its surroundings . if there are large numbers of rsgs in this area that have ages @xmath20 myr , and so are younger than those that populate most of the ifu field ( section 6 ) , then this area might be expected to have deeper co bands and redder colors . kcont , hcont kcont , and co images were also constructed for the central regions of ngc 7793 . apart from an obvious nucleus in the kcont image , these images do not reveal significant structure , and so are not shown here . in particular , the co index of the ngc 7793 nucleus is comparable to that in the circumnuclear area , which is contrary to what is seen in ngc 253 . the emission lines in the ngc 253 spectra probe a diverse range of environments , and so may not track light from the stars that dominate the nir continuum . in this section various emission line indices normalized to kcont are used to examine differences in the distribution of the line - emitting sources and the underlying stellar continuum . comparisons of this nature provide insights into the mix of stellar types throughout the area sampled with the ifu , including the identification of areas that have very young luminosity - weighted ages . spectral indices were computed for four emission lines using the wavelength intervals listed in table 1 . the line measurements were made in the wavelength interval between the points where the wings of the line merges with the adjacent continuum in these data . the continuum is measured on both sides of each line , and the mean continuum contribution was subtracted to obtain an index that tracks only line emission . lccc feature & line & continuum 1 & continuum 2 + & ( @xmath40 m ) & ( @xmath40 m ) & ( @xmath40 m ) + br13 & 1.61361.6449 & 1.60881.6131 & 1.64541.6499 + @xmath61feii@xmath62 $ ] & 1.64131.6485 & 1.63741.6408 & 1.64901.6524 + hei & 2.05532.0637 & 2.05082.0547 & 2.06432.0675 + br@xmath4 & 2.16122.1705 & 2.14292.1606 & 2.17112.1796 + the ratio of the br@xmath4 and kcont indices is shown in the upper left hand corner of figure 9 . if br@xmath4 and kcont have the same projected distributions then the image would be featureless . however , the br@xmath4 to kcont ratio peaks in the outer areas of the irc , with the highest ratios defining an arc that girds the western half of the irc . br@xmath4 emission thus has a more extended spatial distribution than kcont in and around the irc . the br@xmath4 to kcont ratio also peaks in the north east corner of the ifu field . the hei / kcont and [ feii]/kcont ratios are also high in this area ( see below ) . there are pockets with blue f128n f164n colors in this area ( figure 1 ) , and these may be areas of star - forming activity powering the br@xmath4 emission . the photons that excite br@xmath4 emission likely originate from o stars , and kornei & mccrady ( 2009 ) find that the irc contains a large number of stars with ages @xmath63 myr based on the strength of br@xmath4 emission . to the extent that br@xmath4 emission tracks young stars then the arc to the west of the irc is where the ratio of hot young stars to the stars that dominate the nir stellar continuum is locally highest . it is thus worth noting that the co index in the western arc of br@xmath4 emission is not weaker than in the main body of the irc the continuum from hot stellar photospheres evidently does not veil the co features in the outer regions of the irc . however , veiling is likely an issue in the main body of the irc ( section 6 ) . the relative strengths of lines in the brackett series can be used to investigate reddening , although there are caveats . while the brackett lines sample wavelengths where the effects of extinction are greatly diminished when compared with visible wavelengths , there may be areas of extremely high extinction that are largely opaque in the nir , and measuring the total line - of - sight extinction towards these areas will be problematic . moreover , while wide wavelength coverage provides the greatest leverage for estimating extinction , the lines progressively weaken towards higher - order transitions , resulting in a trade - off between sensitivity to reddening variations and noise in line strength measurements . in addition , lines that are at different wavelengths may not originate at a common optical depth , skewing extinction estimates . finally , relative line strengths depend on the electron temperature and density , and if these vary substantially across a region then the brackett decrement will not be a faithful probe of extinction . the ratio of the br13 and br@xmath4 indices is shown in the upper right hand corner of figure 9 . br13 was selected for comparison with br@xmath4 by balancing the desire for a wide wavelength baseline against line strength . there are also well - defined continuum intervals on either side of br13 . while the br13 to br@xmath4 ratio shows only subtle variations and there is considerable noise in this ratio outside of the irc , aperture measurements reveal a systematic tendency for the br13 to br@xmath4 ratio to decrease with increasing distance from the irc . if the electron temperature and density are more - or - less constant across the area sampled then this behaviour suggests that extinction becomes lower with increasing distance from the irc . this runs counter to the trend for color to become redder outside of the irc ( e.g. figure 1 ) . we speculate that the brackett decrement near the irc is tracking differences in ionization properties , rather than differences in line - of - sight extinction . in fact , the ratio of hei line emission strength to kcont discussed in the next section changes with distance from the center of irc , indicating that very hot stars are not uniformly distributed near the irc . prominent hei @xmath5 m emission suggests that w - r stars are present in and around the irc . the presence of these stars indicates that there has been large - scale star formation near the center of ngc 253 within the past few myr . the ratio of the hei and kcont indices is shown in the lower left hand corner of figure 9 . the w - r stars that presumably power the hei emission have a distribution that does not follow the stars that dominate kcont . there is an area of localized high hei emission with respect to kcont to the north west of the irc , and this is likely where the ratio of w - r stars to the stars that dominate the continuum near the irc is highest . in addition to being an area of excess ( when compared with stellar light ) br@xmath4 and [ feii ] emission ( see below ) , the pocket of enhanced hei emission is close to an area with blue f128n f164n color in figure 1 , which is consistent with there being a concentration of hot , blue stars . [ feii ] emission is excited in shocks , and is usually associated with snrs in star - forming environments . while the @xmath64 m [ feii ] line is one of the strongest emission features in the ngc 253 nir spectrum , it is blended with br12 in the gnirs data . based on the peak intensity of the [ feii]@xmath26br12 blend , coupled with the relative strengths of br11 and br13 , it appears that br12 contributes @xmath65 of the total signal in the [ feii ] index as defined in table 1 . this level of contamination will bias the [ feii ] index to follow the distribution of brackett hydrogen emission . the [ feii]/kcont ratio is mapped in the lower right hand corner of figure 9 . as with the other emission indices , [ feii ] does not follow the kcont light . there are four areas where there are localized peaks in the [ feii/]/kcont ratio . one is in the north east corner of the ifu field , where br@xmath4/kcont and hei / kcont also peak . the other three areas surround the irc . two of these are along the western edge of the irc and are close to areas that have a high br@xmath4/kcont ratio , and one of these in turn coincides with the peak in the hei / kcont ratio . the fourth peak in the [ feii]/kcont ratio is located to the south east of the irc . while the [ feii]/kcont ratios in the four regions are comparable , this is not the case for the br@xmath4/kcont ratio , indicating that the [ feii ] index does not parrot br@xmath4 emission . the distributions of br@xmath4 , hei , and [ feii ] emission are compared in figure 10 , where contours obtained from the br@xmath4/kcont and hei / kcont images are superimposed on the [ feii]/kcont image . the local peaks in the [ feii]/kcont ratio around the irc are offset slightly from those in br@xmath4/kcont and hei / kcont , in the sense that the [ feii ] emission peaks at slightly larger distances from the center of the irc when compared with br@xmath4 or hei . such an offset could result if the [ feii ] emission is tracking a shock front presumably powered by sne activity that is propogating outwards from the star - forming regions . we note that gunthardt et al . ( 2015 ) conclude that there is an outflow from the irc based on the distribution of [ neii ] @xmath66 m and si @xmath67 m emission . comparisons with model spectra can provide insights into the stellar content in and around the irc . highly evolved red stars are the dominant contributors to the photospheric nir light from simple stellar populations ( ssps ) that have ages @xmath68 myr . uncertainties in the physics associated with the most advanced stages of stellar evolution ( e.g. review by herwig 2005 ) thus play a significant role in interpreting the spectra of such systems . the well - known age - metallicity degeneracy adds additional uncertainties , as the color ( and hence spectroscopic properties ) of evolved stars depend on both age and metallicity . spectra constructed from ssp models described by leitherer et al . ( 1999 hereafter starburst99 ) and cordier et al . ( 2007 hereafter basti ) are compared with the observed spectra of ngc 253 and ngc 7793 in this section . the model spectra have a resolution @xmath69 , and so even prominent features such as the first and second - overtone co bands are blended with other molecular and/or atomic transitions . the baseline models assume solar metallicities and constant sfrs ( csfrs ) from the present day to a time in the past . given that the ngc 253 and ngc 7793 spectra were not flux calibrated , the comparisons focus on matching the absorption characteristics in continuum - corrected spectra , and no effort is made to estimate a sfr from the models . in any event , an estimate of the sfr depends on the star - forming history ( sfh ) , which is not well - constrained by the nir spectrum . the models are also used to examine the effect of veiling from very young populations , as this can have a significant impact on the depth of absorption features . csfr models constructed from starburst99 ssp spectra are shown in the left hand panels of figures 11 ( @xmath39band ) and 12 ( @xmath25band ) . the models have had the continuum removed , and have been shifted vertically for display purposes . the models are based on the solar metallicity padova isochrones described by bertelli et al . ( 1994 ) , with tp - agb evolution implemented using the procedures described by vazquez & leitherer ( 2005 ) . a two - component kroupa ( 2001 ) mass function is adopted . csfr models constructed from ssp spectra in the basti compilation are shown in the right hand panels of figures 11 and 12 . the model spectra are based on the solar metallicity isochrones described by pietrinferni et al . ( 2004 ) , and include tp - agb evolution as described by cordier et al . a chabrier ( 2001 ) mass function is adopted . the basti models do not include systems with ages @xmath70 myr , and so the starburst99 ssp models in that age range were used when constructing the basti csfr models . the first and second - overtone co bands are the dominant features in the models . these are deepest in the 50 and 100 myr spectra , reflecting the large contribution to nir light made by luminous low surface - gravity rsgs at these ages . the differences in the co band strengths due to age are not large , and the first - overtone co bands are only a few percent deeper in the 50 and 100 myr models when compared with models in which star formation continues over a longer time frame . the depths of the co bands do not change significantly in the csfr models with ages @xmath71 gyr , and so the co bands contain little information about the sfh of older composite systems . that these features are not sensitive to the sfh makes them of potential interest as metallicity indicators . there are age - related differences in the strengths of weaker absorption features in both spectral regions . these weaker features are blends of molecular and atomic transitions and their depths vary by a few percent for ages @xmath72 myr . the peak between the ( 2,0 ) and ( 3,1 ) co bands is also noticeably smaller in the 10 myr model than in the other models . significant differences between the starburst99 and basti models can be seen , with the basti models having larger age - related variations and deeper absorption features . these differences are greatest in the @xmath39band , and the most striking difference is associated with the ballick - ramsey c@xmath3 band . this feature is abscent in the starburst99 models , but is clearly present in the basti csfr models with ages @xmath73 myr . while c stars are usually associated with systems that have sub - solar metallicities , moderate numbers of c stars might be expected in solar metallicity systems . karakas ( 2014 ) finds that solar metallicity stars with masses between 2 and 4.5m@xmath13 will become c stars . the ratio of the number of c stars to the number of o - rich agb stars of spectral - type m c / m in intermediate age solar - metallicity systems is @xmath74 in her models . studies of c stars in the galactic disk suggest that the lower mass limit for c star formation may even extend to lower values than predicted by the karakas models ( e.g. feast , whitelock , & menzies 2006 ) , potentially increasing the possible constribution from c stars to the integrated nir light . the basti csfr models predict that c@xmath3 absorption should be present if star - forming activity has occured over a time interval of at least a few hundred myr . these models further suggest that an exceptionally high sfr during intermediate epochs is not required to produce detectable c@xmath3 absorption . rather , c@xmath3 might be detectable if the sfr during intermediate epochs is simply comparable to that at other epochs and the noise level in the spectrum is no larger than a few percent if the basti models are correct then c@xmath3 absorption with a depth of a few percent should be common in the spectrum of solar and sub - solar metallicity systems that have experienced extended periods of star formation during intermediate epochs . that c@xmath3 absorption is present in the model spectra of systems that have experienced a csfr over a large fraction of the hubble time is probably due to c stars being intrinsically brighter in the nir than older tp - agb stars . c stars then contribute significantly to the nir light of composite systems even though they form only over a limited range of stellar masses . the ngc 253 east and west spectra differ in the strengths of some emission features , but have very similar absorption features . the means of the ngc 253 east and west spectra were thus computed , and the results will be referred to as the circumirc ( circ ) spectra . the spectra of the irc , the circ , and the nucleus of ngc 7793 are compared with various models in figures 13 ( @xmath0 ) and 14 ( @xmath1 ) . the gnirs spectra have been smoothed to match the @xmath75 m resolution of the models . the ngc 7793 spectrum is considered first , as it is free of obvious emission features . there is mixed success in matching the ngc 7793 spectrum with the models . the absorption feature in the ngc 7793 spectrum near @xmath76 m overlaps with the c@xmath3 feature in the basti model spectra . however , many of the weaker features in the ngc 7793 spectra do not have obvious counterparts in the models . this might be attributed in part to binning , as the method by which binning is done can alter the appearance of low resolution spectra ( e.g. figure 6 of davidge et al . binning will presumably be less of a factor for very strong features , such as the depths of the first - overtone co bands , and the co bands in the ngc 7793 spectrum are slightly weaker than in the solar metallicity models . that the @xmath25band model spectra have deeper co bands than in the ngc 7793 nuclear spectrum might suggest a slightly sub - solar metallicity for the nucleus of ngc 7793 . this being said , bibby & crowther ( 2010 ) find a central o abundance of @xmath77log(o / h)@xmath78 for ngc 7793 , which is consistent with solar ( e.g. asplund et al . however , [ o / fe ] has not yet been measured in ngc 7793 . in addition , if the nir light from the nucleus is dominated by stars that formed during intermediate epochs then the nucleus might have a lower luminosity - weighted metallicity than the present - day gas . weak line emission may become evident when the ngc 7793 spectra are smoothed , as this increases the s / n ratio . in fact , subtle features appear in the smoothed ngc 7793 nuclear spectrum that coincide with emission features in the ngc 253 spectra . a modest population of hot young stars that power this emission may then be present in the ngc 7793 nucleus . that only a modest number of young or intermediate age stars are present near the center of ngc 7793 is perhaps not surprising . the frequency of blue nuclei among nearby disk galaxies suggests that the duty cycle for central star formation is on the order of @xmath32 gyr ( e.g. davidge & courteau 2002 ; ho et al . 1997 ) , and the weak emission line contribution might then indicate that the nucleus of ngc 7793 is currently between large - scale star - forming episodes . the basti and starburst99 models both underestimate the total amplitude of the ( 2,0 ) co band head in the circ spectrum . the comparisons in figure 14 thus suggest that the brightest rsgs in the circ have a metallicity that is slightly super - solar . the situation is different for the irc , as the first overtone co bands in that system are weaker than in the solar metallicity models . given that the overall depths of the co bands do not change greatly with sfh ( figure 12 ) , and that veiling by nebular emission does not change the depths of absorption features by huge amounts ( section 4 ) , then the difference in the co depths between the models and the irc spectrum might suggest that the irc has a sub - solar metallicity . however , such a conclusion assumes that the nir light is dominated by red stars . veiling of the co bands could be significant if a large fraction of the nir light originates from hot young stars , which have a near - featureless nir spectrum . previous nir spectroscopic studies of the central regions of ngc 253 have concluded that it is an area of intense recent star formation ( e.g. engelbracht et al . the presence of strong hei and br@xmath4 emission throughout the central regions of ngc 253 indicates that hot , young stars are present in large numbers , and the areas of the irc that have blue nir colors in figure 1 indicate that hot stars contribute significantly to the nir light there . the fraction of the mass of recently formed stars that can veil the absorption spectrum depends on the assumed age of the irc . consider the effect of adding a population with an age of 4 myr ( i.e. a population with massive stars that might power hei emission ) to that of the csfr starburst99 models with ages of 50 myr , 500 myr , and 10 gyr . these ages were selected to represent different evolutionary cases for ngc 253 . the 50 myr age explores the hypothesis that the irc is relatively young , as might be the case if the starburst was only recently triggered . the 500 myr model is included to represent the case in which the irc formed as part of a mature starburst that has been continuing for a few disk crossing times . finally , the 10 gyr model explores the case in which the irc is an old structure possibly the nucleus of ngc 253 that has experienced periodic levels of elevated star - forming activity throughout the age of the universe , and is once again the site of renewed star - forming activity . if the absorption feature in the irc spectrum near @xmath2 m is due to c@xmath3 then this favors the 500 myr and 10 gyr scenarios . assuming that the red stars in the irc have solar metallicities , then the fractional contributions that a 4 myr population must make to the total mass of the irc to diminish the depth of the ( 2,0 ) co band in the solar metallicity model spectrum to that seen in the irc are 70% ( 50 myr assumed age for the irc ) , 30% ( 500 myr assumed age ) , and 10% ( 10 gyr assumed age ) . the fractional contribution from the 4 myr population drops with increasing irc age because the m / l ratio in @xmath1 grows with increasing age , so that a smaller mass of 4 myr stars is required to alter the spectrum of the underlying system as its age increases . still , the assumed age of the irc notwithstanding , the irc is evidently being observed at a special time when it is experiencing rapid growth in stellar mass . if ( 1 ) the current level of star formation has remained constant during recent times and does not change over the near future , and ( 2 ) the stars that form in the irc stay bound to that structure , then the times required to double the original stellar mass of the irc are 3 myr ( 50 myr irc age ) , 9 myr ( 500 myr age ) , and 36 myr ( 10 gyr age ) . when compared with the irc , there is smaller fractional contamination by hot stars in the circ spectrum , and so the absorption features are less affected by veiling . the hei line in the circ spectrum is only one - third as strong as in the irc spectrum , while the br@xmath4 line is roughly half as strong as in the irc . if the light from the 4 myr component in the models discussed above is decreased by a factor of 3 then the ( 2,0 ) co band strength is diminished by only @xmath79 , as opposed to the 7% change that is predicted for the irc . if the circ spectra sample a larger mix of stars with ages @xmath80 myr than the irc spectra then a larger fraction of the nir light in the circ will come from bright rsgs . a larger fractional contribution from very young rsgs with a resulting smaller fractional contribution from hot stars could explain ( 1 ) why the co bands in the east and west spectra have similar depths despite differences in the strength of br@xmath4 emission , and ( 2 ) why the co bands in the pockets of strong hei and br@xmath4 emission seen in figures 8 and 9 are not weaker than in the surrounding areas . based on nir colors , engelbracht et al . ( 1998 ) argue that there is no significant veiling of the spectrum of the center of ngc 253 by a featureless continuum , and this is more - or - less consistent with the subdued hot stellar continuum level inferred here for the circ region . in contrast , the f128n f164n color data in figure 1 clearly shows that the irc contains regions with blue nir colors , where veiling is expected to be significant . we close this section by noting that the addition of a hot stellar continuum will weaken other absorption features , including the trough at 1.78@xmath40 m that might belong to c@xmath3 . adding a hot stellar continuum to the circ @xmath39band spectrum so as to reproduce the depths of the first overtone co bands in the irc will weaken c@xmath3 absorption by 50% . this is more - or - less consistent with the relative depths of the candidate c@xmath3 features in the circ and irc spectra in figure 13 . veiling of this size will also bring the depth of c@xmath3 absorption in the basti models into better agreement with what is seen in the circ and irc spectra . the stellar content near the center of the starburst galaxy ngc 253 has been investigated using nir ifu spectra recorded with gnirs on gs during sub - arcsec seeing conditions . the gnirs ifu samples a @xmath81 arcsec@xmath82 area , which corresponds to @xmath83 kpc@xmath82 at the distance of ngc 253 . the irc , which defines the photometric center of the galaxy , and several nearby sub - structures are sampled . the analysis of integrated nir light is at present one of the few means of examining the stellar content of the crowded central regions of ngc 253 . the use of an ifu allows the spectrum of the irc to be de - coupled from that of the surroundings , and the spatial distribution of stellar types to be mapped . it is clear that the irc is not an ssp . as noted by kornei & mccrady ( 2009 ) , the presence of deep co absorption bands are indicative of rsgs , which form in systems that have an age of at least 8 10 myr . however , the presence of br@xmath4 and hei emission indicate that irc has been the site of highly elevated levels of star formation within the past @xmath20 myr , the irc thus contains stars with an age spread of at least a few myr . barring a termination of star - forming activity then the age spread will increase with time . indeed , the total stellar mass of the irc is expected to double over time scales of a few myr to a few tens of myr if the pace of star formation that occured during the past 10 myr continues unabated . the observations presented here indicate that the youngest stars are not well - mixed with the stars that contribute to the nir continuum in and around the irc . excess br@xmath4 emission with respect to light from the stars that dominate the nir continuum is found in an arc along the western boundary of the irc , while concentrated hei emission is seen to the north of the irc . recent episodes of star formation have thus not been restricted to the center of the irc , but have also occured in its outer regions . a key issue is the age of the irc . is it a young probably transient stellar ensemble , or is it a much older structure that may be the nucleus of ngc 253 ? in section 7.1 we discuss the recent sfh of the irc , and consider its fate if it is a young structure . in section 7.2 the possibility that the irc is the old nucleus of ngc 253 , as suggested by gunthardt et al . ( 2015 ) , is considered . that recent star formation has occured at different locations around the irc suggests that it is likely not a ` cluster ' , but rather a compact star - forming complex containing distinct sub - structures . only the sub - structures in the outer region of the system are detected with gnirs , due to crowding in the denser inner regions . sub - structuring on finer scales is seen in the wfc3 images in figure 1 , which have angular resolutions that are @xmath84 finer than the gnirs data . an age dispersion of at least a few myr is expected in young systems . there is evidence for periods of star formation in young local group clusters that can last up to at least a few myr . one of the largest young systems in the galaxy is westerlund 1 , and it has an age dispersion of a few myr ( kudryavtseva et al . 2012 ; lim et al . a similar age spread is seen in westerlund 3 ( bik et al . 2014 ) . such a short time scale for the formation of a second stellar generation in large star clusters is consistent with simulations of thermally stable clumps that can cool and form stars as a result of self - shielding ( palous et al . 2014 ) . star formation may extend over much longer timescales than is seen in young galactic clusters . the largest nearby star - forming complex to be surveyed in depth is 30 doradus , and de marchi et al . ( 2011 ) find an age dispersion of @xmath85 a few tens of myr . signatures of extended star formation are found in the spectra of some extragalactic clusters ( e.g. peacock et al . 2013 , but see also bastian et al . conroy & spergel ( 2011 ) argue that star clusters with masses @xmath86 m@xmath13 can retain star - forming material for extended periods of time , although the escape velocity prior to the onset of mass segregation may provide a better predictor of gas retention than total cluster mass ( goudfrooij et al . 2014 ) . indeed , while there is photometric evidence for a range of stellar ages in some massive lmc clusters ( e.g. keller et al . 2012 ; li et al . 2014 ) , there are exceptions ( e.g. bastian & silva - villa 2013 ) . multiple episodes of star formation have also been suggested as a means of explaining the chemical properties of stars in globular clusters , and cmds indicate that many globular clusters are not ssps ( e.g. review by gratton et al . 2012 ) . if the irc has a young age then it may not be long - lived . fall & chandar ( 2012 ) present evidence that the majority of clusters are disrupted very early in their lives , and the tidal environment near the center of ngc 253 may not be conducive to the long term survival of even a massive cluster . it is tempting to compare the irc with the arches cluster , which has an age of only a few myr but shows signs of tidal disruption in its global mass function ( stolte et al . 2005 ) and radial structure ( habibi et al . however , the arches is roughly an order of magnitude less massive than the irc . that only modest levels of structuring are seen in the circ region does not indicate that the young stars found there at the present day formed _ in situ _ as part of a diffusely distributed field population . rather , these young stars more likely formed in clusters that are now either defunct or largely disrupted . stars that form in the outer regions of the irc may be susceptible to being calved off , and so these areas may also be a source of future stellar generations in the circ environment . gunthardt et al . ( 2015 ) argue that the irc is the nucleus of ngc 253 . indeed , the integrated brightness and angular size of the irc are similar to those of other galactic nuclei . a complicated sfh also does not preclude the irc from being the nucleus , as it is not uncommon for the nuclei of late - type galaxies to host star clusters ( e.g. boker et al . 2002 ; georgiev & boker 2014 ) . this being said , there is a tendency for the photometric sizes of the nuclei of late - type spirals to vary with wavelength , in the sense that most systems tend to have smaller sizes towards shorter wavelengths , suggesting a tendency for younger stars to be more centrally concentrated than the main stellar body ( e.g. georgiev & boker 2014 ) . this is not the case in the irc . if the irc is the nucleus of ngc 253 then a substantial age spread would be expected . deep cn and c@xmath3 bands are nir spectroscopic signatures of an underlying intermediate age population that would indicate that the irc has survived disruption and formed stars for at least many hundreds of myr . it is thus interesting that there is an absorption feature in the irc spectrum near 1.78@xmath40 m that may be due to c@xmath3 . in section 4 it is argued that this absorption is likely not due to dibs , and this could be checked using higher spectral - resolution observations of the center of ngc 253 in the 1.7 @xmath51 m wavelength interval . the depths of the co bands in the circ spectrum are consistent with a near - solar metallicity , and so the detection of c stars in significant numbers near the center of ngc 253 might be surprising . however , models of c star evolution investigated by karakas ( 2014 ) indicate that c stars can form at solar and super - solar metallicities , although the results for super - solar metallicities depend critically on the helium content . the karakas ( 2014 ) models predict that c stars in solar metallicity systems have progenitor masses between 2 and 4.5 m@xmath13 , which corresponds to main sequence turn - off ages @xmath87 gyr . based on the relative lifespans of c stars when compared with the entire agb phase , then the ratio of c stars to m giants predicted by karakas ( 2014 ) for a solar metallicity system in this age range is @xmath74 . if the m giants and c stars in such a system have comparable intrinsic brightnesses then a modest c@xmath3 feature would be expected in the integrated nir spectrum . davidge ( 2010 ) used nir colors to identify c star candidates throughout the disk and extraplanar regions of ngc 253 , and find that c stars may be present in large numbers throughout the galaxy . near the tip of the agb the c star candidates contribute @xmath88 of the @xmath25band light at galactocentric radii 6 10 kpc . however , a problem when identifying c stars from broad - band photometry alone is that the threshold color that separates c stars from m giants is a function of metallicity ( e.g. discussion in boyer et al . 2013 ) , and so metal - rich m giants may have colors that overlap with those of more metal - poor c stars . the c star contribution to the total light estimated from photometric surveys alone is thus an upper limit . mir photometry may provide a means of distinguishing between very cool m giants and c stars , as the sed of the most evolved stars in the @xmath89 m wavelength interval is sensitive to the mineralogy of the circumstellar envelope ( dellagli et al . 2014 ) . in section 6 it was demonstrated that hot young stars likely contribute significantly to the @xmath25band light from the irc . the models used in section 6 involve only two components a very young ssp coupled with an idealised system that has experienced a csfr . still , these models suggest that if the irc is the nucleus of ngc 253 , and thus contains stars that formed over a large fraction of the age of the universe , then very young stars account for 10% of its stellar mass . this opens the possibility that the nucleus is currently experiencing an episode of very rapid growth in terms of stellar mass . if the nucleus contains a massive black hole then the present day might also be a time during which the black hole is also accreting substantial mass . while gunthardt et al . ( 2015 ) find no signs of nuclear activity , the orientation of ngc 253 on the sky may mean that any acccretion activity is hidden .

the redshifted cm brightness distribution from neutral hydrogen is a promising probe into the cosmic dark ages , cosmic dawn , and re - ionization . lofar s low band antennas ( lba ) may be used in the frequency range mhz to mhz ( ) to measure the sky averaged redshifted cm brightness temperature as a function of frequency , or equivalently , cosmic redshift . these low frequencies are affected by strong galactic foreground emission that is observed through frequency dependent ionospheric and antenna beam distortions which lead to chromatic mixing of spatial structure into spectral structure . using simple models , we show that ( i ) the additional antenna temperature due to ionospheric refraction and absorption are at a% level to orders of magnitude higher than the expected cm signal , and have an approximate dependence , ( ii ) ionospheric refraction leads to a knee - like modulation on the sky spectrum atplasma frequency . using more realistic simulations , we show that in the measured sky spectrum , more than% of the cm signal variance can be lost to confusion from foregrounds and chromatic effects . we conclude that foregrounds and chromatic mixing may not be subtracted as generic functions of frequency as previously thought , but must rather be carefully modeled using additional priors and interferometric measurements . latexl-.36em.3ex-.15em t-.1667em.7ex-.125emx [ section ] [ firstpage ] cosmology : observational , first stars radiolines : general atmospheric effects methods : observational , numerical

1306.2172

neutral hydrogen ( hi ) interacts with @xmath0 cm photons through a spin - flip transition @xcite . observing the redshifted @xmath0 cm brightness temperature cm signal or @xmath0 cm brightness temperature we , really mean _ redshifted _ @xmath0 cm signal ] against a background of the cosmic microwave background is a promising tracer of the cosmic dark ages , cosmic dawn , and the epoch of reionization @xcite . detecting the spatial fluctuations of @xmath0 cm brightness requires many hundreds of hours of integration with large radio synthesis telescopes , owing to its faintness as compared to galactic and extragalactic foregrounds @xcite . on the other hand , the sky averaged @xmath0 cm brightness also called the global signal , is bright enough to be measured within a day s worth of integration based on a signal to noise ratio argument @xcite . since the received frequency of redshifted @xmath0 cm photons corresponds to cosmic redshift , accurately estimating the sky averaged brightness temperature as a function of frequency will provide insights into the evolution of hi during the dark ages , cosmic dawn , and the epoch of reionization @xcite . + thermal uncertainties are not the limiting factor in global @xmath0 cm experiments , and spectral contamination due to systematic artifacts have impeded a reliable detection thus far @xcite . in particular , since the signal in such experiments is the variation of @xmath0 cm brightness temperature with frequency , any instrumental or observational systematic that affects the measured bandpass power poses a severe limitation . these systematics are especially limiting since the foregrounds are @xmath10 orders of magnitude larger than the expected @xmath0 cm signal . consequently , measuring the @xmath0 cm signal spectrum requires precise understanding of frequency dependent effects of instrumental gain , instrumental noise contribution , antenna beam shape , and ionospheric effects , coupled with spatially and spectrally varying foregrounds ( galactic and extragalactic ) . the effects of these parameters are not always mutually separable , further complicating calibration and signal extraction efforts . + ongoing global @xmath0 cm experiments have primarily focused on frequencies ranging from @xmath11 mhz to @xmath12 mhz ( @xmath13)@xcite . dark ages and cosmic dawn experiments at lower frequencies are being planned , or are being commissioned dark ages radio explorer ( dare ) @xcite , large - aperture experiment to detect the dark ages @xcite , and broadband instrument for the global hydrogen reionization signal ( bighorns ) experiment @xcite , among others . global @xmath0 cm work especially in the lower frequency band requires a strict assessment of systematic chromatic corruptions . the reasons for this are threefold . ( i ) firstly , the magnitude of ionospheric effects such as refraction and absorption increase rapidly with decreasing frequency . ( ii ) secondly , galactic foreground brightness temperatures increase with decreasing frequency as a power law with a relatively steep spectral index of about @xmath14 . consequently , any systematic corruptions which are multiplicative , which many of the ones in such experiments are , may undermine @xmath0 cm signal detection efforts more severely at lower frequencies . ( iii ) finally , the increased fractional bandwidth in the lower frequency band leads to an increased variation of antenna beams across the measurement bandwidth , giving larger chromatic effects . + receiver gain and noise temperature may be calibrated by switching the receiver between the sky and known man - made noise sources . such techniques have been demonstrated with moderate success in global @xmath0 cm experiments @xcite . however , little attention has been paid in the literature to chromatic ( frequency dependent ) antenna beam and ionospheric effects . these effects have thus far been assumed to be ` spectrally smooth ' and possibly fitted away along with the foregrounds . they have thus escaped qualitative and quantitative treatment one of the primary aims of this paper . + chromatic effects must be studied in conjunction with algorithms that are used to separate the measured sky spectrum into foregrounds and the @xmath0 cm signal . due to the lack of sufficiently accurate foreground models at these frequencies , such algorithms must rely on some priors on the differential properties of foregrounds and the @xmath0 cm signal . these algorithms typically exploit ( i ) the spectral smoothness of power - law - like foregrounds in comparison to less smooth structure expected in the @xmath0 cm signal , and/or ( ii ) the angular structure of foregrounds as opposed to isotropic nature of the global @xmath0 cm signal . spectral smoothness of foregrounds may be exploited by casting the measured sky spectrum in a basis where foregrounds have a sparse representation unlike the @xmath0 cm signal . we may call such techniques spectral - basis methods , since they only use spectral smoothness as a prior . one such basis set suggested in literature , which we will call ` logpolyfit ` , uses polynomials in logarithmic space as basis to represent the time averaged spectrum @xcite . exploiting priors on the angular structure of foregrounds for global @xmath0 cm experiments has not received due attention in literature , save a recent effort by @xcite , who in light of their simulations , recommend measurements with an angular resolution of @xmath10 degrees . practical implications of a narrow beam ( highly chromatic sidelobes etc . ) remain to be evaluated . moreover , ongoing and proposed global @xmath0 cm experiments have near - hemisphere fields of view and lack any meaningful angular resolution . it is then instructive to place limits on the extent to which beam and ionospheric chromatic effects can confuse @xmath0 cm signatures in the context of spectral - basis algorithms another primary aim of this paper . + in this paper , we simulate the contribution of foregrounds ( with chromatic effects ) to the measured antenna temperature and evaluate an optimal set of basis functions for a sparse representation of foregrounds . by casting the foregrounds and the expected @xmath0 cm signal in this basis , we place limits on the amount of @xmath0 cm signal power that will be lost to foreground confusion in any spectral - basis technique . since the optimal basis functions are not known apriori in real measurements , we may resort to predefined analytic basis functions such as polynomials . in this paper , we show that polynomials in logarithmic space ( ` logpolyfit ` ) are incapable of even detecting the presence of a template @xmath0 cm signal in simulated data in the frequency range of @xmath1 mhz to @xmath15 mhz . + in this paper , we propose an alternative spectral - basis method which we call ` svdfit ` , that evaluates a suitable basis using the measured data itself . despite hemispherical fields of view of ongoing experiments , earth rotation couples angular structure of the foregrounds into the time domain , while the global @xmath0 cm signal being isotropic , has no temporal structure . ` svdfit ` uses the time variable component of the measured dynamic spectra to compute an efficient basis in which the foregrounds and chromatic effects have a sparse representation , but not the @xmath0 cm signal itself . we will show that a ` svdfit ` is better than ` logpolyfit ` in ascertaining the presence of a template @xmath0 cm signal in our simulated data . nevertheless , we argue that for complete reconstruction of the @xmath0 cm signal spectrum , spectral smoothness is an inadequate prior in the above frequency range , and ultimately extracting the @xmath0 cm signal spectrum will require modeling of the foregrounds , antenna beam , and ionospheric effects via a full measurement equation . + the rest of the paper is organized as follows . details of the simulations used herein are described in section [ sec : simu ] . in section [ sec : iono ] we describe our two - layered ionospheric model ( f- and d - layers ) . we derive approximate expressions for chromatic effects from these two layers , and also quantify the level at which we expect these effects . in section [ sec : chrom ] we use the results of our simulations to compute an optimal basis to represent the foreground induced antenna temperature , and quantify the extent to which foregrounds and chromatic effects confuse the @xmath0 cm signatures in spectral - basis methods . we then describe a novel foreground removal technique which we call ` svdfit ` , and also evaluate the efficiency with which ` logpolyfit ` and ` svdfit ` remove foregrounds and chromatic effects . finally in section [ sec : concl ] we draw conclusions and recommendations for future work . this section describes the simulations used in this paper . we assume perfect bandpass calibration of receiver gain and receiver noise . we will discuss bandpass calibration in a forthcoming paper . we also assume that the antenna beam does not vary with time . this is a reasonable assumption since a dipole beam is a function of its mechanical shape , and hence , we do not expect noticeable variations in the antenna beam so long as the dielectric environment of the antenna does not change considerably . we build our simulations from smaller modules , each incorporating a different stage of signal corruption . the end result of our simulations is a dynamic spectrum measuring sky brightness temperature as a function of time and frequency . the following enumeration along with figure [ fig : simu ] describe the parameters in our modular simulations . 1 . _ location _ : we assume the observation location to be one of the lofar stations ( de602 ) near munich , germany , for which we have data in hand . the de602 station is built on slightly sloping land . we assume the latitude and longitude of observations ( @xmath16n , @xmath17e ) to be that point on the locally flat ellipsoid which ` sees ' the same sky as de602 at any instant of time . though this corresponding point and de602 have slightly different horizons , we discount this fact since we are primarily concerned with the chromatic effects of the beam and ionosphere in this paper . sky model _ : our simulations can use either of two sky models : ( i ) the @xmath18 mhz all sky map by @xcite scaled with a global temperature spectral index of @xmath14 , or ( ii ) the all - sky model by @xcite that is a linear combination of observations at different frequencies . while the second sky - model presents a more realistic scenario , we use the first sky model as a reference model to study the spectral nature of chromatic effects due to the ionosphere , lofar lba dipole beam , and the foreground itself . _ antenna beam _ : to study the effects of the chromatic lofar lba dipole beam while facilitating comparison with previous work , we present simulations with two beams : ( i ) a frequency independent @xmath19 beam ( @xmath20 is the zenith angle ) considered by @xcite in their simulations , and ( ii ) a realistic frequency dependent lofar lba beam obtained from electromagnetic simulations of the antenna geometry including the finite ground - plane ( see figure [ fig : lba_pattern ] ) . the frequency dependence of the antenna beam results in the sky being weighed differently at different frequencies and hence couples spatial structure on the sky into the frequency domain . + + 4 . _ @xmath0 cm signal _ : the main feature of the global @xmath0 cm signal expected in the @xmath1 mhz to @xmath15 mhz range is a relatively broad absorption feature @xcite @xcite . since we do not address a full signal reconstruction here , we approximate this absorption feature as a negative gaussian centered at about @xmath21 mhz , with fwhm of @xmath22 mhz . spatial fluctuations of the signal are expected to be on small angular - scales ( @xmath23 degree ) , and are averaged away by the broad dipole beam . recent work has shown that relative velocity between baryons and dark matter may imprint fluctuations on the observed @xmath0 cm brightness on @xmath24 mpc scales @xcite , that correspond to several degrees in the sky . however , single dipoles typically have fields of view spanning several tens of degrees , and hence , we safely ignore any observable brightness fluctuations in the @xmath0 cm signal . _ ionospheric model _ : ionospheric effects may be divided into static effects and dynamic effects . dynamic effects include time variant phenomenon such as ( i ) scintillation induced by turbulence in the ionospheric plasma @xcite , and ( ii ) refraction from large scale traveling ionospheric disturbances @xcite . while we expect these effects to average away for long integration time scales , there has not been a comprehensive study thus far on their effects on global @xmath0 cm experiments , and we defer a discussion on these effects to a future paper . in this paper , we only study static ionospheric refraction and absorption . in section [ sec : iono ] , we describe these static effects in detail . in particular , we show in section [ sec : iono ] that a static ionosphere causes ( i ) frequency dependent deviation of incoming electromagnetic rays ( chromatic refraction ) which we model by ` stretching ' the antenna beam accordingly , and ( ii ) frequency and direction dependent absorption due to electron collision with air molecules which we model as a multiplicative loss factor on the antenna beam . these two effects , when applied on the fiducial antenna beam , give us an effective antenna beam which we use to compute the observed dynamic spectra . _ gridding and computation _ : the sky temperature at a given time and frequency is computed by pixel wise multiplication of the sky model and the effective antenna beam , followed by integration of this product over all pixels while taking their solid angle into account . for a given epoch @xmath25 , and frequency @xmath26 , this computation may be represented as @xmath27 where @xmath28 is the simulated antenna temperature , @xmath29 is the sky brightness temperature ( sky model ) which is a function of zenith angle @xmath20 and azimuth angle @xmath30 , and @xmath31 is the antenna beam as a function of frequency and sky position . note that due to earth rotation , @xmath29 changes with time . as mentioned before , ionospheric effects may be absorbed into the beam term , and if the beam in equation ( [ eqn : mmeq ] ) is replaced by an effective beam @xmath32 , then equation ( [ eqn : mmeq ] ) is the measurement equation which describes the computations in our simulation . + usually , the antenna beam and the sky model are specified in different co - ordinate systems , and have to be brought to a common grid to numerically compute a discretized form of equation ( [ eqn : mmeq ] ) . it is easier to re - grid and interpolate a smoothly varying antenna beam as opposed to the global sky model that has more complex structure due to the galactic disk and point - like sources . we thus work in the co - ordinate system of the sky model ( ra , dec ) and interpolate the effective antenna beams at each frequency and time epoch to the sky grid . an example sky model , antenna beam , and ionospheric refraction induced deviation angle on the sky grid is shown in figure [ fig : simu ] for a single frequency channel at a sidereal time of @xmath33h@xmath33m@xmath33s . for different values of sidereal time , the effective beam will be shifted along the right ascension axis . the end product of the simulations is a dynamic spectrum ( @xmath34 from equation [ eqn : mmeq ] ) in time - frequency domain . an example dynamic spectrum is shown in figure [ fig : typical_spectrum ] . cm are too faint to be discerned by eye on this image . also plotted on the left and bottom are averages along frequency and time axes respectively.[fig : typical_spectrum ] ] this section describes the static model we use for the ionosphere in greater detail , and the origin and nature of chromatic refraction and absorption . the bulk of refraction and absorption occur at two separate layers of the ionosphere the f - layer and the d - layer respectively . we thus evaluate two refractive index values for typical conditions in these two layers , and then use the earth - ionosphere geometry to compute the refractive ray deviation and the absorptive loss factor for different frequencies and directions . the ionosphere is a magnetized plasma whose complex refractive index is given by the appleton - hartree equation @xcite that relates the refractive index to the electron density , magnetic field , and the geometry of wave propagation . we computed the change in refractive index due to the earths magnetic field to be less than @xmath35 part in @xmath36 for the f - layer and less than @xmath37 percent for the d - layer . we ignore this effect since it is smaller than the refractive index variations induced by day to day changes in ionospheric electron density , and we present results for a broad range of electron densities . additionally , the change in refractive index due to a magnetic field is different for left - hand and right - hand circularly polarized radiation , and results in an effect called faraday rotation . in this paper , we ignore faraday rotation by assuming the sky to be unpolarized on scales comparable to our antenna beam . we thus model the ionospheric refractive index using a simplified form of the appleton - hartree equation that does not include the magnetic field term : @xmath38 where @xmath39 is the electron plasma frequency , and @xmath40 is the electron collision frequency . the electric field of a plane wave traveling in a homogeneous ionospheric layer is given by @xmath41 where @xmath42 is the speed of light in free - space , and @xmath43 is the distance measured along the direction of propagation in the ionosphere . + the real part of the refractive index @xmath44 is mostly sensitive to the electron density , and causes a change in the phase velocity ( from that in free space ) resulting in refraction . the imaginary part of @xmath44 ( negative in our case ) is mostly sensitive to the electron collision rate that in turn depends on the electron density , atmospheric gas density , and temperature . the imaginary part exponentially dampens the wave amplitude causing absorption . + due to its low atmospheric gas density ( giving a low collision rate ) and high electron density , we model the f - layer with a real , frequency dependent refractive index @xmath45 , to account for ionospheric refraction . due to its high atmospheric gas density ( giving a high collision rate ) and low electron density , we model the d - layer with an imaginary , frequency dependent refractive index ( @xmath46 ) to account for ionospheric absorption . + we expect most of the global @xmath0 cm observations to take place during night - time when ionospheric electron density is at its lowest . additionally , this avoids the need to model the complex and time - variant spectrum of the sun as a foreground source . we will thus assume values for a typical mid - latitude night - time ionosphere in the absence of intense solar activity . we refer the reader to @xcite , @xcite , and references therein from which we have drawn parameter values for typical ionospheric f- and d - layer conditions . the following subsections compute the refractive and absorptive effects of such an ionosphere . most of the ionospheric electron column density is accounted for by the f - layer that extends between a height of @xmath47 km and @xmath48 km from the earth s surface . the electron density outside of this layer is known to fall off rapidly . though the electron density does vary within the f - layer , to first order , we model the f - layer as a homogeneous shell between the heights of @xmath49 km and @xmath50 km . we assume a constant electron density of @xmath51 m@xmath52 which gives a typical mid - latitude electron column density of @xmath53 tec units tec units equals a column density of @xmath54 electrons per m@xmath55 . ionospheric tec is routinely monitored by measuring the propagation delay in gps signals . see for instance http://iono.jpl.nasa.gov/latest%5frti%5fglobal.html ] . this value is typical of winter - time in mid - latitudes where lofar is situated . tec values are typically higher ( i ) during daytime , ( ii ) closer to the equator , and ( iii ) during summer . additionally , ionospheric tec is sensitive to solar and sunspot activity . due to the above reasons , we will also present results for higher tec values . + ] figure [ fig : refrac_illus ] depicts the refractive effect of the ionospheric f - layer . any incoming ray suffers snell s refraction at the upper and lower boundaries of the f - layer . if the earth were flat , there would be no net deviation in the ray . due to the curvature of the earth ( and hence the ionosphere ) there is a net deviation @xmath56 . this deviation is zero for a source at zenith , and increases as we move towards the horizon . hence , the ionosphere acts like a spherical ` lens ' that deviates incoming rays towards zenith . since the ionospheric refractive index is a strong function of frequency , @xmath56 is also a function of frequency . consequently , the ionosphere is a chromatic lens . figure [ fig : refrac_illus ] also depicts a horizon ray that marks the radio horizon , that is below the geometric horizon . this radio horizon is different at different frequencies . this chromatic lensing of the sky due to ionospheric refraction is an important effect for global @xmath0 cm experiments that use dipoles with near hemispherical fields of view . + it is difficult to derive a closed form expression for @xmath57 , and hence we compute it numerically by applying snell s law at the two interfaces . nevertheless , we may use an analytical approximation to study the dependency of @xmath58 , on @xmath20 and @xmath26 @xcite : @xmath59 where @xmath60 is the mean height of the f - layer ( @xmath61 km in our case ) and @xmath62 is the radius of the earth which we assume to be @xmath63 km . km to @xmath64 km with electron density @xmath65m@xmath52 . the deviation angle is a strong function of both incidence angle and frequency . also shown in black + symbols is the percentage increase in sky area due to extension of radio horizon due to refraction.[fig : devangle ] ] figure [ fig : devangle ] shows a plot of @xmath56 as a function of frequency for different elevation angles . these curves are obtained from simple ray tracing and not from equation ( [ eqn : devangle ] ) . the curves approximately follow a @xmath7 dependence as expected . we also use the value of @xmath56 for the horizon ray ( see figure [ fig : refrac_illus ] ) to compute the percentage increase in visible sky area as compared to the geometric horizon for each frequency . this percentage is also plotted in figure [ fig : devangle ] , and is not only frequency dependent , but also of the order of a few percent even for favorable ionospheric conditions . since the foregrounds are @xmath66 orders of magnitude higher than the @xmath0 cm signal , a frequency dependent additional sky area of a few percent adds an amount of power which is @xmath67 orders of magnitude higher than the expected @xmath0 cm signal . hence , it is important to consider the chromatic effects of ionospheric refraction at the horizon and rapidly decreasing as we move towards zenith @xcite . however this refraction is expected to be non - chromatic @xcite and hence , we disregard it for the purposes of this paper . ] . + the refractive lensing effect of the ionosphere may be absorbed into the antenna beam . we do this by ` stretching ' the antenna beam by an amount @xmath56 to form an effective beam that now includes the effects of chromatic f - layer refraction . if the antenna beam is represented as @xmath68 , then the new effective antenna beam may be represented as @xmath69 which gives the instantaneous measured antenna temperature of @xmath70 note that the effective beam does not integrate to unity like the original antenna beam . it integrates to a value larger than unity due to the additional sky area added by refraction of sub - horizon rays into the original antenna beam . since the dipole beam is ` stretched ' by an amount equal to @xmath56 which is a strong function of frequency , the sky is weighted differently at different frequencies in equation ( [ eqn : mmeq ] ) . this couples spatial structure in the sky to frequency structure in the measured antenna temperature spectrum . it is important to note that this chromatic mixing happens even if the original antenna beam itself is frequency independent . + finally , we quantify the approximate extent and nature of chromatic refraction by considering a simple case where ( i ) the sky brightness temperature is a power law with the same spectral index @xmath71 everywhere ( @xmath72 ) , and ( ii ) the original antenna beam is frequency independent and is given by @xmath73 . the effective antenna beam ( due to chromatic refraction ) is then given by @xmath74 which on taylor - expansion about @xmath20 gives @xmath75 note that the effective beam is now chromatic , while the original antenna beam is not . additionally , we have shown in equation ( [ eqn : devangle ] ) that @xmath76 has a form that is separable in @xmath26 and @xmath20 , and may be expressed as @xmath77 , where @xmath78 is a function independent of @xmath26 . substituting this and equation ( [ eqn : taylor_beam ] ) into equation ( [ eqn : anttemp ] ) , we find that the antenna temperature evaluates to the form @xmath79 where @xmath80 and @xmath81 are independent of frequency , and depend only on the sky brightness , antenna beam , and geometric terms . equation ( [ eqn : f_final ] ) shows that chromatic refraction will add a new component to the original @xmath82 sky brightness temperature . this new component has a spectral shape given by @xmath83 , and as argued before , is at a few percent level . the chromatic foregrounds can now be fit away by the basis functions @xmath82 and @xmath84 . however , since the sky brightness and the lofar lba beam are both more complicated , we will resort to more realistic simulations in section [ sec : chrom ] to accurately evaluate the nature and extent of chromatic refraction . the d - layer is the lowest layer of the ionosphere extending from @xmath85 km to @xmath86 km from the earth s surface . high electron densities in the d - layer are expected to persist only during daytime due to solar insolation . however , residual electron densities of the order of @xmath87 m@xmath52 exists even during night - time . we will use a fiducial value for d - layer electron density of @xmath88 m@xmath52 in our simulations . we choose this value to obtain an absorption of @xmath89 db at @xmath90 mhz that agrees with values quoted in literature @xcite . while this electron density is too low to cause appreciable refraction , due to a high atmospheric gas density at these heights , the electron collision frequency is high enough to cause considerable absorption . we use a typical value of @xmath53 mhz for the electron collisional frequency @xcite . we model the d - layer as a homogeneous layer between the heights of @xmath91 km and @xmath92 km . + if the path - length through the d - layer is @xmath43 , then the multiplicative loss factor due to ionospheric absorption may be computed as @xmath93 where @xmath94 is the imaginary part of the d - layer refractive index , @xmath42 is the speed of light in vacuum , and the approximation holds for small values of the exponent ( expected in the d - layer ) . note that since @xmath94 is negative , @xmath95 . we compute @xmath43 numerically in our simulations , but to understand its dependencies , we provide an approximate expression : @xmath96 where @xmath62 is the radius of the earth , @xmath97 is the mean d - layer height ( @xmath98 km in our case ) , and @xmath99 is the width of the d - layer . in figure [ fig : iono_abs ] we show the computed values of d - layer absorption , @xmath100 , as a function of frequency for different elevation angles . zenith absorption increases from @xmath101 db at @xmath90 mhz to @xmath102 db at @xmath103 mhz . x@xmath104m@xmath52 , and electron collisional frequency of @xmath53 mhz , extending from a height of @xmath91 km to @xmath92 km.[fig : iono_abs ] ] as seen in the plot , absorption changes the incoming brightness temperature by @xmath105 percent . though this is small , it is still large in comparison to the dynamic range between the foregrounds and the expected @xmath0 cm signal . hence , studying the nature of d - layer absorption is important . + the imaginary part of the refractive index from equation ( [ eqn : appl ] ) approximately behaves as @xmath106 this along with equation ( [ eqn : dlayer_abs ] ) gives @xmath107 since @xmath40 is comparable to frequency @xmath26 , @xmath108 does not have a simple power - law like structure as a function of frequency ( also evident in figure [ fig : iono_abs ] ) . as we did with refraction , we may define an effective beam @xmath109 that takes into account ionospheric absorption . this beam will now integrate to less than one , and is given by @xmath110 for a frequency independent @xmath19 beam , using equation ( [ eqn : dlayer ] ) , the effective beam is given by @xmath111 if the sky brightness temperature had a direction independent @xmath82- type power law behavior , the measured antenna temperature in presence of absorption will be of the form @xmath112 where @xmath113 and @xmath114 are independent of frequency , and depend only on the sky brightness distribution , d - layer plasma and collision frequencies , and some geometric terms . this equation shows that , in case of d - layer absorption , we also have an additional component in the antenna temperature . this component has a spectral shape given by @xmath115 , is negative , and as shown earlier , is at the @xmath116% level . because we expect @xmath40 to be @xmath117 mhz , discounting some error near the lowest end of our bandwidth , we may assume @xmath118 , and the additional component may be approximated as having a spectral shape @xmath83 , as in the case of f - layer refraction : @xmath119 consequently , for the case of a sky with a global spectral index measured with a frequency independent @xmath19 beam , ionospheric effects ( refraction plus absorption ) will introduce an additional spectral contribution which approximately has a @xmath84 shape . this is confirmed in figure [ fig : diff_spectrum ] , where we show the excess antenna temperature from our simulations due to the inclusion of ionospheric effects . cm signal . [ fig : diff_spectrum ] ] we have plotted the absolute value of the excess , since absorption over - compensates the additional sky coverage due to refraction , making the excess temperature negative . the additional antenna temperature due to f - layer refraction varies from @xmath120 k at @xmath103 mhz to @xmath4 k at @xmath2 mhz . the excess temperature due to d - layer absorption on the other hand varies from about @xmath121 k at @xmath103 mhz to @xmath122 k at @xmath2 mhz . as expected , the net differential temperature is at a @xmath4% level . also shown in the bottom panel is the time averaged excess temperature along - with ( the negative of ) the expected @xmath0 cm signal . our simulations confirm that the chromatic effects due to a static ionosphere alone are @xmath5 orders of magnitude larger than the expected @xmath0 cm signal and hence must be studied with care . in figure [ fig : diff_spectrum ] we have neither included the chromatic effects of the lba beam , nor a more comprehensive ( and more complex ) sky model by @xcite . nevertheless , the figure stresses the fact that chromatic contamination that comes from just the ionosphere is larger than our signal by more than @xmath5 orders of magnitude . in section [ sec : chrom ] we show results from simulations that also include the above two effects . ] figure [ fig : iono_tir ] depicts the effects of curvature of the ionosphere in cases when the f - layer electron density is considerably larger resulting in high deviation angles . we have thus far assumed that incoming rays that are refracted at the upper f - layer boundary are always incident on its lower boundary . due to the earth s curvature , this need not be true in cases when the f - layer electron density is particularly high . consider the blue ray in figure [ fig : iono_tir ] . the ray is at grazing incidence at the lower f - layer interface marked as point b. any ray that comes from space at a lower elevation angle will not be incident at the lower f - layer interface at all and will escape back into space without reaching the telescope at point t. hence , in this case , the blue ray is ` critical ' and represents the horizon ray . from the point of view of the observer at t , all incoming rays have a zenith angle that is greater than some critical zenith angle @xmath123 . we will thus call this phenomenon low elevation cutoff will suffer total internal reflection at the f - layer an important consideration for communication links . ] + we now compute the conditions under which a low elevation cut - off is relevant , and discuss its effect on the measured sky spectrum . for the case of the horizon ray , the angle of incidence at the lower f - layer interface is equal to @xmath124 , and hence snell s law at the interface becomes ( see figure [ fig : iono_tir ] ) @xmath125 which upon using the sine rule in triangle obt gives @xmath126 where @xmath127 is the height of the lower interface of the f - layer ( @xmath49 km in our case ) . we may use the appleton - hartree formula ( equation [ eqn : appl ] ) with @xmath128 in the above equation to get @xmath129 setting @xmath130 will then give us an expression for the frequency below which low elevation cut - off becomes relevant . above this frequency , we may simply follow our earlier discussion of f - layer refraction from section [ subsec : flayer ] . below this frequency , the increasing visible sky area due to increasing @xmath58 ( chromatic refraction ) is partly compensated by the decreasing sky area due to low elevation cut - off . setting @xmath130 , gives @xmath131 for our values of @xmath132 km and @xmath133 km , this implies @xmath134 . typical mid - latitude night - time f - layer plasma frequencies are below several mhz , and hence @xmath135 typically lies outside our bandwidth . in such cases , our discussion about the low elevation cut - off is not relevant . however , f - layer conditions where @xmath39 exceeds @xmath53 mhz do occur data may be found at http://www.ips.gov.au/hf%5fsystems/6/5 and links therein . ] , and unless due care is taken , some exposure to such conditions may persist in long integrations . + to illustrate the effect of low elevation cut - off , in figure [ fig : radio_horizon ] , we plot the elevation angle of the horizon ray as a function of frequency for different values of f - layer electron density . . for lower electron densities , the knee is below our minimum observation frequency , but not so for higher electron densities . such an abrupt knee will introduce highly undesirable modulation in the foregrounds spectra.[fig : radio_horizon ] ] the electron density values for the different curves are @xmath65 , @xmath136 , @xmath137 , @xmath138 , and @xmath139m@xmath52 . these correspond to column densities of @xmath53 , @xmath140 , @xmath103 , @xmath91 , and @xmath90 tec units respectively . the corresponding values of @xmath141 are @xmath142 , @xmath143 , @xmath144 , @xmath145 , and @xmath15 mhz respectively . as expected , the first two curves have their @xmath141 values outside of our bandwidth , and hence , do not suffer from the effects of low elevation cut - off within our bandwidth . the other curves , have a ` knee ' in their values of radio horizon positions versus frequency within the bandwidth . this knee exists because , as we go below the knee frequency ( @xmath146 ) , the increasing visible sky area due to chromatic refraction is partly compensated by the decreasing visible sky area due to low elevation cut - off . hence , the slope of the curves below and above the knee are markedly different . such abrupt modulation of the foreground spectra is undesirable , given that the foreground subtraction algorithms rely on the smooth spectral nature of foregrounds at all times . it is also important to note that if such high electron density conditions persist even for a fraction of the total integration time , the time - averaged spectra will have components of foregrounds that are heavily modulated . it is thus important to monitor the ionospheric conditions throughout the observation duration , and perhaps even flag data acquired during times of high f - layer plasma density . in section [ sec : iono ] , we provided insight into the nature and extent of chromatic effects due to the ionosphere using simplistic dipole beams and sky models . chromatic effects in presence of a more realistic sky model and dipole beam model are difficult to evaluate analytically . in this section , we evaluate beam and ionospheric chromatic effects using the lofar lba beam and more realistic sky models from @xcite . we will however also include results for a frequency independent beam for comparison with previous work , and for the benefit of experiments that use dipoles with ( approximately ) frequency independent beams ( see @xcite for instance ) . + how severe chromatic effects are depends on our prior knowledge of their nature and extent . if we have accurate enough models of foreground brightness and chromatic mixing , we may simply subtract the foreground contribution to antenna temperature ( taking into account all chromatic effects ) to expose the @xmath0 cm signal . this may not be the case in practice , and we are left to making certain simplifying assumptions about the differential properties of foregrounds ( along with chromatic effects ) and the @xmath0 cm signal . + the simplest assumption that we may make is that the foregrounds and chromatic effects have a smooth spectral behavior unlike the @xmath0 cm signal . we may then express the measured antenna temperature spectra in some optimal basis ( polynomials for instance ) wherein the foreground contaminants ( but not the @xmath0 cm signal ) are sparsely represented . we will refer to algorithms making only this assumption as spectral - basis methods . these methods assume no additional cognizance of the actual antenna beam , ionospheric effects , or constraints from other measurements such as interferometric visibilities . + for such spectral - basis methods , we will evaluate the chromatic effects at two levels . the first level inquires how similar the spectra of chromatic distortions are to the expected @xmath0 cm signal . the relevance of this question comes from the fact that chromatic effects are insidious if they confuse @xmath0 cm signatures in frequency space . to evaluate a best case scenario for spectral - basis methods , we will use the dynamic spectra from our simulations ( excluding the @xmath0 cm signal contribution ) to find an optimal basis to represent foreground contribution to the antenna temperature spectrum . we will then evaluate the relative efficiency with which the foreground contribution and the @xmath0 cm signal contribution to antenna temperature is fit away by these optimal basis vectors . this first level of inquiry merely evaluates the extent of foreground confusion . it does not give us a recipe to find an optimal basis in practice . at the end of this section we will briefly explore our second level of inquiry , that tries to evaluate two different spectral basis : ( i ) ` logpolyfit ` that uses polynomials in logarithmic space as basis function and ( ii ) ` svdfit ` ( described later ) which is a novel way to evaluate near - optimal basis functions from the dynamic spectrum itself . we compute an optimal non - parametric basis using the simulated antenna temperature without the inclusion of the @xmath0 cm signal . note that in practice , with single dipole experiments , we do not have access to the antenna temperature without the inclusion of the @xmath0 cm signal , and hence we can not compute an optimal basis from the data itself . + non - parametric basis functions derived from the data itself have been explored in the literature , albeit in the context of interferometric observations of the spatial fluctuations of the @xmath0 cm signal . examples include the independent component analysis ( ica ) @xcite , singular value decomposition ( svd ) @xcite , and smoothing techniques @xcite . more recently , @xcite have used the svd technique in the context of global @xmath0 cm cm signal extraction . we will use the svd technique to analyze chromatic effects in our simulations , due to its desirable orthogonality property . + our simulations provide a dynamic spectrum , @xmath147 , which is a matrix of dimensions @xmath148 x @xmath149 where @xmath148 and @xmath149 are the number of time and frequency bins in the data respectively . we have used the subscript @xmath150 to note that the dynamic spectrum used here is due to the foregrounds , and does not include the @xmath0 cm signal itself . we will use the subscript @xmath151 to denote the measured spectrum due to the presence of the @xmath0 cm signal . to find an optimal basis where the foregrounds are sparsely represented , we compute the singular value decomposition ( svd ) of the dynamic spectrum @xmath147 : @xmath152 matrix @xmath153 $ ] is an orthonormal matrix of size @xmath149 x @xmath149 whose rows ( @xmath154 ) provide an orthonormal basis to represent the spectral variability in @xmath147 . the vectors @xmath154 are simply the eigen vectors of the correlation matrix @xmath155 . hence , we are treating the spectra measured at different epochs as different realizations of snapshot measurements of sky spectrum , and through eigen decomposition , finding a set of basis vectors that efficiently describe any linear combination of these snapshot spectra . since each snapshot spectrum has contributions from a large part of the sky , we expect the spectra to have some underlying ensemble properties that are efficiently described by the basis vectors @xmath154 . + the representation of the time averaged spectrum @xmath156 in terms of a basis vector @xmath154 is given by @xmath157 and a representation of @xmath156 in terms of the first @xmath158 basis vectors is given by @xmath159 where @xmath160 $ ] . because the vectors @xmath154 form an efficient basis to describe the foregrounds , we expect the residual rms given by @xmath161 to decrease rapidly as we increase @xmath158 . due to its contrasting spectral behavior , the time averaged @xmath0 cm spectrum , @xmath162 is not expected to be efficiently represented by the above basis computed from the covariance matrix @xmath155 due to the foregrounds alone . consequently , we expect the residuals of its representation @xmath163 to fall - off less rapidly with increasing @xmath158 . the last two statements basically reiterate the sparseness assumption mentioned before . the value of @xmath158 for which the residual rms of foregrounds is lower than that of the @xmath0 cm signal then gives us the minimum number of parameters required to fit the foregrounds away . the difference in the rms of the @xmath0 cm signal and the rms of the @xmath0 cm residuals also gives us the amount of power in the @xmath0 cm signal that is fitted away along with the foregrounds . it might also come to pass that the rms of residual of foregrounds are always larger than that of the @xmath0 cm signal . in such cases , the foregrounds and chromatic effects are expected to introduce sufficient non - smooth structure , so as to inhibit their separation from the @xmath0 cm signal using only the information contained in the dynamic spectrum . in other words , the foreground spectral signatures will completely confuse our efforts to detect the @xmath0 cm signal with spectral basis methods . + figure [ fig : svd_demo_lba ] demonstrates the computation of basis functions and residuals using the svd . for this figure , we have used the sky model from @xcite , the simulated lofar lba beam , and @xmath164 hour observation over the frequency range @xmath103 mhz to @xmath2 mhz . we have included f - layer chromatic refraction ( @xmath65m@xmath52 ) and d - layer chromatic absorption ( @xmath165m@xmath52 , @xmath166 mhz ) . the left - hand side panels show the first five basis vectors obtained from svd . the basis vectors are arranged from top to bottom in decreasing order of dominance . the right hand panel shows the residuals when the time averaged antenna temperature spectrum due to foregrounds alone ( red ) and due to the expected @xmath0 cm signal ( blue ) are represented in terms of the first @xmath167 and @xmath168 dominant basis vectors ( from top to bottom ) . the sky model from @xcite at our frequencies of interest is constructed from three principal components , and as such , may be fully expressed as a linear combination of three spectral basis vectors . the beam and ionospheric chromatic effects however , add additional complexity to the time - averaged antenna temperature spectrum and as a consequence , need at least @xmath168 basis vectors to be described to the required level . this is clear from the right - hand panel plots in figure [ fig : svd_demo_lba ] and the enumerated rms residual levels for the red and blue curves . clearly , only for the case of representation with @xmath168 basis vectors does the rms residuals for foregrounds reduce to levels significantly below those of the @xmath0 cm signal . the expected @xmath0 cm signal has a residual of @xmath169 mk after fitting with @xmath168 basis vectors . the original @xmath0 cm signal that was used in the simulations had an rms ( mean subtracted ) of @xmath170 mk . this means that if we use spectral - basis methods , then not only we need a minimum of @xmath168 components to fit the foregrounds , we also end up loosing at least @xmath2% of the variance ( not rms ) in the @xmath0 cm signal to foreground subtraction . we now repeat the analysis of section [ subsec : opt_basis ] for the two dipole beams ( i.e. achromatic and chromatic ) and for different f - layer electron densities . the results are shown in figure [ fig : svd_rms ] . for brevity , we have not shown the basis vectors and residual spectra , but rather shown the rms of residuals when ( i ) the foreground contribution to time - averaged antenna temperature ( red ) and ( ii ) the @xmath0 cm contribution to antenna temperature ( blue ) are fit with increasing number of spectral basis vectors . the top panels show the rms residuals for a frequency independent @xmath19 beam , for different values of f - layer electron density . the bottom panels show the same for the fiducial lofar lba beam obtained from electromagnetic simulations . as argued before , the abscissa for which the red curve dips below the blue curve denotes the minimum number of spectral basis vectors required to separate the foregrounds from the @xmath0 cm signal . note however that this does not necessarily mean that the residuals are a representation of the @xmath0 cm signal as is evident from figure [ fig : svd_demo_lba ] . each panel also show the levels denoting the @xmath90% ( solid black ) , @xmath9% , and @xmath142% ( broken black ) of the variance in the original @xmath0 cm signal spectrum ( black line ) to give intuition into the amount of variance ( not rms ) in the @xmath0 cm spectrum that is lost to foreground subtraction . figure [ fig : svd_rms ] thus also quantifies the minimum extent to which chromatic effects confuse the @xmath0 cm signature in spectral basis methods . non - optimal basis functions ( including analytic functions ) not only require more parameters to model the foregrounds , but will also result in higher co - subtraction of the @xmath0 cm signal along with the foregrounds . + as mentioned before , in the absence of chromatic effects , the @xcite models may be fully described by just the first @xmath6 spectral basis vectors . figure [ fig : svd_rms ] shows that for a frequency independent beam , if the f - layer electron density is low ( @xmath171m@xmath52 ) , then the first @xmath6 spectral basis vectors are still sufficient to separate foregrounds and chromatic effects from the @xmath0 cm signal . in case of the lofar lba beam , at least @xmath172 spectral basis vectors are necessary . any generic basis functions ( polynomials for instance ) if employed , will only require more than the minimum number of parameters thus obtained from figure [ fig : svd_rms ] . for instance , for the case of the lofar lba beam with @xmath173m@xmath52 ( bottom - left panel ) , a minimum of @xmath172 parameters , or at least a @xmath6rd order polynomial is required . indeed , there is no guarantee that a polynomial basis will represent the foregrounds sufficiently , but figure [ fig : svd_rms ] gives us a hard lower limit on the number of independent parameters required for foreground subtraction , and also the minimum amount of @xmath0 cm signal variance that will be lost to foreground subtraction with a spectral - basis method . panels , ` svdfit ` succeeds in the case of a frequency independent beam . [ fig : fit2mthd ] ] the analysis of section [ subsec : limits ] presented an estimate of how much foregrounds and chromatic effects confuse @xmath0 cm signatures for spectral - basis methods . in reality , for single dipole experiments , we do not have access to the optimal basis functions simply because , we can not obtain a measurement of antenna temperature that includes foreground contribution but not the global @xmath0 cm signal contribution . if the dipole is embedded into an array , then we will have independent measurements of spatially varying foregrounds . for single dipole experiments however , we have to adopt other techniques to construct the spectral basis vectors . in this subsection , we explore two such techniques . we will call the first technique ` logpolyfit ` . this method has been proposed in literature @xcite , and attempts to use polynomials in logarithmic space as basis vectors to represent foregrounds . in other words , we seek to fit the time - averaged antenna temperature due to the foregrounds with the function @xmath174 where @xmath175 denotes logarithm to base @xmath53 , and @xmath176 are parameter values to be estimated . to evaluate ( in simplistic terms ) how well ` logpolyfit ` performs , we will fit the model @xmath177 to two time - averaged antenna temperature spectra which we obtain from our simulations : ( i ) @xmath178 , and ( ii ) @xmath179 . here @xmath162 is the antenna temperature due to the fiducial @xmath0 cm signal , and @xmath180 is the total antenna temperature due to both foregrounds and the @xmath0 cm signal . hence the two models may be represented as @xmath181 and @xmath182 assuming that @xmath183 fits out the foregrounds but not the @xmath0 cm signal , model 1 must perform significantly better compared to model 2 in terms of goodness of fit . this is because , model 1 acknowledges the presence of the @xmath0 cm signal contribution in @xmath180 , whereas model 2 does not . note that we are not trying to fit for or solve for the @xmath0 cm signal spectrum . we are assuming its form and are merely trying to ascertain its presence in the simulated antenna temperature spectrum . in reality , we have to estimate the shape of the @xmath0 cm spectrum . but the above goodness of fit test provides a necessary condition for any spectral - basis algorithm to work in the first place . we evaluate the goodness of fit by computing the rms of the residuals of fit . + the second technique which we will call ` svdfit ` is a novel technique which we propose herein . ` svdfit ` derives basis vectors to describe the antenna temperature spectrum using the data itself ( output of the simulation in our case ) . since as argued before , the data contain the @xmath0 cm signal , evaluating the basis vectors from the data itself will give us a set of basis vectors that will remove the foregrounds and the @xmath0 cm signal with equal efficiency , thereby impeding their separation . we instead subtract the time averaged spectrum from the dynamic spectra prior to svd evaluation . since the antenna temperature due to a global @xmath0 cm signal is unchanging with time , the new mean subtracted dynamic spectra has no contribution from the @xmath0 cm signal , and we may compute a set of optimal basis vectors that will not efficiently describe the @xmath0 cm signal . we may then use these basis vectors to fit model 1 and model 2 . + subtracting the time - averaged antenna temperature from the dynamic spectra also removes any contribution from the global component of the foregrounds . in essence , we are extracting a set of basis from the spatially structured foregrounds , and using them to fit the away the global foreground component . since the basis vectors are evaluated from only a fraction of the sky brightness temperature , they will be sub - optimal . however , our antenna beam averages over large parts of the sky , and since the sky brightness temperature spectrum is more or less stationary on such large scales , we expect these basis vectors to well represent the foreground antenna temperature contribution from the entire sky . + figure [ fig : fit2mthd ] shows the rms residuals for model 1 and model 2 for ` logpolyfit ` and ` svdfit ` . on the x - axis is the number of basis vectors used in case of ` svdfit ` , or the order of the polynomial used in case of ` logpolyfit ` . model 1 and model 2 rms for ` svdfit ` are plotted in red and green respectively , whereas , model 1 and model 2 rms for ` logpolyfit ` are plotted in blue and magenta respectively . the different panels in the figure are for different f - layer electron densities , and the two different beams mentioned before . a successful algorithm must show a clear difference in the goodness of fit ( represented by rms residuals here ) for the two models . note that in addition , the difference in rms between the models must be larger than the uncertainties due to noise , which are not considered here . in any case , ` logpolyfit ` ( blue and magenta curves ) is not able to separate the two models for any of the cases , and is hence an inadequate basis . ` svdfit ` on the other hand is able to ascertain the presence of the @xmath0 cm signal for the case of an ideal frequency independent beam . it however fails for the case of the fiducial lofar lba beam . this is perhaps because of the highly chromatic nature of the lofar lba beam which mixes spatial structure of the foregrounds into spectral structure in the antenna temperature spectrum . consequently , basis vectors evaluated from the spatially structured foregrounds are rendered inefficient in fitting the spectrum due to the global foreground component . hence , ` svdfit ` may be an interesting technique only for experiments in our frequency range that use a ( more or less ) frequency independent antenna . the simulations presented in this paper make simplistic assumptions about a more complicated reality . in particular , the real sky brightness may have much more complex spectrum than the one we have assumed in our simulation . similar is the case for the ionospheric effects which in reality are dynamic and originate from an ionosphere structure that is more complicated than our simple layered model . the results presented herein are therefore expected to be lower limits to the chromatic mixing due to the foregrounds , beam , and the ionosphere . consequently , spectral - basis methods may not be the proper way forward for foreground subtraction in our frequency range , despite being attractive due to the simplicity of their only assumption that the foregrounds have a smooth spectrum compared to the @xmath0 cm signal . the way forward warrants for techniques that use stronger priors on the properties of foregrounds and chromatic effects . @xcite have reached a similar conclusion in their analysis , and have suggested that spatial structure in the foreground be exploited to separate them from the @xmath0 cm signal that has no spatial features . this however requires higher spatial resolution unafforded by single dipole experiments . antennas or arrays that have higher resolution also have highly chromatic beams with chromatic sidelobes that will cause a high level of chromatic mixing on many spectral scales . algorithms that overcome this limitation have not been demonstrated till date . + experiments with wide field - of - view dipoles may still benefit considerably from using stronger priors . this is especially true since we have explicit prior knowledge of ( i ) the sky brightness temperature from various surveys , ( ii ) fairly accurate antenna beam models from simulations and measurements , and ( iii ) knowledge of ionospheric conditions from gps satellite based measurements or dipole - dipole cross - correlations . the lofar lba dipoles are part of an array and inter - dipole visibilities may provide strong priors on all three of the above parameters . we thus propose to improve on spectral - basis methods in our future work , wherein we plan to model the data using a full measurement equation similar to equation ( [ eqn : mmeq ] ) . single antenna wide field of view experiments that measure the sky spectrum to high degree of accuracy are interesting probes of the cosmic dark ages , cosmic dawn , and the epoch of reionization . such measurements not only require unprecedented accuracy in receiver gain and noise temperature calibration , but also require accurate modeling and removal of contamination along the frequency axis . ionospheric refraction and absorption are highly frequency dependent and are thus a potential limitation in such experiments . additionally , a frequency dependent antenna beam mixes spatial structure of foregrounds into spectral structure providing an additional source of signal contamination . in this paper , we have studied the nature and magnitude of the above spectral contaminants in the frequency range of @xmath103 mhz to @xmath90 mhz that is particularly interesting for cosmic dawn studies . + we have arrived at the following results / conclusions : 1 . a simple ionospheric model that accounts for static chromatic effects consists of two homogeneous layers : the f - layer that accounts for chromatic refraction , and the d - layer that accounts for chromatic absorption . in case of a sky with a global spectral index @xmath184 , and a frequency invariant beam , chromatic refraction due to typical f - layer electron densities ( @xmath185x@xmath186m@xmath52 ) adds a component to the measured antenna dynamic spectrum that has a spectral index of @xmath187 , and is about @xmath120 k at @xmath1 mhz and @xmath4 k at @xmath2 mhz . likewise , typical d - layer absorption for an electron density of @xmath165m@xmath52 , and collisional rate of @xmath188 mhz , adds a ( negative ) component to the dynamic spectrum at the @xmath105% level with spectral shape given by @xmath115 . typical values for additional signal due to d - layer absorption range from @xmath121 k at @xmath103 mhz to @xmath189 k at @xmath2 mhz . we have also shown that high f - layer electron densities lead to low elevation cut - off of incoming rays , and that this leads to a knee - like feature in elevation angle of the horizon ray versus frequency . we have identified the knee to be at @xmath190 times the plasma frequency . we have shown that if the f - layer electron densities approach or exceed @xmath191m@xmath52 during the measurement duration , low elevation cut - off leads to undesirable modulation of the measured spectrum within the bandwidth of interest ( @xmath103-@xmath2 mhz ) . we thus recommend monitoring of ionospheric tec throughout observations with an intention of flagging data during periods with high tec values . to evaluate chromatic effects from more realistic sky and beam models , we have set - up simulations that accept a variety of sky , beam , and ionospheric parameters and produce dynamic spectra of the measured antenna temperature . we have evaluated chromatic effects for the @xcite sky models , for ideal frequency independent , and simulated lofar lba beams . using the results of these simulations we have placed limits on the efficiency of spectral basis methods algorithms that separate foregrounds from the @xmath0 cm signal based on the spectral smoothness of foreground and chromatic effects as compared with the @xmath0 cm signal . in doing so we have shown that even under ideal ionospheric conditions a minimum of @xmath172 parameters are required to sufficiently describe chromatic effects of the lofar lba beam and the ionosphere , and that a minimum of @xmath192% of the variance in the @xmath0 cm signal is typically lost due to confusion with foreground and chromatic effects , rendering such a method ineffective . + 3 . we have also evaluated the efficiency of two practical algorithms : ( i ) ` logpolyfit ` that uses polynomial in logarithmic space as basis functions , and ( ii ) ` svdfit` a novel algorithm proposed herein that uses the dynamic part of the antenna temperature spectrum to compute a near - optimal set of basis vectors that may then be used to separate foreground from the time averaged ( static part ) antenna temperature spectrum . we show that ` logpolyfit ` fails as a spectral basis method in our frequency range of interest , but ` svdfit ` has potential to succeed in case of a near frequency independent beam . we however conclude that ` svdfit ` fails as a spectral basis method in case of the highly chromatic lofar lba . moreover , dynamic ionospheric effects will only decrease the efficiency of any spectral basis method . we conclude that spectral basis methods , though attractive due to their simplicity , do not use many of the strong priors ( that almost always exist ) from independent measurements of the sky brightness temperature , ionospheric conditions , and the dipole beam , all three of which have a large impact on the dynamic spectrum . as part of future work , we plan to improve upon spectral basis methods for single dipole experiments by modeling the observed data using a full measurement equation . we have recently concluded a pilot project with data from the de602 lofar station ( near garching , germany ) , and plan to present results from a measurement equation based modeling of this pilot data in a forthcoming paper . additionally , embedding the dipole into an array gives dipole - dipole visibilities that provide strong constraints on not only receiver bandpass calibration , but also on all the above factors . array based measurements / constraints may be an effective way forward for global @xmath0 cm experiments . consequently , as part of future work , we also plan to develop a framework that exploits priors derived from dipole - dipole visibilities , to address the challenges posed by large foregrounds and chromatic mixing . while we only address single dipole measurements in this paper , our analysis of chromatic lensing in the ionosphere ( see sections [ subsec : flayer ] and [ subsec : leco ] ) may have implication for interferometric @xmath0 cm experiments . this is especially important for observations at epochs when the galactic plane is at low elevations where refraction is at its strongest . refraction may cause bright radio sources to cross the spectrally varying radio - horizon within the observation bandwidth giving insidious frequency structure in the visibilities . as part of future work , we plan to evaluate these effects on techniques presented by @xcite and @xcite that propose to filter out discrete foreground sources in the frequency domain . hkv and lvek acknowledge the financial support from the european research council under erc - starting grant firstlight - 258942 . the authors also thank michel arts from the netherlands institute for radio astronomy ( astron ) for providing the simulated lofar lba beam patterns . 99 bailey , d. k. , terrestrial magnetism atmospheric electricity , 53(1 ) , 4150 , 1948 , doi:10.1029/te053i001p00041 . beardsley , a. p. , hazelton , b. j. , morales , m. f. , et al . 2013 , mnras , 429 , l5 bougeret , j. l. 1981 , a&a , 96 , 259 bowman , j. d. , rogers , a. e. e. , & hewitt , j. n. 2008 , apj , 676 , 1 bowman , j. d. , & rogers , a. e. e. 2010 , nature , 468 , 796 burns , j. o. , lazio , j. , bowman , j. , et al . 2011 , bulletin of the american astronomical society , 43 , # 107.09 chapman , e. , abdalla , f. b. , harker , g. , et al . 2012 , mnras , 423 , 2518 chippendale , a. p. . 2009 , ph.d . thesis , univ . of sydney crane , r. k. 1977 , ieee proceedings , 65 , 180 de oliveira - costa , a. , tegmark , m. , gaensler , b. m. , et al . 2008 , mnras , 388 , 247 evans , j. v. , & hagfors , t. 1968 , new york : mcgraw - hill , 1968 , edited by evans , john v. ; hagfors , tor field , g. b. 1959 , apj , 129 , 525 furlanetto , s. r. 2006 , mnras , 371 , 867 furlanetto , s. r. , oh , s. p. , & briggs , f. h. 2006 , physics reports , 433 , 181 greenhill , l. j. , & bernardi , g. 2012 , arxiv:1201.1700 harker , g. j. a. , pritchard , j. r. , burns , j. o. , & bowman , j. d. 2012 , mnras , 419 , 1070 harker , g. , zaroubi , s. , bernardi , g. , et al . 2009 , mnras , 397 , 1138 haslam , c. g. t. , salter , c. j. , stoffel , h. , & wilson , w. e. 1995 , astronomy data image library , 1 jeli , v. , zaroubi , s. , labropoulos , p. , et al . 2008 , mnras , 389 , 1319 lassen , h. , i. zeitschrift fr hochfrequenztechnik , 1926 , volume 28 , page 109113 liu , a. , pritchard , j. r. , tegmark , m. , & loeb , a. 2013 , phrevd , 87 , 043002 liu , a. , & tegmark , m. 2012 , mnras , 419 , 3491 madau , p. , meiksin , a. , & rees , m. j. 1997 , apj , 475 , 429 mcquinn , m. , & oleary , r. m. 2012 , apj , 760 , 3 nicolet , m. , journal of atmospheric and terrestrial physics , volume 3 , issue 4 , may 1953 , pages 200 - 211 parsons , a. , pober , j. , mcquinn , m. , jacobs , d. , & aguirre , j. 2012 , apj , 753 , 81 parsons , a. r. , pober , j. c. , aguirre , j. e. , et al . 2012 , apj , 756 , 165 patra , n. , subrahmanyan , r. , raghunathan , a. , & udaya shankar , n. 2012 , arxiv:1211.3800 pritchard , j. r. , & loeb , a. 2010 , phy . d , 82 , 023006 pritchard , j. r. , & loeb , a. 2008 , phy . d , 78 , 103511 sethi , s. k. 2005 , mnras , 363 , 818 shaver , p. a. , windhorst , r. a. , madau , p. , & de bruyn , a. g. 1999 , aap , 345 , 380 shkarofsky , i. p. , proceedings of the ire , vol.49 , no.12 , pp.1857 - 1871 , dec . 1961 subrahmanyan , r. & ekers , r. , private communication , 2012 sunyaev , r. a. , & zeldovich , i. b. 1975 , mnras , 171 , 375 sunyaev , r. a. , & zeldovich , y. b. 1972 , a&a , 20 , 189 thomspon , a. r. , moran , j. m. , & swenson , g. w. , interferometry and synthesis in radio astronomy , john wiley & sons , 2004 . tingay , s , private communication , 2012 h. c. van de hulst , ned.tijd.natuurkunde , vol.11 , p210 , 1945 ( english translation : w. t. sullivan , d. reidel , classics in radio astronomy , 1982 , 302 vedantham , h. , udaya shankar , n. , & subrahmanyan , r. 2012 , apj , 745 , 176

in these notes , we study a class of grand - canonical partition functions with a kernel depending on a small parameter . this class is directly relevant to nekrasov partition functions of susy gauge theories on the 4d-background , for which is identified with one of the equivariant deformation parameter . in the nekrasov - shatashvili limit , we show that the free energy is given by an on - shell effective action . the equations of motion take the form of a tba equation . the free energy is identified with the yang - yang functional of the corresponding system of bethe roots . we further study the associated canonical model that takes the form of a generalized matrix model . confinement of the eigenvalues by the short - range potential is observed . in the limit where this confining potential becomes weak , the collective field theory formulation is recovered . finally , we discuss the connection with the alternative expression of instanton partition functions as sums over young tableaux . apctp pre2014 - 001 confinement and mayer cluster expansions jean - emile bourgine + + _ pohang , gyeongbuk 790 - 784 , republic of korea _

1402.1626

_ `` lesson : if you are stupid , do nt get discouraged but try a different way to get the result . '' _ + madan lal mehta the recent developments in the study of @xmath1 susy gauge theories in four dimensions have driven a lot of progress in several fields of theoretical physics . this is particularly the case for the field of random matrix models . the alday - gaiotto - tachikawa ( agt ) correspondence @xcite relates partition functions to summations over young tableaux of a deformed plancherel measure @xcite . this relation has been used extensively to verify the agt proposal in various limits ( see for instance @xcite ) . similarly , the correspondence with quantum integrable systems proposed in @xcite hints for a connection between the standard matrix models methods ( such as loop equations @xcite and collective field theory @xcite ) and the mayer cluster expansion @xcite . the latter is a statistical physics technique employed in @xcite to treat the gauge theory partition functions . an investigation of this possible connection was performed in @xcite . there , the mayer expansion of a grand - canonical ( generalized ) matrix model is compared to the canonical model at large @xmath4 . is the size of the matrix , it is the number of integration variables after diagonalization . it corresponds here to a number of instantons . it should not be mistaken with the rank @xmath5 of the gauge group , which will remain finite in this paper . ] however , the results obtained in @xcite are not directly applicable to susy gauge theories . the goal of this paper is to extend some of these results to a larger class of models relevant to the gauge / integrability correspondence . more precisely , the models considered here are defined as @xmath6 they will be studied in the limit @xmath3 , with a @xmath7 a sum of two terms : @xmath8 the function @xmath9 presents two single poles at @xmath10 with residue @xmath11 . on the other hand , @xmath12 is a symmetric function independent of @xmath0 , and such that @xmath13 is finite . the parameter @xmath0 is assumed to have a small positive imaginary part , and integrations in [ def_cz ] are contour integrals over the real axis . following @xcite , contours are closed in the upper half - plane , and avoid possible singularities at infinity . ] ] furthermore , both @xmath14 and @xmath12 are supposed to have no singularities on @xmath15 . the presence of @xmath9 was neglected in @xcite , and this particular case may be recovered by setting @xmath16 . in the limit @xmath3 , the poles of @xmath9 pinch the integration contour , drastically changing the behavior of the model . physically , @xmath9 gives rise to a strong integration at short distance @xmath17 , which can no longer be treated perturbatively as @xmath18 becomes of order @xmath19 . instanton partition functions of @xmath1 @xmath20 gauge theories reduces after localization to coupled one - dimensional integrals of the form [ def_cz ] with @xmath21 @xcite . this computation is regularized by considering the gauge theories on the @xmath2-background which depends on two equivariant deformation parameters @xmath22 . the euclidean background @xmath23 is recovered in the limit @xmath24 . connections with quantum integrable systems appear when @xmath25 is sent to zero while keeping @xmath26 finite @xcite . this limit is now referred as the nekrasov - shatashvili ( ns ) limit . under the identification @xmath27 , it coincides with the limit where the kernel becomes close to one considered here . the remaining deformation parameter @xmath26 appears in the definition of the function @xmath12 . its expression depends on the matter content of the gauge theory . it is given below in the limit @xmath28 for @xmath1 super - yang - mills ( sym ) with fundamental flavors , and @xmath29 with an adjoint hypermultiplet of mass @xmath30 , @xmath31 let us also mention that setting @xmath32 while keeping @xmath21 , we recover the model proposed by j. hoppe s in @xcite , and further studied in @xcite . this paper is organized as follows . in the first part , we focus on the grand - canonical model and show that the free energy at first order in @xmath0 takes the form of an on - shell effective action . for @xmath21 , we recover the action proposed in @xcite , and the corresponding equations of motion take the form of a tba - like relation . after a brief reminder on bethe equations and thermodynamical bethe ansatz ( tba ) , the identification between the free energy and the yang - yang functional is done . in the second section , we consider the associated canonical model . it is shown that the previous results can be derived from a confinement hypothesis for the eigenvalues of this model . we further study the limit @xmath33 and recover the collective field action of a dyson gas @xcite . eventually , we compare our results to the alternative approach based on the expression of instanton partition functions as sum over young tableaux . appendices gather the most technical details . _ as the present paper was in preparation , we received the preprint @xcite where similar results are derived for nekrasov partition functions . however , the method presented here is different from the one employed in @xcite , and we believe it is interesting in its own right . _ the mayer cluster expansion provides an expression of the grand - canonical partition function [ def_cz ] as a sum over clusters . these clusters consist of a set of vertices connected by at most one link . to each cluster is associated a set of coupled integrals , with a measure @xmath34 at each vertex @xmath35 and a kernel @xmath36 for each link @xmath37 . taking the logarithm reduces the summation to connected clusters , and the free energy writes @xcite @xmath38 we denoted @xmath39 the connected clusters with @xmath40 vertices , and @xmath41 ( resp . @xmath42 ) their set of vertices ( resp . the symmetry factor @xmath43 takes into account the possibilities of re - arranging the vertices , it is the cardinal of the group of automorphisms that preserve the cluster @xmath39 . since @xmath44 is a sum of two terms @xmath45 and @xmath46 , the previous cluster expansion can be re - written as a sum over clusters @xmath47 with two types of links , referred as p- and g - links @xcite , @xmath48 with the shortcut notation @xmath49 . let us emphasize that clusters @xmath47 have the same structure as the previous clusters @xmath39 , but with links bearing an additional label p or g. this labeling modify the symmetry factors , @xmath50 because automorphisms must preserve the edges labels . the set of edges @xmath51 is obviously the direct union of the sets @xmath52 and @xmath53 of p- and g - links respectively . in order to compute @xmath54 at the leading order in @xmath0 , we need to determine the structure and order of clusters that give the minimal contribution for each term of the @xmath55-expansion . such clusters will be referred shortly as minimal clusters. our strategy is to work recursively on the number of links . connected clusters with the minimal number of links at fixed number of vertices have a tree structure . we will show that trees with @xmath40 vertices are of order @xmath56 , and then discuss the possibility to add either g- or p - links . adding a link can not decrease the order in @xmath0 , and we conclude that minimal clusters are of the order of the trees , i.e. @xmath56 . it implies that @xmath54 defined in [ sum_clusters ] is of order one . clusters made of g - links were considered in @xcite . each g - link brings a factor @xmath0 : g - trees are of order @xmath57 , and all g - clusters having cycles are sub - dominants . trees made of p - links have integrals that are also of order @xmath56 . at this order , they can be computed recursively , this is done in appendix . their contribution only depends on the number @xmath40 of vertices , and read @xmath58 trees involving both p- and g - links can be constructed recursively starting from a g - tree or a p - tree , and adding a number of vertices with g- or p - links . adding g - links to a p - tree ( or a mixed tree ) will obviously bring a new factor @xmath0 for each link . adding p - links to g - trees ( or mixed trees ) is similar to adding p - links to p - trees ( see section of the appendix ) , and also brings a factor @xmath0 per link . thus , any tree with @xmath40 vertices is of order @xmath56 . any cluster can be build from a tree by adding p- or g - links . adding a g - link always cost a factor @xmath0 . it implies that @xmath45-links can not belong to the cycles of minimal clusters . otherwise they could be removed , leading to a connected cluster of strictly smaller order . naively , adding a p - link brings a factor @xmath59 . but kernel residues are proportional to @xmath60 , reducing the degree to @xmath61 which is still subleading . however , there is a specific case where adding a p - link is costless : when the vertices @xmath62 and @xmath63 between which the p - link is inserted are already connected by a path of p - links , thus forming a p - cycle . it is easy to convince oneself that this occurs by computing the integral associated to the first diagrams , such as the triangle , for a simple potential . in this case , the additional p - link is costless because if @xmath62 and @xmath63 are related through a path of p - links , taking the residues of the kernel poles along this path leads to the relation @xmath64 , with @xmath65 an integer . the additional p - link is then evaluated at a fixed distance @xmath66 which is of order @xmath67 . let us emphasize that this phenomenon can not happen with g - links because @xmath13 is finite . note also that if the path between @xmath62 and @xmath63 had involved a g - link , the relation @xmath64 would no longer holds , and the cluster with additional p - link would be subleading . it is not possible to obtain a negative power of @xmath0 by adding a p - link , and we conclude that minimal clusters are of order @xmath56 . to summarize , minimal clusters have the structure depicted in figure , with g - links connecting sets of vertices related by p - links . cycles involve p - links but no g - links . to derive the expression of the grand - canonical free energy at first order in @xmath0 , we employ the technique presented in @xcite . this method relies on generating functions of rooted clusters . rooting a cluster @xmath47 into @xmath68 consists in marking the vertex @xmath62 , such that it is left invariant under preserving automorphisms ( @xmath69 ) , and that its associated integration variable @xmath70 is fixed . with a slight abuse of notation , this integration variable will also be denoted @xmath62 . the generating function of rooted clusters is defined as @xmath71 this expansion of @xmath72 as a sum over rooted clusters @xmath68 involves both p- and g - links . in order to treat separately the action of g and p kernels , it is useful to introduce two additional generating functions , denoted @xmath73 and @xmath74 , by constraining the root @xmath62 to be tied to other vertices only through p - links and g - links respectively . all these generating functions are of order @xmath19 in @xmath0 . we have defined three functions @xmath75 , @xmath76 and @xmath77 , and three relations among them will be derived : [ eom1 ] , [ eom_yg ] and [ eom_y ] . in the next subsection , these functions will be used to express the free energy at small @xmath0 . the limit @xmath3 brings two main simplifications at the basis of our derivation : the tree structure for g - links , responsible for [ eom1 ] and [ eom_yg ] , and the short range of p - interactions leading to [ eom_y ] . before giving the derivation of the three identities , we need to introduce some terminology . vertices connected to the root by a single link are called _ direct vertices_. we further distinguish _ direct g - vertices _ and _ direct p - vertices _ by the type of link between the vertex and the root . vertices related to the root through a path involving only p - links are called _ fixed vertices_. this set includes all the direct p - vertices . finally , the _ descendants _ of a vertex are the vertices connected to this vertex through a path that do not involve the root . the various types of vertices of a rooted clusters are highlighted in the figure . the simplest identity among @xmath75 , @xmath76 and @xmath77 is a factorization property at first order due to the fact that @xmath45-links can not form cycles in minimal clusters , @xmath78 indeed , this identity expresses the fact that rooted clusters can be decomposed into two disconnected sub - cluster , as shown in figure . the first sub - cluster involves the root , direct g - vertices and their descendant , whereas the second one contains the root , fixed vertices and their descendants . the identity [ eom1 ] will be used to eliminate @xmath76 , allowing to work only with @xmath75 and @xmath77 . the second identity is very similar to the one derived in @xcite for @xmath16 . it follows from the fact that the root in minimal clusters of @xmath74 can be connected to @xmath30 direct g - vertices , each being the root of a new cluster formed by their descendants . these new clusters are not connected to each other , otherwise we would form a cycle containing a g - link which would be subdominant . at first order , contributions of these clusters factorize , leading to @xmath79 the symmetry factor @xmath80 takes into account the possibility to permute the direct g - vertices . to establish the last identity , we focus on the minimal clusters contributing to @xmath81 . the set of fixed vertices form together with the root a rooted sub - cluster involving only @xmath46-links , that we denote @xmath82 where @xmath40 is the number of fixed vertices plus the root @xmath62 . taking apart the p - links of @xmath82 , each fixed vertex becomes the root of a sub - cluster formed by itself , its direct g - vertices and their descendants . these sub - clusters are now disconnected , otherwise a cycle involving a g - link would appear . it implies again a factorization property at leading order , @xmath83 the next step is to argue that the factors @xmath84 within the integrals can be approximated by @xmath74 and taken out of the integrations . this seems reasonable because the kernel poles fix the relative value of the variables @xmath85 with respect to the root @xmath62 as @xmath86 with @xmath87 a finite integer . in the limit @xmath3 , it is possible to approximate the residue factor @xmath88 with @xmath74 however , it turns out that this way of reasoning is too simple as it overlooks the contributions from the poles of @xmath84 . these contributions are not negligible and even play a major role in the proper arguments given in appendix . these arguments heavily rely on the hypothesis that the potential @xmath14 has only pole singularities within the contour of integration . it also requires that the kernel @xmath45 preserves this property under convolutions . both requirements are satisfied in the application to instanton partition functions of @xmath1 susy gauge theories . replacing @xmath84 with @xmath74 in [ ap_exp ] , we obtain @xmath89 the coefficients @xmath90 will turn out to be independent of @xmath62 . they can be related to the following integrals , @xmath91 where we identified @xmath92 . to derive this relation , we apply the mayer expansion to the integrals @xmath93 . in the product , terms with @xmath94 act as a potential for the @xmath95 variables @xmath96 . but we can take an alternate point of view and consider the expansion of the full product with @xmath40 variables over rooted clusters with root @xmath97 . this expansion contains terms associated to disconnected clusters . such terms factorize into connected part contributions . but contributions of clusters with no root vanishes because of the absence of poles in the potential . ] where @xmath98 is a cluster of @xmath65 vertices and with only p - links . poles in the kernel fixes the relative values of the variable @xmath85 in terms of one variable @xmath99 , but there is no way to fix the remaining variable . this last integral is vanishing , due to the zero volume property [ zero_volume ] . ] thus , only connected clusters remain , they can be identified to the clusters @xmath82 considered above , and @xmath100 the integer @xmath101 is the number of configurations ( i.e. labeled clusters ) producing the same cluster @xmath82 . by definition , the symmetry factor @xmath102 is the number of labellings of @xmath95 vertices ( the root is fixed ) divided by the number of equivalent configurations @xmath101 , and thus @xmath103 . the integral @xmath93 is a polynomial in @xmath104 of degree @xmath105 times @xmath106 , starting with the term since @xmath107 , we have shown that minimal clusters with p - cycles are of order @xmath56 ] @xmath109 this expression is derived in appendix . for the specific value @xmath21 relevant to nekrasov partition functions , we have the simple result @xmath110 also obtained in appendix . for general @xmath104 , we introduce the function @xmath111 inserting this expression in [ factor_yp ] , and using the factorization property [ eom1 ] , we conclude that @xmath75 and @xmath77 satisfy the identity @xmath112 together with [ eom_yg ] , these two relations allow in principle to find the expression of the generating functions @xmath72 and @xmath74 . at @xmath16 , there are no p - links , @xmath113 , and we recover from [ eom_yg ] the result of @xcite . . blue vertices are dressed by g - links ( in dashed ) , and related to each other with p - links ( in plain ) . they form meta - vertices , drawn as black circles.,width=302 ] it remains to relate the free energy to the generating functions @xmath75 and @xmath77 . as can be seen in figure , the minimal clusters that contribute to the free energy at first order have a tree - structure in terms of g - links . we call meta - vertices the sets of vertices tied together by p - links . at large distance , figure resemble a tree with vertices replaced by meta - vertices . thanks to this tree structure , the basso - sever - vieira formula @xcite still applies . indeed , the combinatorial argument given in the proof @xcite ( appendix a ) for the case @xmath16 can be re - used for the general case since it is possible to associate uniquely a g - link to each meta - vertex . the inner structure of meta - vertices play no role in our discussion , and the symmetry factors are such that the free energy [ sum_clusters_ii ] is again given by a difference of two terms , @xmath114 this equation is graphically represented in figure . in the rhs , the second term @xmath115 is obtained by attaching two rooted clusters through a g - link . since the roots belong to two different meta - vertices , @xmath116 to obtain the term @xmath117 , focus on a meta - vertex . taking apart its p - links , the inner vertices ( in blue on figure ) are the roots of disconnected sub - clusters with direct vertices of type g only . it means that they should be dressed with @xmath74 , and the first term is @xmath118 where @xmath119 is the connected cluster with only p - links associated to a meta - vertex . one possibility to compute this expression is to again approximate @xmath120 and evaluate the remaining integrals , but the justifications may be tedious . fortunately , it is possible to make use of a result demonstrated in @xcite ( appendix b ) which relates the symmetry factors of rooted clusters with non - rooted ones , @xmath121 in this formula , the summation is over the class of equivalence of rooted clusters obtained by rooting a vertex @xmath122 of the original cluster @xmath119 . this identity allows to single out one of the integration variable , and rewrite @xmath117 as @xmath123 we recognize the integrals that appeared in the expression [ ap_exp ] of @xmath81 . using the same trick to evaluate them , we get @xmath124 where we introduced the primitive @xmath125 at @xmath21 , @xmath110 and @xmath126 is equal to the dilogarithm function @xmath127 . the additional factor @xmath128 necessary to transform @xmath129 into @xmath126 , and thus generate the @xmath127 at @xmath21 , comes from the choice of a root among the @xmath40 inner vertices of a meta - vertex . the expression of the grand - canonical free energy takes a nice form if we re - introduce the relation [ eom_yg ] between @xmath75 and @xmath77 , to write it as @xmath130\dfrac{dx}{2i\pi}}.\ ] ] instead of @xmath75 and @xmath77 , let us define @xmath131 and @xmath132 as @xmath133 the function @xmath134 is interpreted as the grand - canonical instanton density ( see appendix or @xcite ) . this change of variables allows to re - write @xmath135 as an on - shell action , i.e. @xmath136,\quad\text{with}\quad \left.\dfrac{\d\sgc}{\d\rho}\right|_{\rho=\rho^\ast}=0,\quad,\left.\dfrac{\d\sgc}{\d\vphi}\right|_{\vphi=\vphi^\ast}=0,\ ] ] and @xmath137=\hf\int{\rho(x)\rho(y)g(x - y)dxdy}+\int{\rho(x)\vphi(x)dx}+\int{l_\a(qq(x)e^{-\vphi(x)})\dfrac{dx}{2i\pi}}.\ ] ] the equations of motion reproduce the two relations [ eom_y ] and [ eom_yg ] obtained previously . at the special value @xmath21 , we recover the results presented in @xcite ( formula 6.1 ) . at @xmath21 , the equations of motion [ eom_yg ] and [ eom_y ] imply that the function @xmath138 defined in [ def_rp ] obeys a non - linear integral equation ( nlie ) , @xmath139 this equation appears in the thermodynamical bethe ansatz method developed in @xcite , usually written in terms of the pseudo - energy @xmath140 with rapidity @xmath62 . it does not pertain to a specific integrable model , but , on the opposite , provides a way to unify the description of different models @xcite . it appears here with an arbitrary potential @xmath14 and kernel @xmath12 , provided they satisfy the proper analyticity conditions necessary to derive [ eom_y ] . in this subsection , we recall how the nlie can be obtained from a general set of bethe equations . from the comparison with our previous results , a dictionary will be established with several integrable model quantities . in particular , we will identify the free energy @xmath135 with the yang - yang functional introduced in @xcite . although these results are not new and can be found in the existing literature on integrability , we present them here for completeness . we restrict ourselves to the kernel given in [ def_g ] for @xmath1 sym . consider the variables @xmath141 , @xmath142 , satisfying a set of @xmath143 equations , @xmath144 by analogy with integrable systems , the variables @xmath141 will be called _ bethe roots _ and the equations [ bethe_equ ] _ bethe equations_. depending on the form of the potential @xmath14 , and the value of @xmath0 , these equations may be relevant to the system of bosons in 1d with @xmath145-interaction ( or quantum non - linear schrdinger equation ) , or to the xxx ( or @xmath146 ) spin chain @xcite . the parameter @xmath55 is interpreted as a twist of the periodic boundary conditions . to establish the nlie associated to [ bethe_equ ] , we employ a trick that goes back to @xcite . sym . for a nice review , see @xcite . ] first , by taking the logarithm , we find @xmath147 where @xmath148 is an integer . it leads to define the counting function @xmath149 where @xmath150 is the baxter q - function , a monic polynomial with zeros at the bethe roots position . from [ log_bethe ] , we deduce that @xmath151 is an integer @xmath148 at @xmath152 . but it may also be an integer for other values of @xmath62 , and we should introduce a contour @xmath153 that surrounds only the set of bethe roots . the function @xmath154 has poles for @xmath151 integer , with residue @xmath155 so that for any function @xmath156 without singularity in the domain delimited by the contour @xmath153 , we have @xmath157 applying this result to [ def_eta ] , we get after integration by parts the integral equation @xmath158 with @xmath12 given in the first line of [ def_g ] . it must emphasized that this nlie has been obtained without taking the thermodynamical limit @xmath159 . in order to compare with the gauge theory result in [ nlie ] , assumptions of continuity for @xmath151 are superfluous , and it is perfectly safe to work with a discrete set of roots . actually , the condensations of roots only appear when we further send @xmath160 @xcite . comparing [ nlie ] with [ tba ] leads to identify : @xmath161 in addition , the contours of integration must coincide , implying that all the bethe roots satisfy @xmath162 , and that they are the only singularities of [ func_eta ] in the upper half - plane . then [ nlie ] corresponds to the nlie of tba in the bosonic case , for a scattering amplitude @xmath163 such that @xmath164 @xcite . the even parity of @xmath12 is a consequence of the unitarity of @xmath163 . we may further observe that the density of bethe roots is ( minus ) the derivative of @xmath134 , and the logarithm of the baxter q - function is the resolvent associated to @xmath134 , @xmath165 the relations found previously between the gauge theory densities and the bethe roots system allow to interpret the free energy [ fe ] as a yang - yang functional . indeed , from the identification [ ident ] , we find after integration by parts , @xmath166 taking the difference , and using the relation [ rel_densities ] with the bethe roots density leads to @xmath167\dfrac{dx}{2i\pi}}=-2i\pi\sum_{i=1}^m{\eta_iu_i}.\ ] ] coming back to the expression [ fe ] of the free energy , the bethe roots density can be introduced using again the integration by parts , and @xmath168 where we used the primitives @xmath169 this is indeed the yang - yang functional first defined in @xcite . interaction . for simplicity , we consider an odd number of bosons , and an even number of bethe roots , satisfying the equations @xmath170 it corresponds to [ bethe_equ ] with @xmath171 and @xmath172 where @xmath173 is the volume . up to a factor @xmath35 , [ yy ] is equivalent to the yang - yang functional : @xmath174 ] in this section , we turn to the canonical ensemble and study the partition function @xmath175 defined in [ def_cz ] . the results obtained previously for the free energy at small @xmath0 will be recovered from the expression of @xmath175 using a simple confinement hypothesis . this phenomenon of confinement was called _ instanton clustering _ in @xcite . it is the underlying cornerstone of the derivation presented in the previous section , it notably appeared in the step leading to [ factor_yp ] . the use of generating functions in this derivation brought a stronger justification for the validity of this hypothesis . the study of the canonical partition function also provides a nice interpretation of the fields @xmath131 and @xmath132 in the action [ action ] . furthermore , this action reduces to the collective field action of the canonical partition function @xmath175 in the appropriate limit . this is an interesting check of the results obtained before . finally , we shall comment on an alternative approach @xcite based on the evaluation of the canonical partition function as a sum over residues that are in one - to - one correspondence with boxes of a set of young tableaux . to study the canonical partition function , we start from the results obtained on the grand - canonical side , and invert the discrete laplace transform with the formula @xmath176 since confinement originates from the p - term of the kernel , it is better understood when the g - interaction is turned off . we will first examine this simpler case , and later re - introduce the g - term . setting bluntly @xmath32 in [ action ] , the equations of motion imply @xmath177 and only remains @xmath178 the definition [ def_la ] of the function @xmath126 provides the @xmath55-expansion of the free energy . after exponentiation , the @xmath55-expansion of the grand - canonical partition function can be plugged into the inversion formula [ invert_laplace ] to give @xmath179 where we denoted @xmath180 the following integral , @xmath181 this integral equals to @xmath182 which is actually independent of @xmath62 . we would like to compare the expression [ exp_zc_nog ] with the original definition [ def_cz ] of @xmath175 for @xmath32 . in the dyson gas interpretation @xcite , [ def_cz ] describes a gas of particles in one dimension , with position @xmath85 , in an external potential @xmath183 , and interacting through the kernel @xmath46 . for standard matrix models , these particles are associated to the eigenvalues of the matrix . here , it is more suitable to call them _ quarks _ , by ( rough ) analogy with the confinement in qcd . contrary to the g - interaction which remains weak for any distance @xmath184 , the p - interaction becomes strong at small distance , i.e. when @xmath185 . thus , the particles are expected to be confined , forming hadrons of an arbitrary number of quarks . since hadrons are supposed to be of size @xmath186 , quarks in the same hadrons experience an equal potential ( at first order in @xmath0 ) . then , the total potential associated to a hadron of @xmath65 quarks at position @xmath62 is approximately @xmath187 . the factor @xmath188 is simply the product of the potentials that each quark feels , taken at the center of mass @xmath62 . the additional factor @xmath180 reflects the inner structure of hadrons , it takes into account the interactions between quarks through p - links . assume that @xmath46 hadrons are formed , each containing @xmath87 quarks . using the confinement approximation for [ def_cz ] , we find the integral in [ exp_zc_nog ] . in this approximation , the p - interaction between hadrons is neglected because @xmath189 is of order @xmath59 for a finite distance @xmath184 . to fully obtain [ exp_zc_nog ] , it only remains to multiply by the appropriate combinatorial coefficients ( @xmath190 and @xmath191 for respectively the indistinguishability of hadrons and quarks within a hadron ) , and sum over the possible configurations . this shows that the expression [ exp_zc_nog ] , and by extension the action [ action ] for the grand - canonical ensemble , can be derived from the definition [ def_cz ] using the confinement hypothesis . the same conclusions are reached when the g - interaction is turned back on . at first order in @xmath0 , the grand - canonical partition function @xmath192 can be expressed as a path integral over the field @xmath131 and @xmath132 , with the action [ action ] , @xmath193\exp\dfrac1\e\left(\hf\int{\rho(x)g(x - y)\rho(y)dxdy}+\int{\rho(x)\vphi(x)dx}+\int{l_\a(qq(x)e^{-\vphi(x)})\dfrac{dx}{2i\pi}}\right)}.\ ] ] expanding the exponential of the function @xmath126 , but keeping the other terms , we find @xmath194e^{\frac1{2\e}\rho g\rho+\frac1\e \rho\vphi}\left(1+\sum_{p=1}^\infty{\dfrac{\e^{-p}}{p!}\sum_{\ { k_i\}}\prod_{i=1}^p\dfrac{i_{k_i}}{k_i}\int{\left(qq(x_i)e^{-\vphi(x_i)}\right)^{k_i}\dfrac{dx_i}{2i\pi}}}\right)},\ ] ] where we used the simplified notations @xmath195 for the g - kernel ( first term in the exponential [ zgc_coll ] ) and @xmath196 for the source term ( second term in [ zgc_coll ] ) . using again the inversion formula [ invert_laplace ] , we deduce the expression of the canonical partition function , @xmath197e^{\frac1{2\e}\rho g\rho}e^{\frac1\e\int{\vphi(x)\left[\rho(x)-\e\sum_ik_i\d(x - x_i)\right]dx}}}}.\ ] ] the field @xmath138 appears to be a lagrange multiplier enforcing the equality @xmath198 the function @xmath134 is interpreted as a density of quarks where the positions of confined quarks are replaced by their center of mass , @xmath199 this interpretation is not a surprise , given the definition [ def_rp ] and the results of @xcite , briefly summarized in appendix : @xmath134 coincides with the grand - canonical density @xmath200 defined in [ def_rgc ] as the vev of the quarks density operator . at first order , this density can be replaced by a density of hadrons weighted by their number of quarks . this density enters in the expression of the kernel term @xmath195 , implying that g - interactions of quarks can be approximated by an interaction between hadrons . indeed , plugging the expression [ rho_k ] of @xmath134 into [ zc_g ] gives @xmath201 this expression can be obtained from the definition [ def_cz ] using again the confinement hypothesis . ] i.e. to neglect the term @xmath202 . it is possible because this term is a multiplicative correction of order @xmath61 to @xmath203 which is of order one only within cycles where @xmath61 corrections are subleading . ] it is natural to wonder how the results of @xcite concerning the comparison with matrix model techniques extend to the present model . here we briefly discuss the collective field theory of the canonical model @xcite . the collective action presented in @xcite is not adapted to the treatment of the model [ def_cz ] . indeed , it only reproduces the sum over clusters with tree structures in the mayer expansion , thus overlooking the cycles involving p - links . furthermore , it assumes that @xmath203 can be treated perturbatively , so that @xmath204 at first order , which is not valid within a p - cycle where @xmath205 . in order to retrieve the collective field theory of @xcite , we need to impose @xmath206 for any link , which can be achieved by sending @xmath33 . to have the g - links of same order , we rescale @xmath207 . it is also necessary to renormalize the fugacity @xmath208 , the free energy @xmath209 and the density @xmath210 . as a result , the grand - canonical free energy at first order in @xmath104 ( and @xmath0 ) is given by a summation over clusters with a tree structure , involving both p- and g - links , which can be compared with the collective field theory of the canonical model . using the asymptotic [ il_1st ] for @xmath93 , the function @xmath129 as @xmath33 can be approximated by a tree function , @xmath211 the tree function @xcite is related to the principle branch of the lambert w function through @xmath212 . it satisfies the following properties , @xmath213 that can be used , together with the identity [ eom_yg ] , to eliminate @xmath77 from the expressions @xmath214 the free energy [ fe ] simplifies as @xmath33 into @xmath215dx},\ ] ] where we used [ def_rp ] to replace @xmath75 with @xmath131 . this expression should be compared with the collective action associated to the partition function @xmath216 we analyze the expression [ fgc_a0 ] term by term . the first term is obviously associated to the double sum of @xmath217 in the exponential . remember that the grand - canonical density @xmath134 is equal to the canonical density defined in [ rel_rho ] , up to a normalization factor @xmath218 which also appears in the relation [ rel_fe ] between free energies . the second term in [ fgc_a0 ] is a potential term generated by the product of @xmath219 in [ zc_order1 ] . the third term comes from the double sum of @xmath220 using the regularization @xmath221 . this approximation can not be made in the general model [ def_cz ] , it is only valid for p - links in tree structures . finally , the last term , also called _ entropic term _ , is a gibbs factor . it can be computed from the change of integration measure @xmath222 $ ] ( see @xcite , appendix c for a derivation ) . we have thus verified that [ fgc_a0 ] coincide , in the sense of @xcite , with the collective field action of the canonical model . the contour integrals involved in the expression of the nekrasov instanton partition functions can be evaluated exactly . the residues are in one - to - one correspondence with the boxes of a set of young tableaux @xcite . this expression of the canonical partition function is at the origin of an alternative approach to the ns limit . this approach was first employed by nekrasov and okounkov in @xcite to recover the seiberg - witten prepotential @xcite from the instanton partition function in the @xmath23 limit @xmath24 of the @xmath2-background . it was then extended to the ns limit @xmath28 in @xcite . for simplicity , here we restrict ourselves to @xmath1 @xmath20 sym with @xmath223 fundamental flavors . however , the method is much more general , and applies to quiver theories as well @xcite . in this approach , the free energy is given in terms of a function satisfying a baxter tq relation . here , we explain how this equation relates to the system of bethe roots considered in the subsection . the purpose of this subsection is not to provide a rigorous derivation , but simply to illustrate the consequences of a relation between densities . accordingly , we will skip the treatment of the perturbative part of the gauge partition function , and only refer to @xcite for more details . super yang - mills with gauge group @xmath20 and @xmath223 hypermultiplets in the fundamental representation is sometimes called super - qcd . its potential is a ratio of mass and gauge polynomials , @xmath224 where @xmath225 denotes the hypermultiplets masses , and @xmath226 the coulomb branch vevs . singularities on the real line are moved away by a small shift @xmath227 such that poles at @xmath228 are inside the integration contour , but not those at @xmath229 . the g - kernel in the limit @xmath28 is given by the first line of [ def_g ] , and @xmath21 . the residues @xmath230 are labeled by the multiple index @xmath231 where @xmath232 is a color index , and @xmath233 denotes a box in the @xmath40th young tableaux @xmath234 . they are interpreted as the instantons position in the moduli space , and are given by @xmath235 we will further denote @xmath236 , @xmath237 , the height of the @xmath35th column for the @xmath40th young tableaux , and @xmath238 the number of columns . in the canonical partition function @xmath175 , @xmath4 is the number of instantons , i.e. the total number of boxes , @xmath239 in the ns limit , the grand - canonical free energy is roughly equal to the canonical one at large @xmath4 , with @xmath240 fixed ( see [ rel_fe ] ) . it leads to consider young tableaux with infinitely many boxes . it is then argued that the summation is dominated by a certain young tableaux profile determined by extremizing the summation . this profile is characterized by a shape function @xmath7 @xcite , or equivalently by a density of instantons defined as @xmath241 we have shown in [ rel_densities ] that the bethe roots density @xmath242 is minus the derivative of the instanton density @xmath134 . in the ns limit , the instanton density is the same for canonical and grand - canonical models , and can be identified with @xmath243 . it was further shown in @xcite that the derivative of @xmath243 is a difference of two densities : @xmath244 in this problem , the numbers of columns @xmath238 is a natural cut - off . it can be sent to infinity , assuming that @xmath236 is vanishing for large @xmath35 . then , the two densities in the rhs of the previous equation are formally given by @xmath245 with @xmath246 and @xmath247 . the density @xmath248 is associated to the perturbative contribution to the gauge theory partition function . it satisfies @xmath249 the relation between densities leads to express ratios of q - functions as @xmath250 a priori , the infinite product in the definition of @xmath251 requires regularization . however , this function only appears here in well - defined ratios . the baxter tq relation associated to the bethe roots system of section is obtained by introducing a function @xmath252 such that @xmath253 introducing the relation [ qpsi ] , and denoting @xmath254 , we recover the tq equation established in @xcite , @xmath255 comparing this tq equation with the previous one , we observe that the potential @xmath14 is replaced by @xmath256 . as explained in @xcite , this is an effect of the perturbative term : the gauge polynomial dependence in @xmath14 cancels with the cross - term between @xmath257 and @xmath258 . we conclude that , up to the perturbative term @xmath259 , bethe roots correspond to the shifted height of young tableaux columns @xmath260 . the equation [ qsw ] is interpreted as a non - perturbative ( or quantized ) version of the seiberg - witten curve @xcite . [ [ confinement-1 ] ] confinement + + + + + + + + + + + confinement can also be understood using the young tableaux representation , where boxes play a role similar to the vertices in clusters . consider a box @xmath62 in a young tableau @xmath234 that will be the analogue of the clusters root . boxes of the same columns correspond to instantons at a microscopic distance @xmath261 of @xmath62 , like the fixed vertices with respect to the root of a cluster , or quarks of the same hadron . other boxes describe instantons at a macroscopic distance @xmath262 of @xmath62 , like the quarks of other hadrons . for @xmath263 , there is only one young tableau and the analogy is actually exact : quarks correspond to boxes , and hadrons to columns where the height gives the number of constituents . for a general @xmath20 gauge group , the picture is a bit more complicated because eigenvalues are scattered between several young tableaux , but a similar sketch can be drawn . in this paper , we presented a derivation of the tba - like equation for the nekrasov instanton partition functions of @xmath1 theories in the ns limit . it is based on the mayer cluster expansion , and makes use of generating functions for rooted clusters . the main result is the expression of the free energy at leading order as an on - shell action [ action ] . the method is applied to a larger class of models with @xmath264 , although the corresponding function @xmath265 remains to be computed . it would be interesting to know if this one parameter generalization of the tba equation is relevant to integrable problems . the extension to quiver gauge theories has not been considered here , but should be rather straightforward . in this case , different types of vertices should be introduced , being associated to each gauge groups . the generating functions of rooted clusters now wear an index to distinguish the types of roots . similar graphical relations can be established among them , thus generalizing the equations of motion obtained here . the phenomenon behind the simplification of the summation over clusters in the ns limit is a confinement of eigenvalues , as shown by the study of the canonical ensemble in the second section . the limit @xmath33 was also investigated , and exhibits agreement with the results from collective field theory . finally , connections with integrable systems was also briefly discussed , as well as the alternative approach based on summations over young tableaux . this paper only initiates the study of models defined by [ def_cz ] , and there is a lot more to be understood . in particular , the connection with matrix model techniques remains mysterious . those were investigated in @xcite for the case @xmath16 , but these results do not extend easily to the general case . new techniques must be developed to handle confinement in loop equations . in this perspective , the derivation of the subleading order in @xmath0 may be a very profitable exercise . at the moment , we are still far from generalizing the topological recursion developed in @xcite to models like [ def_cz ] . such a powerful tool would shed a new light on the agt correspondence : on the liouville side , correlators takes the form of a partition function to which a ( quantized ) topological recursion applies . the method developed here may also be useful in understanding the underlying algebraic structures . it was shown in @xcite ( see also @xcite ) that a central spherical hecke algebra , shc , acts on the instanton partition functions . it is natural to expect that this hopf algebra reduces to the more familiar yangian structures of integrable systems as @xmath28 . the method presented here to handle the ns limit may help to understand the reduction of symmetry algebra . i would like to thank davide fioravanti , yutaka matsuo , daniel ricci - pacifici and dima volin for valuable discussions and comments . i acknowledge the korea ministry of education , science and technology ( mest ) for the support of the young scientist training program at the asia pacific center for theoretical physics ( apctp ) . in this appendix , we provide some arguments for the factorization of the potential leading to the formula [ factor_yp ] . the clusters symmetry coefficients play no role in the discussion , and we focus only on the integral contributions . at this level , rooting a vertex @xmath62 is equivalent to fix the corresponding integration variable @xmath266 . we will assume that the potential @xmath14 can be decomposed into a sum over single poles in the upper half plane ( minus infinity ) denoted @xmath267 , with residues @xmath268 , and a regular part @xmath269 as @xmath270 where @xmath269 has no singularities inside the contour of integration . for simplicity , we also set @xmath271 , as it can be re - absorbed within @xmath14 . this form of the potential is relevant to the case of nekrasov partition functions . later , it will be necessary to replace the potential by a dressing function , in which case double poles may appear . we suggest to treat the double poles at @xmath272 with @xmath273 by shifting upward @xmath267 of a distance @xmath66 ( @xmath274 ) in one of the factors . formally , @xmath275 a similar treatment should be performed for higher order poles . one of the subtle points in the present demonstration is to show that this manipulation is perfectly valid at first order in @xmath0 , and does not influence the final result . this is not guarantee a priori because of the strong sensibility of the kernel @xmath46 in the precise position of the poles , in particular within p - cycles . we shall come back to this point below . we first restrict ourselves to the case of rooted cluster @xmath82 with a tree structure . as a warm - up , consider the simplest case of @xmath276 vertices , i.e. a root attached to a single leaf . the corresponding integral writes @xmath277 to evaluate the integral over @xmath278 , we use the decomposition [ decompose_q ] of @xmath279 , leading to the sum over residues @xmath280.\ ] ] from the identity @xmath281 we can reform @xmath282 and write @xmath283 we notice that the poles of @xmath284 that would contribute to an integration over @xmath62 are shifted upward in @xmath285 , and would still contribute . in a similar way , possible singularities of @xmath269 are shifted downward , and will not contribute to an integration . then , we can safely replace @xmath286 when integrating over @xmath62 . neglecting the poles of @xmath279 in the integration performed previously would have led to the opposite shift of @xmath282 , namely @xmath287 which would have been ambiguous for later @xmath62-integration . thus , those poles are necessary in the intermediate steps , and can not be neglected . only at the final stage , it is possible to write @xmath288 as a side remark , let us also mention that for @xmath289 given by [ decompose_q ] , we can further compute the contribution of a non - rooted cluster with two vertices , @xmath290 it is worth noticing here that the squared terms @xmath291 and @xmath292 have vanishing contribution , and only the cross - term remains . since by definition @xmath269 is regular at @xmath272 , everything is well - defined . vertices.,width=302 ] as a next step , we should explain how the previous factorization property propagates to any rooted tree by recursion . to do so , let us consider a rooted tree @xmath82 , and pick up one of the deepest leaf . we remind the reader that a leaf is a vertex attached to only a single other vertex , and the depth is the minimal distance between the vertex and the root , i.e. the minimal number of intermediate vertices plus one . this leaf is attached to another vertex , that we denote @xmath63 , with a depth @xmath293 . the vertex @xmath63 is attached to a number of leaves @xmath294 ( including the previous one ) , of maximal depth @xmath295 , and a single other vertex of depth @xmath296 ( see the configuration on the figure ) . the corresponding integral is @xmath297 where @xmath298 denotes the contribution of the bi - rooted tree obtained from @xmath82 by removing the leaves @xmath299 and rooting the vertex @xmath63 . integrals over @xmath300 are decoupled , and correspond to @xmath65 copies of the integral [ i - single ] . performing the integrations , we get @xmath301 thus , the potential of the leaves can be transposed to the vertex @xmath63 , at the price of small shifts in @xmath0 . we have checked before that those shifts are not harmful since no pole would cross the integration contour , and it is safe to replace @xmath302 ( although this replacement will be done only at the end of the computation ) . we may assume , after proper treatment of the multiple poles , that the new potential @xmath303 associated to the vertex @xmath63 has still the form [ decompose_q ] . then , the operation of integration over the deepest leaves can be repeated until all vertices are removed , and only the root remains . at the end of the process , the potential appears at the root , modulo @xmath0-shifts that can be neglected at first order , @xmath304 thus , we have shown that the potential factorizes out of the integrations . a slightly more general result can be established , allowing a different potential @xmath305 for each vertex @xmath35 , @xmath306 at first order . as a corollary , the result of the integration is independent of the exact distribution of the potentials among the vertices . it implies that the potential can be concentrated on any vertex of the tree . this stronger result will be necessary in the next subsection where the presence of cycles is investigated . the possibility to consider a different potential at each vertex is a key point to justify the manipulation used to remove the multiple poles that appeared in [ mpoles ] . indeed , it is possible to tune the poles @xmath267 with upward @xmath0-shifts at each vertex such that double poles never happen in the computation , at each order of the recursion . for such a distribution of potentials , the property [ p1 ] holds . taking the limit @xmath3 as a final step , these infinitesimal shifts drop and we conclude that [ p1 ] holds for any distribution of potentials , with multiple poles at initial or intermediate steps . to extend the result of the previous subsection to any rooted cluster @xmath82 , we need to investigate the effect of cycles . to begin with , we consider the necklace , a non - rooted cluster of @xmath40 vertices and @xmath40 links , each vertex forming two bonds ( figure ) . we will focus on a specific vertex @xmath63 and denote the other integration variables @xmath307 . the associated integration is @xmath308 the integrand can be identified with the contribution of a rooted chain of @xmath40 vertices , with root @xmath63 , and a deformed potential for @xmath99 : @xmath309 considering @xmath63 as a fixed variable , this potential for @xmath99 is perfectly valid , and the pole at @xmath310 simply plays the role of an additional variable @xmath267 in the decomposition [ decompose_q ] . there is however an important difference with the initial potential , which is that shifts of @xmath0 become meaningful when integrating over @xmath63 and can not be simplified . however , since we are only willing to move the @xmath311 part of the @xmath99-potential , we can still apply our result [ p1 ] to this rooted chain to get vertex with deformed potential [ deformed_pot ] for @xmath99 . integrating over @xmath99 , we obtain the potential @xmath312 for @xmath313 . integrating successively over @xmath314 , we obtain a potential for @xmath63 containing various shifts of each vertex potential . at this stage , it is not possible to simplify this set of shifts because of the presence of the additional factor in [ deformed_pot ] . however , we may perform a similar operation for the rhs of [ pot_in_y ] . simplification of @xmath0-shifts is possible in all the @xmath289 factors , leading to the equality [ pot_in_y ] at first order in @xmath0 . ] @xmath315 as in the case of rooted trees , the issue of multiple poles is resolved by the possibility of choosing a different potential at each vertex , thus allowing infinitesimal tuning of the poles . we observe again in [ pot_in_y ] that , at first order in @xmath0 , the potential can be concentrated on any vertex of the necklace . it is immediate to generalize this result to any ( non - rooted ) cluster with only one cycle . there always exists a vertex @xmath63 such that if removed , the cycle is broken and the cluster becomes a tree . as before , the integral associated to the initial cluster can be written as the integral over @xmath316 of a rooted tree with a deformed potential [ deformed_pot ] for one of the vertices attached to @xmath63 . using the property [ p1 ] for this tree , we show that the potential can be concentrated on @xmath63 , as in [ pot_in_y ] . this remains true for any distribution of the potentials over the vertices . thus , the potential can be concentrated on any vertex of the cluster . more precisely for 1-cycle clusters @xmath119 and any vertex @xmath317 , we have at first order @xmath318 as a corollary , the potential can be distributed arbitrarily on the vertices of the cluster . in order to transfer this results to rooted cluster , we may perform a functional derivation of the previous formula with respect to @xmath319 where @xmath62 is the root . concentrating the potential in @xmath62 before performing the derivation , we obtain the equivalent of [ p1 ] for rooted clusters with a single cycle . clusters with several 1-cycles can be treated recursively , choosing the integration variable @xmath63 such that removing it leads to a rooted cluster with one cycle less to which we may apply [ p1 ] . for arbitrary clusters , it may happens that @xmath63 is attached to more than two other vertices , but the argument extends smoothly , and any cluster can be reached by the recursion . we conclude that [ p1 ] is true for any rooted cluster @xmath82 with only p - links . in order to apply the previous results [ p1 ] and [ p2 ] to [ ap_exp ] , we need to replace @xmath14 with @xmath74 . this replacement is possible if @xmath74 obeys the property [ decompose_q ] , which depends on the exact form of the kernel @xmath12 . in the case of an even kernel with only poles singularities in the upper half - plane , such as those given in [ def_g ] , the replacement is possible . let us for instance consider a rooted cluster integral @xmath320 and attach a g - link @xmath321 to @xmath63 , such that the new root is @xmath62 . assume that @xmath320 satisfies [ decompose_q ] , and take the kernel relevant to @xmath1 sym in [ def_g ] , . ] one obtains @xmath322 i.e. that poles of @xmath320 in the upper half - plane are moved upward and those of the lower half - plane downward , leaving the result well - defined , and obeying again the relation [ decompose_q ] . attaching several vertices by g - links may bring multiple poles , as in [ mpoles ] , but these can be regularize in a similar way . since @xmath74 is a sum over an infinite number of clusters , it may however happen that @xmath74 does not obey the property [ decompose_q ] although we have the invariance of [ decompose_q ] under convolution with @xmath45 , as in [ int_g ] . this issue can be treated by recursion on the number of vertices , which provides a natural cut - off . from [ eom_yg ] it is seen that each step simply involves a g - convolution . finally , let us briefly mention that the possibility to replace @xmath14 with contributions of @xmath45-trees in the section [ a1 ] allows to show that mixed trees of @xmath40 vertices are also of order @xmath56 . to evaluate @xmath93 at leading order in @xmath104 , it is better to consider the expression [ factor_yp ] of @xmath90 . at first order , only trees contribute . we have already shown by recursion that these integrals give a contribution independent of the trees exact structure . this is the result [ ptrees ] with @xmath323 . it remains to compute the sum over symmetry factors , which can be done using a formula derived in @xcite , @xmath324 where the last equality uses the cayley formula for the number of labeled trees . putting everything together , we find [ il_1st ] for @xmath93 at first order in @xmath104 . to evaluate the integrals @xmath93 in the case @xmath21 , we use a method similar to what was done in @xcite . the cauchy determinant formula @xmath325 allows to expand @xmath93 on the permutations of the symmetric group @xmath326 , @xmath327 these integrals are non - vanishing only if the permutation is a cycle of maximal length @xmath40 . indeed , permutations can be decomposed into cycles and the integral factorizes into cycle contributions . cycles of length @xmath30 that do not contain the fixed variable @xmath328 give , after re - labeling the variables , the contribution ( @xmath329 ) @xmath330 which is vanishing , according to [ zero_volume ] . thus , only cycles of maximal length @xmath40 remain . there are @xmath331 such cycles , and their signature is @xmath332 . by re - labeling the variables , we easily show that all these cycles give the same contribution , which can be evaluated by choosing @xmath333 modulo @xmath40 . residues give @xmath334 , and @xmath335 the paper @xcite presents several results obtained for our model [ def_cz ] in the simplifying limit @xmath16 . at present , it is not known how to extend the matrix model techniques studied there to @xmath336 . nevertheless , many results obtained in @xcite does not require a strong assumption on the form of the kernel @xmath44 , and easily generalize to our case . they are summarized in this appendix . the action of @xmath337 on the grand - canonical free energy @xmath338 is a purely combinatorial result that applies for any kernel and potential , @xmath339 introducing generating function of @xmath340-rooted clusters , it can be generalized into @xmath341 the generating functions @xmath75 are related to the grand - canonical densities . introducing the grand canonical vev of an operator @xmath342 , @xmath343 in terms of the canonical vevs , @xmath344 the one - point grand - canonical density is defined as @xmath345 it is simply related to @xmath72 as @xmath346 . a similar result holds for the 2-points density , @xmath347 the relations of this subsection are valid at all order in @xmath0 . the relations between grand - canonical and canonical ensemble derived in @xcite are obtained from the inversion [ invert_laplace ] of the discrete laplace transform . as such , they are not model dependent and still valid in our case . thus , at large @xmath4 the canonical free energy is related to the grand - canonical one through the legendre transform @xmath348\simeq\frac1\e\fgc(q,\e)-n\log q,\quad\text{when}\quad q\p_q\fgc(q,\e)=n\e,\ ] ] where free energies are defined as @xmath349 in the scaling limit @xmath350 with @xmath351 fixed , this relation becomes at first order @xmath352 where @xmath353 and @xmath218 are conjugated variables , and @xmath354 a similar relation also applies to densities at leading order in @xmath0 : at one - point , @xmath355 and at two - points , @xmath356

we prove local estimates of solutions for the parallel refractor and reflector problems under local assumptions on the target set , and no assumptions are made on the smoothness of the densities .

1404.4011

suppose we have a domain @xmath2 and a domain @xmath1 contained in an @xmath3 dimensional surface in @xmath4 ; @xmath1 is referred as the target domain or screen to be illuminated . let @xmath5 and @xmath6 be the indexes of refraction of two homogeneous and isotropic media i and ii , respectively , and suppose that from the region @xmath7 surrounded by medium i , radiation emanates in the vertical direction @xmath8 with intensity @xmath9 for @xmath10 , and @xmath1 is surrounded by media ii . that is , all emanating rays from @xmath7 are collimated . a _ parallel refractor _ is an optical surface @xmath11 interface between media i and ii , such that all rays refracted by @xmath11 into medium ii are received at the surface @xmath1 , and the prescribed radiation intensity received at each point @xmath12 is @xmath13 . assuming no loss of energy in this process , we have the conservation of energy equation @xmath14 . under general conditions on @xmath7 and @xmath1 , and when @xmath15 is a radon measure in @xmath16 , the existence of parallel refractors is proved in @xcite . the purpose of this paper is to study local regularity of parallel refractors and reflectors . indeed , under suitable conditions on the target and the measure @xmath15 , we prove local @xmath0 estimates . more precisely , if @xmath17 is a parallel refractor in @xmath7 , the target @xmath1 satisfies the local condition from @xmath18 , and the measure @xmath15 satisfies a local condition at that point , condition , then @xmath19 in a neighborhood of the point @xmath20 . throughout the paper we assume that media @xmath21 is denser than media @xmath22 , that is , @xmath23 . when @xmath24 , the geometry of the refractor changes . one needs to use hyperboloids of revolution instead of ellipsoids as indicated in @xcite . the plan of the paper is as follows . section [ sec : preliminaries ] contains results concerning estimates of ellipsoids of revolution , and subsection [ subsec : assumptiontargetrefractor ] contains basic assumptions on the target . section [ sec : regularityhypothesesonthetarget ] contains assumptions on the target modeled on the conditions introduced by loeper ( * ? ? ? * proposition 5.1 ) . indeed , we assume the target satisfies the local condition . we also introduce the differential condition , similar in form to condition ( a3 ) of ma , trudinger and wang @xcite , and show in theorem [ thm : awforperpendicularvectors ] and remark [ rmk : localequivalenceperpvectors ] that and are equivalent . in section [ subsec : localandglobalrefractors ] , we prove that if an ellipsoid supports a parallel refractor locally , then it supports the refractor globally provided the target satisfies the condition ( aw ) given in . the main result in this section is proposition [ prop : localimpliesglobal ] used later in the proof of theorem [ thm : mainestimate ] . section [ sec : definitionofrefractorandmainresults ] contains the main results , lemma [ lm : mainlemmarefractor ] and theorem [ thm : mainestimate ] , and also proposition [ prop : mu(s)=0 ] used later for the application of these results to show regularity of parallel refractors constructed in @xcite , corollary [ corollary : applicationtorefractors ] . section [ sec : holderregularityofgradient ] contains hlder estimates of gradients of refractors under the assumptions and on the target @xmath1 and the measure @xmath15 on @xmath1 . section [ sec : example ] contains examples of targets verifying the assumptions , see condition . up to this point in the paper , refractors are defined with ellipsoids supporting the refractor from above . refractors can also be defined with ellipsoids supporting from below , and in section [ sec : alternativedefinitionofrefractor ] we obtain the same regularity results for refractors with this definition . in section [ sec : regularityforthereflectorproblem ] we obtain similar regularity results for the _ near field parallel reflector _ problem . in this case the proofs are simpler because the differential condition implies the global inequality . * acknowledgements*. it is a pleasure to thank neil trudinger and philippe delano for useful comments and suggestions . we briefly review the process of refraction . our setting is @xmath4 . points will be denoted by @xmath25 . we consider parallel rays traveling in the unit direction @xmath8 . let @xmath26 be a hyperplane with upper unit normal @xmath27 and @xmath28 . we assume that the region below @xmath26 has refractive index @xmath5 and the region above @xmath26 has refractive index @xmath6 and @xmath29 , e.g. , air to glass . in such case , by snell s law of refraction , a ray coming from below in direction @xmath8 that hits @xmath26 at @xmath30 is refracted in the unit direction @xmath31 where @xmath32 since @xmath33 . , width=192 ] in particular , if @xmath34 and the hyperplane @xmath26 is so that the unit upper normal @xmath35 , then the refracted unit direction is @xmath36 with @xmath37 and @xmath38 . ] . the refracted ray is @xmath39 , for @xmath40 . here we have in mind that @xmath26 is the tangent plane to a refractor @xmath17 at @xmath41 and so @xmath42 . given @xmath43 and @xmath44 , consider @xmath45 @xmath46 is an ellipsoid of revolution with foci @xmath47 , @xmath48 . the projection of @xmath46 over @xmath49 is the ball @xmath50 . we define the lower and upper parts of @xmath46 by @xmath51 respectively . we can regard @xmath52 as the graph of the function @xmath53 for @xmath54 . if @xmath55 , @xmath56 , then @xmath57 . a ray with direction @xmath8 that hits from below the graph of @xmath58 at the point @xmath55 is refracted along the ray @xmath59 with @xmath60 , and therefore it passes through the focus @xmath61 . if @xmath55 and the focus @xmath61 can be written as @xmath62 for some @xmath40 and @xmath34 , then @xmath60 . given @xmath63 , let us define @xmath64 and we have @xmath65 given @xmath66 with @xmath67 , then @xmath68 is the unique ellipsoid passing through @xmath69 and having upper focus at @xmath61 . we have that @xmath70 and @xmath71 we fix a bounded domain @xmath72 and a cylinder set @xmath73 ; points in @xmath74 are denoted by the letter @xmath30 . given @xmath75 we define the region @xmath76 the region @xmath77 is open and unbounded . notice that if @xmath78 , then @xmath79 , for all @xmath80 . also , if @xmath81 and @xmath82 , then @xmath83 and so from we get @xmath84 we have that the function @xmath85 @xmath86 , defines the lower part of the ellipsoid @xmath68 for @xmath87 . we then have for @xmath82 , @xmath81 and @xmath88 , @xmath25 , with @xmath10 that @xmath89 with @xmath90 . we show that the function @xmath91 is differentiable of any order in all variables for @xmath92 , and @xmath82 . let us calculate first @xmath93 , @xmath94 . we have @xmath95 we have @xmath96 , with @xmath97 , @xmath10 , then @xmath98 for all @xmath10 from . therefore @xmath99 let us now calculate @xmath100 , @xmath101 . we have @xmath102 so @xmath103 for @xmath10 , @xmath81 and @xmath82 by . next we estimate the derivatives @xmath104 , @xmath105 , @xmath106 . we have @xmath107 now @xmath108 . hence @xmath109 for @xmath10 , @xmath81 and @xmath82 by . therefore , we obtain @xmath110 for all @xmath10 , @xmath82 , and @xmath111 . moreover , these estimates also hold for any @xmath69 such that there exists @xmath112 with @xmath113 . continuing in this manner , we get that the function @xmath91 is @xmath114 in all its @xmath115 variables on @xmath116 and @xmath117 , for any multi - index @xmath118 with a constant @xmath119 depending only on @xmath120 and @xmath121 . we will frequently use the following fact : @xmath122 here @xmath123 is the unit vector given by . we assume the convex hull of the target @xmath1 is contained in @xmath77 , where @xmath77 is given by . for each fixed @xmath124 , we assume each @xmath125 can be represented parametrically with respect to @xmath30 by the equation @xmath126 , with @xmath127 , where the function @xmath128 varies with the point @xmath30 , and @xmath129 is lipschitz in @xmath130 for each @xmath131 . [ lm : estimatesofpointsintarget ] for each @xmath132 , there exists a constant @xmath133 such that if @xmath134 are such that there exist @xmath135 and @xmath136 with @xmath137 , and @xmath138 , with @xmath130 defined by , then @xmath139 this implies that if @xmath140 , are both in @xmath1 , then the points @xmath141 can not be aligned , in other words , from each point @xmath69 one can see at most a point in @xmath1 on any straight line from @xmath69 . since the function @xmath142 is lipschitz in @xmath130 , then the left inequality in follows . indeed , @xmath143 to show the right inequality in , from we can write that ( @xmath144 ) @xmath145 @xmath94 . we write @xmath146 with @xmath147 on the segment @xmath148 $ ] joining @xmath149 and @xmath150 . this segment is contained in the convex hull of @xmath1 , and if this convex hull is bounded and contained in @xmath77 , then the desired estimate follows from the estimates for the derivatives of @xmath58 proved in subsection . the following two lemmas are a consequence of the inequalities for the derivatives of @xmath58 . [ lm : lipschitzinx0 ] let @xmath112 , @xmath81 and @xmath144 . let @xmath151 and set @xmath152 . assume that @xmath153 . then there is a constant @xmath154 such that @xmath155 for all @xmath156 we have @xmath157 for some @xmath158 $ ] . since @xmath159 for some @xmath160 , it follows from the estimates in subsection that @xmath161 for all @xmath156 , where @xmath119 depends on @xmath162 and @xmath121 . [ lm : estimateslipinyandx ] there exists a constant @xmath154 such that for all @xmath163 with the straight segment @xmath164\subset \mathcal t$ ] , and @xmath165 , we have @xmath166 for all @xmath10 , which is assumed convex . since @xmath167 for all @xmath125 , it follows that @xmath168 . we then write @xmath169$}\\ & = \sum_{j=1}^{n+1}\left(\frac{\partial\phi(x,\xi , x_0)}{\partial y_{j}}-\frac{\partial\phi(x_0,\xi , x_0)}{\partial y_{j}}\right)(y_j-\bar y_j)\\ & = \sum_{j=1}^{n+1}\sum_{i=1}^{n}\frac{\partial^{2}\phi(\zeta_i,\xi , x_0)}{\partial x_{i}\partial y_{j}}(x_{i}-x_{0_{i}})(y_j-\bar y_j)\end{aligned}\ ] ] for some @xmath170 $ ] . then the lemma follows from the estimates of the derivatives of @xmath58 proved in subsection . given @xmath171 and @xmath132 , let @xmath172 , @xmath173 , and @xmath174 , with @xmath175 $ ] . we consider the set of points @xmath176\},\ ] ] where @xmath177 is defined by . this set is a two dimensional wedge - shaped surface , in general non - planar , containing all rays having directions @xmath177 , @xmath178 , emanating from @xmath69 . the curve describing the tip of the vector @xmath177 , pictured in figure [ fig : picofcone ] , is not contained in the plane generated by the rays with directions @xmath179 and @xmath180 . if @xmath181 , then @xmath182 . hence , if @xmath183 , then by @xmath184 . and @xmath180 , with @xmath185 , @xmath186 , @xmath187 , @xmath188,width=192 ] for @xmath171 , and @xmath82 , let us define @xmath189_{x_0}:=\sigma \cap c(x_0,\bar y,\hat y).\ ] ] notice that by each ray @xmath190 intersects @xmath1 in at most one point for each @xmath191 . points in @xmath148_{x_0}$ ] have the form @xmath192 we will introduce the following local definition on the target @xmath1 . [ def : localtargetbis ] if @xmath82 we say that the _ target @xmath1 is regular from @xmath69 _ if there exists a neighborhood @xmath193 and positive constants @xmath194 , depending on @xmath193 , such that for all @xmath134 and @xmath195 we have @xmath196 for all @xmath10 with @xmath197 , @xmath198 , and @xmath199 . here @xmath200 , @xmath201 , and @xmath174 . we introduce below the differential condition , similar in form to condition ( a3 ) of ma , trudinger and wang @xcite . assuming that the function @xmath202 in the parametrization of the target is @xmath203 , and the set @xmath148_{x_0}$ ] is a curve for each @xmath69 , we prove in the following theorem that is equivalent to , see also remark [ rmk : localequivalenceperpvectors ] for the local case . theorem [ thm : awforperpendicularvectors ] is similar in form to a result of loeper , ( * ? ? ? * proposition 5.1 ) . [ thm : awforperpendicularvectors ] suppose that there exists a constant @xmath119 such that for all @xmath204 and @xmath205 , perpendicular vectors in @xmath206 , and for @xmath132 and for @xmath207 , we have @xmath208 where , as before @xmath209 and @xmath210 . then there exist structural constants @xmath15 and @xmath119 such that for @xmath134 and @xmath132 we have for @xmath211 $ ] and for @xmath212 that @xmath213 conversely , implies . let @xmath214 and @xmath132 . let @xmath215 , @xmath216 , and @xmath217 . notice that for @xmath205 perpendicular to @xmath204 we have @xmath218 for all @xmath219 $ ] . indeed , fix @xmath220 and @xmath221 , @xmath222 . applying with @xmath223 , @xmath224 ( notice that @xmath225 by ) , and @xmath226 , we obtain @xmath227 . since @xmath228 , follows . for @xmath229 , let @xmath230 denote the orthogonal projection of @xmath231 on the hyperplane through @xmath232 and normal @xmath204 , so @xmath233 is perpendicular to @xmath204 . we let @xmath234 . we will first show that there exist positive constants @xmath235 and @xmath236 such that @xmath237 for all @xmath10 . in fact , fix @xmath10 and let @xmath238 . from , we have @xmath239 . we claim that @xmath240 for all @xmath241 $ ] . to prove this claim , fix @xmath242 . by taylor s theorem , we have that @xmath243 for all @xmath241 $ ] . applying this inequality , first for @xmath244 and then for @xmath245 , multiplying the first by @xmath246 and the second by @xmath247 , and then adding yields @xmath248 which proves the claim . let @xmath249 . since @xmath202 is @xmath203 , we have that @xmath250 and therefore @xmath251 , since @xmath252 . hence @xmath253 , for @xmath178 . therefore we get @xmath254 since @xmath255 , with @xmath194 positive structural constants . this finishes the proof of . we now prove . let @xmath256 $ ] to be chosen later , and let @xmath211 $ ] . we have @xmath257 for some @xmath258 $ ] . using in @xmath259 we get that the last sum is less than or equal to @xmath260 notice that @xmath261 by orthogonality . hence we get that @xmath262 we will now choose @xmath263 and estimate each of the terms . we have that @xmath264 notice that either @xmath265 or @xmath266 . if the @xmath267 sign holds , we choose @xmath268 , and if the @xmath269 sign holds , we choose @xmath270 , so in either case @xmath271 . notice that since @xmath198 , we have from lemma [ lm : estimatesofpointsintarget ] that @xmath272 , if @xmath273 . hence @xmath256 $ ] . we next estimate the remaining terms . we have @xmath274 , again by lemma [ lm : estimatesofpointsintarget ] . similarly , @xmath275 . we also have from the estimates at the end of subsection [ subsec : estimatesderivativesofphi ] that @xmath276 to estimate the cubic form , let @xmath277 . as in the estimate of @xmath278 above , we have @xmath279 . we have @xmath280 combining all these estimates we obtain @xmath281 where @xmath282 and @xmath119 are structural constants . to obtain our desired estimate we write @xmath283 provided we choose @xmath284 , with @xmath285 sufficiently small depending only on the structure . taking @xmath286 , the first part of the lemma follows . we finally show that implies . fix @xmath132 , @xmath207 and let @xmath204 and @xmath205 be perpendicular vectors in @xmath206 . set @xmath209 and hence we can write @xmath287 . let @xmath288 be given by @xmath289 and notice that @xmath290_{x_0}$ ] for the value @xmath291 and for all @xmath292 . therefore from we have @xmath293 for @xmath294 . let @xmath295 . notice that @xmath296 is a normal vector to @xmath297 at @xmath232 . let @xmath298 be a curve contained in @xmath297 such that @xmath299 and @xmath300 . we then have that @xmath301 for all @xmath302 small enough . let @xmath303 . we have @xmath304 and @xmath305 and that is @xmath306 and this inequality holds for all @xmath307 small enough . letting @xmath308 , we get that @xmath309 , we get that @xmath310 . therefore we obtain that @xmath311 . [ rmk : localequivalenceperpvectors ] the local version of can be stated as follows : the target @xmath1 is regular from @xmath132 if there exists a neighborhood @xmath193 and a constant @xmath119 depending on @xmath69 such that for all @xmath207 and for all @xmath312 and for all vectors @xmath204 and @xmath205 such that @xmath222 we have @xmath313 where @xmath314 and @xmath315 . following the proof of theorem [ thm : awforperpendicularvectors ] , one can show that is equivalent to . let @xmath316 $ ] , and assume that the convex hull of @xmath317 , and @xmath7 is connected . given @xmath144 , set @xmath318 . we define @xmath319 the function @xmath17 is a _ parallel refractor _ if @xmath320 for all @xmath144 . we notice that , from the estimates of the derivatives @xmath321 from subsection [ subsec : estimatesderivativesofphi ] , any refractor is a lipschitz function in @xmath7 with a lipschitz constant depending only on @xmath322 in . suppose that @xmath17 is a parallel refractor in @xmath7 . in general , a local supporting ellipsoid might not support the refractor in all of @xmath7 . for example , rotating the figure [ fig : localnonglobal ] around the @xmath323-axis , is composed of the minimum of the ellipsoids given by @xmath324 , and @xmath325 , with @xmath326 and @xmath327 . the ellipsoid given by @xmath328 with @xmath329 , supports @xmath330 at @xmath331 locally but not globally ; @xmath332.,width=576 ] we obtain a refractor in 3d that has a local supporting ellipsoid that is not global . in this case , the target is composed of the circle @xmath333 with @xmath334 , and the point @xmath335 . the purpose of this section is to see that under condition below , a local supporting ellipsoid is also global . this will be used later in the proof of theorem [ thm : mainestimate ] . the target @xmath1 satisfies the condition aw from @xmath132 if for all @xmath207 , and for all vectors @xmath204 and @xmath205 such that @xmath222 we have @xmath336 where @xmath337 and @xmath338 . clearly implies . we will show in proposition [ prop : global ] that implies that for all @xmath134 we have @xmath339 for all @xmath10 , and @xmath178 , with @xmath340 . here @xmath172 , @xmath173 , and @xmath174 . let @xmath341 see and subsection [ subsec : assumptiontargetrefractor ] . condition means that for all @xmath342 , with @xmath343 , we have @xmath344 with @xmath345 . we are going to rewrite this condition in terms of the function @xmath346 . we will show that this is equivalent to @xmath347 recall that from subsection [ subsec : estimatesderivativesofphi ] we have that @xmath348 , where @xmath349 . let us set @xmath350 from and we have @xmath351 therefore , for @xmath352 we get @xmath353 from , @xmath354 . then @xmath355 therefore , @xmath356 . we will show that @xmath357 . calculating the second derivative with respect to @xmath307 , we have that the last inequality is equivalent to @xmath358 for all vectors @xmath222 . if we set @xmath359 , then @xmath360 . a calculation gives that @xmath361 since @xmath222 . therefore we have shown that condition is equivalent to . for simplicity in the notation we assume that @xmath362 and consider the solid ellipsoids @xmath363 where @xmath364 is defined by . from subsection [ subsec : assumptiontargetrefractor ] , we recall that the target @xmath1 is given parametrically from the origin by @xmath365 where @xmath354 and @xmath366 , with @xmath367 given in . [ prop : global ] suppose holds from @xmath69 ( assumed for simplicity 0 ) . let @xmath214 be given by @xmath368 and @xmath369 , and let @xmath370 for some @xmath220 , and let @xmath371 . set @xmath372 , given in . then @xmath373 and @xmath374 and in particular , @xmath375 for all @xmath231 in their common domain ( in particular for @xmath10 ) . inequality follows from which is equivalent to . we first notice that the set @xmath376 is contained on a hyperplane @xmath377 . indeed , for @xmath378 we have that @xmath379 , so @xmath380 which gives that @xmath381 . also @xmath382 , and so @xmath383 . subtracting these identities yields @xmath384 where @xmath385 , and so @xmath386 . in the same way , @xmath387 , where @xmath388 and @xmath389 . from and the definition of @xmath346 , we can write @xmath390 and @xmath391 . the proposition will follow from the following claims : * claim 1 * : if @xmath392 and @xmath393 , then @xmath394 . * claim 2 * : if @xmath395 and @xmath393 , then @xmath396 . * claim 3 * : if @xmath397 and @xmath398 , then @xmath399 . only the proof of the third claim uses condition aw , i.e. , . we prove claim 1 . let @xmath392 and @xmath393 . since @xmath393 , we have @xmath400 ; and since @xmath392 we have @xmath401 . therefore , @xmath402 it follows that @xmath403 . in this inequality , writing @xmath404 , and completing the squares we obtain , since @xmath405 , that @xmath406 and hence @xmath407 , which means @xmath394 . the proof of claim 2 is exactly the same . we now prove claim 3 . assume that @xmath397 and @xmath398 . notice that @xmath408 , where @xmath409 and hence , it is enough to show that @xmath410 . we first assume that @xmath411 and @xmath412 . we will show that we can write @xmath413 with @xmath414 and @xmath415 , for some @xmath416 . the above equality holds if and only if @xmath417 which holds if and only if @xmath418 which is true if and only if @xmath419 therefore we choose @xmath416 such that @xmath420 . since @xmath421 , we have that @xmath422 and @xmath423 . then it follows that @xmath424 and @xmath425 have both the same sign . from the last identity containing @xmath416 we obtain @xmath426 from we get @xmath427 hence @xmath428 , and therefore also @xmath429 next we assume that @xmath430 and @xmath411 . from , this implies that @xmath431 . if we write @xmath432 , then @xmath433 , and so @xmath434 . finally , if @xmath435 , then @xmath436 , and @xmath437 . so in this case , both inequalities @xmath397 and @xmath398 are impossible . this proves claim 3 and hence the proof of the proposition is complete . [ prop : localimpliesglobal ] suppose that @xmath17 is a parallel refractor in @xmath7 , @xmath144 , and assume that holds from @xmath318 . if there exist @xmath438 and @xmath292 such @xmath439 for all @xmath440 , then @xmath439 for all @xmath10 . we define @xmath441 we prove that @xmath442 , where @xmath443 . since from @xmath444 with @xmath125 if and only if @xmath445 , the inclusion is equivalent to show that @xmath446 . it is then enough to show that the extremal points of @xmath447 are contained in @xmath448 and that @xmath448 is convex . let @xmath449 be a extremal point of @xmath447 , then there exist @xmath450 with @xmath17 differentiable at @xmath451 and @xmath452 , see ( * ? ? ? * theorem 2.5.1 ) . let @xmath453 and let @xmath454 . since @xmath17 is differentiable at @xmath451 it follows that @xmath455 and then @xmath456 . we have that @xmath457 for all @xmath10 . letting @xmath458 yields @xmath439 for all @xmath10 , i.e. , @xmath459 and @xmath460 . to show that @xmath448 is convex , let @xmath461 and let @xmath462 , @xmath463 . consider @xmath464 and @xmath465 . since @xmath466 , we need to show that @xmath467 . we have @xmath468 for all @xmath10 by proposition [ prop : global ] . this completes the proof . [ lm : estimateconvexcombinationrefractor ] if @xmath17 is a refractor , then for each @xmath469 and each @xmath470 $ ] we have @xmath471 , with @xmath119 is a structural constant . we use the fact from subsection [ subsec : estimatesderivativesofphi ] that @xmath472 for any @xmath156 , @xmath473 and @xmath474 . given @xmath475 , let @xmath476 and set @xmath477 . we have @xmath478 for all @xmath156 . next , given @xmath479 $ ] , let @xmath480 , @xmath481 , and @xmath482 . applying the previous inequality @xmath483 , @xmath484 , and @xmath485 , we get @xmath486 for all @xmath156 . let @xmath487 . we have @xmath488 . therefore @xmath489 and @xmath490 . proceeding as in theorem [ thm : awforperpendicularvectors ] , we obtain @xmath491 and the lemma is proved . we will next prove our main lemma . [ lm : mainlemmarefractor ] suppose @xmath17 is a parallel refractor and the target @xmath1 is regular from @xmath492 in the sense of definition [ def : localtargetbis ] . there exist constants @xmath322 , @xmath235 and @xmath236 depending on @xmath493 , such that @xmath494 , and if @xmath495 , @xmath496 with @xmath497 , then there exists @xmath498 ( the segment from @xmath499 to @xmath500 ) such that if @xmath501 , then @xmath502 for all @xmath503_{x_0^{\star}}$ ] , @xmath198 , @xmath504 and for all @xmath505 . we remark that @xmath119 is a structural constant , depending only on the bounds for the derivatives of @xmath58 . since @xmath1 is regular from @xmath493 , there exists a neighborhood @xmath506 of @xmath493 such that holds for all @xmath134 and all @xmath507 . since parallel refractors are uniformly lipschitz in @xmath7 , there exists @xmath32 such that @xmath508 for all @xmath509 . if @xmath510 and @xmath511 , it follows that @xmath512 for all @xmath156 with @xmath513 . by continuity there exists @xmath514 $ ] such that @xmath515 . indeed , setting @xmath516 , we have that @xmath517 and @xmath518 . now set @xmath519 , @xmath152 , and recall we have set @xmath520 and @xmath521 . by definition @xmath522 and @xmath523 for all @xmath156 . hence we can write @xmath524 we now estimate @xmath525 . first , notice that by lemma , we have @xmath526 . we claim that @xmath527 we can write @xmath528 for some @xmath529 . then , by lemma , we have @xmath530 . since @xmath531 and @xmath532 , we get @xmath533 on the other hand , @xmath534 , and @xmath535 by convexity . using that @xmath536 , and @xmath537 we obtain @xmath538 it follows that @xmath539 from the estimates for the derivatives of @xmath58 in @xmath61 . therefore , inserting the last estimate in we obtain @xmath540 , where in the last inequality we have used that @xmath470 $ ] and @xmath541 by assumption . we then obtain claim . this yields @xmath542 if @xmath543 , then @xmath544 . so if @xmath503_{x_0^{\star}}$ ] , then we can apply to get @xmath545 , for @xmath546 where the constants @xmath194 depend on @xmath493 . and so @xmath547 finally , from lemma , we have that @xmath548 , and consequently @xmath549 , where the constant @xmath119 is structural . this completes the proof of the lemma . we are now a position to prove our main theorem . [ thm : mainestimate ] suppose @xmath17 is a parallel refractor , and the target @xmath1 is regular from @xmath492 in the sense of definition [ def : localtargetbis ] . let @xmath550 and @xmath322 be the constants of lemma [ lm : mainlemmarefractor ] . there exists a constant @xmath551 depending on @xmath119 and @xmath235 such that if @xmath552 , @xmath496 are such that @xmath553 then there exists @xmath232 on the straight segment @xmath554 such that we have @xmath555_{x_0^{\star}}:\lambda\in [ 1/4,3/4]\right\}\right)\cap\sigma\subseteq f_u(b_{\eta}(x_0)),\ ] ] with @xmath501 , @xmath556 and @xmath557 . from the assumption , we have @xmath497 and so lemma [ lm : mainlemmarefractor ] is applicable . let @xmath232 be the point in that lemma . fix @xmath558_{x_0^{\star}}:\lambda\in [ \frac{1}{4},\frac{3}{4}]\})\cap\sigma$ ] . so , there exists @xmath559_{x_0^{\star}}$ ] with @xmath560 $ ] such that @xmath561 . we then have from lemma [ lm : mainlemmarefractor ] that @xmath562 for all @xmath156 such that @xmath563 . the right hand side in the last inequality is strictly negative for all @xmath10 , with @xmath564 , if we choose @xmath565 . we pick @xmath556 , and @xmath566 with @xmath567 . since @xmath568 , it follows from that @xmath569 and @xmath570 . therefore @xmath571 for all @xmath572 . we now show @xmath573 . notice that by definition of @xmath574 we have @xmath575 . let @xmath576 denote the graph of @xmath17 in @xmath577 , that is , @xmath578 . consider @xmath579 recalling that @xmath580 . we claim that @xmath581 for @xmath582 with @xmath564 . indeed , let @xmath583 and notice that @xmath584 . since @xmath317 , with @xmath77 given by , we have @xmath585 . since @xmath586 for @xmath564 , we have that @xmath30 is below @xmath587 , and therefore we must have @xmath588 , and the claim follows . therefore , the infimum in is attained at some point @xmath477 with @xmath589 . we show that @xmath590 . indeed , we have @xmath591 for all @xmath592 . writing @xmath582 we have that @xmath593 and noticing that @xmath594 , we get @xmath595 for @xmath505 . since @xmath596 , we have that @xmath597 . we therefore obtain the local estimate @xmath598 for all @xmath599 , with @xmath307 small . since @xmath600 , it follows from the choice of @xmath322 in lemma [ lm : mainlemmarefractor ] that @xmath601 , the neighborhood in remark [ rmk : localequivalenceperpvectors ] . therefore we can apply with @xmath602 , and in particular holds at @xmath603 . hence applying proposition [ prop : localimpliesglobal ] with @xmath604 , we obtain that @xmath598 holds for all @xmath10 obtaining that @xmath590 . let @xmath15 denote the borel measure given on the target @xmath1 and let @xmath605 be a borel measure in @xmath7 . we say @xmath606 if and only if @xmath607 , where @xmath608 is defined by . assuming @xmath609 with @xmath610 and the energy conservation condition @xmath611 it is proved in @xcite the existence of a refractor @xmath17 such that @xmath612 the purpose of this subsection is to show the following proposition . [ prop : mu(s)=0 ] suppose @xmath17 is a refractor solving , and define @xmath613 let us assume the following conditions on @xmath1 and @xmath15 : 1 . @xmath1 is the graph of a @xmath614 function , say @xmath615 with @xmath616 some domain in @xmath49 ; 2 . given @xmath125 , let @xmath617 denote the tangent plane to @xmath1 at @xmath61 . assume that for each @xmath504 and for each @xmath124 , the line @xmath618 is not contained in @xmath617 , that is , this line intersects @xmath617 only at the point @xmath61 . ; because @xmath125 and some @xmath82 the line joining @xmath61 and @xmath69 is contained in @xmath617 , then there is ball @xmath448 centered at @xmath69 with @xmath619 . by the visibility condition the convex hull @xmath620 of @xmath61 and @xmath448 intersects @xmath1 only at @xmath61 . but then the line joining @xmath61 and @xmath69 is contained in @xmath620 and @xmath617 . therefore @xmath1 is not differentiable at @xmath61 . ] 3 . if @xmath621 with @xmath622 , then @xmath623 . then @xmath624 and we have the inequality @xmath625 we first notice that @xmath626 is a borel set for each closed ball @xmath448 . because if @xmath627 is closed , then @xmath628 is closed in @xmath1 . in fact , let @xmath629 with @xmath630 and assume @xmath631 with @xmath125 . there exists a subsequence @xmath632 for some @xmath633 . setting @xmath634 , we have @xmath635 where @xmath636 . since @xmath637 for all @xmath156 and for all @xmath638 , we have @xmath639 . therefore @xmath640 . let us assume for a moment that @xmath624 . it is easy to see that that @xmath641 , and therefore , follows . to prove that @xmath624 , let us consider the set @xmath642 we have that @xmath643 and therefore @xmath644 . under the assumptions ( a ) , ( b ) and ( c ) above , we are going to show that @xmath645 . to this end , we define @xmath646 by @xmath647 where @xmath576 is the graph of @xmath17 . we claim @xmath648 is lipschitz in @xmath649 . indeed , say @xmath650 and @xmath651 , then @xmath652 , since @xmath653 is lipschitz . similarly , we get the other inequality . now fix @xmath654 , so @xmath655 with @xmath656 . write @xmath657 with @xmath658 , @xmath582 and @xmath659 . we claim that @xmath648 is not differentiable at @xmath660 . suppose by contradiction that @xmath648 is differentiable at @xmath660 . it is easy to see that @xmath661 and also @xmath662 . from @xmath661 , it follows that @xmath663 for all @xmath664 with equality at @xmath660 . hence we have @xmath665 also from @xmath662 , we deduce that @xmath666 this implies that @xmath667 . let us set @xmath668 and @xmath669 . so @xmath670 and hence @xmath671 . if @xmath672 , then @xmath673 . hence @xmath674 but this implies that @xmath675 , since @xmath676 where @xmath677 . since by assumption @xmath678 , we obtain a contradiction . therefore , @xmath679 and so we can write @xmath680 . we claim that the line @xmath681 is contained in @xmath682 . indeed , if @xmath683 , then a simple calculation shows that @xmath684 . on the other hand , the line @xmath685 clearly intersects @xmath74 . we then obtain a contradiction with the assumption ( b ) above , and therefore @xmath648 is not differentiable at @xmath660 . if we set @xmath686 , then we proved that @xmath687 implies that @xmath648 is not differentiable at @xmath688 . since @xmath648 is lipschitz in @xmath649 , we get @xmath689 . therefore from ( c ) we obtain @xmath645 which completes the proof of the proposition . we introduce the following local condition at @xmath82 between the measure @xmath15 and target @xmath1 : there exist a neighborhood @xmath193 and a constant @xmath690 depending on @xmath69 such that @xmath691_{z}:\lambda\in [ 1/4,3/4]\right\}\right)\cap\sigma \right)\geq \hat c\ , \mu^{n-1}\,|\bar y-\hat y|\ ] ] for any @xmath214 , @xmath312 and for all @xmath692 small ( depending on @xmath69 ) . here @xmath693 denotes the @xmath605- neighborhood of the set @xmath525 in @xmath694 . [ thm : holdercontinuity ] suppose @xmath17 is a parallel refractor , the target @xmath1 is regular from @xmath492 in the sense of definition [ def : localtargetbis ] , and there exist constants @xmath695 and @xmath696 such that @xmath697 for all balls @xmath698 . suppose in addition that the local condition is satisfied at @xmath699 . then there exist constants @xmath322 , @xmath700 and @xmath701 depending on @xmath493 , such that if @xmath552 , @xmath496 are such that @xmath702 then we have @xmath703 with @xmath704 , where @xmath235 depends only on @xmath705 and @xmath706 in , and therefore from @xmath699 . from theorem [ thm : mainestimate ] we have @xmath707_{x_0^{\star}}:\lambda\in [ \frac{1}{4},\frac{3}{4}]\})\cap\sigma\subseteq f_u(b_{\eta}(x_0))$ ] where @xmath556 and @xmath708 . therefore from and , we obtain @xmath709 and the theorem follows . under all previous hypotheses on the target , we now show interior @xmath0 estimates . [ thm : c1alpha ] with the assumptions of theorem [ thm : holdercontinuity ] , there exist positive constants @xmath322 and @xmath119 , depending on @xmath493 , such that @xmath710 with @xmath711 for all @xmath712 . by theorem [ thm : holdercontinuity ] we get that @xmath713 is a singleton for each @xmath509 . take @xmath714 . we first show that if @xmath510 , then @xmath17 is differentiable at @xmath499 , and @xmath715 where @xmath659 . for @xmath716 we have @xmath717 for some @xmath718 . hence @xmath719 . to prove the inequality in the opposite direction , let @xmath720 , @xmath511 , and @xmath721 . we have that @xmath722 for some @xmath723 $ ] . from the estimates for the derivatives of @xmath58 from subsection , we can also write @xmath724 with @xmath119 a structural constant . on the other hand , since @xmath17 is lipschitz ( with a constant depending only on structure ) , we have @xmath725 . in addition , using theorem [ thm : holdercontinuity ] , we get that @xmath726 where @xmath119 is a structural constant depending also on @xmath493 . this yields @xmath727 , completing the proof that @xmath715 . we finally prove that @xmath728 . let @xmath729 , @xmath510 , @xmath511 , and set @xmath659 , @xmath721 . we then have latexmath:[$|d_{i}u(\bar x)-d_{i}u(\hat x)|=|d_{i}\phi(\bar x,\bar y,\bar x)-d_{i}\phi(\hat x,\hat y,\hat x)|\leq |d_{i}\phi(\bar x,\bar y,\bar x)-d_{i}\phi(\hat x,\bar y,\bar x)|+|d_{i}\phi(\hat x,\bar y,\bar x)-d_{i}\phi(\hat x,\hat y,\bar x)|+|d_{i}\phi(\hat x,\hat y,\bar x)-d_{i}\phi(\hat x,\hat y,\hat x)|\leq c\{|\bar x-\hat x|+|\bar y-\hat y|+ constant @xmath119 depending also on @xmath493 . [ corollary : applicationtorefractors ] suppose @xmath17 is a refractor solving , conditions ( a ) , ( b ) , and ( c ) from proposition [ prop : mu(s)=0 ] hold , and @xmath609 with @xmath731 for some @xmath732 . suppose that the target @xmath1 is regular from @xmath492 in the sense of definition [ def : localtargetbis ] , and the local condition is satisfied at @xmath699 . then there exist @xmath322 and @xmath119 depending on @xmath493 such that @xmath710 with @xmath733 given in theorem [ thm : holdercontinuity ] , and @xmath734 . by hlder s inequality @xmath735 , for all balls @xmath736 . therefore from we obtain condition , and the corollary follows . if we assume the target set @xmath1 is given by the graph of a @xmath737 function @xmath738 , then we are going to find a condition on @xmath653 so that holds locally . we will see that for to hold , the graph of the target needs to satisfy a quantitative condition , see . as in subsection [ subsec : localandglobalrefractors ] , we set @xmath739 . we have for @xmath125 that @xmath740 . then we can write @xmath741 . we therefore get that @xmath346 satisfies the implicit equation @xmath742 proceeding as in subsection [ subsec : localandglobalrefractors ] , setting @xmath359 we need to check that @xmath743 for all @xmath744 . we will prove for @xmath745 and by continuity will hold for @xmath30 in a neighborhood of @xmath745 . indeed , we next compute @xmath746 in terms of @xmath653 . since @xmath359 , we first proceed to calculate the derivatives of @xmath346 . from , @xmath747 and so @xmath748 , @xmath749 and @xmath750 . from , we get @xmath751 , @xmath752 , and @xmath753 now , noticing that @xmath754 and inserting the above expressions yields @xmath755 we therefore obtain that @xmath756 thus is equivalent to @xmath757 for all unit vectors @xmath204 . we have then proved that for @xmath1 the graph of @xmath653 , and for @xmath758,@xmath759 , the condition @xmath760 holds for all unit vectors @xmath222 , provided holds . by continuity we have that @xmath761 for all vectors @xmath222 and for all @xmath30 close to @xmath69 and @xmath61 close to @xmath762 . [ rmk : horizontalplaneisnotarefractor ] we show that if the target @xmath1 is a horizontal plane , then does not hold when @xmath763 . indeed , we show it is not true that @xmath764 . in @xmath765 , let @xmath766 , @xmath767 , @xmath768 and @xmath769 , with @xmath770 sufficiently large . let @xmath771 and we have @xmath772 @xmath773 . we also have that @xmath177 is on the two dimensional plane @xmath774 , and so @xmath148_{x_0}=[\bar y,\hat y]$ ] is the straight segment from @xmath149 to @xmath150 . we have @xmath775 . let @xmath776 , where @xmath777 is chosen so that @xmath778 . that is , @xmath779 . we claim that @xmath780 for @xmath781 , @xmath231 small . let @xmath782 and @xmath783 . then one can check that @xmath784 and @xmath785 which gives the claim . we can define refractor with ellipsoids touching @xmath17 from below , that is , the ellipsoids enclose @xmath17 . in fact , we can define analogously to @xmath786 and we can say @xmath17 is a _ refractor _ if @xmath787 for all @xmath144 . with this new definition , we can obtain the same regularity results as with definition by changing the inequalities accordingly . we indicate the changes . in definition [ def : localtargetbis ] , condition is replaced by : @xmath788 inequality is replaced by : @xmath789 lemma [ lm : estimateconvexcombinationrefractor ] is replaced by the convexity of @xmath17 . the inequality is replaced by : @xmath790 and in the proof of lemma [ lm : mainlemmarefractor ] , @xmath791 is replaced by @xmath792 with the corresponding changes in the inequalities . for the example in section [ sec : example ] , we now get that condition is replaced by @xmath793 for all unit vectors @xmath204 . in this section we shall prove results that are similar to the ones proved in the previous sections but for the reflector problem . since the arguments are similar , we will omit most details . we first review the process of reflection . our setting is @xmath4 , and points will be denoted by @xmath25 . we consider parallel rays moving in the direction @xmath8 . let @xmath26 be a hyperplane in @xmath4 with upper unit normal @xmath27 and let @xmath28 . by snell law of reflection , a ray coming from below with direction @xmath8 that hits @xmath26 at @xmath30 is reflected in the unit direction @xmath794 . in particular , if @xmath34 and @xmath795 , then the reflected direction is the unit vector @xmath796 the reflected ray consists of the points @xmath797 , for @xmath40 . we have in mind here that @xmath42 and @xmath582 , where @xmath17 is a reflector . if @xmath43 , and @xmath798 , then the set of @xmath799 with @xmath800 is a downwards paraboloid with focus at @xmath61 . it can be written as the graph of the function @xmath801 . the ray with direction @xmath8 that hits the graph of @xmath802 at @xmath803 is reflected in direction @xmath804 . if @xmath805 with @xmath69 not in the vertical ray with direction @xmath806 emanating from @xmath61 , then there exists a unique paraboloid with focus at @xmath61 passing through @xmath69 . such a paraboloid is described by the function @xmath807 where @xmath808 . we will frequently use the following fact : @xmath809 where @xmath123 is given by . let @xmath72 be open and bounded , and let @xmath810 be the cylinder @xmath811 $ ] . for a fixed number @xmath812 we define the region @xmath813 the set @xmath77 consists of the points @xmath61 such that the cylinder @xmath810 is contained outside the interior of the paraboloid @xmath814 . we will assume that the target @xmath1 has convex hull bounded and contained in @xmath77 . proceeding as in subsection [ subsec : estimatesderivativesofphi ] , it is easy to see that @xmath815 , @xmath816 are bounded uniformly for all @xmath10 , @xmath817 , and @xmath82 for @xmath105 and @xmath106 , where @xmath818 is compact . hence as in lemmas [ lm : lipschitzinx0 ] and [ lm : estimateslipinyandx ] , we obtain the following . given @xmath134 , @xmath132 , we let @xmath823 , @xmath824 , and @xmath174 , for @xmath175 $ ] . consider the set of points @xmath825\},\ ] ] where @xmath130 is given by . notice that from , if @xmath826 , then @xmath827 . for @xmath132 and for @xmath214 we assume @xmath148_{x_0}:=c(x_0,\bar y,\hat y)\cap\sigma$ ] is a curve joining @xmath149 and @xmath150 . we introduce the following condition on the target @xmath1 , similar to definition [ def : localtargetbis ] for refractors . [ def : reflectorlocaltargetbis ] if @xmath82 we say that the _ target @xmath1 is regular from @xmath69 _ if there exists a neighborhood @xmath193 and positive constants @xmath828 , depending on @xmath193 , such that for all @xmath134 and @xmath195 we have @xmath829 for all @xmath10 , @xmath198 , and @xmath199 . here @xmath200 , @xmath201 , and @xmath174 . [ thm : awforperpendicularvectorsreflectors ] suppose that there exists a constant @xmath119 such that for all @xmath204 and @xmath205 , perpendicular vectors in @xmath206 , and for @xmath132 and for @xmath207 , we have @xmath830 where , @xmath831 and @xmath832 . we first show that if holds for @xmath222 , then it holds for all vectors @xmath834 . in fact , we have @xmath835 so if @xmath836 , we pick @xmath837 with @xmath838 and @xmath839 , and we have @xmath840 . let @xmath841 and @xmath82 , @xmath842 , @xmath843 , @xmath844 , and @xmath845 . we then have @xmath846 for @xmath178 . fix @xmath10 , and let @xmath847 . since @xmath848 , where the second inequality follows from the analogue of for reflectors , it follows that @xmath849 . therefore , @xmath850 for all @xmath10 , and @xmath178 . this completes the proof of the theorem . [ rmk : localequivalenceperpvectorsreflectors ] analogously to remark [ rmk : localequivalenceperpvectors ] , a local version of can be stated as follows : the target @xmath1 is regular from @xmath132 if there exists a neighborhood @xmath193 and a constant @xmath119 depending on @xmath69 such that for all @xmath207 and for all @xmath312 and for all vectors @xmath204 and @xmath205 such that @xmath222 we have @xmath851 where @xmath314 and @xmath852 . we say @xmath853 $ ] is a parallel reflector from @xmath7 to @xmath1 if for each @xmath854 , there exists @xmath504 such that @xmath855 for all @xmath156 , where @xmath318 . in this case , we say @xmath856 . any reflector is lipschitz in @xmath7 with a uniform lipschitz constant depending on the bounds for the derivatives of @xmath802 , which are uniform for @xmath857 , @xmath10 and @xmath82 . we make the following hypothesis on the measures . also similarly to , we introduce the following local condition at @xmath82 between the measure @xmath15 and target @xmath1 : there exist a neighborhood @xmath193 and a constant @xmath690 depending on @xmath69 such that @xmath858_{z}:\lambda\in [ 1/4,3/4]\right\}\right)\cap\sigma \right)\geq \hat c\ , \mu^{n-1}\,|\bar y-\hat y|\ ] ] for any @xmath214 , @xmath312 and for all @xmath692 small ( depending on @xmath69 ) . here @xmath693 denotes the @xmath605- neighborhood of the set @xmath525 in @xmath694 . let @xmath17 be a parallel reflector such that the target @xmath1 is regular from @xmath492 in the sense of definition [ def : reflectorlocaltargetbis ] . there exist constants @xmath322 and @xmath235 , depending on @xmath493 such that if @xmath860 , @xmath496 and @xmath497 , then there exists @xmath498 such that if @xmath501 , then we have for all @xmath503_{x_0^{\star}}$ ] and for all @xmath504 and for all @xmath156 that the following inequality holds since @xmath1 is regular from @xmath493 , there exists a neighborhood @xmath506 of @xmath493 such that holds for all @xmath134 and all @xmath507 . since parallel reflectors are uniformly lipschitz in @xmath7 , there exists @xmath32 such that @xmath508 for all @xmath509 . for @xmath156 we have @xmath862 , and there exists @xmath514 $ ] such that @xmath863 . set @xmath152 , and @xmath864 and notice that @xmath865 . similarly as in lemma [ lm : mainlemmarefractor ] we get @xmath866 using lemmas [ lm : lipschitzcontinuityofparaboloids ] and [ lm : estimateconvexcombinationreflector ] , and proceeding as in the proof of claim , we get @xmath867 . hence @xmath868 , where we have used . now using the second inequality in lemma [ lm : lipschitzcontinuityofparaboloids ] , we get @xmath869 , which proves the lemma . we then obtain results similar to theorems [ thm : mainestimate ] and [ thm : holdercontinuity ] for parallel reflectors . we remark that from theorem [ thm : awforperpendicularvectorsreflectors ] , inequality implies for all @xmath10 , and therefore for parallel reflectors the argument in section [ subsec : localandglobalrefractors ] is not needed . [ thm : reflectorsmainestimate ] suppose @xmath17 is a parallel reflector , the target @xmath1 is regular from @xmath492 in the sense of definition [ def : reflectorlocaltargetbis ] . there exist a ball @xmath870 , and a constant @xmath700 depending on @xmath493 such that if @xmath871 , @xmath496 are such that @xmath872 then there exists @xmath873 such that we have @xmath555_{x_0^{\star}}:\lambda\in [ 1/4,3/4]\right\}\right)\cap\sigma\subseteq f_u(b_{\eta}(x_0)),\ ] ] with @xmath501 , @xmath556 and @xmath557 . [ thm : holdercontinuityreflectors ] suppose @xmath17 is a parallel reflector , the target @xmath1 is regular from @xmath492 in the sense of definition [ def : reflectorlocaltargetbis ] , and there exist constants @xmath695 and @xmath696 such that @xmath874 for all balls @xmath698 . suppose in addition that the local condition is satisfied at @xmath699 . then there exist a ball @xmath870 , and a constant @xmath700 depending on @xmath493 , such that if @xmath871 , @xmath496 are such that @xmath875 then we have @xmath703 with @xmath704 , where @xmath235 depends only on @xmath705 and @xmath706 in , and therefore from @xmath699 . [ rmk : planetargetreflector]we show here that when the target @xmath1 is contained in a non - vertical hyperplane , then condition holds . notice that from , it is sufficient to show that @xmath876 for all @xmath877 and @xmath744 ( @xmath352 ) . let @xmath878 , and @xmath879 . suppose @xmath880 for all @xmath131 . we calculate @xmath881 . from we have @xmath882 hence @xmath883 , and then @xmath884 . [ rmk : definitionofreflectorenclosing]similarly to section [ sec : alternativedefinitionofrefractor ] , we can define parallel reflector by taking paraboloids enclosing the solution . in other words , we can say @xmath853 $ ] is a parallel reflector from @xmath7 to @xmath1 if for each @xmath854 , there exists @xmath504 such that @xmath889 for all @xmath156 , where @xmath318 . the same regularity results hold with this definition of solution by changing accordingly the inequalities in the conditions on the target . indeed , with the notation in subsection [ subsect : hypothesesontargetrflector ] , condition is replaced by @xmath890 the analogue of condition is now @xmath891 and as before this is equivalent to . the concavity of @xmath17 now replaces lemma [ lm : estimateconvexcombinationreflector ] , the inequality in lemma [ lm : mainlemmareflector ] is replaced by @xmath892 and in the proof the @xmath792 is replaced by the @xmath791 . in contrast with remark [ rmk : planetargetreflector ] , in this case a target contained in a hyperplane can not satisfy .

we use the theory of lexicographic shellability to provide various examples in which the rank of the homology of a rees product of two partially ordered sets enumerates some set of combinatorial objects , perhaps according to some natural statistic on the set . many of these examples generalize a result of j. jonsson , which says that the rank of the unique nontrivial homology group of the rees product of a truncated boolean algebra of degree and a chain of length is the number of derangements in . linusson is a royal swedish academy of sciences research fellow supported by a grant from the knut and alice wallenberg foundation . ] supported in part by nsf grants dms 0604233 and 0902142 ] supported in part by nsf grants dms 0604562 and 0902323 ]

1203.0922

rees products of posets were defined and studied by a. bjrner and v. welker in @xcite . while the main results in @xcite provide combinatorial analogues of constructions in commutative algebra , it has turned out that rees products of certain posets are connected with permutation enumeration and permutation statistics . the first indication of this connection is provided by a conjecture in @xcite , which says that the reduced euler characteristic of the order complex of the rees product of the truncated boolean algebra @xmath3 and a chain of length @xmath1 is the number of derangements in the symmetric group @xmath4 . this conjecture was proved by j. jonsson in @xcite . as we shall describe below , generalizations of jonsson s result , along with similar results have been proved . our purposes in this paper are 1 . to give additional examples of rees products whose order complexes have reduced euler characteristics that enumerate certain classes of combinatorial objects , possibly according to some natural statistic , and 2 . to show how the theory of lexicographic shellability applies to certain rees products , in particular relating the homology of the order complex of the rees product of a lexicographically shellable poset @xmath5 with a poset whose hasse diagram is a rooted @xmath6-ary tree to the homology of the order complexes of some rank - selected subposets of @xmath5 . these two purposes are in fact intertwined . we prove all of our results on reduced euler characteristics of order complexes of rees products using lexicographic shellings . all posets studied in this paper are finite . we call a poset @xmath5 _ semipure _ if for each @xmath7 , the lower order ideal @xmath8 is pure , that is , any two maximal chains in @xmath9 have the same length . the _ rank _ @xmath10 of such an element @xmath11 is the length of a maximal chain in @xmath9 . given semipure posets @xmath12 with respective rank functions @xmath13 , the _ rees product _ @xmath14 is the poset whose underlying set is @xmath15 with order relation given by @xmath16 if and only if all of the conditions * @xmath17 , * @xmath18 , and * @xmath19 hold . in other words , @xmath20 covers @xmath21 in @xmath14 if and only if 1 . @xmath22 covers @xmath23 in @xmath5 , and 2 . either @xmath24 or @xmath25 covers @xmath26 in @xmath27 . in figure 1 , the rees product of the truncated boolean algebra @xmath28 and the chain @xmath29 is given . the element @xmath30 is written as @xmath31 with the set brackets and commas omitted . * figure 1 . * @xmath32 for any poset @xmath5 , the _ order complex _ @xmath33 is the abstract simplicial complex whose @xmath34-dimensional faces are chains ( totally ordered subsets ) of length @xmath34 from @xmath5 . a simplicial complex @xmath35 is _ cohen - macaulay _ if for each face @xmath36 ( including the empty face ) , the reduced ( integral , simplicial ) homology of the link @xmath37 is trivial in all dimensions except possibly @xmath38 . every cohen - macaulay complex is pure , that is , all maximal faces of a cohen - macaulay complex have the same dimension . a poset is said to be cohen - macaulay if its order complex is cohen - macaulay . we will say that a poset has a particular topological property if its order complex has that property . the ( reduced ) homology of @xmath5 is given by @xmath39 . for further information on cohen - macaulay posets , see the surveys given in @xcite , @xcite , @xcite . bjrner and welker ( * ? ? ? * corollary 2 ) prove that the rees product of any cohen - macaulay poset with any acyclic cohen - macaualy poset is cohen - macaulay . it is known that both @xmath40 and the chain @xmath41 of length @xmath0 are cohen - macaulay , and @xmath41 is acyclic . thus the result of jonsson mentioned above says that , with @xmath42 denoting the number of derangements in @xmath4 , @xmath43 generalizations of ( [ jon ] ) appear in the paper @xcite of shareshian and wachs . for a poset @xmath5 with unique minimum element @xmath44 , @xmath45 will denote @xmath46 . for a prime power @xmath47 and a positive integer @xmath0 , the poset of all subspaces of an @xmath0-dimensional vector space over the @xmath48-element field @xmath49 will be denoted by @xmath50 . also , @xmath51 will denote the set of all derangements in @xmath4 . it is shown in @xcite that @xmath52 where @xmath53 and @xmath54 are , respectively , the major index and the excedance number , introduced by macmahon in ( * ? ? ? 135,186 ; vol . 2 , p. viii ) , @xcite in the early part of the 20th century and extensively studied thereafter . a generalization of ( [ qan ] ) appears in @xcite . for positive integers @xmath55 , let @xmath56 be the poset whose hasse diagram is a complete @xmath6-ary tree of height @xmath0 with root at the bottom . to put it more formally , @xmath56 consists of all sequences of elements of @xmath57:=\{1,\ldots , t\}$ ] that have length at most @xmath0 , including the empty sequence . given two such sequences @xmath58 and and @xmath59 , we declare that @xmath60 if @xmath61 and @xmath62 for al @xmath63 $ ] . note that @xmath64 . it is shown in @xcite that if @xmath5 is cohen - macaulay of length @xmath0 then so is @xmath65 . equation ( [ gtreeintro2 ] ) below is proved in @xcite , and equation ( [ gtreeintro1 ] ) follows quickly from ( [ gtreeintro2 ] ) and ( * ? ? ? * corollary 2.4 ) . we have @xmath66 and @xmath67 one can also find in @xcite type bc analogues of the results mentioned above , where @xmath68 and @xmath50 are replaced , respectively , by the poset of faces of the @xmath0-crosspolytope and the poset of totally isotropic subspaces of a @xmath69-dimensional vector space over @xmath49 equipped with a nondegenerate alternating bilinear form , and @xmath51 is replaced by the set of elements of the weyl group of type bc that act as derangements on the set of vertices of the crossplytope . in @xcite , p. muldoon and m. readdy prove an analog of ( [ jon ] ) that involves the poset of faces of the @xmath0-cube . as was mentioned above , the results of bjrner and welker @xcite are concerned with cohen - macaulayness of rees products . it turns out that analogous results for lexicographic shellability can be obtained and utilized to obtain enumerative results . definitions and basic facts about lexicographic shellability are given in section [ lexshellsec ] . let @xmath5 be a pure poset of length @xmath0 . for @xmath70:= \{0,1,\dots , n\}$ ] , the _ rank - selected subposet _ @xmath71 is the subposet of @xmath5 consisting of all @xmath72 satisfying @xmath73 . if @xmath5 is lexicographically shellable then @xmath5 is cohen - macaulay , as is every rank - selected subposet of @xmath5 ( cf . thus , for all @xmath70 $ ] , the homology of @xmath71 is determined by the betti number @xmath74 let @xmath75 be the minimum element of @xmath56 . note that if @xmath5 has a unique minimum element @xmath76 then the poset @xmath65 has a minimum element @xmath77 , but no maximum element . write @xmath78 for the poset @xmath65 with a maximum element appended . in section [ lexshellsec ] we show that if @xmath5 is lexicographically shellable then so is @xmath78 for all @xmath6 . ( in fact , we prove a stronger result , see theorem [ elth ] . ) we call @xmath79 _ stable _ if there is no @xmath80 such that @xmath81 . for @xmath82 , we write @xmath83 for the set of all stable @xmath84 . we use the lexicographic shellings described in section [ lexshellsec ] to prove in section [ ascfreesec ] that , for pure , lexicographically shellable @xmath5 of length @xmath0 , @xmath85)}\beta(p_{[n-1 ] \setminus s})t^{|s|+1}(1+t)^{n-2|s|-1}.\ ] ] in fact , we prove in section [ ascfreesec ] several formulae similar to ( [ introbettieq ] ) involving either @xmath86 or @xmath87 . ( more general versions of these formulae in which the only requirement on p is that it be pure will appear in a forthcoming paper . ) in sections [ chainsec ] , [ qbnsec ] , [ psec ] and [ noncrosssec ] we apply our results from sections [ lexshellsec ] and [ ascfreesec ] to obtain enumerative results . the boolean algebra @xmath68 is the direct product of @xmath0 copies of the chain @xmath88 . in section [ chainsec ] we prove generalizations of the @xmath89 cases of ( [ gtreeintro1 ] ) and ( [ gtreeintro2 ] ) in which we replace @xmath68 with an arbitrary product of finite chains . let @xmath90 be a weak composition of @xmath0 into @xmath34 parts , that is a @xmath34-tuple of nonnegative integers whose sum is @xmath0 . the product poset @xmath91 is pure of length @xmath0 . it is well known that @xmath92 is lexicographically shellable . let @xmath93 be the multiset in which each @xmath63 $ ] appears with multiplicity @xmath94 . a _ multiset permutation _ of @xmath93 is a @xmath95 array @xmath96 such that * the multisets @xmath97\}$ ] and @xmath98\}$ ] are both equal to @xmath93 , * @xmath99 for all @xmath100 $ ] . let @xmath101 be a multiset permutation of @xmath93 . we say @xmath102 is _ multiset derangement _ of @xmath93 if * @xmath103 , for all @xmath104 $ ] . we say @xmath102 is a _ smirnov word _ on @xmath93 if * @xmath105 for all @xmath100 $ ] . an _ excedance _ of @xmath102 is any @xmath100 $ ] such that @xmath106 . a _ descent _ of @xmath102 is any @xmath100 $ ] such that @xmath107 . we write @xmath108 for the set of excedances of @xmath102 , @xmath109 for the set of descents of @xmath102 , and @xmath110 and @xmath111 , respectively , for @xmath112 and @xmath113 . our main results in section [ chainsec ] say that if @xmath114 and @xmath115 are , respectively , the sets of multiset derangements and smirnov words on @xmath93 then , for all @xmath116 , @xmath117 and @xmath118 when @xmath119 is the set @xmath120 $ ] , equation ( [ mddth ] ) is the @xmath89 case of ( [ gtreeintro1 ] ) . since @xmath121 and @xmath54 are equidistributed on the symmetric group @xmath122 , equation ( [ swth ] ) is the @xmath89 case of ( [ gtreeintro2 ] ) . in section [ qbnsec ] we revisit the rees products @xmath123 that were studied in @xcite . comparing ( [ gtreeintro2 ] ) with a formula for @xmath124 obtained using the techniques developed herein , we exhibit a permutation statistic called @xmath125 such that the pair @xmath126 is equidistributed on @xmath4 with the pair @xmath127 . in section [ psec ] we aim for @xmath128-analogues of the results in section [ chainsec ] . given a weak composition @xmath90 of @xmath0 , a natural choice for a @xmath128-analogue to the poset @xmath92 is the lattice @xmath129 of subgroups of the abelian @xmath128-group @xmath130 . we examine @xmath131 and @xmath132 . here our results are less than optimal . we show that there exist statistics @xmath133 such that @xmath134 and @xmath135 however , we lack natural combinatorial interpretations for @xmath136 and @xmath137 . in section [ noncrosssec ] we consider the lattice @xmath138 of nonncrossing partitions of @xmath120 $ ] , which is known to be lexicographically shellable . we show that @xmath139^{n - k}}t^{{{\rm des}}(w)+k},\ ] ] and @xmath140 @xmath141^{n - k - r } } t^{{{\rm des}}(w)+k } .\ ] ] equation ( [ nc1 ] ) reduces to a particularly nice enumerative formula when @xmath6 is set equal to @xmath142 , namely @xmath143 proofs of the various identities stated above involve symmetric function formulae for generating functions for words with no double descent , words with no double ascent , smirnov words , and multiset derangements , keeping track of descents , ascents , descents and excedances , respectively . the formula involving smirnov words follows from work in @xcite , while the remaining formulae are due to ira gessel . we give all of these formulae in section [ symmsec ] . after reviewing some basic facts from the theory of lexicographic shellability ( cf . @xcite ) , we will present our main results on lexicographic shellability of rees products . let @xmath5 be a _ bounded _ poset , i.e. , a poset with a unique minimum element and a unique maximum element , and let @xmath144 be the set of pairs @xmath145 such that @xmath146 covers @xmath11 in @xmath5 . let @xmath147 be another poset and let @xmath148 be the set of all finite sequences of elements of @xmath147 . the given partial ordering of @xmath147 induces a lexicographic ordering @xmath149 on @xmath148 , which is also a partial order . an _ edge labeling _ of @xmath5 by @xmath147 is a function @xmath150 . given such a function @xmath151 and a saturated chain @xmath152 from @xmath5 , we write @xmath153 for @xmath154 . an _ ascent _ in @xmath155 is any @xmath156 $ ] satisfying @xmath157 . we say @xmath151 is _ weakly increasing _ on @xmath155 if each @xmath156 $ ] is an ascent in @xmath155 . the edge labeling @xmath151 is an _ el - labeling _ of @xmath5 if whenever @xmath158 in @xmath5 there is a unique maximal chain @xmath155 in the interval @xmath159 $ ] on which @xmath151 is weakly increasing and for all other maximal chains @xmath160 in @xmath159 $ ] we have @xmath161 . a bounded poset that admits an el - labeling is said to be el - shellable . the notion of el - shellability for _ pure _ posets was introduced by bjrner in @xcite . a more general concept called cl - shellability , introduced by bjrner and wachs in @xcite , also associates label sequences with maximal chains of a poset . we will not define cl - labelings here . both notions were subsequently extended to all bounded posets by bjrner and wachs in @xcite . all of our results in this section and the next section hold for cl - labelings as well as el - labelings . for the sake of simplicity we state and prove them only for el - labelings . the proofs for cl - labelings are virtually the same as those for el - labelings . given an el - labeling @xmath151 on @xmath5 , we call a maximal chain @xmath155 from @xmath5 _ ascent free _ if its label sequence contains no ascent . the _ descent set _ of a maximal chain @xmath162 is defined to be the set @xmath163 : \lambda(x_{i-1},x_i ) \not\le \lambda(x_{i},x_{i+1 } ) \}$ ] . thus a maximal chain is ascent free if and only if its descent set is @xmath164 $ ] . one of the main results in the theory of lexicographic shellability is the following result . let @xmath151 be an el - labeling of a bounded poset @xmath5 with minimum @xmath165 and maximum @xmath166 . then @xmath167 is homotopy equivalent to a wedge of spheres , where for each @xmath168 the number of spheres of dimension @xmath169 is the number of ascent free maximal chains of length @xmath34 . [ elth ] we will also need the following basic result . given a pure poset @xmath5 of length @xmath0 and a set @xmath70 $ ] , recall that the rank selected subposet is defined by @xmath170 [ rankselth ] let @xmath151 be an el - labeling of a bounded pure poset @xmath5 of length @xmath0 . for @xmath171 $ ] , let @xmath172 be the number of maximal chains in @xmath5 having descent set @xmath173 with respect to @xmath151 . then @xmath174 has the homotopy type of a wedge of @xmath172 spheres of dimension @xmath175 . given a poset @xmath5 , by @xmath176 we mean the poset @xmath5 with a new minimum element @xmath165 and a new maximum element @xmath166 attached even if @xmath5 already has such elements . given a poset @xmath5 with a minimum element @xmath165 , we say that an edge labeling @xmath177 is a semi - el - labeling if @xmath178 $ ] is an el - labeling for each maximal element @xmath179 of @xmath5 . note if @xmath5 is bounded then @xmath151 is a semi - el - labeling if and only if it is an el - labeling . recall that we defined @xmath56 to be the poset whose hasse diagram is the complete @xmath6-ary tree of height @xmath0 with the root at the bottom . the edge labeling in which all the edges in @xmath180 are labeled with @xmath142 is clearly a semi - el - labeling of @xmath5 . [ elth ] let @xmath181 and @xmath182 be semipure posets of the same length . assume also that @xmath182 has a minimum element @xmath183 . let @xmath184 be an el - labeling of @xmath185 and let @xmath186 be a semi - el - labeling of @xmath182 . let @xmath187 denote the minimum element of @xmath185 and let @xmath188 denote the maximum element . let @xmath189 denote the minimum element of @xmath190 and let @xmath166 denote the maximum element . define the edge labeling @xmath191 by @xmath192 for @xmath193 , and @xmath194 then @xmath151 is an el - labeling of @xmath190 . [ elrees ] * case 1 : * @xmath195 in @xmath196 . then @xmath197 in @xmath185 and @xmath198 in @xmath182 . it follows that there is a unique maximal chain @xmath199 in @xmath159 $ ] on which @xmath200 is weakly increasing and a unique maximal chain @xmath201 in @xmath202 $ ] on which @xmath203 is weakly increasing . let @xmath204 the labeling @xmath151 is weakly increasing on the maximal chain @xmath205 of the interval @xmath206 $ ] . to establish uniqueness of the maximal chain with weakly increasing labels , suppose that @xmath151 is weakly increasing on the maximal chain @xmath207 in the interval @xmath208 . then @xmath200 is weakly increasing on the chain @xmath209 , which implies that @xmath210 for all @xmath211 . moreover , if @xmath212 then we must have @xmath213 for @xmath214 and @xmath215 in @xmath216 for @xmath217 . if @xmath218 then @xmath219 , and if @xmath220 then @xmath221 is covered by @xmath222 in @xmath182 and @xmath223 . it follows that if @xmath224 then @xmath225 and @xmath226 is the unique maximal chain of the interval @xmath202 $ ] in @xmath227 for which @xmath203 is weakly increasing . @xmath228 for all @xmath229 . next we show that the maximal chain @xmath155 of @xmath208 has a label sequence that lexicographically precedes the label sequences of all maximal chains of @xmath208 . let @xmath207 be another maximal chain in @xmath208 . assume that @xmath230 for @xmath231 but @xmath232 . we need to show that @xmath233 in @xmath234 first we handle the case in which @xmath235 . in this case we have @xmath236 , which implies @xmath237 if @xmath238 then @xmath239 , which implies that @xmath240 covers @xmath241 in @xmath182 . since @xmath242 and @xmath243 ( [ lexfirsteq ] ) holds . now assume @xmath244 . we have @xmath245 in @xmath246 . indeed , it is a basic property of el - labelings that if @xmath5 is a poset with el - labeling @xmath151 then for each interval @xmath159 $ ] , if @xmath247 covers @xmath11 in the unique maximal chain of @xmath159 $ ] with weakly increasing labels and @xmath248 is an atom of @xmath159 $ ] other than @xmath247 , then @xmath249 ( cf . * proposition 2.5 ) , ( * ? ? ? * lemma 5.3 ) ) . since @xmath250 , for some @xmath251 , the desired inequality ( [ lexfirsteq ] ) follows from ( [ lameq ] ) and ( [ lam1eq ] ) . now assume @xmath252 . in this case we have @xmath253 and @xmath254 cover @xmath255 in @xmath185 , and @xmath256 and @xmath240 cover @xmath257 in @xmath182 . it now follows from the basic property of el - labelings mentioned in the previous paragraph that either ( [ lam1eq ] ) and @xmath258 or @xmath259 and @xmath260 hold , which yields the desired conclusion ( [ lexfirsteq ] ) . * case 2 : * @xmath261 in @xmath196 . then @xmath262 in @xmath185 and there is a unique maximal chain @xmath263 in @xmath264 $ ] on which @xmath200 is weakly increasing . the labeling @xmath265 is weakly increasing on the maximal chain @xmath266 of the interval @xmath267 $ ] . to establish uniqueness of the maximal chain with weakly increasing labels , note that the top label of every maximal chain of @xmath267 $ ] is of the form @xmath268 , where @xmath269 is a maximal element of @xmath181 . hence if @xmath270 is a maximal chain of @xmath267 $ ] with weakly increasing labels then @xmath271 where @xmath272 in @xmath246 . it follows that @xmath273 for all @xmath274 and @xmath275 is the unique maximal chain of @xmath276 $ ] with weakly increasing labels . hence @xmath277 and @xmath278 for all @xmath279 . now let @xmath270 be a maximal chain in @xmath267 $ ] that is different from @xmath155 . we show that the label sequence of @xmath155 is lexicographically less than that of @xmath160 . assume that @xmath280 for @xmath281 but @xmath282 . we need to show that @xmath283 in @xmath234 . if @xmath284 then @xmath285 covers @xmath286 in @xmath182 . we have @xmath287 if @xmath288 then by the basic property of el - labelings mentioned above , @xmath289 . it follows that @xmath290 let @xmath5 be a semipure poset of length @xmath0 . let @xmath292 be an el - labeling of @xmath176 and let @xmath293 be the semi - el - labeling of @xmath56 in which each edge has label 1 . in this section we count the ascent free maximal chains of @xmath294 under the el - labeling @xmath295 described in theorem [ elth ] . for @xmath296 , let @xmath297 be the set of sequences @xmath298 such that @xmath299 . given any maximal chain @xmath300 of @xmath294 , we have that @xmath301 is a maximal chain of @xmath5 and @xmath302 , for some @xmath303 . conversely , given any maximal chain @xmath304 of @xmath5 and any @xmath305 , there is a maximal chain @xmath300 of @xmath294 such that @xmath306 for all @xmath307 $ ] . let @xmath308 $ ] be the set of all such maximum chains of @xmath294 . the following propositions clearly hold . [ propnochains ] the sets @xmath308 $ ] , where @xmath309 is a maximal chain of @xmath5 of length @xmath179 and @xmath310 for @xmath296 , partition the set of maximum chains of @xmath294 . moreover if @xmath310 then @xmath311| = t^j$ ] . [ propascfree ] let @xmath312 be a maximal chain of @xmath5 and let @xmath313 . then for each maximal chain @xmath314 $ ] we have @xmath315 consequently , @xmath160 is ascent free if and only if @xmath316 and @xmath317 , \ , \,\,\ , \lambda_p(x_{i-1},x_i ) \le \lambda_p(x_{i},x_{i+1 } ) \implies d_i = 1 \mbox { and } d_{i+1 } = 0\ ] ] holds . here we have set @xmath318 and @xmath319 . given a word @xmath320 over a partially ordered alphabet @xmath321 , we say @xmath322 $ ] is an ascent of @xmath102 if @xmath323 and that @xmath324 $ ] is a double ascent if @xmath325 . let @xmath326 denote the number of ascents of @xmath102 and @xmath327 we are now ready to count the ascent free maximal chains . we begin with the case in which the semipure poset @xmath5 has a unique maximum element . in this case @xmath5 must necessarily be a pure poset of length @xmath0 . all maximal chains of @xmath176 have length @xmath328 and must have an ascent at @xmath329 under the el - labeling . we leave it to the reader to observe that propositions [ propnochains ] and [ propascfree ] imply the following result . [ maxth ] if @xmath5 has a unique maximum element then the number of ascent free maximal chains of @xmath294 of length @xmath328 under the el - labeling of theorem [ elth ] is given by @xmath330 where @xmath331 is the number of maximal chains of @xmath332 with label sequence @xmath102 . in the general case in which it is not assumed that @xmath5 has a unique maximum element , we have the following result which also is a consequence of propositions [ propnochains ] and [ propascfree ] . let @xmath333 . then the number ascent free maximal chains of @xmath294 of length @xmath334 under the el - labeling of theorem [ elth ] is given by @xmath335 @xmath336 where @xmath337 is the number of maximal chains of @xmath176 of length @xmath334 with label sequence @xmath102 . note that if @xmath5 has a unique minimum element then @xmath291 has unique minimum element , which implies that @xmath291 is contractible . hence the number of ascent free maximal chains of @xmath294 has to be 0 . this is corroborated by @xmath338 if @xmath339 , which follows from the fact that there is only one maximal chain in each interval @xmath340 $ ] of @xmath176 , where @xmath247 is an atom of @xmath5 . therefore in the case that @xmath5 has a unique minimum element , it is more interesting to consider the number of ascent free chains of the interval @xmath341 of @xmath342 . the following results also follow from propositions [ propnochains ] and [ propascfree ] . [ minmaxth ] if @xmath5 has both a unique minimum element and a unique maximum element then the number of ascent free maximal chains of @xmath343 under the el - labeling of theorem [ elth ] is given by @xmath344 where @xmath331 is the number of maximal chains of @xmath345 with label sequence @xmath102 . let @xmath333 . if @xmath5 has a unique minimum element then the number of ascent free maximal chains of @xmath343 of length @xmath346 under the el - labeling of theorem [ elth ] is given by @xmath347 @xmath348 where @xmath331 is the number of maximal chains of @xmath349 of length @xmath346 with label sequence @xmath102 . for pure @xmath5 we can restate the above results by applying theorem [ rankselth ] . we need to recall the following terminology and notation . a set of integers is _ stable _ if it contains no two consecutive integers . for @xmath350 , the set of all stable subsets of @xmath351 is denoted by @xmath83 . for @xmath352 , let @xmath353 : = \{i , i+1,\dots , j\}$ ] and @xmath354:=[1,j]$ ] . if @xmath5 is a poset of length @xmath0 let @xmath355 if @xmath5 has a unique minimum element @xmath165 let @xmath356 [ mobcor1 ] let @xmath5 be a pure poset of length @xmath0 such that @xmath176 is el - shellable . assume that @xmath5 has a unique maximum element . then @xmath357 ) } \beta(p_{[0,n-1]\setminus s } ) t^{|s|+1 } ( t+1)^{n-2|s|-1 } .\ ] ] if @xmath5 also has a unique minimum element then @xmath358 ) } \beta(p_{[n-1]\setminus s } ) t^{|s|+1 } ( t+1)^{n-2|s|-1 } .\ ] ] [ mobcor2 ] let @xmath5 be a pure poset of length @xmath0 such that @xmath176 is el - shellable . then @xmath359 @xmath360 ) } \beta ( p_{[0,n]\setminus s } ) t^{|s| } ( t+1)^{n-2|s| } + \sum_{s \in \psx([n-2 ] ) } \beta ( p_{[0,n-1]\setminus s } ) t^{|s|+1 } ( t+1)^{n-2|s|-1 } .\ ] ] if @xmath5 has a unique minimum element then @xmath361 @xmath360 ) } \beta ( p_{[n]\setminus s } ) t^{|s| } ( t+1)^{n-2|s| } + \sum_{s \in \psx([n-2 ] ) } \beta ( p_{[n-1 ] \setminus s } ) t^{|s|+1 } ( t+1)^{n-2|s|-1 } .\ ] ] let @xmath362 denote the complete homogenous symmetric function of degree @xmath0 in indeterminants @xmath363 and @xmath364 denote the elementary symmetric function of degree @xmath0 in indeterminants @xmath365 . that is @xmath366 also let @xmath367_t : = 1+t+ \dots + t^{n-1}.\ ] ] in this section we will discuss various combinatorial interpretations of variations of the symmetric function @xmath368_t h_i z^i } } \ , , \ ] ] which play a key role in the proofs of the results in the subsequent sections . these and other interpretations are discussed in ( * ? ? ? * section 7 ) . let @xmath369 . recall that we say @xmath322 $ ] is an ascent of @xmath102 if @xmath323 and that @xmath324 $ ] is a double ascent if @xmath325 . recall that @xmath326 denotes the number of ascents of @xmath102 and @xmath370 similarly , @xmath322 $ ] is a descent of @xmath102 if @xmath371 and @xmath324 $ ] is a double descent if @xmath372 . let @xmath110 denote the number of descents of @xmath102 and @xmath373 we write @xmath374 for @xmath375 . we begin by presenting the following interpretations due to gessel , see theorem 7.3 of @xcite . ( gessel s original proofs will appear in @xcite . ) @xmath376_t e_i z^i } } \,\ , , \ ] ] @xmath377_t e_i z^i } } \,\ , , \ ] ] @xmath378_t h_i z^i } } \,\ , , \ ] ] @xmath379_t h_i z^i } } \,\ , .\ ] ] next we present an interpretation due to shareshian and wachs @xcite . a barred word of length @xmath0 over alphabet @xmath321 is an element of @xmath380 . we visualize barred words as words over @xmath321 in which some of the letters are barred ; @xmath381 is a barred letter and @xmath382 is an unbarred letter . if @xmath102 is a barred word then @xmath383 denotes the word @xmath102 with the bars removed similarly , let @xmath384 . if @xmath385 is a barred or unbarred letter , we refer to @xmath386 as the _ absolute value _ of @xmath385 . for a barred word @xmath102 , let @xmath387 denote the number of barred letters of @xmath102 . let @xmath388 be the set of barred words @xmath320 of length @xmath0 over @xmath389 satisfying 1 . @xmath390 is unbarred 2 . for all @xmath391 $ ] , if @xmath392 then @xmath393 is unbarred 3 . for all @xmath391 $ ] , if @xmath394 then @xmath395 is barred . elements of @xmath396 are called _ banners _ in ( * ? ? ? * section 3 ) , where it is shown that @xmath397_t h_i z^i } } \,\ , .\ ] ] we will also need an interpretation due to askey and ismail @xcite and one due to stanley ( personal communication , see theorem 7.2 of @xcite ) . given a finite multiset @xmath398 over @xmath389 , let @xmath399 denote the set of multiset permutations of @xmath398 . recall that we can write @xmath400 in two - line notation as a @xmath401 array @xmath402 whose top row is a weakly increasing arrangement of the multiset @xmath398 and whose bottom row is an arbitrary arrangement of @xmath398 . by supressing the top row , we write @xmath102 in one - line notation as the word , @xmath403 , where @xmath404 . if @xmath400 we say that @xmath102 has length @xmath405 . an excedance of a multiset permutation @xmath406 , written in two - line notation , is a column @xmath303 such that @xmath407 . let @xmath111 be the number of excedances of @xmath102 . recall that @xmath408 is a multiset derangement if each of the columns of @xmath102 have distinct entries , i.e , . @xmath409 for all @xmath410 . for example , if @xmath411 then @xmath102 is a multiset derangement in @xmath412 and @xmath413 . now let @xmath414 be the set of all multiset derangements of length @xmath0 . askey and ismail @xcite ( see also @xcite ) proved the following @xmath6-analog of macmahon s ( * ? ? ? iii ) result on multiset derangements @xmath415_t e_i z^i}.\ ] ] recall from section [ secintro ] that a multiset permutation @xmath416 is called a smirnov word if it has no adjacent repeats , i.e. @xmath417 for all @xmath418 . let @xmath419 be the set of all smirnov words of length @xmath0 . stanley ( see theorem 7.2 and ( 7.7 ) of @xcite ) observed that the following @xmath6-analog of a result of carlitz , scoville and vaughan @xcite @xmath420_t e_i z^i}}\ ] ] is equivalent to ( [ bannereq ] ) by p - partition reciprocity ( ( * ? ? ? * section 4.5 ) ) . in this section we generalize the @xmath89 case of ( [ gtreeintro1 ] ) and ( [ gtreeintro2 ] ) by utilizing the results of the previous section . given a weak composition @xmath421 of @xmath0 , let @xmath92 denote the product of chains @xmath422 . recall that @xmath93 denotes the mulitset @xmath423 . given a multiset @xmath398 , let @xmath424 be the set of multiset derangements of the multiset @xmath398 and let @xmath425 be the set of smirnov words that are multiset permutations of @xmath398 . [ multirees ] let @xmath426 be a composition of @xmath0 . then @xmath427 and @xmath428 have the homotopy type of a wedge of @xmath429-spheres . the numbers of spheres in these wedges are , respectively , @xmath430 and @xmath431 we begin by applying theorem [ maxth ] to @xmath432 , which has length @xmath1 . let @xmath433 . there is a well - known el - labeling of @xmath92 in which the edge @xmath434 is labeled by @xmath229 . here @xmath435 is the totally ordered set @xmath436 . hence @xmath437 has an el - labeling as described in theorem [ elth ] . the label sequence of each maximal chain of @xmath438 is a permutation of the multiset @xmath93 . moreover each mulitset permutation occurs exactly once as the label sequence of a maximal chain . so @xmath439 if @xmath440 and @xmath338 if @xmath441^n - \mathfrak s_{m(\mu)}$ ] . it follows from theorem [ maxth ] that the number of ascent free maximal chains of @xmath442 under the given labeling is @xmath443 similarly by theorems [ elth ] and [ minmaxth ] with @xmath444 , the poset @xmath445 has an el - labeling for which the number of ascent free maximal chains is @xmath446 hence by theorem [ elth ] , the posets @xmath427 and @xmath428 have the homotopy type of a wedge of @xmath429-spheres and the top betti numbers are given by @xmath447 and @xmath448 by combining ( [ macderang ] ) and ( [ ges1 ] ) we obtain @xmath449 and by combining ( [ smirnov ] ) and ( [ ges2 ] ) we obtain @xmath450 for all multisets @xmath398 on @xmath389 of size @xmath451 . equations ( [ deraneq ] ) and ( [ smireq ] ) now follow from ( [ gesmac ] ) and ( [ gesstan ] ) , respectively . when @xmath452 , equation ( [ gesstan ] ) reduces to a result of foata and schtzenberger @xcite , which is used to show that the eulerian polynomials are palindromic and unimodal . we see from ( [ gesmac ] ) and ( [ gesstan ] ) , respectively , that the polynomials @xmath453 and @xmath454 are palindromic and unimodal for all multisets @xmath398 . the lattice @xmath50 of subspaces of an @xmath0-dimensional vector space over the finite field @xmath49 is bounded and pure of length @xmath0 . it is well known that @xmath50 is el - shellable ( see @xcite ) . using ( 3.3 ) to compute @xmath455 and equating the resulting formula with the formula given in ( [ gtreeintro2 ] ) , we obtain a new mahonian permutation statistic , which we call @xmath456 , and we show that the pairs @xmath126 and @xmath127 are equidistributed on @xmath2 . let @xmath457 . recall that an _ inversion _ of @xmath458 is a pair @xmath459 such that @xmath460 and @xmath461 . an _ admissible inversion _ of @xmath458 is an inversion @xmath462 that satisfies either * @xmath463 and @xmath464 or * there is some @xmath34 such that @xmath465 and @xmath466 . we write @xmath467 for the number of inversions of @xmath458 and @xmath468 for the number of admissible inversions of @xmath458 . for example , if @xmath469 then there are 11 inversions , but only @xmath470 and @xmath471 are admissible . so @xmath472 and @xmath473 . now let @xmath474 it turns out that @xmath456 is equidistributed with the mahonian permutation statistics @xmath475 and @xmath53 on @xmath2 . we give a short combinatorial proof of this in proposition [ mahonaid ] below . first we prove the following more general joint distribution result . [ equidistth ] for all @xmath476 , @xmath477 it is well - known ( see ( * ? ? ? * theorem 3.12.3 ) ) that for all @xmath171 $ ] , @xmath478 hence by ( [ minmaxeq2 ] ) we have @xmath479 ) } \sum_{\scriptsize\begin{array}{c } \sigma \in { { \mathfrak s}}_n \\ { { \rm des}}(\sigma ) = [ 1,n-1]\setminus s \end{array}}\hspace{-.2 in } q^{{{\rm inv}}(\sigma)}t^{|s|+1 } ( t+1)^{n-1 - 2|s| } \\ \label{qbetaeq } & = & \sum_{\scriptsize\begin{array}{c } \sigma \in { { \mathfrak s}}_n \cap { { \rm nda}}_n \\ \sigma_{n-1 } > \sigma_n \end{array } } q^{{{\rm inv}}(\sigma)}t^{{{\rm asc}}(\sigma)+1 } ( 1+t ) ^{n-1 - 2{{\rm asc}}(\sigma)}. \end{aligned}\ ] ] we will rewrite the expression ( [ qbetaeq ] ) as the enumerator of barred permutations . given a set @xmath351 of size @xmath0 , a barred permutation of @xmath351 is a word @xmath480 with @xmath0 distinct letters in @xmath351 , in which some of the letters are barred . let @xmath481 denote the letter @xmath395 with the bar removed if there is one and let @xmath482 , where @xmath483 is the set of ordinary permutations of @xmath351 . let @xmath387 denote the number of bars of @xmath102 . let @xmath484 be the set of barred permutations @xmath102 of @xmath351 satisfying * @xmath390 is barred * if @xmath485 $ ] and @xmath486 then @xmath395 is barred and @xmath487 is not barred . it is not hard to see that the expression ( [ qbetaeq ] ) equals @xmath488 } } q^{{{\rm inv}}(|w| ) } t^{{{\rm bar}}(w ) } , \ ] ] which by lemma [ bijlem ] below equals @xmath489 hence @xmath490 the result now follows from ( [ gtreeintro2 ] ) . given barred permutations @xmath491 and @xmath492 , where @xmath321 and @xmath493 are disjoint sets , let @xmath494 denote the barred permutation in @xmath495 obtained by concatenating the words @xmath385 and @xmath496 . also let @xmath497 denote the empty word . we define a map @xmath498 recursively as follows . if @xmath102 is in the domain of @xmath499 and @xmath179 is the maximum letter of @xmath383 then @xmath500 [ bijlem ] the map @xmath499 is a well - defined bijection which satisfies 1 . @xmath501 , 2 . @xmath502 3 . @xmath503 for all finite nonempty subsets @xmath351 of @xmath389 and all @xmath504 . by ( b ) of the definition of @xmath484 , if letter @xmath505 is barred in the word @xmath504 then it is the first letter of @xmath102 . by ( a ) , if @xmath179 is unbarred it can not be the last letter . hence the three cases of the definition of @xmath499 cover all possibilities . it is also clear from the definition of @xmath484 that if @xmath506 and @xmath507 then @xmath491 and @xmath508 for some subset @xmath509 . hence by induction on @xmath510 we have that @xmath499 is a well - defined map that takes elements of @xmath484 to @xmath511 . to show that @xmath499 is a bijection satisfying ( 1 ) we construct its inverse . define @xmath512 recursively by @xmath513 where @xmath179 is the maximum letter of @xmath458 . let @xmath514 . one can see that conditions ( a ) and ( b ) of the definition of @xmath484 hold for @xmath515 whenever they hold for @xmath516 and @xmath517 . hence by induction on @xmath510 , @xmath518 is a well defined map . one can easily also show by induction that @xmath499 and @xmath518 are inverses of each other . we also prove ( 2 ) by induction on @xmath510 , with the base case @xmath519 being trivial . we do the third case of the definition of @xmath499 and leave the second to the reader . let @xmath520 with @xmath507 . if @xmath521 then @xmath522 by the induction hypothesis . since @xmath179 is the largest element of @xmath351 and is not the last letter of @xmath523 , we have @xmath524 hence ( 2 ) holds in this case . our proof of ( 3 ) proceeds by induction on @xmath525 , the case @xmath526 being trivial . if @xmath527 then @xmath528 indeed , the first two equalities follow immediately from the definitions and the third follows from our inductive hypothesis . next , say @xmath529 with @xmath530 and @xmath492 , where @xmath531 . set @xmath532 it follows quickly from the inductive hypothesis and the definitions that @xmath533 now @xmath534 and a straightforward calculation shows that @xmath535 hence @xmath536 as desired . we pose the question of whether there is a nice direct bijective proof of theorem [ equidistth ] . our proof of theorem [ equidistth ] relies on ( [ gtreeintro2 ] ) , whose proof , in turn , relies on a @xmath48-analog of euler s formula for the eulerian polynomials derived by shareshian and wachs in @xcite . a considerable amount of work in symmetric function theory and bijective combinatorics went into the proof of this @xmath48-analog of euler s formula . since the steps in deriving theorem [ equidistth ] from the @xmath48-analog of euler s formula are reversable , a nice direct combinatorial proof of theorem [ equidistth ] would provide an interesting alternative proof of the @xmath48-analog of euler s formula . here we give a simple combinatorial proof that @xmath456 is mahonian . [ mahonaid ] let @xmath537 . then @xmath538 satisfies the following recurrence for all @xmath539 , @xmath540_q q^j f_{j-1}(q ) f_{n - j } ( q ) .\ ] ] consequently @xmath541_q!.$ ] the terms on the right side of the recurrence @xmath48-count permutations according to the position of @xmath0 in the permutation . that is for each @xmath303 , @xmath542_q q^j f_{j-1}(q ) f_{n - j } ( q ) & \mbox { if } j=2,\dots , n-1\\ & \\ f_{n-1}(q ) & \mbox { if } j = 1 \\ q f_{n-1}(q ) & \mbox { if } j = n.\end{cases}\ ] ] it is easy to see that @xmath120_q!$ ] also satisfies the same recurrence relation . a more natural mahonian permutation statistic whose joint distribution with @xmath121 is the same as that of @xmath456 is discussed in @xcite . this statistic is a member of a family of mahonian statistics introduced by rawlings @xcite . given a prime @xmath128 and a weak composition @xmath421 of @xmath0 , let @xmath129 denote the lattice of subgroups of the abelian @xmath128-group @xmath543 . the poset @xmath129 is a natural @xmath128-analog of @xmath92 . it is pure and bounded of length @xmath0 . moreover , it provides the following @xmath128-analog of theorem [ multirees ] . [ multireesp ] let @xmath426 be a weak composition of @xmath0 and let @xmath128 be a prime . then @xmath544 and @xmath545 have the homotopy type of a wedge of @xmath429-spheres . the numbers of spheres in these wedges are , respectively , @xmath546 and @xmath547 where @xmath548 are statistics on words over @xmath389 . it is well - known that @xmath129 is el - shellable . hence by theorem [ elth ] , @xmath549 and @xmath550 are el - shellable . it is also known ( see ( * ? ? ? * ( 1.30 ) ) ) that for all @xmath171 $ ] , @xmath551 where @xmath552 is a statistic on words introduced by lascoux and schtzenberger for the purpose of showing that the kostka polynomials have nonnegative integer coefficients . ( we will not need the precise definition of cocharge here . ) now by ( [ minmaxeq1 ] ) we have @xmath553 ) } \sum_{\scriptsize\begin{array}{c } w \in { { \mathfrak s}}_{m(\mu ) } \\ { { \rm des}}(w ) = [ 1,n-1]\setminus s \end{array}}\hspace{-.2 in } p^{{{\rm cocharge}}(w)}t^{|s|+1 } ( t+1)^{n-2 - 2|s| } .\ ] ] hence @xmath554 where @xmath555 $ ] . by ( [ gesmac ] ) , @xmath556 since @xmath557 is the sum of the coefficients of @xmath558 , we can assign a nonnegative integer @xmath559 to each word @xmath102 in @xmath560 so that @xmath561 by plugging this into ( [ ppderangeeq ] ) , we obtain the desired result ( [ pderaneq ] ) . the proof of ( [ psmireq ] ) follows along the lines of that of ( [ pderaneq ] ) with ( [ minmaxeq2 ] ) and ( [ gesstan ] ) used instead of ( [ minmaxeq1 ] ) and ( [ gesmac ] ) . * problem : * it would be interesting to find nice combinatorial descriptions of the coefficients of the polynomials @xmath562 and @xmath563 . that is , find natural statistics @xmath136 and @xmath137 for which ( [ pderaneq ] ) and ( [ psmireq ] ) hold . when @xmath564 , we see from ( [ gtreeintro1 ] ) that @xmath559 can be defined to be @xmath565 and from ( [ aibeta ] ) that @xmath566 can be defined to be @xmath567 . a set partition @xmath568 is said to be _ noncrossing _ if for all @xmath569 , whenever @xmath570 are in a block @xmath493 of @xmath568 and @xmath571 are in a block @xmath572 of @xmath568 then @xmath573 . let @xmath138 be the poset of noncrossing partitions of @xmath120 $ ] ordered by reverse refinement . this poset , known as the _ noncrossing partition lattice _ , was first introduced by kreweras @xcite , who showed that it is a pure lattice with mbius invariant equal to the signed catalan number @xmath574 . bjrner and edelman ( cf . @xcite ) gave the first el labeling of @xmath138 and later stanley @xcite gave a different el - labeling in which the maximum chains are labeled with parking functions . a word @xmath575 is said to be a _ parking function _ of length @xmath0 if its weakly increasing rearrangement @xmath576 satisfies @xmath577 for all @xmath578 $ ] . let @xmath579 be the set of parking functions of length @xmath0 . recall that for @xmath580 , @xmath581 : w_i > w_{i+1}\ } \mbox { and } { { \rm des}}(w ) : = |{{\rm des}}(w)|.\ ] ] stanley uses his el - labeling to prove that for all @xmath171 $ ] , @xmath582\setminus s})= |\ { w \in { { \rm pf}}_n : { { \rm des}}(w ) = s\}|.\ ] ] [ noncross ] for all @xmath583 , the posets @xmath584 and @xmath585 have the homotopy type of a wedge of @xmath429-spheres . the numbers of spheres in these wedges are , respectively , @xmath586^{n - k } } t^{{{\rm des}}(w)+k } .\ ] ] and @xmath587 @xmath141^{n - k - r } } t^{{{\rm des}}(w)+k } .\ ] ] [ noncrosscor ] for all @xmath589 , the posets @xmath590 and @xmath591 have the homotopy type of a wedge of @xmath429-spheres . the numbers of spheres in these wedges are , respectively , @xmath592 and @xmath593 @xmath595 ) } |\ { w \in { { \rm pf}}_n : { { \rm des}}(w ) = s\}| t^{|s|+1 } ( t+1)^{n-2|s|-1 } \\ \nonumber & = & \sum_{\scriptsize\begin{array}{c}w \in { { \rm pf}}_n \cap { { \rm ndd}}_n\\ w_{n-1 } \le w_n\end{array}}t^{{{\rm des}}(w)+1}(t+1)^{n-2{{\rm des}}(w ) -1 } .\end{aligned}\ ] ] let @xmath596 be the set of all weak compositions of @xmath0 into @xmath34 parts . it is straightforward to show that @xmath597 if and only if @xmath598 for some @xmath599 such that @xmath600 for all @xmath601 . we will call a weak composition of @xmath0 into @xmath0 parts that satisfies this condition a _ parking composition _ of @xmath0 , and let @xmath602 be the set of all parking compositions of @xmath0 . it now follows from ( [ noncrossbetti1 ] ) that @xmath603 note that every parking composition @xmath426 of @xmath0 can be viewed as an element of @xmath604 by adjoining a @xmath605 to the end of @xmath426 . for @xmath606 , we say that @xmath426 and @xmath607 are cyclically equivalent if @xmath608 can be obtained by cyclically rotating the parts of @xmath426 . since all elements of @xmath604 are primitive words , i.e. , they are not equal to a power of a shorter word , the equivalence classes of @xmath604 under cyclic equivalence all have size equal to @xmath329 . moreover , each equivalence class has exactly one parking composition @xmath426 , i.e. @xmath609 where @xmath610 . given a weak composition @xmath426 of @xmath0 , let @xmath611 and let @xmath612 for @xmath613 . it follows from ( [ ges4 ] ) that @xmath614 , is a polynomial in @xmath6 whose coefficients are symmetric polynomials in the variables @xmath615 . hence @xmath616 whenever @xmath607 is a rearrangement of @xmath426 ; so @xmath617 is constant on cyclic equivalence classes of @xmath618 . we can therefore choose a representative of each cyclic equivalence class of @xmath619 to compute the sum of @xmath617 over the weak compositions @xmath426 in @xmath619 . by letting the parking compositions be the chosen representatives , we arrive at @xmath620 it now follows from ( [ noncrossbetti ] ) that we claim that for all @xmath179 and @xmath0 , @xmath626^n \end{array } } t^{{{\rm bar}}(w ) } = \sum_{k=0}^{n-1 } { n-1 \choose k } t^k \sum_{u \in [ m]^{n - k } } t^{{{\rm des}}(u)}.\ ] ] to prove this claim first note that there are two types of barred letters in @xmath627 . the type i barred letters are those that are followed by a letter equal to it in absolute value and the type ii barred letters are those that are followed by a letter that is smaller than it in absolute value . the summand on the right side of the equation enumerates banners that have exactly @xmath34 barred letters of the first type . to obtain such a banner first choose the @xmath34 positions from the first @xmath1 positions in which the type i barred letters are to appear and leave them blank , then fill in the remaining @xmath628 positions with an arbitrary word @xmath576 in @xmath629^{n - k}$ ] , then fill in the @xmath34 barred positions that were left blank from right to left so that the letter equals its successor in absolute value , and finally put bars over the descent positions of the resulting word , the number of which is clearly equal to @xmath630 . _ proof of ( [ noncross1])_. it follows from ( [ strank ] ) and ( [ minmaxeq1 ] ) that @xmath633 where @xmath634 using a similar argument to that which was used to derive ( [ bettifu ] ) ( with ( [ ges3 ] ) now playing the role of ( [ ges4 ] ) ) we obtain @xmath635 for @xmath636 , let @xmath637 and @xmath638 also let @xmath639 . by ( [ ges4 ] ) and ( [ ges3 ] ) we have @xmath640 hence @xmath641 equating coefficients of @xmath642 yields @xmath643 by applying to ( [ gneq ] ) , the specialization that sets @xmath644 \\ 0 & \mbox { otherwise } , \end{cases}\ ] ] we obtain p.a . macmahon , _ the indices of permutations and the derivation therefrom of functions of a single variable associated with the permutations of any assemblage of objects _ , amer . j. math . * 35 * ( 1913 ) , no . 3 , 281322 . j. shareshian and m.l . chromatic quasisymmetric functions and hessenberg varieties _ , proceedings of program on configuration spaces , centro di ricerca matematica ennio de giorgi . pisa , italy , to appear .

in this paper , we describe an experimental test of the he - mckellar - wilkens ( hmw ) topological phase with our lithium atom interferometer . the expected value of the hmw phase shift in our experiment is small and its measurement was difficult because of stray phase shifts due to small experimental defects . we start by describing our setup and we characterize the effects of the electric and magnetic fields needed to observe the hmw effect . then , we develop a model of our interferometer signals including all the defects we have identified . after various tests of this model , we use it to suppress the largest part of the stray phase shifts . we thus obtain a series of measurements of the hmw phase : the results are% larger than expected and this discrepancy is probably due to some limitations of our model . * keywords : * atom interferometry ; high phase sensitivity ; stark effect ; zeeman effect ; paschen - back effect ; fringe visibility ; fringe phase shift ; interferometer defects .

1310.1272

the topological he - mckellar - wilkens ( hmw ) phase introduced in 1993 by x.g . he and b.h.j . mckellar @xcite and in 1994 by m. wilkens @xcite was never tested since its theoretical discovery . we have recently published such a test @xcite using our lithium atom interferometer . in a companion paper @xcite quoted here as hmwi , we have recalled the theory of this topological phase and its relations with the aharonov - bohm @xcite and aharonov - casher phases @xcite . we have also discussed the effects of phase dispersion on interferometer signals and we have considered in detail the phase shifts induced by electric and magnetic fields , namely the dynamical phase shifts due to the stark and zeeman hamiltonian and the topological phase shift due to the aharonov - casher effect . the present paper is devoted to a detailed presentation of the experiment , of its results and of the analysis of the stray effects which have complicated the test of the hmw phase . in the following sections , we first describe the experiment , the data recording procedure and the interferometer signals ( section [ expset ] ) . then , we discuss the effects of the electric fields ( section [ elec ] ) and of the magnetic fields ( section [ mag ] ) . section [ data ] presents the data set for the hmw phase measurement and the raw results . the model describing the stray effects due to phase shift dispersion , introduced in hmwi and developed in the appendix ( section [ app ] ) , is tested thanks to numerous and sensitive measurements of the fringe phase and visibility ( section [ test ] ) . thanks to this model , we have been able to reject most of the stray effects and to measure the hmw phase , as detailed in section [ hmwm ] . a conclusion ( section [ conc ] ) summarizes what we have learnt from this experiment , recalls the open questions ( in particular a phase shift presently not understood ) and discusses how to improve this experiment . in this part , we briefly describe our atom interferometer and , with greater details , the interaction region used to observe the hmw effect . we also describe the compensator coil used to produce a magnetic field gradient opposite to the one due to the hmw interaction region . finally , we explain our data recording procedure which rejects the interferometer phase drifts . our separated arm atom interferometer ( see figure [ fig1 ] ) , has been previously described @xcite . here , we present only its main features and some recent improvements . the atomic source is a supersonic beam of lithium seeded in argon , with a mean velocity @xmath1 m / s . once collimated by two @xmath2 @xmath3 m wide slits , this beam has a transverse velocity distribution with a width comparable to the recoil velocity of lithium , @xmath4 cm / s . this beam is then diffracted by three quasi - resonant laser standing waves in the bragg regime : the present experiment uses first order diffraction , with only two diffracted beams of orders @xmath5 and @xmath6 ( or @xmath7 ) . we thus get a mach - zehnder atom interferometer with two output beams carrying complementary interference signals . one output beam is selected by a slit and its intensity @xmath8 , measured by a surface ionization detector , is given by : @xmath9\ ] ] @xmath10 is the mean intensity and @xmath11 the fringe visibility . the phase is the sum of the phase @xmath12 due to perturbations and the phase @xmath13 due to the diffraction process : @xmath14 , where @xmath15 is the laser wavevector and @xmath16 the @xmath17-position of mirror m@xmath18 . the choice of the laser frequency , at about @xmath19 ghz on the blue side of the @xmath20s@xmath21 - @xmath20p@xmath22 transition of @xmath23li , and the @xmath24% natural abundance of @xmath23li ensure that the interferometer signal is almost purely due to this isotope @xcite . to record interference fringes , we sweep the phase @xmath13 by varying @xmath25 with a piezoelectric actuator . we measure the variations of @xmath25 with a michelson interferometer @xcite . intense signals , with a mean intensity @xmath26 atoms / s and a high fringe visibility @xmath27% provide a large phase sensitivity , with a practically achieved value @xmath28 mrad@xmath29 . a hmw phase is induced when an atom propagates in crossed electric and magnetic fields , both transverse to the atom velocity . our experimental arrangement is inspired by the ideas of h. wei _ et al . _ @xcite , the electric fields are horizontal , in the interferometer plane , and opposite on the two interferometer arms , while the common magnetic field is vertical and as homogeneous as possible . the electric fields are produced by a double capacitor with a septum @xcite located between the interferometer arms ( see figure [ fig1 ] ) . each of the two capacitors is similar to the one we used for the measurement of the electric polarizability of lithium @xcite . outer electrodes are made of polished @xmath30 mm - thick glass plates with evaporated aluminium electrodes . a central high - voltage electrode of length @xmath31 mm is separated from two @xmath30 mm - long grounded guard electrodes by @xmath32 mm - wide gaps : these gaps withstand a voltage larger than @xmath32 kv . the septum is a @xmath33 @xmath3 m thick aluminium foil . the capacitors are assembled by gluing together the electrodes and the glass spacers ( thickness @xmath34 mm ) with epotex 301 glue ( epoxy technologies ) . the septum must remain well stretched , even if the capacitor temperature varies . with some advice given by a. cronin , we have acquired the know - how to glue a pre - stretched septum on the electrode - spacer assembly heated near @xmath35c and , due to differential thermal expansion , the septum is fully stretched when the assembly has cooled down @xcite . the capacitors are as symmetric as possible and they are powered by slightly different voltages issued from the same power supply , with an adjustable voltage ratio thanks to potentiometers . this arrangement minimizes stark phase noise due to voltage fluctuations of the power supply . figure [ fig2 ] presents the calculated @xmath36-variation of the electric field @xmath37-component , which is relevant for the hmw phase . @xmath37 is calculated at the septum surface whereas the atom - septum mean distance is near @xmath38 @xmath3 m but the associated correction is very small @xcite . the capacitors assembly is inserted in a brass support on which we have coiled @xmath39 mm - diameter enameled copper wires to produce the vertical magnetic field needed for the hmw phase . we use two rectangular coils , located below and above the interferometer plane , each coil being made of @xmath19 layers and each layer of @xmath40 windings , glued together and to the brass support with a high thermal conductivity glue ( stycast 2850 ft ) . a @xmath19-mm internal diameter copper pipe is also glued on the brass support at mid - distance between the two coils and with a @xmath41 @xmath42/s flow of tap water ( a low flow rate chosen to minimize vibrations ) , the temperature rise is about @xmath43 k / w . usually , we apply a current @xmath8 in the coils @xmath44% of the time so that the maximum current @xmath45 a induces a @xmath46 w mean joule power and a temperature rise near @xmath47 k. in figure [ fig2 ] , we have plotted the calculated @xmath36-variation of the magnetic field @xmath48-component which is the one relevant for the hmw phase . as discussed in hmwi , a zeeman phase shift appears if the magnetic field modulus @xmath49 is different on the two interferometer arms , and we have minimized this difference by careful coiling and design of the connection wires geometry . the hmw interaction region is placed just ahead the second laser standing wave , where the distance between the center of the interferometer arms is largest , close to @xmath50 @xmath3 m . in order not to induce vibrations of the standing wave mirrors , the interaction region is suspended from the top of the vacuum chamber . initial adjustments of the rotation around the horizontal @xmath36-axis and the vertical @xmath51-axis are performed with optical methods . rotation around @xmath51-axis as well as translation in the @xmath17-direction can be operated under vacuum , and the ultimate tunings are done with the atom interferometer running . after optimization of the interferometer signal , the mean intensity @xmath10 and the fringe visibility @xmath11 are not modified by the presence of the septum between the two arms . the magnetic field produced by the hmw coil was measured with a 3d hall probe and compared to the field calculations , showing a good agreement . concerning the electric field , calibration measurements using the atom interferometer ( described in part [ elec1 ] ) yield an accurate knowledge of each capacitor geometry needed for electric field calculations . with the electric and magnetic field components @xmath37 and @xmath48 as a function of @xmath36 , we can calculate the integral @xmath52 and thanks to the very accurate knowledge of the electric polarizability of lithium atom @xcite , we can predict the slope of the hmw phase as a function of @xmath53 product : @xmath54 where the error bar is due to the uncertainty on the geometrical parameters of the capacitors and of the hmw coil . in spite of our efforts , the magnetic field of the hmw coil is slightly different on the two interferometer arms , with a mean relative difference @xmath55 . this difference is most probably due to a bad centering of the septum in the hmw coil , with a distance between the coil symmetry plane and the septum of the order of @xmath56 @xmath3 m . for @xmath57 a , the induced zeeman phase shift is equal to @xmath58 rad for the @xmath59 sublevels . we compensate these phase shifts thanks to a supplementary coil producing an opposite magnetic field gradient along the @xmath17-axis . this so - called compensator coil is made of @xmath60 turns of copper wire on a @xmath33 mm - diameter aluminium cylinder . it is located at mid - distance between the first and second laser standing waves , with a mean distance between the compensator coil and the interferometer arms near @xmath47 mm . this coil is cooled by conduction through its support and temperature rise limits its current @xmath61 to @xmath30 a , if applied only @xmath44% of the time . then , the magnetic field seen by the atoms is below @xmath62 t , a range for which zeeman effect is linear . in our previous experiments @xcite , we deduced the effect of a perturbation by comparing fringe signals successively recorded with and without this perturbation . the phase measured in the absence of perturbation , which should be constant , drifts with time , typically by several tens of mrad over the few minutes needed for recording a high - quality fringe signal . these drifts are not linear in time and they are due to minute distortions of the rail supporting the standing wave mirrors . their magnitude is due to the high sensitivity of the diffraction phase @xmath63 to the mirror positions , with @xmath64 rad/@xmath3 m for m@xmath65 or m@xmath66 and @xmath67 rad/@xmath3 m for m@xmath68 . they were the main limitation of our phase shift measurements . for the present experiment , we have almost canceled the sensitivity to these drifts by applying several field configurations during each fringe recording : a field configuration is defined by the @xmath69 values , where @xmath70 is the capacitor mean voltage and @xmath8 the current in the hmw coil ( this current is accompanied by a current @xmath61 in the compensator coil , as explained below ) . we have used either 4 configurations , @xmath71 , @xmath72 , @xmath69 and @xmath73 or 6 configurations , by adding @xmath74 and ( @xmath75 to this list . a typical fringe recording with @xmath41 configurations is shown in figure [ fig3 ] . a fit of the different fringe signal systems is made using equation ( [ a0 ] ) , where the fringe systems of all configurations share the same value of the diffraction phase @xmath63 . we thus get the mean intensity @xmath76 , the fringe phase @xmath77 and the fringe visibility @xmath78 for each field configuration . in this way , we deduce the effects of the application of the electromagnetic field corresponding to each configuration . @xmath10 is independent of the field configuration , but the visibility and the phase are both modified . we define a relative visibility and a fringe phase shift for each field configuration by : @xmath79 @xmath80 and @xmath81 are the fringe relative visibility and phase shift with the electric field only . @xmath82 and @xmath83 are the fringe relative visibility and phase shift with the magnetic field only and @xmath84 and @xmath85 are the fringe relative visibility and phase shift with the electric and magnetic fields applied simultaneously . a fringe scan such as shown in figure [ fig3 ] lasts about @xmath46 s , a duration sufficiently small to ensure quasi - linearity of the interferometer phase drift with time ( an exactly linear phase drift only alters @xmath63 and leaves the results of eqs . ( [ a1 ] ) unchanged ) the error bars are about @xmath19% on the relative visibility and @xmath33 mrad on the induced phase shifts . we repeat about @xmath50 successive fringe scans , taking care that the fringe scan period and the field configuration period are not commensurate , in order to avoid any possible bias in the fits . the error bars on the averages of such a scan series are near @xmath86% for the relative visibility and near @xmath87 mrad for the phase shifts , small enough to detect fine perturbations of the interference fringe signals and to understand systematic effects . during calibration measurements , we applied a voltage @xmath70 to one capacitor only , the other one being grounded . figure [ fig4 ] presents typical results for the fringe visibility @xmath88 and the induced phase shift @xmath89 as a function of @xmath90 . these measurements were fitted using equations ( 19 - 20 ) and ( 31 ) of hmwi , which yields the value of the parallel speed ratio @xmath91 and the values of the stark phase shifts induced by each capacitor for the mean atom velocity : @xmath92 rad / v@xmath20 and @xmath93 rad / v@xmath20 . using the very accurate theoretical value @xcite of lithium atom electric polarizability @xmath94 , we may deduce the geometrical parameter @xmath95 for both capacitors ( @xmath96 is the capacitor effective length and @xmath97 the plate spacing @xcite ) . the effective length is the same for both capacitors with a good accuracy , @xmath98 mm , so that these experiments provide measurements of the mean values of the capacitor spacings @xmath99 mm and @xmath100 mm . when we apply electric fields to both capacitors , with the voltage ratio tuned to cancel the stark phase shift @xmath81 , we observe a residual phase shift due to imperfect tuning : @xmath81 is small and approximately proportional to @xmath90 , but with large fluctuations of the measured value ( see figure [ fig5 ] ) . we have observed that @xmath81 drifts with time when the interaction region temperature varies : this behavior can be explained by a delay of the expansion of one capacitor with respect to the other one , delay due to the low thermal conductivity of glass . for @xmath101 v , the stark phase induced on each interferometer arm can reach @xmath102 rad . then , a typical deviation for @xmath81 of @xmath103 rad from its mean value corresponds to a @xmath104 relative variation of the geometrical parameter @xmath95 of one capacitor with respect to the other one . this variation is somewhat larger than expected for a simple thermal expansion effect with a temperature variation smaller than @xmath47k . the conclusion is that , because of dispersion and drift , the residual stark phase shift @xmath81 does not carry much useful information . in the experiments with 6 field configurations , we can measure the difference of the stark phase shifts for opposite @xmath70 values , with an error bar close to @xmath32 mrad . figure [ fig6 ] plots the quantity @xmath105 as a function of @xmath106 . 58 ) of hmwi predicts that the only @xmath70-odd term in @xmath107 is the contact potential phase @xmath108 so that : @xmath109 we have fitted the measured values of @xmath110 by a function @xmath111 . the fitted @xmath112-value , @xmath113 mrad is not explained by eq . ( [ ex0 ] ) but its presence might be due to the use of different power supplies , one per polarity , to produce opposite voltages . the fitted slope @xmath114 mrad / v can be explained by a difference of the mean contact potentials @xmath115 mv : we may conclude that contact potentials play a very minor role in our experiment and this idea will be supported by further results . we now discuss the measurements of the relative visibility @xmath116 . the residual stark phase shift @xmath117 is sufficiently small to neglect any effect of the velocity dispersion on the visibility . the relative visibility showed highly dependent of the standing wave mirrors alignment . therefore the collected data for @xmath116 was partitioned into 7 different sets : within a given set , this interferometer alignment is identical for all the data points . [ fig7 ] presents two of these data sets . all the sets exhibit different behaviors , but are well independently fitted using the following equation : @xmath118 as illustrated by fig . [ fig7 ] , the relative visibility can become larger than @xmath32 , a result apparently surprising . this happens when phase dispersions which exist when no interaction is applied , are partially canceled by the phase dispersion due to the application of the electric fields . the pre - existing phase dispersions originate from the zeeman phase shifts due to the inhomogeneity of the laboratory magnetic field when @xmath119 and from the diffraction phase which presents a spatial dispersion because of an imperfect alignment of the laser standing wave mirrors m@xmath18 . the application of the electric fields induces an aharonov - casher phase shift and a stark phase shift : the aharonov - casher phase shift is dispersed because of its dependence with the @xmath120 sublevel and the stark phase shift is dispersed because of capacitors defects . we must describe all these effects , in order to explain the behavior of @xmath116 . assuming a balanced hyperfine population i.e. @xmath121 ( @xmath122 is defined in appendix a of hmwi ) , we use equations ( 23 ) , ( 42 ) and ( 46 - 47 ) of hmwi to evaluate @xmath123 : @xmath124 \frac{\mathcal{v}_{b0}}{\mathcal{v}_0 } \nonumber \\ \mbox{with } \frac{\mathcal{v}_{b0}}{\mathcal{v}_0 } & = & \frac{1 + \cos\left(j_0\right ) + 2 \cos\left(j_0/2\right)}{4}\end{aligned}\ ] ] @xmath125 is the dispersion of the diffraction phase and @xmath126 is the visibility modified by the inhomogeneity of the laboratory magnetic field . because of these two effects , the observed visibility @xmath123 is smaller than its optimum value @xmath127 . with electric fields on both arms , the stark phase shift @xmath128 is a function of @xmath51 , and the aharonov - casher phase shift @xmath129 is given by @xmath130 ( this formula is valid because zeeman effect is linear in the laboratory field ) . we deduce the fringe visibility @xmath131 : @xmath132 \nonumber \\ & - & \frac{\varphi_{ac}(2,2)}{4 } \left[\sin\left(j_0\right ) + \sin\left(\frac{j_0}{2}\right)\right]\end{aligned}\ ] ] the last term of the r.h.s . is a first - order taylor expansion of the trigonometric functions of @xmath133 , valid because @xmath134 mrad in the present experiment . we get the relative visibility @xmath135 : @xmath136\end{aligned}\ ] ] the dispersion @xmath137 of the stark phase shift is given by @xmath138 with the geometrical defect term @xmath139 and the contact potential term @xmath140 . @xmath141 while @xmath142 and @xmath143 are independent of @xmath70 . we thus deduce the values of the @xmath144 coefficients : @xmath145 \nonumber \\ k_{v2 } & = & \left < \left(\delta \varphi_{s , c}\right)^2/2\right > + \left<\delta \varphi_d \delta\varphi_{s , g}\right > \nonumber \\ k_{v3 } & = & \left < \delta \varphi_{s , g}\delta \varphi_{s , c}\right > \nonumber \\ k_{v4 } & = & \left < \left(\delta \varphi_{s , g}\right)^2/2\right>\end{aligned}\ ] ] discussed below is a comparison of eqs . ( [ ex5 ] ) with the results of fits of @xmath146 for the 7 available data sets : at the same time , we test the validity of our description of experimental defects and we get some insights on the nature of the systematic effects . all @xmath147 values are positive and compatible with their mean , @xmath148 v@xmath149 . this is in agreement with eqs . ( [ ex5 ] ) which predicts that @xmath147 is positive and depends solely of the geometrical defects of the capacitors and not of the interferometer alignment . from this result , we may estimate the geometrical defects of the capacitors if we assume that the spacing difference @xmath150 is the main defect and that it varies linearly with @xmath51 . we then find that @xmath151 varies by about @xmath152 @xmath3 m over the @xmath51-range sampled by the atoms ( about @xmath19 mm ) . this @xmath153 value appears to be quite small for capacitors assembled by gluing parts together but , when @xmath154 v , this small defect is sufficient to induce a total dispersion of the stark phase shift along the atomic beam height equal to @xmath155 rad . all @xmath156 values ( excepted one ) are compatible with @xmath5 , with a very small mean value , @xmath157 v@xmath158 , corresponding to @xmath159 rad@xmath20 for @xmath101 v. the dispersions @xmath160 and @xmath161 are not correlated , in agreement with the idea that contact potentials fluctuate on small scales and that geometrical defects are smooth functions of @xmath51 . each @xmath162 value has a small error bar but @xmath162 varies strongly from one set of data to the next , covering the range from @xmath163 to @xmath164 v@xmath165 . these large variations prove that the dominant contribution comes from the interferometer alignment i.e. from the @xmath166 term . when @xmath142 and @xmath167 have opposite variations , the electric fields increase the visibility , as observed in figure [ fig7 ] . all @xmath168 values are compatible with their mean , @xmath169 v@xmath170 . the first term @xmath171 , which involves the contact potential term , is expected to be very small for the same reasons which explain the weakness of @xmath156 and , if this term was not negligible , @xmath168 should vary with the interferometer alignment like @xmath162 . the second term , which is due to the aharonov - casher phase shift in the laboratory magnetic field , must be dominant . assuming that the laboratory magnetic field @xmath172 is constant over the capacitor length , and that the electric fields are equal to @xmath173 on a length @xmath174 mm , we estimate the aharonov - casher phase shift given in eqs . ( 2 ) and ( 49 ) of hmwi : @xmath175 where @xmath176 points in the direction of @xmath172 . the measurements presented in the next section give access to @xmath177 rad and to @xmath178 . we thus deduce @xmath179 i.e. @xmath172 points downward , at about @xmath180 from the vertical , in agreement with direct measurements of the local laboratory field . we have measured the relative visibility @xmath181 and the phase shift @xmath182 of the interference fringes as a function of the compensator coil current @xmath61 . the results are plotted in fig . [ fig8 ] with fits based on eqs . ( 42 ) and ( 45 ) of hmwi , with @xmath183 and assuming balanced sublevel populations ( see appendix a of hmwi ) . the relative visibility @xmath184 and the phase shift @xmath83 were measured as a function of the hmw coil current @xmath8 . some of the results are plotted in fig . [ fig9 ] . equation ( 41 ) of hmwi gives @xmath185 as a function of @xmath186 , @xmath187 and @xmath188 and we use @xmath189 , @xmath190 , @xmath191 to fit the data . the hyperfine population unbalance parameter @xmath122 is also fitted , with a different value for each data set corresponding to a slightly different laser frequency . with a hmw coil current @xmath8 and a compensator coil current @xmath61 , optimum compensation of the linear part of the zeeman phase shift is obtained with @xmath192 . when @xmath193 a , it is impossible to use @xmath194 because the compensator coil temperature rises too much and , then , we have used @xmath195 a. fig . [ fig10 ] presents the relative visibility @xmath196 as a function of @xmath8 . comparison with fig . [ fig9 ] proves the efficiency of the compensator : when @xmath197 , @xmath198 vanishes for @xmath199 a while , with the compensator in operation , it remains larger than @xmath200% if @xmath201 a and vanishes only for @xmath202 a. the revival observed for @xmath203 a , with a relative visibility near @xmath204% and a phase shift close to @xmath205 , is explained in @xcite . in order to have the best estimate of the zeeman phase shifts induced during the hmw effect measurements , we performed a single global fit of all the data recorded while testing the effects of magnetic fields . this data set was collected during the hmw effect measurements using both coils ( with @xmath61 related to @xmath8 for optimum compensation ) as well as during calibration measurements using either both coils ( with different relative tuning of @xmath8 and @xmath61 ) or only one coil . as introduced in hmwi , the zeeman phase shifts are calculated using @xmath206 , with @xmath207 . the data set for @xmath208 and @xmath209 includes about 150 data points which belong to @xmath0 series corresponding to slightly different laser frequencies and a different @xmath122 value is fitted for each series . here are the fitted values of @xmath143 , @xmath10 , @xmath210 , @xmath211 and @xmath212 provided by this global fit : @xmath213 fig . [ fig11 ] presents the data set collected for the hmw phase measurement . as @xmath214 is very small and proportional to the @xmath53 product , we have chosen to record data with large values either of @xmath70 or of @xmath8 , so that we have no data point near the origin . the measured values of the phase shift @xmath215 are plotted as a function of the @xmath53 product in fig . [ fig12 ] : these results do not agree with the predicted variations of @xmath214 and we explain this discrepancy by stray phase shifts which appear when the electric and magnetic fields are simultaneously applied . the origin of these stray phases have been explained on general grounds in hmwi and the detailed calculation is presented in the appendix of the present paper . we are going to test these calculations first on the relative visibility @xmath216 and afterwards on the phase shift @xmath215 . the various stray effects differ by their symmetry with respect to the reversal of the electric and/or magnetic fields and , in order to test separately these effects , it is necessary to extract the even / odd parts of these quantities with respect to field reversals by combining measurements for opposite @xmath70 or @xmath8 values . for any quantity @xmath217 , the mean @xmath218 and the half - difference @xmath219 for opposite values of @xmath70 ( then @xmath220 ) or of @xmath8 ( then @xmath221 ) are equal to : @xmath222/2 \nonumber \\ \delta_ef(v , i)&= & \left[f(v , i ) -f(-v , i)\right]/2 \nonumber \\ \mathcal{m}_bf(v , i)&= & \left[f(v , i ) + f(v ,- i)\right]/2 \nonumber \\ \delta_bf(v , i)&= & \left[f(v , i ) - f(v ,- i)\right]/2\end{aligned}\ ] ] most experiments were done with 6 field configurations and they provide simultaneous measurements of @xmath216 and @xmath215 for opposite voltages , with exactly the same current @xmath8 and the same value of the population unbalance parameter @xmath122 : we thus have very sensitive tests of the effects of electric field reversal . we are going to test the predictions of the calculations described in the appendix of the present paper . following eq . ( [ a6 ] ) , four combinations of @xmath223 separate the contributions of the four @xmath224 terms . however , as shown by eq . ( [ a14 ] ) , the quantity @xmath225 also includes a contribution due to the aharonov - casher effect in the corresponding @xmath72 field configuration . it is given by the term @xmath226 , with a value close to @xmath227% for @xmath101 v. because it involves the ac phase , this effect is an odd function of the voltage @xmath70 . we eliminate this contribution by using the measured values of @xmath228 to calculate @xmath229 , from which we deduce a corrected fringe visibility given by @xmath230 . this quantity is now simply expressed by eq . ( [ a3 ] ) ( obviously , this correction is necessary only when studying @xmath70-odd terms ) . the variations of @xmath231 give a test of the @xmath232 term . we do not plot these results here because all the values are very small , in the @xmath233 range , with error bars near @xmath234 , with one exception , for @xmath235 a ( the visibility is then very low and some approximations of our calculations of the appendix are no more valid ) . the variations of @xmath236 give a test of the @xmath237 term . we do not plot these results here because all the values are also very small , in the @xmath238 range , with error bars near @xmath234 . these two results prove that the @xmath232 term and the @xmath239 term are very small , in good agreement with our calculations which predict that these terms should vanish if the contact potential terms are negligible . having verified that the @xmath240 terms are negligible , the quantity @xmath241 reduces to @xmath242 ( see eq . ( [ a5 ] ) ) . the leading terms of @xmath243 given by eq . ( [ aa6 ] ) are proportional to the aharanov - casher phase : @xmath244 thanks to our knowledge of the zeeman phases ( eqs . ( [ ex13 ] ) ) , we can evaluate all the terms of eq . ( [ a15 ] ) and we compare its prediction to our measurements in fig . [ fig13 ] and fig . [ fig14 ] . the good agreement , obtained without any fitted parameter , proves that the dominant @xmath70-odd effect is due to the ac phase shift , and confirms the validity of our calculations . we first discuss the combination @xmath245 given by : @xmath246 as @xmath247 is non - zero only if contact potentials are not negligible , we expected this quantity to be negligible . @xmath248 is plotted as a function of @xmath8 on fig . [ fig15 ] and as a function of @xmath70 on fig . [ fig16 ] . these experimental results are surprising : @xmath245 is almost independent of the current @xmath8 and it rapidly increases with the voltage @xmath70 . the measured values are well fitted as the sum of two terms , one term in @xmath70 and one in @xmath249 ( odd powers of @xmath70 have been chosen because this quantity is @xmath70-odd ) . contact potentials can in principle explain non zero values of @xmath248 ( the calculation is made in ref . @xcite ) , but the predicted effect depends of the current @xmath8 with divergences similar to those visible in fig . [ fig13 ] , in complete disagreement with the measurements plotted in fig . [ fig15 ] . moreover , the observed magnitude of @xmath245 would require values of contact potentials that are ruled out by the measurements of @xmath146 and @xmath250 previously presented . this effect is strange because @xmath215 given by eq . ( [ a1 ] ) is already a difference of phase shifts measured with and without the magnetic field , so that @xmath245 must vanish when the applied magnetic field goes to zero : as a consequence , the independence of @xmath245 with the current @xmath8 can not extend to @xmath251 . however , if the transition occurs for instance when the laboratory field and the hmw field are comparable in magnitudes , it should be observed with a current @xmath8 of the order of @xmath252 a , a range of @xmath8-values we have not studied . we have investigated several possible explanations which revealed unsatisfactory for different reasons : usually , either the symmetry with respect to @xmath70 and @xmath8 reversals or the order of magnitude of the observed phase are not in agreement with our observations . moreover , most explanations can not explain why the effect is sensitive to the presence of the magnetic field but independent of its value in the studied range in figure [ fig15 ] . we will not discuss here these failed explanations , for lack of space . the origin of this phase shift remains mysterious but thanks to its independence with regards to @xmath8 , it can be easily eliminated by combining data with opposite @xmath8-values . we now discuss the quantity @xmath253 which vanishes if contact potentials are negligible . if they are taken into account , this quantity is given by @xcite : @xmath254 because of the presence of a contact potential term @xmath161 , @xmath255 is expected to be small , and we have not included higher order terms in eq . ( [ a17 ] ) , because they should be even smaller . the measured values of @xmath253 are plotted in fig . [ fig17 ] , with different symbols for data points depending if @xmath256 is smaller or larger than @xmath257 a. when @xmath258 a , the measured values are very small and compatible with @xmath5 : in this range of @xmath8 values , the zeeman phases @xmath259 are small thanks to the compensator , the systematic effects are weak and the approximations done in our model are good . when @xmath260 a , the zeeman phases increase rapidly with @xmath261 , and several effects decrease the accuracy of our model . first , the polynomial expansion of the zeeman phases in powers of the current @xmath8 is poorly convergent for some sublevels ( see hmwi ) while the systematic effects are very sensitive to the value of the zeeman phases . secondly , with increasing zeeman phases , the systematic effects which involve the dispersion @xmath262 increase ( this point is discussed below ) . finally , increasing zeeman phases induce a rapid decrease of the visibility which cancels for @xmath263 a and higher order terms in @xmath264 or @xmath265 are no more negligible . the tests on the fringe phase presented up to now have detected stray phase shifts not larger than @xmath266 mrad . we end this part by considering the quantity @xmath267 , which includes the largest stray phase shifts . @xmath268 is an even function of the current @xmath8 , because we have just shown that @xmath269/2 $ ] is negligibly small . @xmath267 is given by : @xmath270 where we have neglected higher - order terms ( see eq . ( [ aa6 ] ) ) . the measured values of @xmath268 for @xmath101 v are plotted in fig . [ fig17sl18 ] . the stark phase dispersion @xmath271 has been characterized thanks to the study of @xmath116 ( part [ elec3 ] ) . the evaluation of the variations of @xmath272 with @xmath8 is done at the expense of a supplementary approximation , assuming a rectangular profile for the field of the hmw coil along the atom trajectory ( compare fig . [ fig2 ] ) . it then becomes possible to perform a fit of the measured values of @xmath267 ( taking into account the terms neglected in eq . ( [ ajoutsl ] ) ) . the result of this fit is also shown in figure [ fig17sl18 ] : a good agreement is found for the behavior of this quantity , and the fitted parameter values are compatible with the expected dispersion of @xmath273 along the atomic beam height according to the calculations of the magnetic field . this result confirms the importance of the spatial dispersion @xmath272 of the zeeman phase shifts and it proves that the main systematic effects are due to these spatial phase dispersions . here are the main results of our study of these effects : * the effects of the spatial dispersions @xmath274 and @xmath262 of the stark and zeeman phase shifts respectively are well identified ; * the effects of the the dispersion @xmath275 of the stark phase due to contact potentials appear to be below our experimental sensitivity ; * our model provides a qualitative understanding of the systematic effects for all values of the current @xmath8 . the visibility decreases rapidly and vanishes for @xmath276 a : this circumstance has been used to enhance the sensitivity to certain terms but clearly , as soon as in @xmath260 a , our model describing the systematic effects is less accurate . * as the visibility presents a revival for @xmath277 a with the zeeman phases @xmath259 being close to @xmath278 , we have made several series of measurements in this range of @xmath8-values but we can not expect our model to be accurate ; * we have observed an unexpected phase shift which is independent of the current @xmath8 in the range @xmath279 a and which is odd with respect to @xmath70-reversal . we presently have no explanation for this effect and we continue our investigations on its possible sources . it may be either a systematic effect forgotten in our analysis or a fundamental physical effect , for instance such as the effects discussed by j. anandan @xcite but , as far as we can judge , these fundamental effects are either too small or they have not the correct symmetry with respect to @xmath70 and @xmath8 . we now use our knowledge of the stray phase shifts in order to eliminate their contributions to the measurement of the hmw phase . the hmw phase @xmath214 is proportional to the @xmath53 product , i.e. it is odd with respect to @xmath70- and @xmath8-reversals . the main contribution in the stray phase shifts on the measurements of @xmath280 are even with respect to @xmath70 and @xmath8 , but because of the existence of a @xmath70-odd phase of unknown origin , we choose to use the @xmath8-odd character of @xmath214 to cancel the maximum amount of systematic effects . accordingly , we plot the quantity @xmath281 as a function of the @xmath53 product . we have used different symbols for the measurements depending if @xmath256 is smaller or larger than @xmath257 a and we have made separate fits of these two sets of data using @xmath282 . @xmath283 and @xmath284 in both fits , the intercept @xmath285 for @xmath286 is compatible with a vanishing value . the error bar on the slopes @xmath94 is substantially smaller when @xmath201 a than when @xmath287 a : this is visible on the data which is more dispersed when @xmath287 a. for both fits , the fitted slopes are larger ( in modulus ) than the predicted value @xmath288 rad / va . the discrepancy is equal @xmath289% if @xmath201 a and @xmath290% if @xmath287 a. our model predicts that there are two contributions to @xmath281 : @xmath291 @xmath292 given by equation [ aa6 ] is the product of the ac phase by correlation terms . thanks to the knowledge of the experimental defects , it is possible to evaluate all the terms involved in @xmath293 . the only quantities which are not directly measured are the correlations @xmath294 and @xmath295 which are evaluated from the measurement the correlation with @xmath296 replacing @xmath297 , assuming that both effects are linear functions of @xmath51 . the calculated value of @xmath293 never exceeds @xmath87 mrad for the data set with @xmath201 a , and we have made this correction to get @xmath298 which is plotted in fig . 3 of our letter @xcite . the fitted slope @xmath299 rad / v.a is still too large but the discrepancy with the theoretical value is reduced to @xmath0% . we have described a measurement of the he - mckellar - wilkens topological phase by atom interferometry . this experiment was feasible with our atom interferometer because the interferometer arms are well separated in space and the interferometer signal is intense , with a large fringe visibility , near @xmath300% . the arm separation is needed in order to insert a septum between the two interferometer arms without any degradation of the signal . the signal intensity and the large value of the fringe visibility both contribute to enhance the phase sensitivity : its practically achieved value near @xmath301 mrad/@xmath302 is needed for the present measurement . the hmw phase shift is rather small , at most @xmath303 mrad under our experimental conditions , and appears as the combination of four phase measurements for which @xmath304 s of data recording were needed to reduce the uncertainty near @xmath87 mrad . the analysis of the experiment revealed more complex than expected , because of stray phases . the complexity of the signal is due to several factors : * the signal is the sum of the contributions of 8 sublevels which are not exactly in phase because of the zeeman phase shifts due to the slightly different values of the magnetic field on the two interferometer arms ; * we have built a compensator coil to produce an opposite gradient of the magnetic field at another place in the interferometer . the use of this compensator has been very fruitful as it has enabled us to apply substantially higher fields with a limited loss of fringe visibility . however the compensator produces a low field , so that it can correct only the part of the phase shifts due to linear zeeman effect ; * the weights of the various @xmath305 sublevels are functions of the laser frequency and power density in the standing waves used for atom diffraction . we had to control these parameters rather tightly in order to keep these weights almost equal and constant ; * the main phase shifts ( diffraction phase shift , stark and zeeman phase shifts ) present a dispersion with the atomic trajectory described in our calculations by the @xmath51-coordinate . in the presence of several dispersed phase shifts , the visibility of the contribution of a given sublevel to the total fringe signal is better or worse , depending if the dispersions of the different phase shifts subtract or add their effects . we have developed a model taking into account all these effects and this model has been very successful in explaining the variations of the observed phase shifts and visibility with the capacitor voltage @xmath70 and the hmw coil current @xmath8 . however , an extra - phase has been observed and characterized : this phase is odd with the capacitor voltage @xmath70 ; it behaves roughly like @xmath249 ; it appears only when the magnetic field is applied but its value is independent of the magnetic field magnitude in a wide range . we continue our investigations to understand the effect which produces this phase . by combining measured phase shifts with opposite values of the current @xmath8 , we have eliminated this phase and we have obtained a first measurement of the hmw phase shift . the observed effect is larger than its expected value by @xmath290% if we use all the collected data points and only by @xmath289% if we consider only the data points with @xmath306 a. finally , there is a small stray contribution of the aharonov - casher phase to the measured phase shift , and using our model , it was possible to evaluate this contribution and to correct the measured values accordingly . the discrepancy between our corrected measurements and the expected hmw phase shift is then reduced to @xmath0% . it is necessary to improve this experiment in order to reduce the uncertainty on the hmw phase - shift . here are the main possibilities : * reduction of stray effects by a better construction of the hmw interaction region . the present construction has two main defects : the difference of the capacitor thicknesses varies with the @xmath51-coordinate and the septum does not coincide with the symmetry plane of the hmw coils . the construction of capacitors with a better controlled geometry is probably possible but quite difficult , because of the need of using a stretched septum . a better centering of the septum with respect to the hmw coils is probably rather easy and this would reduce substantially the zeeman phase shifts which are the largest source of complication . * reduction of stray effects by optical pumping of the atomic beam . if all the atoms are in one @xmath120 sublevel only , the signal is no more an average on the hyperfine sublevels populations . moreover , the trajectory - averaged zeeman phase shift can be exactly canceled by the compensator if the pumping is done in the @xmath307 , @xmath308 ( or @xmath309 ) sublevel for which zeeman effect is exactly linear . as a consequence , this arrangement , which should reduce most of the stray phase shifts , is feasible with minor modifications of our setup and experiments are in progress . * reduction of stray effects by using another atom : this requires the development of a completely new atom interferometer with separated arms . most of the difficulties are due to the paramagnetic character of lithium and an atom with a @xmath310s@xmath311 non - degenerate ground state ( i.e. with a zero nuclear spin ) would be ideal because there would be no zeeman phase shift and no aharonov - casher phase shift . we may consider either a thermal beam of a light atom or a laser - cooled atomic source . in the case of a thermal beam , the most obvious choice is ground state helium , with which a very nice interferometer using diffraction by material gratings was developed by j.p . toennies and co - workers @xcite . because helium electric polarizability is small ( @xmath312 ) , larger electric and/or magnetic fields will be needed . among atoms which have been laser - cooled , magnesium , calcium , strontium or ytterbium all have a @xmath310s@xmath311 ground state and at least one stable isotope with a nuclear spin equal to @xmath5 . we thank the laboratory technical staff for their help , a. cronin for fruitful discussions , g. trnec , a. miffre and m. jacquey for all the work done on our atom interferometer . we are greatly indebted toward cnrs inp , anr ( grants anr-05-blan-0094 and anr-11-bs04 - 016 - 01 hipati ) and rgion midi - pyrnes for supporting our research . we describe here the main points of our calculation of the fringe signal from which we deduce the stray phase and the fringe visibility . the fringe phase @xmath313 is the sum of the diffraction phase @xmath63 , the sagnac phase @xmath314 due to earth rotation , the stark phase @xmath315 , the zeeman phase @xmath185 , the hmw phase @xmath214 and the aharonov - casher phase @xmath129 : @xmath63 , @xmath214 and @xmath317 are independent of the atom velocity @xmath318 ; @xmath314 and @xmath315 vary like @xmath319 and @xmath185 like @xmath320 . these velocity - dependent phases are small ( @xmath321 rad , @xmath322 rad and @xmath323 rad when @xmath324 a - only these data points will be retained for the final analysis ) . as the parallel speed ratio @xmath325 of the lithium beam is large , @xmath326 , we may forget the velocity average and , as a consequence , the sagnac phase @xmath314 which is constant . we consider the spatial dispersion of @xmath63 , @xmath315 and @xmath185 only and we neglect this dispersion for @xmath214 and @xmath129 , because they are small , @xmath327 mrad and @xmath328 mrad , for our largest fields . the total phase dispersion @xmath329 is the sum of three terms only : from now on , the @xmath120 dependence of @xmath331 and @xmath133 is not explicit and , for @xmath63 , @xmath315 and @xmath331 , we note @xmath332 the spatial average of @xmath333 given by @xmath334 . the average over the @xmath120 sublevels is taken with equal weights , @xmath335 . this is a good approximation because in the experiments devoted to the hmw phase measurement , we have kept @xmath122 small ( @xmath336 ) and randomly distributed around @xmath5 ( its main effect is to induce a supplementary dispersion of our phase measurements ) . with these approximations , following the discussion of section iv of hmwi , the signal due to one @xmath120 sub - level is given by : @xmath337/8 \nonumber \\ \mathcal{v}_m & = & \mathcal{v}_0 \left [ 1 -\left < { \left(\delta \varphi\right)}^2/2\right > \right ] \nonumber \\ \varphi_m & = & \phi- \left < { \left(\delta \varphi \right)}^3/6\right>\nonumber \\ \mbox{with } \phi & = & \phi_d + \phi_s+\phi_{z}(f , m_f)+ \varphi_{hmw}+ \varphi_{ac}(f , m_f ) \nonumber \\ \mbox{and } \delta \varphi & = & \delta\varphi_d + \delta \varphi_s+ \delta\varphi_{z}(f , m_f)\end{aligned}\ ] ] if we neglect nuclear magnetism , the @xmath120 sublevels form 4 pairs with exactly opposite zeeman energy shifts : three pairs of levels with the same @xmath338 value and the pair @xmath339 . we label these pairs by an @xmath338 value going from @xmath7 to @xmath340 and we note @xmath341 the value of @xmath185 for the sublevel @xmath342 . because of numerous terms , these calculations are rather complicated and we first present a tutorial calculation in which we cancel @xmath343 and @xmath129 , and we forget the cubic term in @xmath329 . we first calculate the signal @xmath344 of a pair of sublevels : @xmath345/ { \mathcal{v}}_0 & \approx & \left[1 -\frac{\left<\left(\delta \varphi_s\right)^2\right>+\left<\left(\delta \varphi_{z}\right)^2\right>}{2}\right ] \nonumber \\ & \times & \cos\left(\phi_{z}\right ) \cos\left ( \phi_d + \phi_s + \varphi_{hmw } -\theta \right ) \nonumber \\ \mbox{with } \tan \theta & \approx & \theta \approx \left<\delta \varphi_s\delta \varphi_{z}\right > \tan\left(\phi_{z}\right)\end{aligned}\ ] ] the important point is the phase shift @xmath346 proportional to the correlation term @xmath347 and this effect is due to the fact that the contributions of the two levels of the pair have different visibility : the term @xmath348 $ ] modifies these visibility in a different way because the dispersions @xmath349 and @xmath350 have the same sign for one level of the pair and opposite signs for the other one . because of the @xmath351 factor , @xmath346 is very sensitive to @xmath352 value , especially when @xmath352 is close to @xmath353 . @xmath355/ { \mathcal{v}}_0 & \approx & \frac{1}{4}\left[d \cos\left ( \phi_d + \phi_s + \varphi_{hmw } \right ) \right . \nonumber \\ & & + \left . n \sin\left ( \phi_d + \phi_s + \varphi_{hmw } \right)\right ] \nonumber \\ & \approx & \frac{\sqrt{d^2 + n^2}}{4 } \cos\left ( \phi_d + \phi_s + \varphi_{hmw } -\theta \right ) \nonumber \\ \mbox{with } \tan \theta & \approx & \theta \approx \frac{n}{d}\end{aligned}\ ] ] @xmath360\nonumber \\ d_{+/- } & = & \sum\left [ d_{+,+ } + d_{-,+ } + d_{+,- } + d_{-,-}\right]\nonumber \\ n_z & = & \sum \left[\left<\delta \varphi_d \delta \varphi_z\right > \sin\left(\phi_{z}\right)+ \left<\delta \varphi_d\left ( \delta \varphi_z\right)^2\right > \cos\left(\phi_{z}\right)\right ] \nonumber \\ n_{+/- } & = & \sum \left [ n_{+,+ } + n_{-,+ } + n_{+,- } + n_{-,-}\right]\end{aligned}\ ] ] in these equations , @xmath361 is the sum over the 4 pair of levels labeled by the @xmath338 value as defined after equation ( [ aa1 ] ) and this index is omitted everywhere . @xmath362 represents the effect of the zeeman phase shifts @xmath352 on the visibility , neglecting their spatial dispersion . @xmath363 and @xmath364 represent the effects of the dispersions of the diffraction phase shift @xmath297 and of the zeeman phase shift @xmath350 . the effects of @xmath362 and @xmath363 are independent of the application of the electric field . @xmath365 and @xmath366 are the sum of four terms which involve the simultaneous application of the electric and magnetic field : the first index is the parity with respect to voltage reversal and the second index is the parity with respect to current reversal . in ref . @xcite , we have developed the calculations of the @xmath367 and @xmath368 terms including the contributions of the dispersion @xmath161 due to contact potential ( see hmwi ) and the presence of this @xmath70-odd phase largely increases the number of terms in these equations . as the contact potential terms appear to be extremely small , we do not take them into account in the present discussion but we refer the reader to ref . @xcite for a more complete discussion . with this simplification , @xmath369 is reduced to the geometrical defect term which is @xmath70-even and the @xmath367 and @xmath368 terms are given by : @xmath370 \sin \left ( \phi_z\right ) \nonumber \\ d_{-,+ } & = & d_{+,- } = 0\nonumber \\ d_{-,- } & = & \left [ -1 + \frac{\left<\left(\delta \varphi_d + \delta \varphi_s \right)^2 \right>+ \left<\left(\delta \varphi_z\right)^2\right>}{2}\right]\nonumber \\ & \times & \varphi_{ac } \sin \left ( \phi_z\right ) + \left[\frac{\left<\left(\delta \varphi_z \right)^3 \right>}{6 } + \left<\delta \varphi_d \delta \varphi_s \delta \varphi_z \right>\right . \nonumber \\ & + & \left.\frac{\left<\left(\delta \varphi_s\right)^2 \delta \varphi_z + \left(\delta \varphi_d\right)^2 \delta \varphi_z \right>}{2 } \right]\varphi_{ac } \cos \left ( \phi_z\right ) \nonumber \\ n_{+,+ } & = & \left<\delta \varphi_s \delta \varphi_z\right > \sin \left ( \phi_z\right ) + \frac{\left<\delta \varphi_s \left(\delta \varphi_z\right)^2\right>}{2 } \cos \left ( \phi_z\right ) \nonumber \\ n_{-,+ } & = & n_{+,- } = 0 \nonumber \\ n_{-,- } & = & \left[\left<\delta \varphi_s \delta \varphi_z\right > + \left<\delta \varphi_d \delta \varphi_z\right>\right ] \varphi_{ac } \cos \left ( \phi_z\right ) \nonumber \\ & - & \left[\frac{\left<\left(\delta \varphi_s + \delta \varphi_d \right)^3 \right>}{6 } + \frac{\left<\left(\delta \varphi_s + \delta \varphi_d \right ) \left ( \delta \varphi_z \right)^2 \right>}{2 } \right]\nonumber \\ & & \times \varphi_{ac } \sin \left ( \phi_z\right)\end{aligned}\ ] ] we have used a third - order approximation of the sine and cosine function in eq . ( 23 ) of hmwi but we use only a first - order approximation to get @xmath372 and @xmath373 . this first order approximation is good if @xmath374 . for a practical use of these results , it will be necessary to assume that @xmath362 is considerably larger that the other terms appearing in @xmath357 and that @xmath356 is small with respect to @xmath362 so that we will further simplify the expression of @xmath375 . we are going to use the following equations for the analysis of our experimental results : [ [ theory2 ] ] calculation of the phase shift and the visibility neglecting the effects of the laboratory magnetic field ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ to evaluate the visibility @xmath216 and the phase @xmath377 defined by eqs . ( [ a1 ] ) , we use eq . ( [ a2 ] ) to calculate the terms corresponding to the different field configurations . as a first simplified approach , we consider that the laboratory magnetic field is homogeneous . in the case of fields configurations for which @xmath378 , apart from canceling the zeeman phase shifts , this also enables neglecting the effect of the aharonov - casher phase . at first order in the stray terms , several cancelations appear : using the definitions of eqs . ( [ a4 ] ) , we separate the contributions in @xmath380 and in @xmath381 following their even / odd characters with respect to @xmath70 and @xmath8 . here are the results for the visibility : @xmath382 \nonumber \\ \delta_e\mathcal{v}_{eb}&= & \sum\left ( d_{-,+ } + d_{-,-}\right)/d_0 \nonumber \\ \mathcal{m}_b\mathcal{v}_{eb}&= & 1 + \left[\sum\left ( d_{+,+ } + d_{-,+}\right)/d_0 \right ] \nonumber \\ \delta_b\mathcal{v}_{eb}&= & \sum\left ( d_{+,- } + d_{-,-}\right)/d_0\end{aligned}\ ] ] @xmath383 \nonumber \\ \mathcal{m}_b\delta_e\mathcal{v}_{eb}&= & \sum d_{-,+}/d_0 \nonumber \\ \delta_b\mathcal{m}_e\mathcal{v}_{eb}&= & \sum d_{+,-}/d_0 \nonumber \\ \delta_b\delta_e\mathcal{v}_{eb}&= & \sum d_{-,-}/d_0\end{aligned}\ ] ] [ [ effect - of - the - inhomogeneity - of - the - laboratory - magnetic - field - on - the - measurements ] ] effect of the inhomogeneity of the laboratory magnetic field on the measurements ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ we now take into account the inhomogeneity of the laboratory magnetic field . its main effect is to induce weak zeeman phase shifts , and we neglect their spatial dispersions ( @xmath387 ) . this brings corrections only to the @xmath388 and @xmath389 terms , i.e. to the visibility terms in the field configurations for which @xmath378 . it is straightforward to calculate @xmath388 : when the electric field is applied , the residual zeeman phase shifts are still present ( @xmath391 ) . with nonzero zeeman phase shifts , the ac effect modifies the visibility independently of the presence of spatial phase dispersion @xmath392 : this modification is described by the leading term @xmath393 in the expressions of @xmath243 , eqs . ( [ aa6 ] ) . we thus obtain : @xmath394 d_{0,b0 } + d_{ac , b0}\nonumber \\ \mbox{with } d_{ac , b0}&= & - \sum \varphi_{ac } \sin\left(\phi_z\right)\nonumber \\ & = & - \varphi_{ac , b0 } \left [ \sin\left(j_0\right ) + \sin\left(\frac{j_0}{2}\right)\right]\end{aligned}\ ] ] here @xmath395 is the ac phase of the @xmath396 sub - level in the presence of the laboratory magnetic field : this phase shift is proportional to the applied voltage @xmath70 . in this way , we regain the results of eq . ( [ ex4 ] ) for the relative visibility @xmath397 , and we express the asymmetry @xmath398 of the visibility with voltage reversal : in the field configurations with @xmath400 , the calculations of the type of eqs . ( [ a2 ] ) are not further modified . therefore , the presence of the laboratory magnetic field brings a correction only to @xmath225 in equations ( [ a3 ] ) , in the following form :

it is shown that in a model in which up - type and down - type fermions acquire mass from different higgs doublets , the anthropic tuning of the higgs mass parameters can explain the fact that the observed masses of the and quarks are nearly the same with slightly heavier . if yukawa couplings are assumed not to scan " ( vary among domains ) , this would also help explain why is much heavier than . it is also pointed out that the existence of dark matter invalidates some earlier anthropic arguments against the viability of domains where the standard model higgs has positive , but makes other even stronger arguments possible .

hep-ph0703219

the mass parameter of the higgs field in the standard model ( @xmath4 ) gives the appearance of being anthropically tuned " @xcite . that is , if one imagines the other parameters of the standard model to be fixed , and considers what the universe would look like for different values of @xmath4 , one finds that organic life may only be possible if @xmath4 is negative and has a magnitude very close to the value actually observed . from the physics point of view this might be just a coincidence , though a remarkable one . on the other hand , as pointed out by ref . @xcite , it might have a physical explanation in the context of many - domain " , multiverse " , landscape " scenarios . ( these are equivalent names for the same idea . for recent reviews see @xcite . ) an explanation of that sort would account for the closeness of the strong - interaction and weak - interaction scales ( @xmath5 and @xmath6 ) , something not yet explained by any other scenario . ( the nearness of the weak - scale to the strong scale results in the lightest quarks having masses small compared to @xmath7 , which in turn leads to the existence of nonperturbative bound states of quarks , pseudogoldstone pions , and the the richness of hadronic and nuclear physics , and consequently of chemistry . ) in this paper we extend the analysis of @xcite to a slightly more general higgs structure and show that a possible anthropic explanation of the fact that @xmath8 emerges . we also reconsider the crucial case of positive @xmath4 , and point out that one of the anthropic arguments proposed in @xcite to exclude most of the @xmath9 region is invalidated by the existence of dark energy ( which had not yet been discovered when @xcite was written ) . however , we show that the existence of dark energy makes possible a very different set of arguments that reach even stronger conclusions . the papers of ref . 1 assumed that the only parameter of the standard model that varies among domains and is anthropically tuned is @xmath4 . several facts suggest the possibility that the observed value of @xmath4 may be acounted for in a different way than the values of the other parameters of the standard model . first , @xmath4 is the only dimensionful parameter of the standard model lagrangian . second , it is the most highly tuned of the standard model parameters , being @xmath10 of its natural " value . ( the next most tuned parameter is @xmath11 , which is less than @xmath12 of its natural value . ) third , the smallness of @xmath4 is so far the most intractable of the naturalness problems of the standard model . various plausible mechanisms of a more conventional sort have been proposed for explaining the smallness of other standard model parameters . ( for example , the smallness of @xmath11 can be explained by the peccei - quinn mechanism @xcite or by spontaneously broken cp or p @xcite ; and various symmetry schemes have been proposed to explain the smallness of the yukawa couplings of the light quarks and leptons ) . by contrast , attempts to explain the smallness of the weak scale by technicolor or similar ideas are plagued by a variety of well - known difficulties . and low energy supersymmetry does not by itself explain the magnitude of the weak scale , though it protects it from radiative correction . since the focus of ref . 1 was on the mass parameter of the higgs field , the question naturally arises how things would be different if there were two or more higgs doublets . two higgs doublets appear in a wide variety of theoretical contexts : in theories with supersymmetry ; in peccei - quinn models ; and in many grand unified models ( such as @xmath13 , if the @xmath14 of higgs fields is assumed to be complex ) . the simplest possibility , which is typical of the above - mentioned scenarios , and which avoids problems of higgs - mediated flavor - changing processes , is that one doublet ( @xmath15 ) couples to the up - type quarks , while the other doublet ( @xmath16 ) couples to the down - type quarks and charged leptons @xcite . it is this situation we will analyze in this paper . thus , at high scales we have @xmath17 with two higgs doublets , there are several dimensionful higgs - mass parameters that may vary among domains , with the consequence that @xmath18 and @xmath19 may vary independently . this means that there is the possibility of explaining not only the magnitude of the weak scale , but also the relative magnitudes of the masses of the up - type and down - type quarks . in particular , we shall see that an explanation of the size of @xmath20 and thus a partial explanation of @xmath21 emerges . much of our analysis would apply equally to models with or without low - energy supersymmetry , and would apply in fact to any model where in effect both @xmath22 and @xmath23 vary among domains . thus , the anthropic " considerations given in this paper are not proposed as an alternative to low - energy supersymmetry . indeed , they may shed light eventually on supersymmetry - breaking and the @xmath24 problem . however , in this paper we will consider for purposes of concreteness a non - supersymmetric two - higgs - doublet model . in such a model , the parameter @xmath4 is replaced by a two - by - two hermitian mass matrix @xmath25 , @xmath26 , with four real parameters . to make the weak - scale small , still only one tuning is required , namely of the determinant of @xmath25 , since only one higgs doublet needs to remain light to break @xmath27 . the other doublet has no reason to be light , and in fact would be expected to be of superlarge mass . the doublet that is light ( whose mass - squared we shall call @xmath4 ) is a linear combination of @xmath15 and @xmath16 : @xmath28 , where @xmath29 . the mixing angle @xmath30 enters into the relative magnitudes of the up - quark masses and down - quark masses . in particular , the ratios @xmath20 and @xmath21 are proportional to @xmath31 . if @xmath25 varies in the landscape ( or scans " ) , then so does @xmath31 . we shall see that anthropic " considerations strongly favor the observation of values of @xmath20 that are of order ( but somewhat smaller than ) 1 and thus of values of @xmath21 much larger than 1 ( assuming , as we shall , that quark and lepton yukawa couplings are fixed , i.e. do not vary among domains ) . it is noteworthy that in published models or scenarios that do not invoke anthropic " or landscape " considerations the relations @xmath8 and @xmath32 are generally not accounted for but are simply assumed to hold . ( however , there have been a few attempts to explain @xmath33 in more conventional ways , such as @xcite . ) it seems that anthropic and non - anthropic explanations are somewhat complementary : those relationships that seem most resistant to conventional dynamical or symmetry explanations ( which would include not only @xmath8 , but also @xmath34 and @xmath35 ) are also those that are most straightforwardly accounted for anthropically " . in this paper , as in ref . 1 , we assume that the dimensionless parameters of the lagrangian are fixed , i.e. do not vary among domains . ( more precisely , we assume that their values at some high scale are fixed , since their low - scale values will be slightly influenced through renormalization group running by the values of @xmath22 and @xmath23 . ) if this is assumed , then the inter - family ratios of yukawa couplings are known , e.g. the ratios of @xmath36 , @xmath37 , @xmath38 , and @xmath39 . we will also assume that the yukawa couplings of the third family are all of order 1 . this is plausible , given that schemes that unify all the fermions of a family , such as @xmath13 , relate these three couplings ; and the simplest such schemes give @xmath40 . the relative values of the yukawa couplings of the up - type and down - type quarks are shown in a log plot in fig 1 . . the up - type quarks have a much steeper family hierarchy . the scale on the left is @xmath41 . ] since the mass hierarchy among the up - type quarks is much steeper than that among the down - type quarks , normalizing @xmath42 implies that @xmath43 . that is , @xmath44 in the scenario we are discussing , the relative normalization of the up - type masses and down - type masses is free to scan " , i.e. vary among domains . the consequence of this , as we shall see , is that @xmath45 is anthropically selected to be close to but somewhat larger than 1 , and therefore @xmath21 ends up ( in anthropically viable domains ) being much larger than 1 , as shown in fig 2 . is anthropically set so @xmath8 , then @xmath46 due to the steeper hierarchy of the up - type quarks . the scale on the left is @xmath47 . ] it is in this sense that one can anthropically explain " the largeness of @xmath21 . of course , this assumes the pattern of interfamily mass ratios shown in fig 1 , i.e. the greater steepness of the hierarchy for up - type quarks than for down - type quarks . there exist many interesting and plausible theoretical schemes for explaining that pattern . one example is the model of @xcite , where the hierarchy among the up - type quarks is quadratic and the hierarchy among the down - type quarks is linear in the same small parameters . in this paper , we shall call domains of the universe where organic life based on chemistry is possible viable " , and we shall call the values of parameters in such domains anthropically allowed " . in section 2 of this paper , we shall examine the consequences for viability " of different values of @xmath48 and @xmath49 and graphically display the anthropically allowed and forbidden regions of the @xmath50-@xmath51 plane . in section 3 , we shall look at the probability distributions of @xmath52 that arise from certain simple assumptions about the higgs mass matrix @xmath25 . in section 4 we will use the probability distributions derived in section 3 to analyze the @xmath53 region of parameter space in light of the existence of dark energy and the anthropic requirement that it be fine - tuned to be small . in this section we will discuss anthropic constraints on @xmath48 and @xmath49 . most of these constraints have been discussed in slightly different contexts in @xcite and @xcite , and much of the discussion in this section makes use of their results . the vacuum expectation @xmath55 that breaks the weak interactions and gives @xmath1 and @xmath0 their masses is that of the lighter higgs doublet @xmath56 , whose mass - squared parameter we are calling @xmath4 . we suppose that @xmath57 has a superlarge natural " scale @xmath58 , which is of order @xmath59 ( or @xmath60 , it does not matter for our analysis which ) . there are two qualitatively very different cases , @xmath61 and @xmath62 . if @xmath61 , as in our domain , then @xmath63 and is naturally of order @xmath58 . in this case , smaller @xmath55 corresponds to more fine - tuning of @xmath64 . on the other hand , if @xmath62 , the breaking of the weak interactions is only induced by the coupling of @xmath56 to the quarks , which acquire condensates through the dynamical breaking of chiral symmetry by qcd , so that @xmath65 . in this case , the natural scale of @xmath55 is @xmath66 , and _ larger _ @xmath55 corresponds to more fine - tuning . one sees that the maximum and minimum values of @xmath55 are different by a factor of @xmath67 . in other words , if @xmath68 , @xmath55 can range over 60 orders of magnitude . in fact , @xmath22 and @xmath23 each span the same range . this enormous space of possible values of @xmath48 and @xmath49 is shown in fig . 3 . , @xmath69 , is from @xmath70 to @xmath71 . the blue dot shows the observed values . the contours are of probability density in the landscape , red ( green ) green for the @xmath61 ( @xmath72 ) region . neighboring contours differ by @xmath73 in probability density , increasing in the direction of the arrows . ] in that figure are also plotted the contours of equal probability density in the landscape ( i.e. equal amount of fine - tuning of parameters ) , which will be derived in section 3 under the assumption that the parameters of @xmath64 have a flat probability distribution in the landscape in the interval @xmath74 $ ] . the red curves are contours of equal probability density in the @xmath75 region , and the green curves are contours of equal probability density in the @xmath53 region . neighboring contours differ in probability density by a factor of @xmath73 , with the direction of increasing probability density indicated by the arrows . the position of theanthropically allowed " region in fig . 3 in which our domain is located is indicated by the blue dot . in fact , the viable region is so small that it is smaller than the blue dot in fig . 4 is a blow - up of that region of parameter space . since we are assuming that the yukawa couplings of the quarks and leptons do not vary among domains , and that both the down - type quarks and charged leptons get mass from the same higgs doublet @xmath16 , the electron mass is given by @xmath76 , where @xmath77 is a fixed ratio ( up to logarithmic dependence on @xmath22 and @xmath23 through the running of parameters ) . we take @xmath78 mev and @xmath79 mev , where the subscript zero refers to the observed values of these parameters , i.e. their values in our domain of the universe . thus @xmath80 . first , we will discuss the case of negative @xmath4 . if @xmath4 is negative and smaller in absolute value than in our domain , that involves more fine - tuning and is thus a less probable region of parameter space . our interest is therefore primarily in @xmath4 negative and greater than ( in absolute value ) or comparable to its value in our domain . that corresponds to values of @xmath48 and @xmath49 greater than or comparable to their observed values . if @xmath49 is sufficiently large , then all @xmath0 quarks will beta decay into @xmath1 quarks even inside hadrons , so that the only stable hadron will be @xmath81 ( @xmath82 ) . sufficiently large " in this case means large enough to overcome the extra energy required to put three quarks of the same flavor in a totally antisymmetric state , which energy we call @xmath83 . for @xmath22 and @xmath23 near the observed weak scale @xmath84 , @xmath85 mev ( however , there is a weak dependence of @xmath83 on @xmath22 and @xmath23 through the renormalization - group running of @xmath86 , due to quark thresholds . ) the condition for all @xmath0 quarks to beta decay is @xmath87 , or @xmath88 . this is the region above the upper blue curve in fig . 4 , which is a log - log plot . for @xmath48 and @xmath49 masses very large compared to the qcd scale , this curve asymptotically approaches the straight line @xmath89 , which is a line of slope 1 in fig . 4 . and @xmath49 discussed in the text . the green shaded region is potentially viable . the dot is our domain . ] for @xmath48 and @xmath49 small compared to the qcd scale , the curve asymptotically approaches the horizontal line @xmath90 . in the region where all the @xmath0 quarks decay , not only will the @xmath81 be the only stable hadron , it will surely also be the only stable nucleus . ( in this regime , @xmath49 is large compared to the qcd scale , so that the pions are not pseudo - goldstone particles and the range of the nuclear potential is shorter than in nuclei of our own domain . moreover , there is a strong coulomb repulsion between two @xmath81 . ) these @xmath81 will clothe themselves with a pair of electrons , and act chemically like helium . with the only element being helium - like , it is extremely implausible that life based on chemistry could exist . on the other hand , if @xmath48 is sufficiently large compared to @xmath49 , then all the @xmath1 quarks will decay to @xmath0 , and one will have a domain in which the only stable hadron and nucleus will be a @xmath91 ( @xmath92 ) . these will clothe themselves with positrons ( which will presumably be more numerous than electrons in such a domain , due to the beta decay of all the @xmath1 quarks ) , and thus act chemically like hydrogen . an all-@xmath91 universe is also extremely unlikely to give rise to chemistry - based life . the all-@xmath91 universe corresponds to @xmath93 , which is the region below the lower blue curve in fig 4 . ( for large quark masses this curve it is asymptotic to the line @xmath94 , which is a line of slope 1 in fig . 4 . ) thus we can rule out almost all of the three quadrants in fig . 4 where either @xmath48 or @xmath49 is large compared to its observed value . there is a thin ribbon between the two blue curves where both @xmath48 and @xmath49 are large compared to the qcd scale and @xmath95 , where neither @xmath96 nor @xmath97 can happen . however , this region is unlikely to be viable , since @xmath98 nuclei are not bound there , as shall be seen . more of the parameter space is ruled out as viable if we take into account the conditions that both neutrons and protons should be stable inside nuclei . if @xmath99 denotes a nucleus of mass @xmath100 and atomic number @xmath101 , then the process @xmath102 is energetically allowed if @xmath103 , where @xmath104 is the binding energy of the proton in the nucleus , and @xmath105 is the electromagnetic contribution to the proton - neutron mass difference . following @xcite , we take @xmath106 mev , and we take 10 mev as a typical value of @xmath104 . ( of course , for very large values of @xmath22 and @xmath23 , the values of @xmath105 and @xmath104 would be somewhat different . however , in that limit @xmath105 and @xmath104 are are negligible compared to @xmath48 and @xmath49 anyway , so the shape of the curves is little affected . ) this gives the condition @xmath107 mev . if this is satisfied , all neutrons beta decay , even inside nuclei . since stable nuclei having @xmath108 can not be made up only of protons @xcite , the only stable nucleus in such domains and therefore the only chemical element would be @xmath109h . it is unlikely that chemical life can exist in an such all hydrogen " domains . thus viability requires that one has @xmath110 mev , which rules out the region above the upper green curve in fig . another constraint comes from the requirement that protons should be stable in nuclei . if @xmath111 , then protons even in nuclei would be able to decay through the process @xmath112 . viability thus requires @xmath113 mev , which rules out the region below the lower green curve in fig . a stronger bound comes from the requirement that isolated protons are stable against @xmath114 . if they were not , there would be no @xmath109h in such domains . since almost all organic molecules ( with a few exceptions , such as co@xmath115 ) contain hydrogen , a no - hydrogen domain would presumably have dim prospects for life based on chemistry . this rules out the region below the curve @xmath116 , shown as the orange curve in fig . a slightly stronger bound comes from the condition that the proton in the nucleus of hydrogen not convert to a neutron by electron capture @xmath117 , which rules out the region below the curve @xmath118 , shown as the lower red , concave - up curve in fig . 4 . if the reaction @xmath119 in the sun is to be exothermic one must have @xmath120 , where @xmath121 mev is the binding energy of the deuteron . a very similar condition comes from requiring that the deuteron be stable against the weak decay @xmath122 , namely @xmath123 . in @xcite the binding energy of the deuteron was parameterized by @xmath124 where energies are in mev and the parameter @xmath125 is very sensitive to the model of the internucleon potential . using two different models , @xcite obtained @xmath126 mev and @xmath127 mev . using eq . ( 3 ) , the condition for the deuteron to be stable against weak decay becomes @xmath128 . for @xmath126 mev , this rules out the region above the upper , red , concave - up curve in fig . as noted in @xcite , it may be possible for the hydrogen burning to occur in stars and for nucleosynthesis to proceed even if the deuteron is unstable to weak decay , if its lifetime is sufficiently long . thus , a more clear - cut constraint comes from the condition that the deuteron be stable against the ( much faster ) strong decay @xmath129 . the condition for this is simply that @xmath130 be positive . using @xmath126 mev , this gives @xmath131 mev , which rules out the region above the upper , red , concave - down curve in fig . if the quark masses are too small , then the pion mass becomes small , and the nuclear force becomes sufficiently long - range to bind the diproton " , i.e. the @xmath132he nucleus , as pointed out by dyson long ago @xcite . this would allow hydrogen to burn in stars by the reaction @xmath133he , which is very fast as it does not involve the weak interactions . hydrogen - burning stars would have lifetimes orders of magnitude shorter than the billions of years presumably required for complex organisms to evolve . in @xcite it was estimated that if @xmath134 mev , the diproton is bound . this rules out the region below the lower , red , concave - down curve in fig . 4 . altogether , then , we see that the viable region is squeezed down to the remarkably small shaded green area bounded by the four red curves in fig . 4 . the actual value of @xmath135 observed in our domain is right in the middle of it , as shown in fig . ( for perspective , recall that this shaded green region is smaller than the dot in fig . 3 , which shows the whole parameter space . ) exotic possibilities open up if either @xmath137 and @xmath138 is small . first we shall consider the situation where @xmath22 is so small that @xmath137 becomes comparable to or smaller than @xmath49 . ( of course , since the yukawa coupling ratios are fixed , @xmath48 becomes about two orders of magnitude less than @xmath49 . ) we shall call these light-@xmath136-quark domains " . in such domains the light quarks are @xmath0 , @xmath1 and @xmath136 , and there is a baryon octet made up of valence quarks of these three flavors . this octet , shown in table i , contains baryons of charge 0 , + 1 , and + 2 , but none with negative charges . therefore these domains are similar to ours in the following respects : ( 1 ) all nuclei have @xmath139 . ( 2 ) if a nucleus has large @xmath100 , then its charge @xmath101 is also large , since fermi energy tends to equalize the number of baryons of different species . ( 3 ) for large @xmath101 nuclei , coulomb energy becomes important and gives an upper bound to the size of stable nuclei . @xmath140 * table i * in the light-@xmath136-quark part of parameter space one can distinguish three regions , which are shown in fig . 5 : ( a ) for large enough @xmath137 , the lightest baryons in the octet are those without valence @xmath136 quarks , namely @xmath141 and @xmath142 . for nuclei composed of @xmath141 and @xmath142 the analysis of the previous subsection applies . consequently , case ( a ) does not give any anthropically allowed regions of parameter space except those we have already discussed . ( b ) for small enough @xmath137 , the lightest baryons in the octet are those with only valence @xmath136 and @xmath1 quarks , namely @xmath143 and @xmath144 ( i.e. @xmath145 and @xmath146 , respectively ) . as we shall see , domains of this sort seem unlikely to be viable . ( c ) for intermediate values of @xmath137 , the lightest baryons in the octet are @xmath141 and @xmath143 . some domains of this sort may be viable . the sufficient condition for case ( a ) is that @xmath147 , which gives @xmath148 , where the last term is the electromagnetic contribution to the splitting of @xmath143 and @xmath142 . we may neglect @xmath48 in this expression , since it is less than a percent of @xmath137 . we take @xmath149 mev . ( note that the observed splitting between @xmath143 and @xmath150 is 0.3 mev , so that if the quark mass contribution is @xmath151 mev , it implies an electromagnetic contribution of 6.3 mev . of course , the wavefunctions for the charmed baryons would be different in the light-@xmath136-quark domains . ) the condition is then @xmath152 mev . this is the region to the right of the lower red curve in fig . 5 . -quark " region of parameter space . the green shaded region may be viable . the dot is our domain . ] in this region , the analysis we gave in section 2 applies , and one expects that only anthropically allowed region is that shown in fig . 4 , which includes our domain ( indicated by the blue dot in fig . 5 ) . the condition for case ( b ) is that @xmath153 , which gives @xmath154 , or approximately @xmath155 mev . this is the region to the left of the upper red curve in fig . this region seems unlikely to be viable since all stable nuclei would have @xmath156 . ( it is easily seen that an isolated @xmath141 will beta decay to @xmath143 , as will a @xmath141 bound with neutral baryons . ) consequently , there would be no chemical analogue of hydrogen . the most interesting case is ( c ) , which lies in between the two red curves in fig . the lightest two baryons are @xmath141 and @xmath143 . in such a domain , the only @xmath157 nucleus consists of a single proton , i.e. @xmath109h . in order for hydrogen to be stable , it is required that the decay @xmath158 be energetically forbidden , which gives the condition @xmath159 . this rules out the region above the blue , concave - up curve in fig . a second condition is that diprotons not be bound , as otherwise hydrogen - burning stars would burn too fast , as noted above . this gives the bound that @xmath160 mev , as given previously . this rules out the region below the nearly horizontal blue line that barely misses our domain in fig . a third condition is that the @xmath161 nucleus composed of a @xmath141 and a @xmath162 be bound . let us call this analogue of the deuteron @xmath163 . the potential between the @xmath141 and the @xmath162 at large distances is given by one-@xmath164-meson exchange rather than one - pion - exchange , and is thus controlled by @xmath165 rather than by @xmath166 . ( 3 ) , the binding energy of @xmath163 can be written @xmath167 $ ] , where @xmath168 is the coulomb interaction energy of the baryons in the @xmath163 . the condition that @xmath163 be bound is then @xmath169 , where energies are in mev . if @xmath126 mev and @xmath170 mev , this rules out the region above the blue , concave - down curve in fig . the combination of the three constraints leaves only the very small triangular region shaded green in fig . 5 . in fact , since the three curves depend on quantities that are not well known , such as @xmath125 , @xmath168 , @xmath171 , @xmath172 , it is not really clear that _ any _ area is actually enclosed by them . furthermore , even if there is , it does not necessarily mean that this area is an anthropically allowed region of parameter space . it only means that three of the more obvious disasters for viability do not happen . one other potential problem for viability is that nucleosynthesis may be inhibited in such domains because of large coulomb barriers , due to the fact that nuclei are composed of charge-1 and charge-2 baryons . a much more serious potential problem for viability is that @xmath173 is about four orders of magnitude larger in this region of parameter space than it is in our domain . this would make the @xmath174 reaction " @xmath175 proceed in hydrogen - burning stars very much faster than the ordinary @xmath174 reaction does in our domain . that would presumably be disastrous for the same reason that the diproton being bound would be . as we shall see in section 3 , where we discuss probability distributions in parameter space under some simple and reasonable assumptions , these potentially viable light-@xmath136-quark domains occur less commonly in the universe than domains like our own . therefore , unless a typical domain of this sort is _ more _ viable than ours , the existence of such domains should not affect the anthropic explanation of the observed values of @xmath48 and @xmath49 that we are exploring here . we now consider the situation where @xmath23 is so small that the mass of the @xmath176 quark is comparable to or smaller than @xmath48 . ( of course , then @xmath177 , since we are assuming that the ratios of yukawa coupling are fixed . ) we will call these light-@xmath176-quark domains " . in such domains , the lightest three quarks are @xmath0 , @xmath176 , and @xmath1 , as in our domain , but the corresponding flavor @xmath178 is a very good symmetry , since @xmath138 is small . consequently , there is the familiar baryon octet , shown in table ii , but it is more nearly degenerate than in our domain . @xmath179 * table ii * there are three cases to consider for light-@xmath176-quark domains ( just as there were for light-@xmath136-quark domains ) . these are shown in fig . ( a ) for large enough @xmath138 , the lightest two baryons are the two non - strange baryons @xmath141 and @xmath142 , as in our domain . the condition for this is that @xmath180 , which gives @xmath181 . this is the region above the upper red curve in fig . ( b ) for small enough @xmath138 , @xmath1 is heavier than both @xmath176 and @xmath0 and the lightest two baryons are @xmath182 and @xmath183 , which have no valence @xmath1 quarks . the condition for case ( b ) is that @xmath184 , which gives @xmath185 . this is the region below the lower red curve in fig . ( c ) for intermediate @xmath138 , the lightest two baryons would be @xmath142 and @xmath182 . this is the region between the two red curves in fig . 6 . for case ( a ) , where the lightest baryons are @xmath141 and @xmath142 , as in our domain , the analysis of viability we presented earlier applies , and so the only anthropically allowed region is that shown in fig . 4 , which includes our domain ( indicated by the blue dot in fig . 6 ) . -quark region of parameter space . some of the region below the green line and to the right of the rightmost red curve may be viable . the dot is our domain . ] case ( c ) is interesting because it might lead to domains in which the @xmath182 plays the role that the proton plays in our domain . for example , a hydrogen atom " in such a domain would consist of a @xmath182 and an @xmath186 . ( one would expect antileptons to predominate in such domains because of charge neutrality . ) there are at least three constraints that @xmath48 and @xmath138 must satisfy in case ( c ) . first , the @xmath182 in the nucleus of the hydrogen " atom must be stable against beta decay into @xmath187 . this gives the constraint @xmath188 . this rules out the region to the left of the blue curve that is concave to the right in fig . the second constraint is that the @xmath142 and @xmath182 must be able to bind to make the analogue of the deuteron . the long - distance part of the potential between these nucleons is controlled by the mass of the kaon , and thus by @xmath189 . thus the condition for the @xmath142 and @xmath182 to bind is @xmath190 mev , if @xmath126 mev . this rules out the region above the blue curve that is concave to the left in fig . the third constraint is that the di-@xmath182 not be bound ( the analogue of the diproton not being bound in our domain ) . this gives @xmath191 mev , which rules out the region above the nearly horizontal blue line in fig . altogether , there is no allowed region left for case ( c ) . however , this depends on the value of @xmath125 , and other parameters that are not precisely known . if @xmath125 is less than about 3 mev , there may be a small region that satisfies the three constraints . of course , even if there were , that hardly proves that such a region is anthropically allowed ; it merely means that some obvious disasters for viability are avoided . we turn now to case ( b ) . in this case the lightest baryons are @xmath182 and @xmath183 , and @xmath182 is the lighter of the two . the only obvious constraint is that the analogue of deuterium ( i.e. a bound state of @xmath182 and @xmath183 ) be allowed to form . for @xmath126 mev , this gives the constraint @xmath192 mev . this rules out the region above the horizontal green curve in fig . . there may be some anthropic upper bound on @xmath48 , coming from the fact that very large @xmath48 corresponds to very small @xmath173 ; however , that is not clear . if @xmath22 is superlarge and therefore @xmath193 is supersmall , rather strange situations are possible . even though @xmath173 is supersmall , the decay rates of @xmath1 , @xmath136 , and @xmath2 quarks , being of order @xmath194 will be very fast , and the only quarks around will be @xmath0 , @xmath176 , and @xmath3 . the amplitudes for transitions of @xmath0 , @xmath176 , and @xmath3 into each other will be suppressed by a factor of @xmath195 and therefore , in effect , there will be separate conservation of @xmath0-number , @xmath176-number , and @xmath3-number . ( sphaleron processes will , of course also be highly suppressed . ) the story is similar for the leptons ; there will be separate conservation of @xmath196 , @xmath197 , and @xmath198 . in this case , much depends on the details of primordial lepto / baryogenesis . it is possible that lepto / baryogenesis leads to asymmetries for @xmath0 , @xmath176 , and @xmath3 quarks that differ in sign . for example , there could be an excess of @xmath0 over @xmath199 , but of @xmath200 over @xmath176 and @xmath201 over @xmath3 . in that case , ordinary " hadronic matter would be made up of @xmath0 , @xmath200 , and @xmath201 . in particular , there would be stable @xmath202 , @xmath203 mesons , @xmath182 baryons , and positively charged antibaryons consisting of valence @xmath200 and @xmath201 . these could capture leptons of positive or negative charge ( as the case may be ) and make atoms . in the same way , there could be asymmetries of different sign for different flavors of lepton . for instance , there could be an excess of @xmath204 and @xmath205 but of @xmath206 . thus there could be atoms consisting of an @xmath204 orbiting a @xmath206 . clearly , all sorts of rich and strange possibilities for chemistry might exist . and if the chemistry is rich , then it is plausible that life might be possible . however , in order to have any nuclei with @xmath207 , one still needs two baryons to be able to bind . there is still , therefore , an anthropic upper bound on @xmath208 for case ( b ) , as discussed above . as is clear from the shape of the probability contours shown in fig . 3 and derived in section 3 , this will put an upper bound on the probability density of viable small-@xmath176-quark domains . in sum , there are regions of parameter space with @xmath138 of order a few mev , and in particular case ( b ) , where we have found no very obvious convincing argument against viability . there may also be a small anthropically allowed region where @xmath137 is of order a few mev . as we will see in section 3 , for certain plausible assumptions about the probability distribution in the @xmath209 plane , the possibility of such exotic viable light-@xmath176-quark or light-@xmath136-quark domains would not affect the anthropic explanation of the observed values of @xmath48 and @xmath49 being explored in this paper . in domains where @xmath62 the vacuum expectation value of @xmath210 is of order @xmath211 and therefore extremely small compared to its value in our domain . this means , of course , that in these domains the quark and lepton masses are correspondingly small . ( for example , if @xmath212 and @xmath31 is the same as in our domain , then the electron mass is only about @xmath213 ev . ) this has important consequences for life based on chemistry . the reason is that the energies of typical atomic and molecular transitions are of order , or smaller than , @xmath214 , and therefore so are the temperatures needed for biological processes to occur . ( that is why in our domain the temperature needed for life is about 300 k @xmath215 . ) it follows that before life based on chemistry can appear the universe has to have cooled down to a temperature much smaller than @xmath214 . because @xmath216 is so small in positive-@xmath4 domains , it was argued in @xcite that by the time the universe cools sufficiently for chemistry - based life to exist in those domains various disasters might already have occurred , such as all baryons having decayed away or all stars having burned out . those arguments depended crucially on the assumption that universe expands as a power of @xmath2 . however , it now appears that after dark energy comes to dominate the universe will expand exponentially in @xmath2 . once the expansion becomes exponential , it does not take long for the universe to reach extremely small temperatures . for example , consider a domain where @xmath217 ev . if there were no dark energy , it would take such a domain about @xmath218 years to reach temperatures of order @xmath219 . on the other hand , if there were the observed amount of dark energy , it would take only about @xmath220 years . the difference is even more dramatic if @xmath4 is larger . in a typical domain where @xmath221 , it would take about @xmath222 years to reach biochemical temperatures if there were no dark energy , but only about @xmath223 years if there is the amount of dark energy we observe . the existence of dark energy means that the specific astrophysical and cosmological anthropic arguments used for the positive-@xmath4 case in @xcite are not valid . however , there are other anthropic arguments based on astrophysical and cosmological considerations that become possible precisely because of the existence of dark energy , as we will now see . for the rest of the paper we will assume that the dark energy has @xmath224 and will refer to it as the cosmological constant " . our arguments should not be qualitatively affected if @xmath225 is not exactly @xmath226 . it seems reasonable to suppose that life requires the existence of planets upon which to evolve . a planet is supported against gravitational collapse by the electrons , and has therefore a typical density that goes as @xmath227 , where @xmath100 is the average number of baryons per nucleus . a planet will be ripped apart by the inflation caused the cosmological constant , unless @xmath228 if @xmath229 is truly a constant of nature , so that it the same in every domain , then its measured value of about @xmath230 means that planets can exist only if @xmath231 ev . for @xmath31 having the same value as in our domain , this means that @xmath4 has to be less than about ( 300 gev)@xmath132 . this is significantly stronger than the bound given in eq . ( 6 ) of @xcite ( a bound that came from not having baryons decay away before the universe cooled to temperatures where chemical life could exist ) . however , it is more interesting to suppose that @xmath229 , as well as the higgs mass parameters , varies among domains . indeed , this comports with the idea that all the dimensionful parameters of the lagrangian are determined anthropically . in that case one has to consider simultaneously the probability distributions of @xmath4 , @xmath31 and @xmath229 . this will be done in section 4 . it will be seen that the viability of a significant part of the positive @xmath4 parameter space is left unaffected by the planet argument " . however , most of that part of parameter space can be anthropically ruled out by a second argument , which is based on structure formation . that second argument assumes that structure can not begin to form until recombination occurs . if by the time of recombination the density of the universe is already dominated by the cosmological constant , then structure will never form . the recombination temperature @xmath232 is about @xmath233 , so the constraint is that structure can not form unless @xmath234 this appears to be a much more stringent constraint than the one in eq . however , it is not clear that it can be used for @xmath216 smaller than about 20 ev . the point is that , if @xmath235 ev , then when matter begins to dominate over radiation ( at @xmath236 ev ) the electrons will still be almost relativistic , and the thermal number density of electrons and positrons will be larger than that of the baryons by a factor of about @xmath237 . there will then be of order @xmath238 electrons and positrons for each baryon . the charges of the baryons will be debye screened , and it would seem that baryons can already begin to condense at that point , since @xmath239 would not have a significant effect on the density of the leptons and photons . we will therefore only use eq . ( 5 ) when @xmath240 ev . the anthropic bounds coming from eqs . ( 4 ) and ( 5 ) will be analyzed in section 4 . since anthropic explanations of relations among parameters are based on probabilities , one must be able to make reasonable hypotheses about probability distributions within the parameter space or landscape " . the parameters that we are assuming to vary among domains of the universe are those in the higgs mass - squared matrix @xmath25 . this matrix acts in the space of @xmath242 . it is convenient to parameterize this matrix as follows @xmath243 first , we will assume that the four real parameters @xmath244 ( which have dimensions of mass squared ) are naturally " superlarge ( either @xmath245 or @xmath246 ) and of the same order . call their natural scale @xmath247 . the simplest assumption is that the four @xmath244 are independent and that each has a flat probability distribution in the interval @xmath248 $ ] . ( however , symmetry may somewhat suppress the off - diagonal elements of @xmath64 . we will discuss the effects of this later . ) we will take @xmath249 to be positive . ( if it is negative and very large the weak interactions are broken at very large scales , whereas if it is negative and small then breaking the weak interactions at small scales would require that @xmath250 also be small , which is an extra fine tuning . ) it is convenient to think of the parameters @xmath251 , @xmath252 , and @xmath250 as forming the components of a vector @xmath253 in a three - dimensional space . as these three parameters have been assumed all to have flat probability distributions in the interval @xmath248 $ ] , the probability distribution is rotationally invariant in the space of @xmath253 . we shall denote the larger eigenvalue of @xmath25 by @xmath254 and the smaller eigenvalue by @xmath4 . ( by our assumption about @xmath249 , @xmath255 is positive . ) since @xmath254 is generally much larger than @xmath4 , the heavier higgs doublet decouples ; so the parameter @xmath4 is in effect just the @xmath4 of the standard model and controls the breaking of the weak interactions . because the cases of anthropic interest are those where @xmath256 , we may write @xmath257 @xmath258 where @xmath30 is the angle between the mass basis of the higgs doublets @xmath259 and the original flavor basis @xmath260 ( i.e. @xmath261 and @xmath262 ) , and @xmath263 is the angle between @xmath253 and the @xmath250 axis . there are two cases to consider , @xmath75 and @xmath62 . in the case of negative @xmath4 , the vacuum expectation value that breaks the weak interactions is @xmath264 @xmath265 , where @xmath125 is the magnitude of @xmath253 , and @xmath266 is the quartic self - coupling of the lighter higgs doublet . then we may write @xmath267 @xmath268 , and similarly @xmath269 @xmath270 . defining @xmath271 and recalling the definition of @xmath272 in eq . ( 2 ) , one has @xmath273 since , by our assumptions , the probability distribution is rotationally invariant in the space of the vector @xmath253 , one may write the probability distribution in spherical coordinates @xmath274 in that space : @xmath275 where @xmath276 is the appropriate normalization factor . we would like to compute the probability distribution in the space of @xmath48 and @xmath49 . thus we write @xmath277 , and compute the jacobian for the transformation of coordinates : @xmath278 this yields the result @xmath279 > from eq . ( 9 ) , one has @xmath280 and @xmath281 this yields @xmath282 for the regions of parameter space that are anthropically interesting ( i.e. where life is likely to be possible ) one has @xmath283 , so that the previous equation can be written @xmath284 . one then simply integrates over the variable @xmath249 to get @xmath285 it is more interesting to express the probabilities in terms of the logarithms of the quark masses . defining @xmath286 , @xmath287 , and @xmath288 , one has @xmath289 the contours of constant probability given by this expression are plotted as the concave - up curves in fig . 3 . note the shift represented by the fact that @xmath290 depends on @xmath291 , where @xmath292 . this shift means that domains where @xmath293 and @xmath294 are just as likely to occur as domains like ours . the reason for this is simple . in our domain , one has @xmath295 , even though @xmath296 . this means that @xmath297 , so that @xmath298 . reducing @xmath48 from its value in our domain ( @xmath171 ) while holding @xmath49 fixed , means reducing @xmath22 while holding @xmath23 fixed . this has the effect of reducing @xmath55 and thus @xmath4 , which increases the fine tuning . however , it also has the effect of making the ratio @xmath299 closer to one , which _ reduces _ the fine - tuning . ( the factor of @xmath300 in @xmath290 prefers equal values of @xmath22 and @xmath23 . ) these effects cancel , so that it does not cost probability " to move to the left in fig . 2 from the point representing our domain down to values @xmath301 . decreasing @xmath48 more than that does cost probability , however , because it means making @xmath55 and @xmath4 more fine - tuned without any compensating effect from @xmath31 , which indeed starts to move away again from 1 towards 0 . thus , to explain why we do nt observe a value of @xmath48 as small as @xmath302 , one can not argue that such domains are less common . rather , one has to argue that they are less viable . in fact , we have presented such arguments in section 2 . what makes the region of parameter space with @xmath303 less viable is a combination of constraints coming from reactions in hydrogen - burning stars like the sun , as one sees from fig . if @xmath48 is even slightly smaller than @xmath171 , it either makes the @xmath174 reaction @xmath119 endothermic ( since protons become lighter relative to neutrons ) , or it makes the diproton a bound state , allowing the reaction @xmath304 @xmath132he , or it has both effects . another effect that is probably deleterious to the chances of life is that as @xmath22 gets smaller @xmath55 gets smaller , as mentioned , and so @xmath173 gets larger . this would make the @xmath174 reaction go much faster ( since it involves the weak interaction ) , presumably making stars like the sun burn much faster . however , we have not calculated this effect , since the reaction @xmath304 @xmath132he being allowed is an even more drastic effect . if @xmath4 is positive , then the light higgs doublet @xmath56 only acquires a vacuum expectation value due to quark - antiquark condensation at @xmath305 , the chiral - symmetry - breaking scale of qcd . specifically , the lighter higgs doublet has the coupling @xmath306 . after qcd chiral symmetry breaking this produces a linear term in the effective potential for @xmath56 that is of the form @xmath307 . since we have assumed that @xmath308 , the coefficient @xmath309 is a slowly varying function of @xmath30 that is of order 1 . we will denote this @xmath310 . then instead of eq . ( 9 ) one has @xmath311 thus , the jacobian is given not by eq . ( 11 ) but by @xmath312 there may be values of @xmath30 for which @xmath313 due to an almost exact cancelation between the @xmath314 term and the @xmath315 term . however , that simply makes @xmath48 and @xmath49 small . since smaller values of @xmath48 and @xmath49 are the the most probable ones anyway in the case of @xmath53 , the possibility of such cancelations has virtually no effect on the probability distribution we shall derive . thus we will get qualitatively the right behavior for @xmath316 if we treat @xmath317 as a constant , which we will now do for simplicity . then following the same steps as before one ends up with @xmath318 and , in terms of the variables @xmath319 and @xmath320 , @xmath321 the probability contours for both positive and negative @xmath4 that we have just derived have been displayed in fig . 3 . for negative @xmath4 the probabilities increase upward and to the left ( i.e. towards larger @xmath48 and @xmath49 ) , whereas for positive @xmath4 the probabilities increase downward and to the left ( i.e. toward smaller @xmath48 and @xmath49 ) . the directions of increasing probability are shown by the arrows and the contours with equal probability to our own domain are shown by the red curves . we see from this that the potentially viable light-@xmath176 quark domains " and light-@xmath136-quark domains " are less common than domains like our own . in deriving the probability distributions given in eqs . ( 16 ) and ( 20 ) , we assumed that all the elements of the higgs mass matrix @xmath25 have the same natural scale @xmath247 . however , the natural flavor conservation " @xcite pattern of yukawa couplings that we have assumed in eq . ( 1 ) ( whereby @xmath15 couples to up - type quarks and @xmath16 couples to down - type quarks and charged leptons ) is not technically natural " unless there is a symmetry @xmath322 under which @xmath15 and @xmath323 transform differently . the off - diagonal elements of @xmath25 necessarily break this symmetry . if @xmath322 is a global symmetry , then one can simply regard the @xmath324 term @xmath325 as a soft breaking of @xmath322 . in that case , the pattern of yukawa couplings in eq . ( 1 ) is still technically natural . it would be reasonable to assume that the natural scale of @xmath326 was the same as that of @xmath327 and @xmath328 . on the other hand , one might imagine that @xmath322 is spontaneously broken . for example , suppose @xmath326 arises from a @xmath322-invariant term @xmath329 , where @xmath330 has a potential @xmath331 . then it would be consistent with our approach to say that @xmath332 scanned and had a natural scale @xmath333 . in this scenario , there are two equally probable cases , @xmath334 and @xmath335 . for @xmath336 , the natural scale of @xmath337 is @xmath338 . if @xmath339 is small , the probability distribution @xmath340 is suppressed when @xmath22 and @xmath23 are comparable ( more precisely , when they are within a factor of @xmath339 of each other ) , i.e. near the line @xmath341 in fig . 3 . this would not qualitatively affect our analysis unless @xmath339 were very small ( less than about @xmath342 ) , in which case the potentially viable small-@xmath136- quark and small-@xmath176-quark regions would start to have probability densities rivaling that of our domain . if @xmath334 , then @xmath343 and quite a different situation results . a single fine - tuning to make the weak scale small would leave either @xmath344 or @xmath345 of order @xmath346 . if @xmath347 , the @xmath1 , @xmath136 , and @xmath2 quarks would have negligible masses , while the down quark masses would be set by the scale of weak interaction breaking . we have not investigated the viability of such domains . in our discussions of probability distributions , we have not considered the renormalization group running of @xmath266 and @xmath4 from the scale @xmath58 down to the scale @xmath55 . this would affect the probability distributions by factors of order 1 . however , it should not qualitatively affect our discussion . in section 2.4 , two anthropic constraints on the relation between the mass of the electron and the cosmological constant were discussed . these eliminate as non - viable the great majority of the positive-@xmath4 parameter space . however , in order to express these as anthropic constraints on the higgs mass parameters ( or equivalently on @xmath349 and @xmath287 ) one must do an analysis that simultaneously takes into account the probability distributions of @xmath319 , @xmath350 and @xmath229 . that is done in this section . both of the anthropic constraints on @xmath229 discussed in section 2.4 and given in eqs . ( 4 ) and ( 5 ) require @xmath229 to be less than bounds that are proportional to @xmath351 and thus to @xmath352 . if we assume that the probability distribution of @xmath229 among domains is flat in the interval @xmath353 $ ] , then the probability distribution given in eq . ( 20 ) gets modified by a changed normalization factor and by an extra factor of @xmath352 to account for the tuning of the cosmological constant required to satisfy eq . ( 4 ) or ( 5 ) : @xmath354 the prime on this probability density indicates that this is a conditional probability : it is the probability density for being at @xmath355 subject to the condition that @xmath229 has been tuned to a small enough value to satisfy either eq . ( 4 ) or eq . if the condition is that eq . ( 4 ) be satisfied , we will call this @xmath356 , if the condition is that eq . ( 5 ) is satisfied , we will call it @xmath357 . the constant of proportionality in eq . ( 21 ) depends on which condition is being enforced . to determine the constants of proportionality , it is convenient to start by finding the points in the positive-@xmath4 and negative-@xmath4 regions of the @xmath358 plane where @xmath359 and where the higgs mass matrix @xmath25 is fine - tuned to the same degree as in our domain ( without any condition on @xmath229 ) . we will call these point @xmath360 and @xmath361 . ( the @xmath330 stands for symmetric , since @xmath359 there , and the subscript tells whether it lies in the @xmath53 or @xmath75 region . ) these points are shown in fig . 7 , where the point representing our domain is labeled d. region ( approximately @xmath362 ) is below and to the left of the blue curve . the arguments of section 2.4 rule out all but the shaded blue region . ] suppose that eq . ( 5 ) implies that at @xmath360 the cosmological constant has to satisfy @xmath363 , where @xmath364 is the value observed in our domain , i.e. that @xmath229 must be fine - tuned at @xmath360 to be less than it is in our domain by a factor smaller than @xmath365 . on the other hand , the higgs mass parameters are equally fine - tuned at @xmath360 and in our domain , by definition of @xmath366 . so the conditional probability density @xmath357 at @xmath360 divided by the conditional probability density @xmath367 in our domain ( where @xmath229 is tuned to @xmath368 ) is just given by @xmath365 . therefore , if we define the shifted variables @xmath369 and @xmath370 , we may write @xmath371 note that at @xmath360 , where @xmath372 , the right - hand side of the equation is just @xmath365 . in order to compute the coefficient @xmath365 , it is necessary to determine the value of @xmath373 and substitute it into eq . ( 5 ) . at the point @xmath164 , i.e. in our domain , one has @xmath374 gev , @xmath375 gev ( if we assume @xmath376 ) , and @xmath377 ( 4 gev ) . since @xmath378 , eq . ( 16 ) implies that @xmath379 gev , and therefore @xmath380 gev ; so that @xmath381 gev ) @xmath382 mev . on the other hand , at the point @xmath360 , the electron mass is given by @xmath383 . but by definition , @xmath384 and therefore @xmath385 gev . taking @xmath386 we have @xmath387 ev . substituting this into eq . ( 5 ) , one has @xmath388 . in other words , @xmath389 the constraint on @xmath229 coming from eq . ( 5 ) only applies if @xmath240 ev . the region with @xmath240 ev lies above the horizontal straight red line in fig . the region of positive @xmath4 corresponds to the area below and to the left of the blue curve in fig . ( this curve arises from the fact that for very small positive @xmath4 , the value of @xmath390 is set by the competition between the quartic term @xmath391 and the linear term for @xmath210 coming from the qcd condensates of @xmath392 . this gives a maximum value for @xmath393 in the positive-@xmath4 region . ) it is easy to see from eqs . ( 22 ) and ( 23 ) that the entire region satisfying both @xmath53 and @xmath240 ev ( shaded gray in fig . 7 ) is much more fine - tuned than our domain , i.e. @xmath394 . turning now to the condition that planets be able to form ( eq . ( 4 ) ) , the calculation is completely parallel to the one just done . to find @xmath395 one substitutes @xmath373 into eq . ( 4 ) and divides by @xmath368 , which gives @xmath396 and @xmath397 where @xmath100 is the average atomic weight of a nucleus in the planet . it is easy to see that the region to the left of the vertical orange line or below the slanted orange line are more fine - tuned than our domain , i.e. @xmath398 . the shape of the orange line arises from the fact that when @xmath399 , the expression in eq . ( 24 ) is approximately proportional to simply @xmath400 , whereas when @xmath401 , the expression in eq . ( 24 ) is approximately proportional to @xmath402 . the two arguments given in section 2.4 have excluded ( as being more fine - tuned than our domain ) the whole of the positive-@xmath4 region except for the region shaded light blue in fig . 7 . however , that is not good enough . it is necessary to exclude the _ entire _ positive-@xmath4 region either as not viable or as more improbable than our domain , if the anthropic explanation of the observed values of @xmath48 and @xmath49 is to be tenable . other arguments that were suggested in @xcite hold some promise of being able to do this . for example , as noted there , it might be that the peculiarities of nuclear physics in the positive-@xmath4 region may lead to runaway nucleosynthesis . however , the situation for @xmath53 has many unclear aspects , and more thought must be given to it . we have investigated the possibility that the entire mass matrix of the higgs fields in a two - higgs - doublet model varies among domains of a multiverse " . this means that both the weak scale @xmath403 and @xmath404 vary among domains . making plausible assumptions about probability distributions of parameters in the landscape " and about the requirements for life , we have found that almost all of the @xmath209 space is either anthropically non - viable or much more fine - tuned ( and therefore more rare in the multiverse ) than our domain . the potentially viable regions are islands in parameter space . one of these islands is ours , and it is a very small island indeed , as seen in figs . 3 and 4 . if this is the only island of viability , there is the possibility of accounting for two otherwise not easily understood facts about the world we observe , namely that the weak scale is comparable to the strong scale , and that @xmath8 . whether this is in fact the only island of viability is not yet clear . there are possible islands of viability in the light-@xmath136-quark " and light-@xmath176-quark " regions . however , even if islands of viability do exist in those parts of parameter space they are less probable in the landscape than the island of viability in which our domain is located , according to the analysis of section 3 . more significant is the possibility of an island of viability in the @xmath53 region of parameter space . an interesting new twist has been given to the discussion of the positive-@xmath4 case by the discovery of dark matter / cosmological constant . this discovery ties together the anthropic analysis of the higgs mass given in @xcite and @xcite and the anthropic analysis of the cosmological constant given in @xcite . we have seen that some of the @xmath53 constraints are invalidated by the existence of dark energy , but other and equally strong constraints that depend on the existence of dark energy come into play . however , the constraints we have discussed are not by themselves enough to rule out the entire @xmath53 region as non - viable . this seems to be the main loophole at present in the attempt to explain @xmath33 and the value of the weak scale anthropically . however , it seems probable that much stronger arguments exist for the case of @xmath53 , and that the entire positive-@xmath4 region of parameter space is non - viable . we have assumed that only the dimensionful parameters of the lagrangian scan " , i.e. vary in the multiverse , and that only they are anthropically tuned . this seems like a comparatively conservative assumption , since the dimensionful parameters of our current theories are the least well understood , are the most ( apparently ) fine - tuned , and are the most amenable to anthropic explanation . however , obviously , other assumptions are possible . the question for us is whether there exists at least some set of plausible assumptions under which the observed relation @xmath33 can be anthropically explained . we acknowledge useful conversations with d. seckel , ilia gogoladze , and tian - jun li . this research was supported by the doe grant de - fg02 - 91er40626 . 999 v. agrawal , s.m . barr , j.f . donoghue and d. seckel , _ phys.rev . _ * d57 * , 5480 ( 1998 ) ; _ ibid . lett . _ * 80 * , 1822 ( 1998 ) . c. hogan , _ rev . phys . _ * 72 * , 1149 ( 2000 ) ; s. weinberg , hep - ph-0511037 , opening talk at the symposium expectations of a final theory " , published in _ universe or multiuniverse ? _ , ed . b. carr , cambridge university press , 2006 . peccei and h.r . quinn , _ phys . lett . _ * 38 * , 1440 ( 1977 ) . m.a.b . beg and h .- s . tsao , _ phys . lett . _ * 41 * , 278 ( 1978 ) ; r.n . mohapatra and g. senjanovic , _ phys . lett . _ * 79b * , 283 ( 1978 ) ; h. georgi , had . j. * 1 * , 155 ( 1978 ) ; s.l . glashow and s. weinberg , _ phys . rev . _ * d15 * , 1966 ( 1977 ) . barr and i. dorsner , _ phys . lett . _ * b566 * , 125 ( 2003 ) . babu and s.m . barr , _ phys . lett . _ * b381 * , 202 ( 1996 ) . c. hogan , _ rev . phys . _ * 72 * , 1149 ( 2000 ) ; _ phys.rev . _ * d74 * , 123514 ( 2006 ) . dyson , _ scientific american _ , * 225 * ( sept . ) , 51 ( 1971 ) ; c. hogan , _ rev . mod . phys . _ * 72 * , 1149 ( 2000 ) s. weinberg , _ phys . lett . _ * 59 * , 2607 ( 1987 ) .

we suggest the statistical spectral theory of oscillations in quasi - optical cavity resonator filled with random inhomogeneities . it is shown that inhomogeneities in the resonator result in intermode scattering leading to the shift and broadening of spectral lines . the shift and broadening of each line essentially depends on frequency distance to adjacent spectral lines . with increasing the distance the influence of inhomogeneities sharply reduces . the solitary spectral lines which have the distance to the nearest lines quite large is slightly changed due to small inhomogeneities . owing to such selective influence of inhomogeneities on the spectral lines the effective spectrum rarefaction appears . both the shift and broadening of spectral lines as well as spectrum rarefaction in quasi - optical cavity millimeter wave resonator were detected experimentally . we found out that inhomogeneities result in stochastization of the resonator spectrum in that mixed state appears , i.e. the spectrum acquires both regular and random parts . the active self - oscillator system based on the inhomogeneous quasi - optical cavity millimeter wave resonator with gunn diode was studied as well . the inhomogeneous quasi - optical cavity millimeter wave resonator ( passive and active ) can serve as a model of semiconductor quantum billiard . based on our results we suggest using such billiards with spectrum rarefied by random inhomogeneities as an active system of semiconductor laser .

cond-mat0606740

the problem of electromagnetic wave propagation in random inhomogeneous media has been the issue of the day for several decades . numerous publications are devoted to the analysis of different aspects of this problem ( see [ 1,2 ] and references therein ) . in this scientific area the subject of research is normally the scattering of electromagnetic ( acoustic ) waves in unbounded or partially confined systems ( for example , in waveguides ) which contain random inhomogeneities . researches on wave propagation in open statistically unregulated systems are stimulated by numerous applications for long - distance signal transmission in both radio and optical wave ranges . the problem of electromagnetic oscillations in confined systems with random inhomogeneities has significant meaning as well . at the same time , by now it has not been developed enough both from theoretical and experimental side . it is caused by , in particular , modern theoretical studies of electromagnetic wave propagation and scattering in random inhomogeneous media use assumption about the isotropy and homogeneity scattering medium filled confined systems [ 1,2 ] . therefore , they can not be applied in general to the study of confined systems such as cavity resonators . the experimental study also requires new approach to study electromagnetic oscillations in inhomogeneous cavity resonators in wide wavelength band . in our recent paper [ 3 ] we suggested a new spectral approach to study confined systems with random inhomogeneous . there spectral properties of cavity spherical quasi - optical millimeter wave resonator filled with random sapphire particles were considered . the sapphire particles with dimensions of the order of operating wavelength affect significantly the resonator spectrum . the spherical frequency degeneration is completely removed and the spectral lines thus have chaotic distribution on the frequency scale . the lines became wider and the quality factor is correspondingly decreased . the analogous broadening of spectral lines caused by random inhomogeneities has been detected in refs . [ 4,5 ] , where resonators with random rough boundaries were studied . recently the question about the influence of random inhomogeneities on the resonator spectrum has attracted a great attention in relation to the design of lasers on open micro - resonators [ 6 ] . in such resonators , the whispering - gallery modes with super - high quality factors can be excited . the possibility to realize such quality factors depends essentially on the number of inhomogeneities ( roughness level ) on the resonator boundaries . it should be realized that cavity resonator with extremely small dissipation loss is the almost hermitian system whose oscillations are formed due to the restricted motion of electromagnetic waves . therefore , the resonator spectrum is discrete . random inhomogeneities whether bulk or on the resonator boundaries not introducing additional dissipative loss in general have not to broaden the spectral lines . nevertheless , the broadening of lines in the spectrum effects and its stochastization have been detected in experiment [ 3 ] . in this connection the question arises : what is the physical mechanism of spectral lines broadening and spectrum stochastization in the random inhomogeneous resonator ? the study of this question is one of the reasons of the present paper . another motivation is relevant to the nanoelectonics problem . recently the great emphasis is the study of a new type of nanoelectron systems ( so - called zero - dimensional ones ) . the charge carrier motion in these systems has space restriction in all three dimensions . the peculiar example of zero - dimensional system is a quantum dot ( qd ) . qd is a semiconductor area by the size of order of 10 nm with electron ( hole ) conductivity and restricted from outer area by potential barrier . owing to finite charge carrier motion the energy spectrum in qd is discrete and the number of spectral levels due to small qd size is relatively few . recently an elegant method of qd array implementation has been designed . this method is based on the self - organization effect in strained double gaas heterostructures [ 7 - 9 ] . the usage of qd array as an active medium permits to design lasers with high performance [ 10 - 12 ] . however , the self - organization process of qd array forming is difficult to manage . the random inhomogeneities that usually exist in heterostructure essentially affect on this process . they lead to inhomogeneous broadening of spectral lines [ 13 ] and , correspondingly , to degradation of laser radiation quality . the qd design method based on the self - organization effect is , in fact , an alternative to electron lithography whose level of development does not permit to implement the ordered array of approximately the same qds with small enough dispersion of their sizes . in the present paper we propose another way to design semiconductor laser system . this way presupposes the usage of the same regular microscopic areas ordered array that can be implemented by lithography as an active laser medium . this proposal is based on the following . at present technology of gaas mono - crystal with super high mobility and big length of phase coherence of charge carriers that achieves 10 @xmath0 and more is well - developed . due to that the system of microscopic regular areas with potential barrier on the margin of each of them made from such materials can be implemented . because of big length of phase coherence , carriers will take part in ballistic motion and reflect back from area margins . the motion of charge carriers is similar to dynamics of billiard systems . since this motion is described by shrodinger equation such a billiard system can be characterized as a quantum billiard ( qb ) . due to finite motion of charge carrier , electron spectrum of qb is discrete . at the same time it is quite dense , because of a qb size is much bigger than the wavelength of quasi - particles . this fact makes complicate the qb usage as an active system for the semiconductor laser , because dense spectrum decreases frequency stability of laser radiation . the frequency jumps appear easily at the small deviation of control parameters in the laser resonator with dense frequency spectrum . as a result there is a problem of `` rarefaction '' of the spectrum : _ can the dense qb spectrum be done much sparser without changing of qb geometrical parameters ? _ a quasi - optical cavity resonator has dense and discrete frequency spectrum as well . the maxwell equation describing electromagnetic oscillations in such a resonator coincides with corresponding scalar shrodinger equation at definite conditions . all that gives an opportunity to use the quasi - optical resonators as model objects to study spectral properties of qb . similar modeling using microwave resonators was done earlier for study of the phenomena that is relevant to quantum chaos [ 14 - 17 ] . in the present paper the nature of spectral lines broadening and their shifting are studied in a quasi - optical millimeter wave cavity resonator with random inhomogeneous inside . to clarify the mechanism of broadening and shifting the original statistical spectral theory based on the separation mode technique was worked out . this technique was designed earlier by one of the author for open waveguide systems [ 18,19 ] . using this technique we found out that the mechanism of spectral lines broadening was caused by intermode scattering . at that the largest broadening are subjected to closely set lines . and solitary lines keep high quality factor and intensity in the random inhomogeneous resonator . it results in so - called spectral `` rarefaction '' . the systematic experimental spectral measurements of numerous realizations of quasi - optical random inhomogeneity cavity resonator filling were carried out for theory results implementation . the experimental results proved that the main mechanism of broadening and shifting of spectral line is inter - mode scattering caused by inhomogeneities resonator filling . we detected theoretically predicted effect of `` rarefaction '' of the resonator spectrum . thus , we determined that the influence of inhomogeneities on the resonator spectrum has both , as known , negative nature ( broaden lines ) and a positive one . this positive influence is essential `` rarefaction '' of the spectrum . this fact can be important in qb usage as an active system in a semiconductor laser approach . here we state that the high quality modes are not subjected , practically , to broadening and can satisfy self - excitation laser condition even under big inhomogeneities in an active resonator medium . exactly the stable laser generation can be implemented under one of such modes . another aspect of influence of inhomogeneities on resonator spectrum relates to its stochastization . in the paper we carried out statistical analysis of the resonator spectrum with a different number of inhomogeneities . we found out that the inter - frequency ( if ) interval distribution in the resonator spectrum is close to poisson distribution in a small number of inhomogeneities . this kind of distribution has systems with non - correlated if intervals . the stochastic spectrum part appears at the increase of the number of inhomogeneities . the spectrum is mixed , i.e. contains regular and stochastic parts . we estimated relationship between regular and stochastic spectrum parts by statistical distribution analysis of if intervals depending on the quantity of inhomogeneities in the resonator . we carried out the modeling of semiconductor laser based on qb with random inhomogeneities . for the purpose of this we used a quasi - optical millimeter wave cavity resonator filled with random inhomogeneities and with an inserted gunn diode as an active element . in such a self - oscillation system we studied self - excitation oscillation conditions . we found out that the generation is unstable and multi - frequency exists in an empty resonator near the excitation threshold because of dense spectrum . such generation can be explained by frequency jumps in dense resonator spectrum . these jumps disappear because of `` rarefaction '' of the spectrum if inhomogeneities are inserted into the resonator , and total number of generating frequencies is decreased at a definite range of control parameters deviation . let us consider a cylindrical quasi - optical cavity resonator of radius @xmath1 and height @xmath2 ( see fig.[fig1 ] ) . inner volume of the resonator , @xmath3 , is assumed to be filled with the material having random inhomogeneous permittivity . we will be interested in oscillations that turn into transverse - electrical resonance mode ( @xmath4-mode ) provided that the resonator is filled homogeneously . the vertical ( @xmath5 ) component of the electrical field of this mode is equal to zero . the geometry of the cylindrical quasi - optical cavity resonator . @xmath6 is the resonator side face , @xmath3 is its volume , @xmath2 is the height of the cylinder , @xmath1 is the radius of its base . [ fig1 ] ] according to ref . @xcite , electromagnetic field of the @xmath4 mode can be calculated trough magnetic hertz vector having only one non - zero component , namely , @xmath5-component @xmath7 . we assume that inhomogeneity of permittivity of the resonator infill is small . in this case to define @xmath7 in the inhomogeneous resonator we can use the approximate wave equation @xmath8\pi _ z ( { \bf{r } } ) = 0\ , \ ] ] where all components of the hertz vector , except z - component , are assumed of zero value taking into account small inhomogeneity . in eq.([eq1 ] ) , @xmath9 is the three - dimensional ( 3d ) laplacian , @xmath10 is the complex permittivity whose imaginary part @xmath11 takes phenomenologically into account ohmic loss in the system ; the function @xmath12 describes random space fluctuations of the permittivity around its average value @xmath13 , @xmath14 is wavenumber . in this paper we are interested in the analogy between electromagnetic resonators and solid - state micro - objects , namely , qds and qbs . in the case of classical resonance system , the problem of excitation by a given point monochromatic source is governed by the equation eq.([eq1 ] ) which should be supplied with @xmath15-term in the right - hand side . the equation thus obtained coincides in form with the equation for green function of quantum particles moving in dissipative medium and being subjected to inhomogeneous scalar potential . having this in mind we will carry out all further analysis relevant to the electromagnetic resonator in terms of the dynamic equation for particles in the quantum dot of cylindrical shape . the equation for green function of quantum particles moving in dissipative medium under the influence of random static potential @xmath16 has the form @xmath17g({\bf{r}},{\bf{r } } ' ) = \delta ( { \bf{r } } - { \bf{r}}')\ .\ ] ] here @xmath18 is the dissipative attenuation time whose inverse value has the same physical meaning as the imaginary part of function @xmath19 in eq . taken with minus sign . the potential @xmath16 in the case of electromagnetic system is given by @xmath20 . boundary conditions for the solution to eq.([eq2 ] ) result from the requirement of vanishing tangential components of the electrical field of _ te _ mode on the resonator interface . at side boundary @xmath6 , the neuman condition [ eq : whole_1 ] @xmath21 should be met whereas at end surfaces @xmath22 the dirichlet condition @xmath23 should be satisfied . for studying oscillations spectrum of the resonator with random inhomogeneities , the poles of green function averaged over realizations of the potential @xmath16 from eq.([eq2 ] ) should be determined . this function can be found , for example , from dyson equation . yet there by now do not exist effectual methods for solving this equation in the case of confined multi - dimensional systems . in this paper , we apply for this purpose the original calculation technique which relies on precise separation of modes in an arbitrary confined system , including the disordered one . the method of mode separation was previously developed for solving transport problems in disordered 2d open systems @xcite and then modified for systems of waveguide geometry in three dimensions @xcite . below we set forth adaptation of this technique for systems of closed geometry , in particular , for cavity resonators and quantum dots . let us at first turn to mode representation of the equation eq.([eq2 ] ) using some set of basis functions . the most appropriate for our purposes seems to be the whole set of eigenfunctions of the laplace operator . for the cylindrical resonator shown in fig.[fig1 ] these functions can be factorized to the form @xmath24 where @xmath25 is the radius - vector in cylindrical coordinates , @xmath26 is the vectorial mode index conjugate to that vector . normalized eigenfunctions of `` transverse '' part of the laplacian , which obey boundary conditions eq.([subeq:1 ] ) , are given by [ eq : whole_2 ] @xmath27 & \hspace{1ex}l = 1,2 , \ldots , \quad n = 0 , \pm 1 , \pm 2 , \ldots\ , \notag\ , \end{aligned}\ ] ] where the coefficient @xmath28 has the form @xmath29^{1/2 } \,j_{|n| } \left ( { \gamma _ l^{(|n| ) } } \right)}}\ .\ ] ] the set of coefficients @xmath30 in eqs.([eq : whole_2 ] ) consists of positive zeros of the function @xmath31 which are numbered by index @xmath32 in ascending order . the eigenvalues corresponding to functions eq.([translapleig ] ) are equal to @xmath33 . basis functions of the `` longitudinal '' laplacian , @xmath34 , which meet boundary conditions eq.([subeq:2 ] ) , are given by @xmath35 \\[3 mm ] & q = 1,2,\ ; \ldots \;\ , \notag\end{aligned}\ ] ] the corresponding eigenvalues being equal to @xmath36 . in the basis of functions , eq . takes the form @xmath37 here , @xmath38 is the green function in mode representation , the parameter @xmath39 is the unperturbed `` energy '' of the mode @xmath40 ( the eigenvalue of 3d laplace operator ) , functions @xmath41 are mode matrix elements of the random potential , @xmath42 attention should be drawn to the fact that the _ intramode _ , i.e. diagonal in mode indices , matrix element @xmath43 is separated in eq.([eq7 ] ) from other terms of the sum where thus only matrix elements corresponding to _ intermode _ scattering are left . it was shown in ref . @xcite that such a separation of intra- and intermode effective potentials provides mathematical correctness of the derivation of closed equations for the diagonal components of green matrix @xmath44 and through those components for that matrix integrally . to solve the infinite set of coupled equations eq.([eq7 ] ) is not less intricate problem than direct solution of multi - dimensional differential equation eq.([eq2 ] ) . the problem would be resolved elementary if the resonator modes allowed for their strict separation . normally , modes are easily separable if the resonator has no inhomogeneities . it will be shown below , based on the technics developed in refs . @xcite , that in fact they can be separated even in the case of arbitrarily inhomogeneous resonator . but the cost of this separation in the general case is the appearance in equations for each of the modes of the effective potential , known as _ t_-matrix in quantum theory of scattering @xcite , whose functional structure is much more involved than that of the initial potential @xmath45 . as a starting point for mode separation we introduce unperturbed ( or trial ) mode propagator @xmath46 by omitting in eq.([eq7 ] ) all intermode potentials @xmath47 , @xmath48 the term `` unperturbed '' will thus be used hereupon with respect to intermode potentials , intramode ones being accounted for precisely . by substituting @xmath49 in eq.([eq7 ] ) we obtain linear non - uniform connection of intramode propagator @xmath50 with all intermode green functions having the particular right - hand mode index @xmath40 , @xmath51 assuming then @xmath52 and performing some necessary re - labellings of mode indices we can reduce eq.([eq7 ] ) to the form @xmath53^ { - 1 } g_{{\bm{\nu \mu } } } - \sum\limits_{\scriptstyle { \bm{\nu } } ' \ne { \bm{\nu } } \hfill \atop \scriptstyle { \bm{\nu } } ' \ne { \bm{\mu } } \hfill } { { \cal u}_{{\bm{\nu \nu } } ' } } g_{{\bm{\nu } } ' { \bm{\mu } } } = { \cal u}_{{\bm{\nu \mu } } } g_{{\bm{\mu \mu } } } \\ & ( { \bm{\nu } } \ne { \bm{\mu } } ) \ .\notag\end{aligned}\ ] ] the latter system of interconnected equations can be solved with respect to intermode elements of the green matrix . in the operator form the solution is given by @xmath54 where the linear operator @xmath55 of intermode scattering is introduced , which acts in mode subspace @xmath56 consisting of the whole set of mode indices but the index @xmath40 ; @xmath57 is the projection operator whose action reduces to the assignment of the value @xmath40 to the nearest mode index of any adjacent operator , no matter standing to the left or to the right of @xmath57 . operators @xmath58 and @xmath59 are specified on @xmath56 by matrix elements [ eq : whole_3 ] @xmath60 correspondingly , matrix elements of the operator @xmath61 are given by @xmath62 by substituting intermode propagators in the form eq.([eq13 ] ) into the relationship eq.([eq11 ] ) we obtain the following rigorous expression for intramode propagator @xmath63 , @xmath64 here @xmath65 is the portion of the mode @xmath40 eigen - energy which relates to the intermode scattering . it should be noted that for determining the disordered resonator spectrum the poles of solely diagonal elements of the green matrix suffice to be found . just these elements determine , in view of relationship eq.([eq13 ] ) , all major analytical properties of the whole green function from eq.([eq2 ] ) . below we will analyze the cavity - resonator spectrum in frames of relatively simple statistical model of random potential @xmath45 . we assume the potential @xmath45 to have zero mean value , @xmath66 , and the binary correlation function @xmath67 bearing in mind the forthcoming numerical analysis we will take the function @xmath68 in the form of gaussian exponent , viz . @xmath69 , where @xmath70 stands for the correlation radius . in the case of electromagnetic resonator the normalization constant @xmath71 in eq.([eq18 ] ) is given by @xmath72 , where @xmath73 is the variance of permittivity fluctuations . the pair of selected statistical parameters , namely , the average random potential and its binary correlation function , are sufficient for making detailed analysis of the system in study if the function @xmath12 is the gaussian - distributed random variable . yet these two parameters suffice for carrying out the asymptotically correct analysis even in the case where statistics of the fluctuations is markedly non - gaussian , provided that the potential @xmath74 is in a certain sense small . as it is conventional in condensed matter physics , we will regard the potential to be small and the resulting scattering , correspondingly , weak provided that the scattering rate calculated in born approximation is small as compared to the unperturbed quasiparticle energy , @xmath75 in our case . the smallness of one - fold scattering probability enables one to regard the potential @xmath16 with parametric accuracy as gaussian random process , whatever in fact its distribution is @xcite . consider now the self - energy operator of the mode @xmath76 which consists , in accord with eq.([eq16 ] ) , of two terms , @xmath77 the first term in the right - hand side of this formula vanishes when being averaged whereas the second term does not . its average value can be easily calculated under the presumption of weak scattering . the strength of the intermode scattering is estimated by the norm of the operator @xmath61 entering eq.([eq17 ] ) . assuming this norm to be small as compared to unity and keeping only two main terms in the expansion of the inverse operator in eq.([eq17 ] ) we come to the necessity of performing the averaging not of the exact operator potential @xmath78 but rather of its relatively compact limiting value @xmath79 when calculating the quantity @xmath80 , one can neglect with parametric accuracy the correlation between intramode and intermode potentials . this enables us to average intramode propagator @xmath81 and its envelopes consisting of intermode potentials @xmath82 independently . to average the function @xmath81 it is worthwhile to present it , at first , in the integral form , @xmath83}\ .\ ] ] then the averaging of the integrand in eq.([eq21 ] ) with the aid of continual integration with gaussian functional weight yields @xmath84 } \nonumber \\ & = & \frac{1}{{k^2 } } \sqrt { \frac{\pi } { { 2\sigma ^2 l_{\bm{\nu } } ( r_c ) } } } \exp \left [ { - \frac{{\left ( { k^2 - \kappa _ { \bm{\nu } } ^2 - { i \mathord{\left/ { \vphantom { i { \tau _ d } } } \right.\kern-\nulldelimiterspace } { \tau _ d } } } \right)^2 } } { { 2k^4 \sigma ^2 l_{\bm{\nu } } ( r_c ) } } } \right]\left\ { { 1 - \phi \left [ { \frac{{i\left ( { k^2 - \kappa _ { \bm{\nu } } ^2 - { i \mathord{\left/ { \vphantom { i { \tau _ d } } } \right . \kern-\nulldelimiterspace } { \tau _ d } } } \right)}}{{\sqrt { 2k^4 \sigma ^2 l_{\bm{\nu } } ( r_c ) } } } } \right ] } \right\}\ . \label{eq22}\end{aligned}\ ] ] here @xmath85 is the probability integral @xcite , @xmath86 is the dimensionless correlator of intramode potential @xmath87 . in the case of gaussian correlation function @xmath68 we readily obtain @xmath88\left\langle { { \bf{r}}',{\bm{\nu } } } \right . \left| { { \bf{r}}',{\bm{\nu } } } \right\rangle \\ \nonumber & = & \frac{8}{\pi } c^4_{l_{\bm{\nu } } n_{\bm{\nu } } } \int\limits_0 ^ 1 { \int\limits_0 ^ 1 { dsds ' \exp \left [ { - \frac{{h^2 } } { { 2r_c^2 } } ( s - s')^2 } \right ] } } \sin ^2 ( \pi q_{n_{\bm{\nu } } } s ) \sin ^2 ( \pi q_{n_{\bm{\nu } } } s ' ) \\ & & \times \int\limits_0 ^ 1 { \int\limits_0 ^ 1 { tt'dtdt ' j_{|n_{_{\bm{\nu } } } |}^2 } } ( \gamma _ { l_{\bm{\nu } } ^ { } } ^{|n_{\bm{\nu } } | } t ) j_{|n_{_{\bm{\nu } } } |}^2 ( \gamma _ { l_{\bm{\nu } } ^ { } } ^{|n_{\bm{\nu } } | } t ' ) \oint { d\varphi \exp \left [ { - \frac{{r^2 } } { { 2r_c^2 } } \left ( { t^2 + t'^2 - 2tt'\cos \varphi } \right ) } \right]}\ . \label{eq23}\end{aligned}\ ] ] although the expression eq.([eq22 ] ) is formally exact , it is not quite convenient for the analysis in view of its bulky structure . asymptotical calculations at large and small values of the probability integral argument permit us to make use of much simpler interpolation expression , @xmath89 which is close in form to the initial unperturbed green function eq.([eq10 ] ) . here we have used the notation @xmath90 for the effective scattering frequency . the latter includes both the initial dissipative term @xmath91 and the addendum originating from wave scattering by random inhomogeneities . interpolation of the average trial green function by the expression eq.([eq24 ] ) permits us to interpret the quantity @xmath92 as the dephasing frequency of the mode state @xmath93 , which is related to scattering by the inhomogeneities in the resonator . to make further comparison of theoretical results with experimental data we will consider below the specific limiting case where the inequalities are fulfilled @xmath94 these restrictions enable us to calculate the integrals in eq.([eq23 ] ) asymptotically , thus resulting in the following estimate for the parameter @xmath95 : @xmath96 . the correlator of intermode potentials in eq.([eq20 ] ) can be represented as @xmath97\sin ( \pi q_{n_{\bm{\mu } } } s ) } \sin ( \pi q_{n_{\bm{\mu } } } s')\sin ( \pi q_{n_{\bm{\nu } } } s)\sin ( \pi q_{n_{\bm{\nu } } } s ' ) \nonumber \\ & & \times \int\limits_0 ^ 1 { \int\limits_0 ^ 1 { tt'dtdt'j_{\left| { n_{\bm{\mu } } } \right| } ( \gamma _ { l_{\bm{\mu } } } ^{(\left| { n_{\bm{\mu } } } \right| ) } t ) } } j_{\left| { n_{\bm{\mu } } } \right| } ( \gamma _ { l_{\bm{\mu } } } ^{(\left| { n_{\bm{\mu } } } \right| ) } t')j_{\left| { n_{\bm{\nu } } } \right| } ( \gamma _ { l_{\bm{\nu } } } ^{(\left| { n_{\bm{\mu } } } \right| ) } t)j_{\left| { n_{\bm{\nu } } } \right| } ( \gamma _ { l_{\bm{\nu } } } ^{(\left| { n_{\bm{\mu } } } \right|)}t')\nonumber \\[3 mm ] & & \times \oint\oint { d\varphi d\varphi ' \exp \left\ { { - i(n_{\bm{\mu } } - n_{\bm{\nu } } ) ( \varphi - \varphi ' ) - \frac{{r^2 } } { { 2r_c ^2 } } \left [ { t^2 + t'^2 - 2tt'\cos ( \varphi - \varphi ' ) } \right ] } \right\}}\ . \label{eq26}\end{aligned}\ ] ] asymptotic calculation of the integrals over @xmath98 and @xmath99 results in the following formula @xmath100 where the factor @xmath101 is given by @xmath102 } \ , } { \kern 1pt } \,{\kern 1pt } \sin \left ( { \pi q_{n_{\bm{\mu } } } s } \right)\sin \left ( { \pi q_{n_{\bm{\mu } } } s ' } \right)\sin \left ( { \pi q_{n_{\bm{\nu } } } s } \right)\sin \left ( { \pi q_{n_{\bm{\nu } } } s ' } \right){\kern 1pt } \nonumber \\ & \times \int\limits_0 ^ 1 \int\limits_0 ^ 1 \sqrt { tt'\ , } dt{\kern 1pt } dt'\exp \left [ { - \frac{{r^2 } } { { 2r_c^2 } } \left ( { t - t ' } \right)^2 - \frac{{r_c^2 \left ( { n_{\bm{\mu } } - n_{\bm{\nu } } } \right)^2 } } { { 4r^2 tt ' } } } \right]\notag\\[3 mm ] & \times j_{|n_{\bm{\mu } } |}^ { } \left ( { \gamma _ { l_{\bm{\mu } } } ^{(|n_{\bm{\mu } } | ) } t } \right)j_{|n_{\bm{\mu } } |}^ { } \left ( { \gamma _ { l_{\bm{\mu } } } ^{(|n_{\bm{\mu } } | ) } t ' } \right)j_{|n_{\bm{\nu } } |}^ { } \left ( { \gamma _ { l_{\bm{\nu } } } ^{(|n_{\bm{\nu } } | ) } t } \right)j_{|n_{\bm{\nu } } |}^ { } \left ( { \gamma _ { l_{\bm{\nu } } } ^{(|n_{\bm{\nu } } | ) } t ' } \right ) \;. \label{eq28}\end{aligned}\ ] ] with the results and , separation of real and imaginary parts of the average @xmath103-matrix , @xmath104 , results in the following expressions for shifting and broadening the @xmath76-th resonant level , [ eq : whole_4 ] @xmath105 @xmath106 the factor @xmath101 , as can be seen from eq.([eq28 ] ) , is a real - valued quantity , its absolute value being estimated as @xmath107 . although the sign of this factor can not be uniquely identified in the general case since it contains the dependence on specific indices of modes between which the scattering is carried out , numerical analysis shows that @xmath108 , which corresponds to the broadening of spectral lines . one can adequately estimate both the shift and the broadening of @xmath109-th resonant level by substituting the function @xmath110 in the interpolated form eq.([eq24 ] ) , instead of its exact expression eq.([eq22 ] ) , into the right - hand sides of formulas eq.([eq : whole_4 ] ) . the result in this case reduces to [ eq : whole_5 ] @xmath111 @xmath112 the structure of the summands in eqs.([eq : whole_5 ] ) indicates unambiguously that both the shift and the broadening of each given resonance @xmath113 are mainly provided by its interactions with the adjacent resonances , whose position on the frequency axis are confined to the region limited by order equality @xmath114 . in the case where only one resonance level with , say , frequency @xmath115 proves to fall into the above indicated interval around the @xmath116-th resonance ( we assume the condition @xmath117 of weak intramode scattering to be met ) its contribution to the shift and the width of the @xmath40-th resonance is estimated as [ eq : whole_6 ] @xmath118 @xmath119 here @xmath120 and @xmath121 are the cyclic spectral frequency of empty resonator and that of the resonator filled with random inhomogeneities , respectively , @xmath122 and @xmath123 are relative widths of their spectral lines , the parameter @xmath124 . if several resonances fall into the indicated vicinity of the @xmath40-th resonance simultaneously , they contribute additively to both the level position and width . the above presented theory shows that the filling of a quasioptical cavity resonator with randomly distributed inhomogeneities results in both the random shift of its resonance lines and the increase of their linewidths . it should be noted that formulas eqs.([eq : whole_4 ] ) and were obtained with the use of some procedure of averaging over the ensemble of macroscopically identical resonance systems whose microscopic realizations are different . what this formally means is that the statistical theory can give only qualitative predictions for the experiment . the situation seems to be quite similar to that one faces when studying problems of wave scattering by randomly rough surfaces @xcite . yet the problem of oscillations in the quasi - optical cavity resonator filled with bulk random inhomogeneities differs substantially from the problem of wave scattering by randomly rough surface . the point is that oscillations in cavity resonators become established as a result of multiple transmissions of waves ( with multiplicity of the order of @xmath125 ) through the random infill of the resonator . with this consideration in mind one might expect that under the restriction @xmath126 the characteristics of resonance lines should be effectively self - averaged @xcite . if this is the case , the agreement between theory and experiment may be achieved either in the course of small number of trials or even for one particular realization of the random system . the confirmation or the denial of this assumption may be achieved through the comparison of theoretical predictions and the data of corresponding measurements . to compare the theory and experiment we calculate the spectrum of @xmath4 modes in the quasi - optical cylinder cavity millimeter wave resonator filled with random dielectric inhomogeneities using our theoretical results . at the resonator spectrum calculation the resonator parameters were taken according to the experiment conditions . the spectrum was calculated by solving the excitation task with pointed external dipole . fig.2 shows the influence of random bulk inhomogeneities on the spectrum at the different values of the parameter @xmath127 calculated using eq.(30 ) . as an example we selected the spectrum in the frequency interval of 36 - 37 ghz . the dependence of relative shifting @xmath128 ( a , c ) of spectral lines of the resonator on frequency @xmath129 caused by bulk inhomogeneities and their relative broadening @xmath130 ( b , d ) at the permittivity dispersion values of inhomogeneities : @xmath131 ( a , b ) and @xmath132 ( c , d ) ; @xmath133 cm . , title="fig:",width=307 ] the dependence of relative shifting @xmath128 ( a , c ) of spectral lines of the resonator on frequency @xmath129 caused by bulk inhomogeneities and their relative broadening @xmath130 ( b , d ) at the permittivity dispersion values of inhomogeneities : @xmath131 ( a , b ) and @xmath132 ( c , d ) ; @xmath133 cm . , title="fig:",width=307 ] it is shown that the spectrum essentially changed at the presence of dielectric inhomogeneities in the resonator . the magnitude of shifting and broadening of spectral lines are caused by resonator filling by inhomogeneities . the solitary spectral lines even at presence of inhomogeneities keep high quality factor , and their broadening is enough small ( for example the spectral lines in the range of 36.2 - 36.6 ghz ) . at the same time the shifting and broadening of adjacent lines are significant ( for example , the spectral lines near of 36.8 ghz ) . this results that the number of excited lines with high quality factor are essentially decreased if the number of inhomogeneities are getting bigger . the reduction of the number of high quality lines we can explain as `` rarefaction '' of the resonator spectrum caused by unequal scattering conditions of different modes on inhomogeneities . fig.3 shows the amplitude - frequency dependence of the resonator spectrum built as the frequency dependence of green function module at different magnitude of parameter @xmath127 . so , at the increase of the number of inhomogeneities in the resonator the adjacent resonances overlapped and the quality factor of combined resonances is getting smaller ( for example , spectral lines near by 36.8 ghz ) . the amplitude - frequency dependence of the empty resonator spectrum ( a ) and the spectrum of resonator with inhomogeneities at @xmath131 ( b ) and @xmath134 ( c ) ; @xmath133 cm . the amplitude is normalized by maximal amplitude of resonances for the empty resonator in the considered wave range.,width=307 ] the main goal of our experiment is the verification of the numerical simulation results of shifting and broadening of spectral lines caused by inhomogeneities and the possibility of resonator spectrum `` rarefaction '' . the another goal is as follows . in the paper [ 3 ] we detected strong spectrum stochastization caused by inhomogeneities such as anisotropic sapphire particles with the dimensions of order of operating wavelength placed into the cavity resonator . the spectrum has mixed state due to both regular and chaotic spectrum part exist . in contrast to [ 3 ] , in the present paper we study the influence on the resonator spectrum relatively small isotropic inhomogeneities . such inhomogeneities can be made by styrofoam particles with the average value of the permittivity closed to one and with small dielectric loss angle . thus , there is a question : is it possible the spectrum chaotization here ? the study of the influence of inhomogeneities on the cavity resonator spectrum we carried out at frequency range of 32 - 37 ghz . we took a quasi - optical cylinder millimeter wave resonator random filled with styrofoam particles ( fig.4 ) . the styrofoam particles have the real part of permittivity is about one , close to the open - air permittivity and small dielectric loss , @xmath135 . the quasi - optical cylinder cavity millimeter wave resonator filled with inhomogeneities , 1,6 are input / output waveguides , 2,5 are holes coupling with waveguides , 3 is the resonator body , 4 is styrofoam particles , @xmath136 is the resonator diameter , @xmath2 is its hight ; @xmath137 mm , @xmath138 mm . , width=307 ] to study the influence of small dielectric particles on the resonator spectrum it is necessary to provide high quality factor for the resonator oscillations without inhomogeneities . for that we excited @xmath4 mode . the magnetic field vector of this mode is directed along the resonator z - axis , and microwave currents do not cross the interface between the flat resonator face and cylinder surface . owing to that the empty resonator has high quality factor up to @xmath139 . to excite the selected mode we used a waveguide diffraction antenna . it is the circular hole with the diameter of 2 mm in a thin diaphragm with the thickness of 0.1 mm closing the input waveguide . the diaphragm surface is flush - mounted with the side - cut cylinder surface of the resonator . the same antenna is used to receive the oscillations on the opposite side of the cylinder surface . the resonator spectrum was detected using `` on pass '' regime in 32 - 37 ghz wave range by the wide frequency standing wave ratio meter . measurement process was automated . signal from the measurement device using analog - digital conversion goes into computer . the further signal processing ( the determination of the spectral line intensity , its quality factor , and frequency ) was done by special software gui application . it gives resource in a short space of time to handle measurement data for a huge number of realizations of the random inhomogeneities distribution in the resonator . owing to that the frequency and quality factor measurement accuracy for spectral lines do not exceed 0.1% and 1% , respectively . the styrofoam particles that are used as inhomogeneities have the size about of 2 - 3 mm . the space distribution of them was arbitrary for each realization . the spectral characteristics were measured depending on the number of these inhomogeneities . the spectrum of the empty resonator is dense enough and consists of 84 narrow spectral lines in the range of 32 - 37 ghz ( fig.5 ) . each line was identified according to mode indexes of corresponding to its eigen resonator oscillation . we found out the oscillations with high quality factor ( @xmath125 ) of order of @xmath140 and higher have small azimuth indexes ( @xmath141 ) and high radial indexes ( @xmath142 ) . for the base oscillation mode ( @xmath143 ) that has the field distribution concentrated inside of the resonator side - cut ( whispering - gallery oscillations [ 5 ] ) , @xmath125 is about @xmath144 . the spectrum of the resonator with inhomogeneities is essentially different from the spectrum of the empty one . by the increase of the number of inhomogeneities spectral lines are getting wider , and @xmath125 is getting smaller . the intensity of broadening lines is decreased as well . the spectrum keeps only few lines with enough high intercity and @xmath125 value . for such lines the @xmath125 value is close to the @xmath125-factor of the spectral lines in the empty resonator . so , the number of spectral lines with enough high intensity and with @xmath145 is 52 in the empty resonator ( fig.5a ) , at filling the resonator by styrofoam particles the number of such lines is 10 ( fig.5b ) , and at filling the resonator by pressed styrofoam particles the number of lines is 3 ( fig.5c ) . the spectrum of the empty resonator ( 84 spectral lines ) ( a ) ; the resonator filled with styrofoam particles ( 77 spectral lines ) ( b ) ; the resonator filled with pressed styrofoam particles ( 57 spectral lines ) ( c ) ; the normalization of amplitude was made on the maximal amplitude value for the empty resonator.,width=307 ] the presence in the resonator with inhomogeneities together with broadening lines few high @xmath125 spectral lines is equivalent to the spectrum `` rarefaction '' ( fig.5 ) . it is necessary to note that the presence of lines with high @xmath125 indicates that the broadening is specified , mainly , by the inhomogeneous of the resonator filling permittivity and is not specified by additional dissipation loss caused by small dielectric loss in the styrofoam . except broadening of spectral lines they have frequency shifting . this frequency shifting has both regular component and random one as well ( fig.6 ) . the dependence of the frequency shifting @xmath146 on frequency @xmath129 for the resonator filled by inhomogeneities . the dotted line is the value of shifting regular component.,width=307 ] the regular shifting happens towards the low frequency direction because of the increase of average dielectric permittivity of inhomogeneous medium in the resonator . for example , it is 150 mgz for the resonator fully filled with styrofoam particles . it is significant that if the distance between resonances is enough small they can both come close and depart from each other because of influence of inhomogeneities . their @xmath125 is essentially changed ( fig.7,8 ) : it can increase for one of them and decrease for another . the relative @xmath125-factor deviation @xmath147 depends upon frequency @xmath129 for the resonator filled by inhomogeneities . for major part of resonances the relative @xmath125-factor deviation is more than zero , i.e. , their @xmath125 is less than for the empty resonator . the relative @xmath125-factor deviation is less than zero for several nearest adjacent resonances that have overlapped spectral lines ( in the case of the empty resonator ) and split into separate resonances at the presence of inhomogeneities , i.e. , the spectral line repulsion takes place here ( see fig.8 ) . , width=307 ] the adjacent resonances with indexes @xmath148 in the empty resonator ( a ) and in the resonator filled by inhomogeneities ( b ) . the normalization of the spectral lines amplitude was made on the maximal spectral line amplitude for the empty resonator . , width=307 ] both the spectral lines broadening and their shifting have quality agreement with described above theory and can be interpreted in the terms of intermode scattering . the maximal broadening is for the adjacent lines that is agreed with theory . we can give the following physical explanation of spectral lines broadening caused by random inhomogeneities based on quantum mechanics interpretation of resonator oscillations . in empty resonator the lifetime of a resonant mode in a given state ( describing by an eigen resonant frequency ) and its spectral line width define by dissipation loss in the resonator body . the uncertainty of mode energy within the spectral line width is caused by this loss . the intermode interaction appears at inserting inhomogeneities into the resonator . owing to that the uncertainty of mode energy and , correspondingly , the spectral line width increase due to the mode transition between nearest states . the mode energy loss due to dissipation loss in the resonator body and transition to the neighboring state are additive to the spectral line width . if the number of inhomogeneities increases the spectrum of the quasi - optical resonator possesses stochastic character that is visualized in distribution of if intervals . in order to define the relation between regular and random spectral components the comparison of if intervals distribution obtained experimentally with different theoretical distributions based on a priori data about statistical process is usually used . in particular , we use the brody function determined the distribution of if intervals probability @xmath149 that is given by @xmath150 where @xmath151 , @xmath152 is the spectral line frequency , @xmath153 is the spectral density - the superposition of regular and random motion density , @xmath154 is the measure of stochastic motion , constants @xmath155 and @xmath156 are defined from the condition of standardization : @xmath157 , @xmath158 , @xmath159 is the gamma function . at @xmath160 if intervals in the spectrum are not correlated and can be described by the poisson distribution , and at @xmath161 we have wigner distribution , when the repulsion effect of spectral lines exists , that is the probability of closest to zero inter - frequency interval is equal zero as well . if stochastization measure in the spectrum is relatively small and the distribution of if intervals is close to the poisson one we can use berry - robnik distribution @xmath162 [ 26 ] that is given by @xmath163\left ( { 1 - \rho } \right)e^ { - \rho s - \frac{\pi } { 4}\left ( { 1 - \rho } \right)^2 s^2 } \\ \end{array}\ ] ] where @xmath164 . @xmath165 is relative phase volume occupied with regular trajectory in mixed systems . the limit @xmath166 corresponds to regular system ; @xmath167 is completely chaotic system . @xmath168 is relative phase volume occupied with chaotic motion . the experimental data show that in the empty resonator the if intervals distribution is close enough to the berry - robnik distribution with @xmath165 ( poisson distribution , @xmath169)(fig.9a ) . by increase of the number of inhomogeneities ( styrofoam particles ) the function @xmath170 is reduced and it has maximum at small @xmath171 . the presence of the maximum @xmath170 indicates the appearance of the repulsion effect of spectral lines , and the random component is increased . we found out that at full filling the resonator by styrofoam particles @xmath172 ( fig.9b ) . the distribution of if intervals , @xmath170 . figure 9a is for the empty resonator spectrum . solid line is for the berry - robnik distribution at @xmath173 or the brody distribution at @xmath160 . figure 9b is for the spectrum of the resonator fully filled with styrofoam particles . the curve 1 is for brody distribution at @xmath174 ; the curve 2 is for berry - robnik distribution at @xmath172 ; the curve 3 is for podolskiy - narimanov distribution at @xmath172 and @xmath175 [ 27].,title="fig:",width=307 ] the distribution of if intervals , @xmath170 . figure 9a is for the empty resonator spectrum . solid line is for the berry - robnik distribution at @xmath173 or the brody distribution at @xmath160 . figure 9b is for the spectrum of the resonator fully filled with styrofoam particles . the curve 1 is for brody distribution at @xmath174 ; the curve 2 is for berry - robnik distribution at @xmath172 ; the curve 3 is for podolskiy - narimanov distribution at @xmath172 and @xmath175 [ 27].,title="fig:",width=307 ] as we can see in fig.9 the function @xmath176 has maximum at small @xmath171 that is not described by the berry - robnik dependence [ 26 ] . the presence of such a maximum can be explained by chaos - assisted tunneling ( cat ) [ 27 ] in the resonator with random inhomogeneities . the proposed in [ 27 ] if distribution function gives the possibility to describe the observed distribution at the parameter @xmath172 corresponding to the classical dynamics in the system and the parameter @xmath175 describing the tunneling between different states of modes . we calculated the spectral rigidity for the resonator with inhomogeneities . the spectral rigidity @xmath177 is an integral characteristic of degree of spectral lines ordering for frequency distances that is much more than inter - frequency interval . it is given by [ 15 ] @xmath178 ^ 2 d\varepsilon } \ ] ] where @xmath179 is the interval length on which the function @xmath180 is determined . the function @xmath181 is built as follows [ 15 ] . for the sequence of the frequencies @xmath182 normalized on unit density ( @xmath183 ) , we introduce a staircase function @xmath184 equals the number of frequencies with @xmath185 . the function @xmath184 has a staircase view with average unit tilt . the function @xmath186 is determined as the minimum of quadratic deviation @xmath184 in the interval @xmath187 from the straight line . the meaning of the averaged in @xmath188 spectral rigidity ( eq.(34 ) ) @xmath189 depends only on @xmath179 and is denoted as @xmath190 . the obtained curve @xmath191 for the resonator with random inhomogeneities is shown in fig.10 . this curve ( 2 ) is placed between the spectral rigidity for the poisson distribution ( curve 1 , @xmath192 ) ) and curve 3 that is corresponded to the spectral rigidity for gauss orthogonal ensemble ( goa ) @xmath193 [ 15 ] . the latter is observed at the modeling of quantum chaos in microwave cavity unstable resonators similar to sinai and bunimovich billiards [ 14 ] . the spectral rigidity for the resonator filled with styrofoam particles ( curve 2 ) . the straight line is for the spectral rigidity of poisson distribution ( curve 1 ) , curve 3 is for goa.,width=307 ] we studied also an active quasi - optical resonator with random inhomogeneities . in comparison with the considered above passive resonator the active one is a self - sustained oscillation system where the presence of random inhomogeneities affects on the excitation of resonator oscillations . for that we used the same resonator as mentioned above ( fig . 4 ) with pointed gunn diode inside . the microwave electrical field for @xmath4 mode was directed along diode s axis . the diode dc power supply was implemented through a filter as a quarter - wave microwave isolator ; owing to spurious microwave radiation was prevented . quasi - optical resonator with gunn diode is an active oscillator with distributed parameters . near threshold of excitation in such an oscillator with the empty resonator was detected unstable multi - frequency generation . such kind of generation we can explain by frequency jumps between adjacent spectral lines with high quality factor . if excitation threshold was highly exceeded , as a result of frequency competition , corresponding to these spectral lines , mono - frequency generation occurs ( fig.11b ) . the active oscillator selects `` itself '' the only frequency to provide maximal regeneration factor [ 28 ] . the random inhomogeneities lead to effective rarefaction of the resonator spectrum . the number of adjacent spectral lines with high quality factor is reduced , and , as a result , multi - frequency generation disappears . at small exceeding of the generation threshold the noise generation appears ( fig.11c ) in the resonator with inhomogeneities . at big exceeding of the generation threshold the stable mono - frequency is observed ( fig.11d ) . owing to inhomogeneities mono - frequency generation possesses greater frequency stability , and the number of generating frequencies is reduced in the range of diode negative resistance . fig.12 shows generating frequencies for the empty resonator and with random bulk inhomogeneities one . oscillogram of the 36 ghz generation obtained by a millimeter wave spectrum analyzer . ( a ) and ( b ) are for the empty resonator ; multi - frequency generation near the generation threshold ( a ) and mono - frequency generation much far from the generation threshold ( b ) . ( c ) and ( d ) are for the resonator filled with inhomogeneities ; chaotic generation near the generation threshold ( c ) and mono - frequency stable generation much far from the generation threshold ( d).,title="fig:",width=144 ] oscillogram of the 36 ghz generation obtained by a millimeter wave spectrum analyzer . ( a ) and ( b ) are for the empty resonator ; multi - frequency generation near the generation threshold ( a ) and mono - frequency generation much far from the generation threshold ( b ) . ( c ) and ( d ) are for the resonator filled with inhomogeneities ; chaotic generation near the generation threshold ( c ) and mono - frequency stable generation much far from the generation threshold ( d).,title="fig:",width=144 ] oscillogram of the 36 ghz generation obtained by a millimeter wave spectrum analyzer . ( a ) and ( b ) are for the empty resonator ; multi - frequency generation near the generation threshold ( a ) and mono - frequency generation much far from the generation threshold ( b ) . ( c ) and ( d ) are for the resonator filled with inhomogeneities ; chaotic generation near the generation threshold ( c ) and mono - frequency stable generation much far from the generation threshold ( d).,title="fig:",width=144 ] oscillogram of the 36 ghz generation obtained by a millimeter wave spectrum analyzer . ( a ) and ( b ) are for the empty resonator ; multi - frequency generation near the generation threshold ( a ) and mono - frequency generation much far from the generation threshold ( b ) . ( c ) and ( d ) are for the resonator filled with inhomogeneities ; chaotic generation near the generation threshold ( c ) and mono - frequency stable generation much far from the generation threshold ( d).,title="fig:",width=144 ] the generating frequencies @xmath129 depending on the number of inhomogeneities @xmath194 . the empty resonator ( @xmath195 ) , the resonator quarter - filled with inhomogeneities ( @xmath196 ) , the resonator half - filled with inhomogeneities ( @xmath197 ) , the resonator fully filled with inhomogeneities ( @xmath198 ) . , width=307 ] the statistical spectral theory of quasi - optical cavity resonator filled with random dielectric inhomogeneities was developed in the present paper . we showed that the presence of inhomogeneities leads to the broadening and shifting of spectral lines . it is found out that the nature of broadening and shifting of spectral lines is relevant to the intermode scattering . the scattering effect for the given spectral line essentially depends on frequency distance between of it and adjacent ones and is sharply decreased for bigger distances . under the influence of ransom inhomogeneities original spectrum modification occurs that can be interpreted as spectrum rarefaction . the spectrum is rarefied because of solitary spectral lines are not practically subjected to the influence of inhomogeneities . the quality factor of such lines and , correspondingly , their intensity stay high at the resonator excitation . the intensity of adjacent lines broadened under the influence of inhomogeneities is essentially reduced . owing to that at the great number of inhomogeneities the resonator spectrum is `` rarefied '' - few solitary high quality factor spectral lines prevail in the spectrum . theoretical prediction of broadening and shifting of resonator spectral lines and spectrum `` rarefaction '' are subject to experimental check . for that purpose we studied experimentally in 8-millimeter wave range the spectrum of the quasi - optical cavity resonator filled with random small - scattered bulk inhomogeneities . these inhomogeneities were styrofoam particles with smaller size than operating wavelength . it is found out that such inhomogeneities lead to broadening and shifting the spectral lines . as experiment showed the maximum of their influence was on frequency adjacent spectral lines . the solitary lines , according to our theory , were subjected to this influence in much smaller degree . we detected the effect of stochastic spectrum `` rarefaction '' . they prove that the main mechanism of broadening and shifting of spectral lines is relevant to inhomogeneities intermode scattering . we studied also chaotic properties of oscillations in our resonator . it is found out that the empty resonator has if intervals distribution similar to poisson distribution that is typical to the spectrum with non - correlated if intervals . even at small number of inhomogeneities the resonator spectrum has random part that increases in proportion to the number of inhomogeneities in the resonator . at that if intervals distribution in random inhomogeneous resonator is described by brody and berry - robnik distributions of if intervals . we obtained the results concerning the influence of bulk inhomogeneities on the process of generation in a self - oscillatory system . the self - oscillatory system was a quasi - optical millimeter wave cavity resonator containing inhomogeneities with an active element as gunn diode . we detected that inhomogeneities leads to the essential `` rarefaction '' of the spectrum and creates conditions for monochromatic stable generation in self - oscillatory system . the inhomogeneous quasi - optical cavity millimeter wave resonator ( passive and active ) can serve as a model of semiconductor quantum billiard . based on our results we suggest using such billiards with spectrum rarefied by random inhomogeneities as an active system of semiconductor laser . 99 s.m . rytov , ya.a . kravtsov , and v.i . tatarskii , _ principles of statistical radiophysics _ ( springer verlag , v.4 . berlin , 1987 ) . f.g.bass , i.m.fuks , _ wave scattering from statistically rough surfaces _ ( pergamon press , n.y . , 1979 ) . ganapolski , z.e . eremenko , phys . e , * 65 * , 056218 ( 2002 ) . braginskii , v.p . mitrofanov , and v.i . panov , _ systems with low dissipations _ ( nauka , moscow , 1981 ) . a.n.oraevskii , m.scally , and v.l . velichanskii , quantum electronics , * 25 * , 211 ( 1998 ) . b.e.little , j - p . laine , s.chu , optics letters , * 22 * , 4 ( 1997 ) . l. goldstein , f. glass , j.y . marzin , m.n . charasse , and g.le . roux , appl . * 47 * , 1099 ( 1985 ) . p.m.petroff , s.p . den baars , superlat . , * 15 * , 15 ( 1994 ) . m. moison , f. houzay , f. barthe , l. leprince , e.andre , o. vatel , appl . lett . , * 64 * , 196 ( 1994 ) . zh.i . alferov et al , fizika i tekhnika poluprovodnikov [ soviet physics semiconductors ] , * 30 * , 351 ( 1996 ) . i. alferov et al , fizika i tekhnika poluprovodnikov [ soviet physics semiconductors ] , * 30 * , 357 ( 1996 ) . m. schernjakov et al , pisma v zhurnal tekhnicheskoi fiziki [ soviet technical physics letters ] , * 23 * , 51 ( 1997 ) . beljavskii , s.v.shevtsov , fizika i tekhnika poluprovodnikov [ soviet physics semiconductors ] , * 36 * , 874 ( 2002 ) . stockmann , _ quantum chaos an introduction _ ( cambridge university press , 1999 ) . elutin , uspekhi fizicheskikh nauk [ soviet physics uspekhi ] , * 155 * , 397 ( 1988 ) . h. alt , h. d.graf , r. hofferbert et al . e , * 54 * , 2303 ( 1996 ) . e.m.ganapolskii , z.e.eremenko , dopovidi akademii nauk ukrayiny [ ukraniain physics doklady ] , no.12 , 93 ( 2000 ) . tarasov , waves in random media , * 10 * , 395 ( 2000 ) . tarasov , low temp . phys . , * 29 * , 45 ( 2003 ) . vainstein , _ electromagnetic waves _ ( radio i svjaz , moscow , 1988 ) . yu.v . tarasov , phys . b , * 71 * , 125112 ( 2005 ) . yu.v . tarasov , phys . b , * 73 * , 014202 ( 2006 ) . r. newton . _ scattering theory of waves and particles _ ( mcgraw - hill , new - york , 1968 ) . lifshitz , s.a . gredeskul , l.a . pastur , _ intriduction in theory of disordered systems _ ( nauka , moscow , 1982 ) . gradshtein , i.m . ryzhik , _ tables of integrals , sums , series and products _ ( nauka , moscow , 1971 ) . m.v . berry , m. robnik , j. phys . , * a17 * , 2413 ( 1984 ) . podolskiy , e.e . narimanov , phys . lett . , * 91 * , 263601 ( 2003 ) ; v.podolskiy , e.narimanov , arxiv : nlin.cd/0310034 . tzarapkin , _ microwave gann oscillators _ ( radio i svjaz , moscow , 1982 ) .

we apply a large - deviation method to study the diffusive trajectories of the quadrature operators of light within a reservoir connected to dissipative quantum systems . we formulate the study of quadrature trajectories in terms of characteristic operators and show that in the long time limit the statistics of such trajectories obey a large - deviation principle . we take our motivation from homodyne detection schemes which allow the statistics of quadrature operator of the light field to be measured . we illustrate our approach with four examples of increasing complexity : a driven two - level system , a ` blinking ' three - level system , a pair of weakly - coupled two - level driven systems , and the micromaser . we discuss how quadrature operators can serve as alternative order parameters for the classification of dynamical phases , which is particularly useful in cases where the statistics of quantum jumps can not distinguish between such phases . the formalism we introduce also allows us to analyse the properties of the light emitted by quantum jump trajectories which fluctuate far from the typical dynamics .

1206.5719

this paper addresses questions about the nature of dynamical phase transitions and crossovers in open quantum systems from the perspective of observers able to make measurements on the environment . we take our motivation from experiments using homodyne detection schemes @xcite and study the time series or trajectories of quadratures of the light emitted from open quantum systems . we go beyond recent work @xcite which focused on trajectories of quantum jumps analyzed from the point of view of the counting statistics of , for example , photons entering the environment . in this approach , identifying the average rate of photon emission as a dynamical order parameter allowed to uncover dynamical phase transitions and crossovers in a number of systems @xcite . a central idea in the works upon which we build was the introduction of a field conjugate to the number of emitted quanta , the so - called `` @xmath0-field '' . in equilibrium statistical mechanics , the observation of phase transitions is dependent on choice of the external field used to tune across a transition and the use of an appropriate order parameter ; in this work we demonstrate that different _ dynamical _ order parameters and conjugate @xmath0-fields allow exploration of different _ dynamical _ phase transitions . specifically we show that crossovers between dynamical phases may be observed from quadrature measurements which could not be found by counting photons , and vice versa . we also extend our study to consider the effects of two @xmath0-fields , so that we can influence both light quadratures and quantum jumps . this enables us to understand the relation between inferences from different dynamical observables ; for example , we can find the typical quadrature measurements which would be made in rare quantum - jump trajectories , when either fewer or more photons are emitted than average . our approach takes as a starting point the basic observation that ( real - time ) dynamics is more than statics . sometimes dynamical behaviour , including transitions and crossovers between different dynamical regimes , can be understood from the properties of the stationary state . but a more common occurrence is that dynamical fluctuations are only revealed via ( often high - order ) time - correlation functions , and not by static or one - time observables . this suggests that for a proper statistical analysis of the dynamics of open systems a ` statistical mechanics of trajectories ' is needed . such an approach is the so - called @xmath0-ensemble method , which has proven useful for the study of classical many - body systems displaying complex cooperative dynamics such as glasses @xcite . the aim is to describe dynamical phases in terms of strictly dynamical order parameters and to classify these dynamical phases and the changes between them using a mathematical and conceptual framework analogous to that of equilibrium statistical mechanics . this ensemble method for dynamics can therefore be thought of as a ` thermodynamics of trajectories ' approach @xcite . dynamical phase transitions are not limited to classical systems ; transitions and crossovers have been discovered in a number of driven open quantum systems . famous examples include the laser @xcite , whose behaviour close to threshold resembles a closed thermodynamic system in the neighbourhood of a continuous phase transition , and the micromaser @xcite , where a flux of atoms is coupled to an optical cavity mode and drives the cavity through a series of crossovers . however more recently , dynamical phase transitions have also been discovered in a broad range of fields , spanning decohering spin channels @xcite , information transport in complex systems @xcite , current fluctuations in isolated diffusive systems @xcite and even the decoherence of tunnelling molecules @xcite . the @xmath0-ensemble method was recently applied to the study of ensembles of quantum - jump trajectories in open quantum systems @xcite . although similar in spirit to ideas in full counting statistics @xcite , the @xmath0-ensemble approaches the problem from a different perspective and has now been developed for a variety of problems where dynamical phases are classified according to the counting statistics of quantum jumps @xcite . , where the operators @xmath1 ( @xmath2 ) are creation ( annihilation ) operators of an harmonic oscillator bath and the operators @xmath3 and @xmath4 are system operators which are related to the lindblad projectors of the system . ] in this paper , we formulate the application of the @xmath0-ensemble to light quadratures by means of reduced characteristic operators . measuring quadratures of emitted light allows the quantum state of light emitted from an open system to be probed in a way not possible by counting photons . for example , it provides an understanding of whether the light is in a coherent state , a squeezed state or another more exotic state . we apply this method to a selection of open quantum systems , ranging from few - level quantum - optical systems to the micromaser , all of which have a stucture illustrated schematically in fig . [ fig : gen ] . in contrast to counting the number of emitted photons , as in studies of quantum - jump trajectories , we consider the temporal accumulation of quadratures of light emitted from a system into its environment . we consider this a _ quadrature trajectory_. unlike the quantum - jump trajectories , the accumlation of light quadratures in the environment is a diffusive process . however we can define a _ quadrature activity _ , defined as a time - averaged light quadrature , and use this as a dynamical order parameter to characterise dynamical phases in the space of quadrature trajectories . our motivation for studying the statistics of quadratures is inspired by the experimental technique of homodyne detection . the @xmath5-quadrature trajectories of emitted light are directly related to the _ homodyne current _ @xcite , as described in fig . [ fig:3 ] . ( it is also sensible for us to study other quadratures , since these are accessible via a change in the driving hamiltonian , which we will discuss later in sec . [ sec : level2 ] . ) the statistics of quadrature trajectories provide a more natural probe of the system dynamics than quantum - jump trajectories and we will show that such measurement schemes allow exploration of dynamical phases in a variety of systems . @xmath6{homodyne.eps } \includegraphics[scale=0.125]{rand1.eps } \end{array}$ ] beyond examining dynamical phases identified by quadrature activity , we construct marginal probability distributions for general quadrature operators at all angles in phase space . using these marginals , we reconstruct wigner distributions @xcite via the inverse radon transform , to find the state of the emitted light . we extend our studies to examine the typical quadrature trajectories of systems biased by the number of photon emissions . this allows us to understand the nature of light emitted from a system for all quantum - jump trajectories , whether _ rare _ or _ typical _ with respect to the number of emitted photons . more generally we study the trajectories of a particular observable after having biased the system towards rare trajectories of another ( generally non - commuting ) observable . this technique also allows us to identify the appropriate dynamical order parameters to understand the dynamical phases in different open systems . the paper is structured as follows . in the following section we introduce the formalism of the @xmath0-ensemble and describe the it calculus methods which we employ for the stochastic description of the environment we use in this paper . we further develop generalised master equations for the dynamics of quadrature - biased systems and systems biased towards rare trajectories of both quantum - jumps and quadratures . we also discuss use of the wigner function to analyse our results . in subsequent sections we present our results . in sec . [ sec : level2 ] we discuss a simple driven two - level system and in sec . [ sec : level3 ] we extend our analysis to a driven three - level system . in sec . [ sec:2l2c ] we study a pair of coupled two - level systems and demonstrate that measuring quadrature trajectories allows identification of dynamical phases which are hidden in photon - counting experiments . finally , in sec . [ sec : micro ] we give a concise account of quadrature trajectories in the micromaser , providing insights via a mean - field theory as well as exact numerical diagonalisation . in sec . [ sec : conc ] we give our conclusions . several recent studies have applied the @xmath0-ensemble method to study the statistics of quantum trajectories in open quantum systems using a thermodynamic description @xcite . in this section we will give an account of the @xmath0-ensemble formalism for quantum trajectories . we discuss the specific examples of counting photons and measuring the quadratures of light entering the environment around an open quantum system . first , in section [ sec : sensemble ] we discuss the general application of the @xmath0-ensemble in the study of the statistics of an observable @xmath7 . we then reformulate these principles so their application to open systems is clear . section [ sec : inout ] discusses the open quantum system methods which we use , which are applied to the @xmath0-ensemble in sections [ sec : s1 ] and [ sec : s2 ] . we begin by considering a particular measurable quantity @xmath7 associated with a quantum trajectory of a quantum system coupled to a reservoir . ( for example , @xmath7 could be the number of emitted photons in a time @xmath8 , or other quantities which we define below . ) projecting the system density matrix @xmath9 on to the subspace where @xmath7 takes a particular value , one may define a reduced density matrix @xmath10 . the probability of such a realisation occuring in time @xmath8 is then given by @xmath11 $ ] and , after long times when the system reaches a steady state , this takes a large - deviation ( ld ) form : @xmath12 where @xmath13 contains all the information about the probability distribution of @xmath7 at long times . alternatively the statistics may be described introducing a moment generating function associated with these probabilities . this method proceeds by introducing another density matrix @xmath14 defined by the laplace transform @xmath15 from which we define the moment generating function @xmath16 in both and , the integrals should be replaced with sums if @xmath7 is a discrete quantity , as is the case when counting photons . a ld form @xmath17 is found at long times , with ld functions @xmath18 and @xmath13 related by legendre transform , @xmath19 . the full statistics of @xmath7 are therefore contained within @xmath18 . the density matrix @xmath20 describes , when @xmath21 , rare trajectories where @xmath7 is far from the mean . in counting processes , where @xmath7 is bounded from below by zero , the rare trajectories can be separated into _ more active _ trajectories when @xmath22 and _ less active _ trajectories when @xmath23 . these correspond respectively to trajectories with a larger or a smaller number of counts than in the @xmath24 physical dynamics . there are two immediate advantages to this @xmath0-ensemble approach . firstly , as the ld function @xmath18 is the largest real eigenvalue of a superoperator which generates the @xmath0-biased dynamics of @xmath20 ( which we derive in the following sections ) , evaluation of the statistics of trajectories can be straightforward . of particular interest are the first and second moments , @xmath25 and @xmath26 , which may be found directly from the ld function via @xmath27 where the @xmath0 subscripts indicated that the expectation values are taken with respect to the ensemble of trajectories biased by @xmath28 . secondly , this method provides a thermodynamic formalism for non - equilibrium processes . the ld functions @xmath18 and @xmath13 are analogues of free energy and entropy densities , with @xmath0 the conjugate _ intensive _ field to the _ time - extensive _ @xmath7 . furthermore , the so - called activity @xmath25 may be used as a dynamical order parameter to distinguish dynamical phases , whose boundaries may be crossed by tuning the parameter @xmath0 , or other system parameters . indeed , such phase boundaries are crossed when an @xmath0-derivative of @xmath18 becomes discontinuous . so far , the @xmath0-ensemble studies of open quantum systems have explored the thermodynamics of quantum - jump trajectories associated with counting quanta emitted from a system into a markovian environment . in these cases the quantity @xmath7 is , for example , the number of photons emitted ( _ i.e. _ the number of quantum jumps ) , @xmath29 , from a system in unit time . defining reservoir ladder operators @xmath2 and @xmath1 in the heisenberg representation , the quantity @xmath29 corresponds integrating over time the observable @xmath30 . biasing these trajectories , we obtain a ld function @xmath31 from which we extract the dynamical order parameter @xmath32 . in this paper , we develop the @xmath0-ensemble further to study the statistics of the quadratures of light entering a markovian bath from various quantum systems . choosing @xmath7 to be the @xmath5- and @xmath33-quadratures of the light , corresponding to bath operators @xmath34 and @xmath35 , we will use corresponding _ quadrature activities _ @xmath36 and @xmath37 when applying @xmath0-fields to bias the dynamics towards rare quadrature trajectories . in much the same way as the photon activity @xmath38 can be associated with the average number of photons emitted in a time @xmath8 , the quadrature activities @xmath39 and @xmath40 are time averages of the @xmath5- and @xmath33-quadratures of the light leaving the system . we will gain yet further insight into the dynamics of the systems we study by examining general quadratures of the form @xmath41 where the angle @xmath42 in phase space is illustrated in fig . [ fig:2 ] . for these general quadratures , we can consider marginal distributions @xmath43 associated with general quadratures , which can be extracted from the relevant ld functions @xmath44 and their @xmath0-derivatives using eqs . and . in the following subsections we will introduce the mathematical formalism which allows us to construct @xmath0-biased ensembles and show how the appropriate ld functions are derived . we will also explore doubly - biased ensembles where we study the typical trajectories of one observable for ensembles of trajectories biased by another observable . we develop this formulation in section [ sec : s2 ] , but we first discuss how the ld functions for ( singly- ) biased ensembles are derived . previous studies on quantum - jump trajectories @xcite have demonstrated that one may identify the ld function @xmath18 as the largest eigenvalue of an @xmath0-modified master equation obeyed by @xmath45 , @xmath46 where @xmath47 is the generator for the @xmath0-biased dynamics . we will outline how to construct formally these generalized master equations in terms of an it calculus and characteristic operators , before proceeding to apply them to the study of quadratures in sec . [ sec : s1 ] . we consider the system to be weakly coupled to a reservoir in the markovian regime @xcite ; within this picture the reservoir is treated as a bath of harmonic oscillators which effectively act as a white noise source to our system . this type of reservoir admits a stochastic description whereby the effects of reservoir ladder operators @xmath48 and @xmath49 , with commutator @xmath50 = \delta ( t - t')$ ] are described by it increments which obey a quantum it calculus @xcite . we next sketch our method within this formalism ; we give full details of the derivation in appendix [ sec : der1 ] . for the rest of the paper we consider the reservoir to be an unsqueezed vacuum @xcite . we introduce temporal increments @xmath51 and @xmath52 which are defined in terms of the reservoir ladder operators @xmath53 and @xmath54 by @xmath55 and are evaluated in the immediate future . the dynamics are found using a stochastic density matrix @xmath56 whose it increment is related to the system hamiltonian , @xmath57 , the reservoir increments defined above and the lindblad operators of the system @xmath58 : @xmath59dt-\frac{1}{2}\sum_{i}\{{l}^{\dag}_{i}{l}_{i},\rho\}dt\\ & + \sum_{i}{db}^{\dag}\left(t \right){l}_{i}\rho\left(t \right){l}^{\dag}_{i}db\left(t \right)\nonumber\\ & + \sum_{i}{db}^{\dag}\left(t \right){l}_{i}\rho(t)\nonumber \\ & + \sum_{i}\rho\left(t \right)db\left(t \right){l}^{\dag}_{i}.\nonumber\end{aligned}\ ] ] tracing out the reservoir degrees of freedom , one obtains the familiar lindblad master equation @xcite for the ( trace - preserving ) system dynamics : @xmath60-\sum_{i}\{{l}^{\dag}_{i}{l}_{i},\rho^0 \}+ \sum_{i}{l}_{i}\rho^0{l}^{\dag}_{i}\label{eq:3e}\end{aligned}\ ] ] where @xmath61 is the anti - commutator and @xmath62 is the density matrix of the system , with @xmath63 denoting the trace over the reservoir hilbert space . to modify this trace - preserving scheme to the desired @xmath0-biased generalized master equation for trajectories of @xmath7 , we introduce a characteristic operator , @xmath64 . the corresponding stochastic increment d@xmath7 is related to the reservoir increments in eq . . the characteristic operator is defined by @xmath65 from this operator we identify @xmath0-biased reduced density matrix as @xmath66 using eq . one may derive eq . by finding the increment of @xmath45 in terms of the increments of @xmath67 and @xmath68 . then , application of the it calculus of d@xmath7 in terms of reservoir increments allows the trace to be taken over the reservoir degrees of freedom ( see appendix [ sec : der1 ] ) . this method of deriving @xmath47 relies upon the existence of a complementary stochastic reservoir operator to the system operator of interest . we discuss these operators next for the cases of quantum jumps and quadratures . we turn our attention to statistics of quantum - jump trajectories , before moving on to quadrature trajectories , where @xmath69 . a quantum - jump trajectory is the time record of the number of jump events @xmath29 over a time @xmath8 . we formulate the stochastic process associated with jump trajectories in terms of an increment @xmath70 . in terms of the reservoir increments this jump increment is defined as @xmath71 which has eigenvalues of 0 and 1 within the interval @xmath72 $ ] . although @xmath70 tells us how many photons were emitted into the bath in the interval @xmath72 $ ] , it tells us nothing else about the form of the light emitted . we address questions about the quantum state of emitted light by looking at the statistics of the quadrature trajectories of the light emitted into the reservoir . we use the @xmath5- and @xmath33-quadratures to define coordinate axes of the optical phase space ( see fig . [ fig:2 ] ) . we will also study the general quadrature operator where the observable is @xmath73 , defined in eq . , where @xmath42 is the polar angle . the stochastic increments associated with these operators , @xmath74 , are defined by @xmath75 evaluation at @xmath76 and @xmath77 yields the increment for the familiar @xmath5- and @xmath33-quadratures . although @xmath74 takes the same form for all @xmath42 , the form of the stochastic process @xmath70 associated with the jump trajectories is very different . therefore , for the rest of this paper we use notation where @xmath29 is biased by a conjugate field @xmath78 , while the quadratures @xmath79 are biased by @xmath0 . with this notation , we define characteristic operators of the form associated with each ensemble of trajectories , @xmath80 with these definitions we will construct generalized master equations for each ensemble of trajectories in sec . [ sec : s1 ] , and we will proceed further in sec . [ sec : s2 ] to ask about the typical statistics of one process in the biased ensemble of the other . we now turn to the statistics of the trajectories associated with the quadrature operators . we consider projections of the density matrix @xmath9 on to the subspace where @xmath81 takes a specific realisation . by this we mean that the time - integrated quadrature for light entering the bath up to time @xmath8 takes a specific value @xmath79 . we denote a corresponding projected reduced density matrix by @xmath82 . as described in section [ sec : sensemble ] , this is related to an @xmath0-biased reduced density matrix @xmath14 , defined in eq . with @xmath83 , via the laplace transform @xmath84 which upon taking the trace gives us the moment generating function associated with these diffusive probabilities @xmath85 , where the ld form is valid at long times . the _ activity _ associated with the quadratures is @xmath86 . this , and the second moment @xmath87 are found from the derivatives of the ld function @xmath44 defined in eqs . and with @xmath83 . this ld function is identified as the largest real eigenvalue @xcite of the generalized master equation @xmath88 we derive this generalised master equation using eqs . and , where tracing out the environment is performed within the it calculus formalism , the details of which are presented in appendix [ sec : der1 ] . when @xmath89 , the superoperator collapses to the trace - preserving liouvillian @xmath90 , eq . . away from @xmath24 the @xmath0-field biases the dynamics towards rare trajectories of the system . the properties of @xmath0-modified master equations are discussed in ref . before presenting our results for various model systems , we discuss doubly - biased ensembles . thus far we have provided the theoretical formalism for an @xmath0-ensemble of quadrature trajectories . before presenting results for a variety of systems , we introduce a further interesting problem in this section concerning the trajectories of already - biased ensembles . consider a system biased towards rare trajectories with , for example , more quantum jumps than the @xmath91 average . we now ask what are the quadratures for these trajectories which are , for example , _ more active _ with respect to the number of emitted photons ? in order to answer this question we need to introduce two counting fields , @xmath78 conjugate to the number of emitted photons , @xmath29 , and @xmath0 conjugate to quadratures @xmath79 . the derivation of doubly - biased ensembles is an extension of the derivation in sec . [ sec : s1 ] , using eq . with characteristic operators and tracing out the reservoir . the characteristic operators associated with the quadrature increment and photon increment are respectively @xmath92 and @xmath93 . with these we can define a new @xmath0 and @xmath78 biased density operator , @xmath94 , which incorporates information on the statistics of both ensembles of trajectories , @xmath95 the ordering of the characteristic operators is important due to the non - commutability of the two observables . as defined , eq . is interpreted as biasing the jump trajectories and within these @xmath78-biased ensembles measuring the quadrature realisations . we show in appendix [ sec : der1 ] how it calculus is used to calculate the it increment of this new double - biased density operator . ( we formally trace out the reservoir using the it tables in eqs . and . ) we find this new doubly - biased master density operator obeys a new generalised master equation which reads @xmath96 notice that if @xmath0 , @xmath97 then the left - hand side reduces to the ( trace - preserving ) liouvillian . once again we may encapsulate the total long time trajectory statistics using the largest eigenvalue , @xmath98 , of the generator of these statistics . furthermore we note that @xmath98 necessarily reduces to the ld function for @xmath78-biased jump trajectories , @xmath99 , in the limit @xmath100 @xcite . to examine the typical quadratures of quantum - jump biased systems we examine the derivatives of @xmath101 , with respect to @xmath0 , evaluated in the limit @xmath100 . note that similar doubly - biased ensembles may be created by choosing different characteristic operators in eq . . for example , switching the characteristic operators in eq . allows us to find the statistics of quantum jumps in quadrature - biased ensembles . next , we put the formalism into practice in the following sections with studies of few - level quantum - optical systems and a system with more complex dynamics , the micromaser . in this section we present our results for a simple quantum - optical system consisting of a dissipative two - level system driven by a laser , shown schematically in fig . [ fig:1 ] . we begin by examining the @xmath5- and @xmath33-quadrature statistics of the system and compare it with @xmath0-ensemble studies for the jump trajectories studied in @xcite . after examining both analytic and numerical forms of the ld function we turn our attention to the doubly - biased statistics of the system . we finish by constructing plots of time - independent probability distributions , at various @xmath78 , associated with the @xmath79 which we refer to as phase portraits before examining the wigner functions of this system . for the driven two - level system , the generalized master operator , @xmath47 , is of the form in eq . , where now we have one pair of lindblad operators @xmath102 and @xmath103 , where @xmath104 , with @xmath105 the decay rate and @xmath3 the lowering operator @xmath106 . we drive the system with a laser polarised in the @xmath107-direction , which introduces a driving term @xmath108 in the two - level system hamiltonian , @xmath109 where @xmath110 is the rabi frequency . here we consider the specific choice @xmath111 as this has been shown to be an interesting parameter choice in previous works @xcite , and for further reasons we discuss below . we study the @xmath5- and @xmath33-quadratures specifically . in this regime although the ld function does not take on a straightforward form , writing @xmath47 in matrix form we numerically diagonalise it and identify its largest real eigenvalue as the ld function . although no concise analytical form exists for the ld function , we may taylor expand the ld functions about @xmath24 to see how the statistics of the physical system behave . for @xmath5- and @xmath33-quadratures we find the expansions @xmath112 firstly we examine the form of the ld functions close to @xmath24 . from the taylor expansion , we see that that @xmath113 to @xmath114 , is symmetric about @xmath24 . immediately we find that the @xmath5-quadrature activity for physical @xmath24 dynamics is @xmath115 . this symmetry is not present in the @xmath33-quadrature statistics as the hamiltonian breaks the symmetry between @xmath5 and @xmath33 ; the odd powers of @xmath0 in the taylor expansion imply that near @xmath24 , @xmath116 is asymmetric such that , close to the @xmath24 physical dynamics , the sign of @xmath0 is relevant . evaluating the physical @xmath33-quadrature activity we find @xmath117 , the absolute value of which corresponds to the value of the physical photon emission rate , @xmath118 , as shown in fig . [ yandk ] . -quadrature activity @xmath119 and minus the quantum - jump activity @xmath120 for the two - level system when biased by the appropriate fields @xmath0 and @xmath78 . note that the order parameters coincide for the physical dynamics at @xmath121 . ] this result is specific to the system parameters here . while the choice @xmath111 was identified as special in ref . @xcite due to a self similarity of quantum - jump statistics , we find this is also the choice of system parameters where the @xmath33-quadrature and quantum - jump activities coincide . that this relation is found for @xmath33-quadratures instead of @xmath5-quadratures is entirely due to the chosen laser polarisation . + [ fig:4 ] switching our focus to the full numerical forms of the ld functions , we show in figs . [ fig:2lx ] and [ fig:2ly ] that the minimum of @xmath122 is shifted away from @xmath24 . however , while @xmath123 and @xmath124 are distinct near @xmath24 , at very large @xmath125 , @xmath126 and @xmath122 both tend towards @xmath127 . while the laser breaks the symmetry between @xmath5- and @xmath33-quadratures , the rare trajectories become equivalent again at large @xmath125 . we can understand this as the @xmath0-field dominating the non - equilibrium dynamics over the self - hamiltonian of the system . we now explore how biasing the photon activities with a counting field @xmath128 affects the @xmath24 measured quadratures . the dynamics of our doubly biased system are described by the superoperator @xmath129 defined in eq . . we diagonalise this superoperator to find the ld function @xmath130 and evaluate the quadrature activities by taking the derivative with respect to @xmath0 at @xmath131 . we look far into both the photon inactive and active regimes discussed in sec . [ sec : sensemble ] by choosing photon counting fields @xmath132 and @xmath133 respectively . to aid visualisation of how the biasing affects the light , we construct a time - independent phase portrait of the light and find the corresponding wigner distribution , which we define in appendix [ sec : radon ] . to construct our phase portrait we determine the ld function for equally spaced values of @xmath42 in the range [ 0,@xmath134 and use the legendre transform to calculate @xmath135 , from eq . . we plot these probability distributions in an activity phase space using coordinates @xmath136 and @xmath137 ( hereafter we drop the @xmath24 subscript ) . these plots , or portraits , illustrate the set of marginal distributions for all @xmath42 : the marginal probability distributions correspond to cuts through the origin at angle @xmath42 to the abscissa . our results are shown in fig . [ fig:5 ] . using the values @xmath135 at each @xmath128 , we also construct the wigner distribution , shown in fig . [ fig:6 ] , with a numerical implementation of the inverse wigner transform . figure [ fig:5 ] shows that the physical dynamics have a `` heart - shaped '' portrait centred on @xmath138 . although it may appear as though there are bi - modal regions , all @xmath135 are in fact non - gaussian unimodal distributions where , as @xmath42 increases from 0 to @xmath77 , the mean shifts from 0 to @xmath139 . interestingly the photon inactive system appears more like a vacuum as it is centred closer to the origin at @xmath140 . however , although this is a photon - inactive region , there is always a finite probability of quantum jump provided @xmath0 is finite . in a complementary fashion the @xmath141 portrait `` blows out '' as the centre of each @xmath135 moves rapidly away from zero , when @xmath42 increases from @xmath142 to @xmath77 . corresponding features are shown in the wigner functions plotted in fig . [ fig:6 ] . the wigner function becomes more concentrated about the origin as we enter the photon - number inactive region , confirming the notion that the output field becomes more vacuum - like as we proceed to more positive @xmath128 . we now turn our attention to a more complicated quantum optical system , the three - level system , depicted in fig . [ fig:3lev ] . we drive this system by two resonant lasers with rabi frequencies @xmath143 and @xmath144 between the @xmath145 level and upper levels @xmath146 and @xmath147 respectively . when @xmath148 the photon emission trajectories are intermittent @xcite ; the @xmath149 transition is the active or light line transition while the @xmath147 is an inactive level . previous studies @xcite found that the @xmath150 exhibits a dynamical crossover at @xmath91 . the crossover is from an active phase @xmath151 , where the dynamics are dominated by the @xmath146 and @xmath145 levels , to an inactive phase @xmath152 , where the system dynamics are dominated by long periods where the @xmath147 level is occupied . with this in mind we will examine firstly the quadrature trajectories , then proceed to examine the effects of double biasing and examine the phase portraits of the system . and when @xmath148 the photon emission trajectories are intermittent ; the @xmath149 transition is the active or light line transition while @xmath147 is an inactive level.,height=75 ] considering the three - level system driven by two lasers , the generalized superoperator @xmath47 is given in eq . , still with one set of lindblad operators @xmath102 and @xmath103 . in this system , @xmath153 and there are no other lindblad operators because of the null decay rate between @xmath147 and @xmath145 , as illustrated in fig . [ fig:3lev ] . however our hamiltonian for this system is slightly different : @xmath154 where @xmath155 and @xmath156 . when @xmath157 typical photon emissions are intermittent displaying both bright " and dark " periods @xcite . the ld function @xmath158 for each @xmath42 is obtained by direct diagonalization of @xmath47 and we examine specifically the case @xmath159 and @xmath160 . we first study the ensembles of trajectories with @xmath5-quadrature biases , shown in fig . [ fig:8](a ) . here , @xmath113 is symmetric in @xmath0 and exhibits no sharp crossovers . on the other hand , the ensemble of @xmath33-quadratures exhibits a sharp peak in its dynamical variance as we cross @xmath24 , as shown in fig . [ fig:8](b ) . this peak corresponds to a crossover between two distinct dynamical phases corresponding to the changeover of the behaviour of @xmath161 in this region . + in the photon emission study in ref . @xcite , the crossover in @xmath162 was attributed to a change in the effective behaviour where the _ active _ side where @xmath151 was argued to be similar to a two - level system . in the _ inactive _ phase found for @xmath152 , the @xmath147 level is occupied for long time periods such that few photon emissions occur . we attribute the crossover observed in @xmath40 to the change in dynamics associated with the crossover in photon emission discussed in ref . @xcite . to support this claim we examine phase - space portraits of the system and the effects of the @xmath78-bias on these portraits , by solving the doubly - biased master equation . first we look at the photon - inactive dynamics with @xmath163 . the phase portraits and wigner functions are shown fig . [ fig:9](a ) : these appear identical to the two - level system biased by the same @xmath78-field , shown in fig . [ fig:5](a ) . moving through to @xmath164 , shown in fig . [ fig:9](c ) , the phase portrait appears almost like a vacuum . this is not a true vacuum ; both photon activity and @xmath40 for this portrait are small but non - zero . however the forms of the portrait and wigner function do suggest that for large @xmath78 , the @xmath147 state is occupied for a large fraction of the time . we illustrate our finding more clearly in figs [ c1 ] and [ c2 ] , where we show how a contour plot of the @xmath24 two - level portrait with three - level shifted @xmath165 contours closely resembles the three - level system @xmath24 plot . this demonstrates that the physical dynamics of the three - level system is effectively made up of an active two - level system plus an inactive two - level system ( this is where the dynamics are dominated by occupations of the @xmath147 state ) with some `` mixing '' of the two . + to determine whether or not it is this @xmath150 crossover which is responsible for the crossover in @xmath40 we examine the statistics of typical jump trajectories for rare quadrature trajectories . we repeat the procedure outlined in sec . [ sec : s2 ] but now we first bias quadratures and then count quantum jumps by introducing the doubly - biased density matrix @xmath166 then , we generate a new master equation @xmath167 solving for the ld function , @xmath168 for both the @xmath5- and @xmath33- quadratures we find @xmath169 for various quadrature biases @xmath0 . the typical activity @xmath170 shows no sharp features as we bias the @xmath5-quadrature , irrespective of the sign of @xmath0;. however , if we bias the @xmath33-quadrature , the activity @xmath170 and variance @xmath171 show photon - number inactivity when @xmath22 . that is to say , more positive @xmath40 leads to smaller @xmath172 . this is demonstrated in fig . [ fig:11 ] . this supports our previous assertion that the activities @xmath173 and @xmath174 may be used as equivalent order parameters in this system . , for different quadrature biases @xmath0 in the three - level system . the jump activity grows as we bias the @xmath5-quadrature regardless of sign of @xmath0 . however the photon count exhibits a dynamical crossover in the rare @xmath33-trajectories about @xmath24 , and shows that the sign of @xmath0 determines the behaviour of @xmath174 . furthermore the photon variance in these quadratures also indicates the inactive / active phase crossover , with a peak at @xmath24 correlating with the dynamical phase transition at this point ] . [ fig:11 ] as a function of photon bias @xmath78 in the three - level system . @xmath175 exhibits a crossover at @xmath91 , confirming that while @xmath174 varies with @xmath0 in a highly correlated way , @xmath173 is similarly correlated with @xmath128 . ] for completeness , we show in fig . [ fig:12 ] how @xmath175 varies as a function of @xmath78 . surprisingly , we find the same correlation at @xmath24 . this result is surprising because , although no inactive phase exists for @xmath173 , it is still a valid dynamical order parameter due to its connection to the quantum - jump activity : for @xmath137 , the region where @xmath176 corresponds to the active quantum - jump phase and @xmath177 corresponds to the inactive phase . this connection between @xmath175 and @xmath174 also exists in the two - level system studied in sec . [ sec : level2 ] , and may be a trait present in many systems . to conclude , the finding that quadrature activities may encode the same dynamical statistics as the photon activity is quite surprising , given the non - commuting nature of the operators . we next ask whether there are systems where a dynamical crossover is only visible through the quadrature activity . we will discuss this question in secs . [ sec:2l2c ] and [ sec : micro ] . we extend our study to a pair of coupled two - level systems driven by lasers with different polarisations , depicted in fig . [ fig:13 ] . in the previous two sections , we have identified that crossovers in quadrature activity mirror crossovers in the number of emitted photons , showing an equivalence between quadrature activity and photon activity as dynamical order parameters . here we show that crossovers can occur in the quadrature activity and not in the photon activity , emphasising the use of quadrature activity as a dynamical order parameter in its own right . we consider two weakly coupled two - level systems as shown in fig . [ fig:13 ] . there are now two lindblad operators associated with each two - level system : @xmath178 and @xmath179 , where these operators are the ladder operators of each two - level system , i and ii respectively . these two - level systems have identical rabi frequencies and decay rates . as previously , we choose @xmath180 , but distinguish the two subsystems by the polarisation of the driving laser . subsystem i is driven by a @xmath181 polarized laser , while subsystem ii is driven by a @xmath182 polarized laser . the hamiltonian is @xmath183 this model allows quantum coherence between the two - level systems to be preserved . we choose the coupling @xmath184 @xmath185 , with @xmath186 in the presented results . the weak coupling along with the similar physical properties of two subsystems is such that dynamical crossovers in quantum jump trajectories of this system can occur . however the different polarizations of the lasers driving each subsystem affords the possibility of a transition in quadrature trajectories . -quadrature ld function as well as the activity and dynamical variance , for the four - level system depicted in fig . [ fig:13 ] . note that the dynamical variance has been rescaled to be a factor of 5 smaller . ] while the statistics of the @xmath33-quadrature is featureless , fig . [ fig:14 ] shows that the @xmath5-quadrature displays a crossover around @xmath24 . this crossover is smeared out if we increase @xmath184 and becomes sharper if we decrease @xmath184 . the @xmath22 region has a positive @xmath5-activity ; this is due to light emitted from subsystem i. when @xmath23 , the light is emitted from subsystem ii and the @xmath5-activity is negative . we expect that the @xmath24 behaviour is a combination of these two distinct dynamical phases and , to demonstrate this is the case , we construct time - independent probability distributions , or phase portraits , for each dynamical regime . we wish to construct the marginal distributions @xmath135 for rare @xmath5-quadrature trajectories . we will bias the @xmath5-quadrature statistics using a field denoted @xmath187 and measure the typical @xmath79 for all @xmath42 using the conjugate field @xmath0 . to this end we apply the double biasing scheme as in eq . where instead of biasing with the jump operator and measuring @xmath188 , or vice - versa , we now bias the @xmath5-quadrature . this leads to an equivalent doubly - biased master equation @xmath189 where the sum is over the two subsystems . using @xmath190 we bias the system to rare @xmath5-quadrature trajectories and we look at derivatives of the ld function with respect to @xmath0 in the limit @xmath191 to measure the quadratures of these biased trajectories . solving for the probability distributions , for @xmath192 , 0 and @xmath193 , we find that the behaviour of typical ( @xmath24 ) is made of two distinct dynamical phases , as is clear in fig . [ fig:15 ] . this crossover of emission from one subsystem to another is not observable in the jump trajectories and highlights that quadrature activities are not just equivalent to the jump activity as a dynamical order parameter . these quadrature activities may reveal extra dynamical phases which are not visible simply by counting photons . the final example we present is a many - body problem consisting of a set of two - level atoms interacting with a single cavity mode , called a micromaser @xcite . we begin with a brief description of this problem before providing a mean - field treatment to determine the ld function . we then proceed to determine the full ld function with exact numerical diagonalisation and produce quadrature activity phase diagrams for the system . finally , we examine the doubly - biased trajectory spectrum of the system . we then summarise our findings , before moving on to present our conclusions in the final section of this work , sec . [ sec : conc ] . the micromaser is a single - mode resonant cavity coupled to a thermal bath and pumped by excited two - level atoms which pass through the cavity . we denote the total number of atoms which pass through the cavity divided by the cavity lifetime by @xmath194 . the steady - state cavity occupation distribution can change from uni - modal to bi - modal , depending on @xmath194 and the atom - cavity coupling . here , we fix @xmath195 . as we increase the coupling between the cavity mode and the atoms , we reach many points where the steady state cavity occupation number exhibits a bistability : at these points , a small increase in the coupling leads to a crossover in occupation number where the occupation number changes dramatically . previously this bistability in the state of the cavity was shown to have an equivalent dynamical bistability in quantum - jump trajectories in ref . a rich dynamical phase structure was found , where the number of atoms which have changed state while traversing the cavity was used to quantify the dynamical activity . we now turn to the model for the micromaser . tracing out the atom and bath degrees of freedom , one obtains the superoperator , @xmath196 , which contains no coherent evolution term and is of lindblad form that is of the form in eq . with @xmath197 and four pairs of lindblad operators . there are two associated with the atom - cavity interaction @xmath198 and two result from the cavity - bath interaction , @xmath199 here @xmath200 is the accumulated rabi frequency which encodes the atom - cavity interaction , @xmath201 and @xmath202 are the cavity raising and lowering operators respectively . choosing a zero - temperature bath , we will set the thermal occupation number of the bath @xmath203 . we study the quadratures of the light emitted into to the bath and therefore deform the superoperator by the non - equilibrium quadrature field to @xmath204 for this system , as the normal dynamics of the system are purely lindbladian in nature , the ld function for the quadratures is identical for all @xmath42 . therefore in the following discussion we will focus solely on the @xmath5-quadrature ( where @xmath76 ) from here on . . also we note some bending of these transition lines occurs as they approach @xmath24 . the form of the mean - field theory is very similar to one obtained for the atom " counting case of ref . @xcite , highlighting the connection between transitions in quadrature and jump - activities . ] although a full analytic solution for the @xmath24 steady state density operator is well known @xcite , away from @xmath24 it is not easily generalised . in a study of quantum - jump trajectories of this system @xcite progress was made by assuming the eigenmatrix of an @xmath78-dependent superoperator was diagonal in @xmath205 . this approach demonstrated that the cavity pump rate controlled the properties of the coexistence line at @xmath91 and the findings agreed well with exact diagonalisation results . mulitple first order transition lines were observed in the photon active region where @xmath151 , and a single transition was observed in the photon inactive regime where @xmath152 , along with a critical point at @xmath206 and @xmath207 . in this work , the effects of quadrature biasing with @xmath0 using @xmath47 in eq . [ eq:19e ] leads to eigenmatrices which are not diagonal in @xmath205 for general @xmath0 . however close to @xmath24 we approximate the non - diagonal term , @xmath208 , with a diagonal one given by @xmath209 here we have assumed that @xmath210 and @xmath211 are approximately @xmath212 , where @xmath213 is the cavity occupation number . thus for @xmath214 we may approximate the non - diagonal piece by an exponential and the appearance of @xmath125 reflects the fact that @xmath123 must be an even function of @xmath0 . through this crude approximation we have restricted our analysis to density operators diagonal in the number basis , and so the generalized quantum master equation reduces to an operator @xmath215 a variational approach to calculating the largest eigenvalue of @xmath216 in eq . can be constructed using a coherent state ansatz . this amounts to setting @xmath217 and @xmath218 and solving the corresponding euler - lagrange equations , @xmath219 and @xmath220 . solving the first equation one obtains , @xmath221 substituting these into @xmath216 one obtains a variational free energy " , @xmath222 , whose minimum with respect to @xmath213 yields an estimate for the ld function , @xmath223 this minimization process was performed numerically and the result is displayed in fig . [ fig:16 ] . minimization reveals that multiple first - order transitions occur in both the cavity occupation number and quadrature activity . these transition lines begin to bend as they approach @xmath24 . the first transition line ends at @xmath224 , @xmath225 . this point is the critical point discovered in ref . @xcite , which controls the photon number dynamics . here it is much more masked , even within this crude diagonal approximation , than in the study of jump - trajectories of the atoms presented in ref . @xcite . within this diagonal approximation , , there are limitations and further implicit assumptions : the approximation becomes less accurate at larger @xmath200 where non - linearities in the operator eq . are more prominent . there is also an implicit assumption of normal ordering and that the averages of products of the raising and lowering operators may be replaced by products of their averages . despite ignoring the off - diagonal behaviour @xmath47 , this free energy predicts multiple first order transitions in the quadrature activity and , moreover , that these transitions are due to transitions in the occupation number , @xmath213 , of the cavity . we now examine the generalized master equation in order to determine the exact form of the ld function . the generalized master equation may be expressed in terms of a matrix and diagonalized numerically . the method we employ is similar to that in ref . @xcite , and outlined in appendix [ sec : diag ] . the full phase diagram is shown in figure [ fig:17](b ) , where we see that the critical point previously discussed manifests itself as the only point where the first order transition lines accumulate at @xmath24 . beyond @xmath207 the transition lines begin to bend in such a manner as not to accumulate at the @xmath24 line . it is clear however that the mean - field theory in fig . [ fig:16 ] predicts values of @xmath200 where the transitions occur up to @xmath226 . these transitions correlate with the cavity - occupation plot shown in fig . [ fig:18](b ) , demonstrating that these first order transitions in quadrature activity are also associated with the bistability of the cavity photon number @xcite . we now examine the statistics of double biasing , using the field @xmath78 to place the system in higher or lower photon number state and then examining the typical quadrature realisations . notice that the transition lines shift such that they either accumulate more at @xmath24 or bend away further before reaching @xmath24 when placing the system in a high ( @xmath151 ) or low ( @xmath152 ) cavity number respectively . placing the system in a @xmath170-active phase removes lower - cavity - number regions near @xmath24 , visible in fig . [ fig:18](b ) , causing the quadrature activity lines to meet at @xmath24 . conversely , biasing the dynamics towards @xmath170-inactive trajectories ensures that the cavity occupation is small unless a large @xmath0-bias is applied . this can be seen in the formation of a band around @xmath24 where both the quadrature activity ( fig . [ fig:17 ] ) and cavity occupation ( fig . [ fig:18 ] ) are small . although sharp crossovers occur in the quadrature activity , corresponding to crossovers in the cavity occupation , it is not clear that the @xmath24 dynamics are made up of two distinct phases . these phases are only apparent by examining @xmath170 . furthermore , even though the critical point exists in both the @xmath170 and @xmath107 phase diagrams , it is much easier to discern in the former . to conclude , we find that quadrature measurements in micromaser can be used to identify dynamical phases , but in sharp contrast to the four - level system in [ sec:2l2c ] , the study of quadrature trajectories does not give significant insights into the system dynamics beyond those available by counting photons . in this paper we developed an @xmath0-ensemble approach to gain insight into the behaviour of light emitted from several quantum - optical systems . previous studies used the @xmath0-ensemble to describe the dynamics of quantum - jump trajectories and understand the statistics of photon emissions in open quantum systems . we reformulated the approach in terms of generating functions and characteristic operators , treating the reservoir stochastically . using this approach we examined the statistics of the quadratures of light emitted from various systems and re - constructed appropriate phase space portraits and wigner functions which describe the state of the emitted light . we extended this scheme to examine the light not only emitted during rare quantum - jump trajectories where , for example , the number of emitted photons is larger or smaller than for average dynamics . we also examined the quantum - jump statistics of the quadrature - biased trajectories , ascertaining the observable photon count when the quadratures of emitted light depart significantly from their dynamical averages . we used these results to understand whether different dynamical order parameters are correlated , and to select the most appropriate @xmath0-field and order parameter with which to characterise dynamical phases . our results show that even simple quantum - optical models such as two- and three - level systems show interesting features in the space of quadrature trajectories . for example , for the case of the blinking three - level system , quadrature phase - space portraits indicate that emitted light may be identified with a mixture of the light from two simple active and inactive subsystems , which is consistent with the dynamical coexistence description in terms of number of emitted photons described in ref . @xcite . for the case of the two - level system , our analysis here shows that the ` special point ' in parameter space found in ref . @xcite is also special for another reason : it is the point where one may measure the @xmath33-quadrature activity @xmath40 in place of the quantum - jump activity @xmath227 of the physical system , as they are identical in magnitude : for this system @xmath173 represents an alternative dynamical order parameter , as biasing trajectories towards larger @xmath173 leads to smaller @xmath228 , and vice versa . examination of a pair of coupled two - level systems showed that , in some circumstances , the quadrature activity reveals a phase structure which is completely unobservable through photon count statistics . this highlights the usefulness of studying quadrature trajectories , and illustrates that greater inference about an open system can be obtained by such measurements . however , it also highlights that the choice of dynamical order parameter is highly system dependent . we further illustrated the difficulty in the choice of order parameter by studying quadratures in the micromaser . the phase diagram for the quadrature dynamics is rich and displays many features of the quantum - jump activity phase diagram studied in ref . @xcite , such as numerous first order transition lines . however , the lack of direct control over the cavity - state bistability means one can not simply describe the @xmath24 dynamics in terms of two distinct quadrature phases . finally , we emphasize the significance of our results in light of homodyne detection schemes , which allow @xmath5-quadrature trajectories to be measured . furthermore , we note that the @xmath33-quadrature statistics discussed are accessible to experimental measurement with a suitable modification to the driving laser polarisation . we also add that quantum tomographic techniques allow reconstruction of the wigner functions of such systems . we further note that the methodology presented in this work can be extended to other more complex measurement schemes with other unravellings of the quantum dynamics for different choices of bath measurement operators with both discrete and continuous spectra . we are grateful to madalin guta for important discussions . this work was supported by the leverhulme trust grant no . f/00114/b6 and epsrc grant no . ep / i017828/1 . using a weakly coupled markovian reservoir one may consider that the ladder lowering operator of the reservoir acts as a driving field for the time - evolution operator of the total system , and so the effects of the reservoir ladder operator fields permit a description in terms of an input - output formalism @xcite . in general , such a formalism works on the foundation that outside the range of interaction , the field which interacts with the system is the sum of the input field , that is the field just prior to interacting with the system , and the output field , the state of the field just after interaction with the system . the reservoir ladder operators , @xmath48 and @xmath229 , are the fields interacting with the system , the time @xmath8 is the initial time when the fields interact with the system and , due to our choice of a reservoir which admits a fock space representation , one may use a stochastic description of the system - plus - bath dynamics . although the input - output formalism is intuitive , especially if one is working with quantum optics where the input / output fields are manifested as light beams , the stochastic description @xcite of the system allows for a concise mathematical representation where the effects of the reservoir on the system are reduced to an it increments which obey a quantum it calculus . before we proceed with the derivations of the generalized master equation for biased quadrature ensembles , it is necessary to outline the calculus of the reservoir increments , as well as the quantum - jump @xmath70 process . working with an unsqueezed vacuum , the only non - zero combination of @xmath230 and @xmath231 is : @xmath232 with this it table we can deduce the appropriate table for the jump trajectory process @xmath70 , @xmath233 now , using the definition of @xmath45 , we examine the increment of the @xmath0-biased density matrix @xmath234 = { \text{tr}}_{\text{res}}\left(d[{v}_{q}^{s}\left(t\right)]\rho + { v}_{q}^{s}\left(t\right)d[\rho ] + d[{v}_{q}^{s}\left(t\right)]d[\rho]\right)\,.\ ] ] the first two terms appear in standard calculus while the final term appears due to the stochastic nature of these increments . we have already defined @xmath235 in eq . , and so we need to examine the increment of the characteristic operator . expanding @xmath236 , defined in , and using the calculus set out in eq . one finds @xmath237 = { v}_{x^\alpha}^{s}\left(t\right)\left(\frac{{s}^{2}}{8}dt -\frac{s}{2}d{x}^{\alpha}\right).\ ] ] we may now evaluate the expansion on the right - hand side of , term by term , to obtain @xmath238\rho\right)= \frac{s^{2}}{8}{\rho}^{s}dt , \nonumber \\ & { \text{tr}}_{\text{res}}\left({v}_{x^\alpha}^{s}\left(t\right)d[\rho]\right)= \mathcal{l}\left({\rho}^{s}\right)dt,\nonumber \\ & { \text{tr}}_{\text{res}}\left(d[{v}_{x^\alpha}^{s}\left(t\right)]d[\rho]\right)=-\sum_{i}\frac{s}{2}\left({e}^{-i\alpha}{l}_{i}{\rho}^{s}+{e}^{i\alpha}{\rho}^{s}{l}_{i}^{\dag}\right)dt , \end{aligned}\ ] ] where the @xmath239 are the lindblad operators of the system coupled to the environment . we can now see the increment equation leads to eq . . this procedure may be extended straightforwardly to the doubly - biased cases . in the case of the biasing photon trajectories and examining the typical quadrature statistics ( eq . ) one introduces the jump - trajectory characteristic operator @xmath240 as in eq . and defines its increment @xmath241 = { v}^{{s}'}_{k}\left(t\right)\left({e}^{-{s}'}-1\right)dk.\ ] ] now we proceed as before using the it calculus set out in eqns . and to arrive at the result eq . . finally we note that , although introduced for a specific set - up , this formalism is not restricted to pure states and allows a freedom of choice for the reservoir provided we stay within the markovian approximation . for example , it may be readily generalized to a squeezed thermal reservoir . in the analysis of quadrature statistics , we consider the set of marginal distributions @xmath135 , associated with the general quadratures defined in eq . , where @xmath42 varies over the range @xmath242 $ ] . we also employ the quasi - probability distribution of the light , known as the wigner distribution @xcite . we use these distributions to study the light in the reservoir coupled to the system . taking the density matrix of the reservoir , @xmath243 , the wigner function in the schrdinger representation is @xmath244 the variable @xmath245 is identified with the quadratures through the relations @xmath246 as direct evaluation of this distribution is difficult we will use the fact that @xmath247 is related to the marginals @xmath135 by the radon transform @xcite , @xmath248 , via @xmath249\\ & = \int^{+\infty}_{-\infty}{w}_{\rho}\left(x\cos\alpha\!-\ ! y\sin\alpha , y\cos\alpha\!+\ ! x\sin\alpha\right)dy\nonumber\end{aligned}\ ] ] and we employ its numerical inverse to the marginals @xmath250 to obtain @xmath247 . this technique is frequently used in quantum tomography @xcite . the wigner distribution allows classification of the light into quantum and classical components based on the sign of the distribution : negative regions are indicative of quantum effects such as entanglement . to determine the ld function of the micromaser numerially , we numerically diagonalized the generalized master operator @xmath47 in matrix form , as in ref . @xcite . it is necessary to truncate the basis of the system . the basis of number states @xmath251 is suitable and , for the @xmath194 value of 100 studied in this work , we restrict the maximum photon number to @xmath252 . the form of @xmath47 introduces coherences between different eigenstates of the cavity and so we must ensure our basis allows these coherences to be preserved . however , since in matrix form the representation of the superoperator is an @xmath253 matrix , we investigated whether further truncation was possible . in practice it is possible to truncate the basis in such a manner that only coherences between number states with occupation number differing by @xmath254 ( with @xmath255 ) are preserved . the value of @xmath254 may be tested numerically to ensure the results are not sensitive to this truncation . in the results presented in this work , the values of @xmath252 and @xmath256 were found to be sufficient for the diagonalization process . the largest real eigenvalue extracted from this matrix is identified as @xmath158 . the matrices were diagonalized using an arnoldi iteration scheme @xcite . the activity was calculated by taking numerical derivatives of the resulting ld function .

this paper develops a geometric model for coupled two - state quantum systems ( qubits ) using geometric ( aka clifford ) algebra . it begins by showing how euclidean spinors can be interpreted as entities in the geometric algebra of a euclidean vector space . this algebra is then lifted to minkowski space - time and its associated geometric algebra , and the insights this provides into how density operators and entanglement behave under lorentz transformations are discussed . the direct sum of multiple copies of space - time induces a tensor product structure on the associated algebra , in which a suitable quotient is isomorphic to the matrix algebra conventionally used in multi - qubit quantum mechanics . finally , the utility of geometric algebra in understanding both unitary and nonunitary quantum operations is demonstrated on several examples of interest in quantum information processing .

quant-ph0004031

quantum mechanics attaches physical significance to representations of the rotation group which differ substantially from those studied in classical geometry . much of the mystery surrounding it is due to this fact . the enormous interest recently generated by proposals to build a _ quantum computer _ @xcite has focussed attention on the simplest possible quantum system : a two - state system or _ qubit_. our understanding of qubits is based on two distinct geometric models of their states and transformations : * a complex projective line under the action of @xmath0 . * a euclidean unit @xmath1-sphere under the action of @xmath2 . the first is used almost exclusively in fundamental quantum physics , while the second ( `` classical '' ) model is used extensively in certain applications , e.g. nuclear magnetic resonance ( nmr ) spectroscopy @xcite . in particular , in quantum computing a qubit represents a binary @xmath3 or @xmath4 as its state corresponds to one of a pair of conjugate ( orthogonal ) projective points @xmath5 or @xmath6 . these in turn correspond to a pair of diametrical points on the unit sphere , which determine the alignment of the qubit with or against the corresponding axis of quantization . formally , these two models are related by stereographic projection of the riemann ( unit ) sphere onto the argand plane , the points of which are the ratios of the homogeneous coordinates of points on the projective line ( see e.g. @xcite ) . while this elegant construction describes the mapping between the two representations in a geometric fashion , it does not unite them in a single mathematical structure . this paper provides an informal account of how this is done by geometric ( aka clifford ) algebra ; in addition , it describes an extension of this formalism to multi - qubit systems , and shows that it provides a concise and lucid means of describing the operations of quantum information processing @xcite . significantly , this extension is most naturally derived via the geometric algebra of minkowski space - time @xcite , which has also been shown to be an efficient formalism within which to study a very wide range of problems in classical , relativistic @xcite and fundamental quantum @xcite physics . more complete and rigorous accounts may be found in these references , and in @xcite . let @xmath7 be a three - dimensional euclidean vector space whose inner product is denoted by @xmath8 . the clifford or _ geometric algebra _ of @xmath7 is the associative algebra generated by @xmath7 over @xmath9 such that @xmath10 for all @xmath11 . this algebra will be referred to in the following as the _ pauli algebra _ , and denoted by @xmath12 . the interesting thing about this algebra is its geometric interpretation , which will now be described . to begin , note that every nonzero vector @xmath13 has an inverse @xmath14 . in addition , a simple application of the law of cosines shows that the inner product of @xmath15 with any other vector @xmath16 is given by the symmetric part of the geometric product : @xmath17 the antisymmetric part , by way of contrast , is called the _ outer product _ , and denoted by @xmath18 . since the outer product of two vectors @xmath19 is invariant under inversion in the origin , it can not itself be a vector . the space @xmath20 therefore carries an inequivalent representation of the orthogonal group @xmath21 , and its elements are accordingly called _ bivectors_. these are most naturally interpreted as oriented plane segments , instead of oriented line segments like vectors in @xmath22 . if we similarly define the outer product of a vector with a bivector and require it to be associative , i.e. @xmath23 ( @xmath24 ) , then it can be shown via straightforward though somewhat lengthy calculations that this product of three vectors is totally antisymmetric , meaning that the outer product generates the well - known exterior algebra @xmath25 ( cf . ) . the outer product of three vectors is called a _ trivector _ , and ( since it changes sign under inversion ) is most appropriately interpreted as an oriented space segment or volume element . the general properties of inner and outer products in the geometric algebras of arbitrary metric vector spaces can be worked out along these lines in a coordinate - free fashion @xcite . the remainder of this section will focus on how the pauli algebra is used to describe the quantum mechanics of qubits . in this application it is more common to work with a fixed orthonormal basis @xmath26 . quantum mechanics , however , views these basis vectors in a very different way from that taken above , in that they are regarded as _ operators _ on a two - dimensional hilbert space @xmath27 ( see e.g. @xcite ) . these operators , in turn , are usually identified with the pauli matrices @xmath28 where @xmath29 is an imaginary unit ( @xmath30 ) , the underline signifies that the associated symbol is a matrix , and throughout this paper the symbol `` @xmath31 '' should be read as `` is represented by '' or `` is equivalent to '' . the connection between the two viewpoints lies the fact that these matrices satisfy the defining relations of the _ abstract _ pauli algebra @xmath12 , namely @xmath32 and hence constitute a faithful matrix representation of it . this shows , in particular , that @xmath12 is @xmath33-dimensional as a _ real _ linear space . is often used for @xmath34 . because @xmath15 is geometrically just a vector in @xmath7 ( _ not _ a matrix for it in a basis - dependent representation of @xmath35 ) , this is an abuse of the dot - notation for the euclidean inner product , which is otherwise perhaps the most consistently used notation in all of science . this abuse of notation will not be perpetrated in this paper . ] in most physical situations , these operators ( times @xmath36 ) represent measurements of the intrinsic angular momentum of the qubits , and hence are regarded as generators of rotations in the lie algebra @xmath37 over @xmath38 satisfying the commutator relation @xmath39 ~=~ \imath { { { \boldsymbol}\sigma\!}}_{{\mathsf{z } } } ~,\ ] ] and its cyclic permutations . in terms of geometric algebra , the left - hand side is just the outer product of the vectors . the right - hand side is somewhat harder to interpret , because the pauli algebra is defined over the real numbers . the trick is to observe that , in terms of the matrix representation , @xmath40 . thus by interpreting the abstract imaginary @xmath29 as the trivector @xmath41 , the angular momentum relations become a triviality : @xmath42 more generally , the vector cross product is related to the outer product by @xmath43 from which it may be seen that multiplication by the unit trivector @xmath44 maps vectors to orthogonal bivectors and vice versa . since they span a one - dimensional space but change sign under inversion in the origin , trivectors can also be regarded as _ pseudo - scalars_. perhaps the most important thing which geometric algebra contributes to physics are geometric interpretations for the imaginary units which it otherwise uses blindly . if we denote the induced bivector basis by @xmath45 it is readily seen that these basis bivectors likewise square to @xmath46 . on multiplying the angular momentum generating relations through by @xmath47 , we obtain @xmath48 this shows that these basis bivectors generate a subalgebra of @xmath12 isomorphic to hamilton s quaternions @xcite , which is also known as the _ even subalgebra _ @xmath49 ( since it is generated by the products of even numbers of vectors ) . it is well - known that the quaternions multiplicative group is @xmath50 , which implies that the even subalgebra should be closely related to rotations . this relationship will now be worked out explicitly . consider the result of conjugating a vector @xmath51 by a vector @xmath15 , i.e. @xmath52 where we have split @xmath53 into its parts parallel and perpendicular to @xmath15 . this shows that @xmath54 is the _ reflection _ of @xmath51 in the plane orthogonal to @xmath15 . from the well - known fact that the composition of two reflections is a rotation by _ twice _ the lessor angle between their planes and about these planes line of intersection , it follows that conjugating a vector by an element of the even subalgebra just rotates it accordingly : @xmath55 let @xmath56 , @xmath57 and @xmath58 be the corresponding _ unit _ quaternion . then @xmath59 where @xmath60 and @xmath61 . moreover , the inverse @xmath62 is now simply the _ reverse _ @xmath63 , which in turn corresponds to the _ conjugate _ quaternion @xmath64 . this reversal operation on @xmath49 extends to a well - defined anti - automorphism of @xmath65 , which corresponds to hermitian conjugation in its representation by pauli matrices . on splitting @xmath51 into its parts parallel @xmath66 and perpendicular @xmath67 to @xmath68 as above , the rotation may now be written as @xmath69{0pt}{1.6ex}}^2 $ ] @xmath70 and so may be viewed as multiplication of @xmath67 by the `` complex number '' @xmath71 in the argand plane defined by the bivector @xmath72 . by collecting even and odd powers in its taylor series , it may be seen that any unit quaternion can be written as the exponential of a bivector orthogonal to the axis of rotation @xmath68 : @xmath73 this is formally analogous to a complex exponential , and is also in accord with our previous observation that the space of bivectors is isomorphic to the lie algebra @xmath74 under the commutator product . the pair @xmath75 $ ] are known as _ euler - rodrigues _ parameters for the rotation group @xmath2 ; since @xmath76 $ ] determines the same rotation , this parametrization is two - to - one . a one - to - one parametrization is obtained from the _ outer exponential _ , i.e. @xmath77 the squared norm of this outer exponential is @xmath78 , so that the normalized outer exponential equals the usual exponential if we set @xmath79 . because @xmath80 is the four - dimensional stereographic projection of @xmath81 $ ] from @xmath82 , it has been called the _ stereographic _ parameter for @xmath2 . note , however , that this parametrization does not include rotations by @xmath83 . another two - to - one parametrization of rotations is given by the _ cayley - klein _ parameters @xmath84 \in { { \mathbb{c}}}^2 $ ] , where @xmath85 the corresponding @xmath0 matrix is simply @xmath86 it follows that the complex column vector @xmath87 $ ] itself transforms under left - multiplication with matrices in @xmath0 , which is commonly described in quantum mechanics by calling it a _ spinor_. in particular , the spinors @xmath88 $ ] and @xmath89 $ ] are those commonly used in quantum computing to store binary information . since the cayley - klein parameters uniquely determine the @xmath0 matrix , however , we can just as well regard spinors as entities _ in _ @xmath0 , e.g. @xmath90 and @xmath91 . the usual action of @xmath0 on spinors then becomes the left - regular action of @xmath0 on itself . the representation of @xmath0 used above depends upon the choice of coordinate system : changing to a different the coordinate system gives a different ( though equivalent ) representation . recalling that @xmath0 is isomorphic to the multiplicative group of unit elements ( quaternions ) in the even subalgebra @xmath49 , a coordinate - free or _ geometric _ interpretation of spinors is obtained by regarding them as elements of @xmath49 itself . this interpretation of spinors as entities in ordinary euclidean geometry was first pointed out by hestenes over thirty years ago @xcite , but physicists persist in putting operators and operands into separate spaces , and in working with a matrix representation instead of directly with the geometric entities themselves . the perceived nonintuitive nature of quantum mechanics is due in large part to the resulting confusion over the geometric meaning of the objects with which it deals , which is spelled out explicitly in geometric algebra . as another example , consider how the _ density operator _ of an `` ensemble '' of qubits can be interpreted in geometric algebra . this operator @xmath92 is usually defined via a matrix representation as @xmath93 , where the overline denotes the average over the ensemble . as first observed by von neumann , this matrix contains all the information needed to compute the ensemble average expectation values of the qubit observables , since @xmath94 ( @xmath95 ) . to translate this into geometric algebra , we set the second column of @xmath96 to zero by right - multiplying it by the idempotent matrix @xmath97 , i.e. @xmath98 this corresponds to projecting @xmath99 onto a _ left - ideal _ in @xmath65 , and allows the dyadic product @xmath100 in eq . ( [ eq : rho1_obs ] ) to be written as : @xmath101 thus the interpretation of the density operator in geometric algebra is @xmath102 ( cf . the vector part @xmath103 is called the _ polarization vector _ ( in optics , its components are known as the _ stokes parameters _ @xcite , while in nmr it is known as the _ bloch vector _ after the pioneer of nmr who rediscovered it @xcite ) . its length is @xmath104 with equality if and only if all members of the ensemble are in the same state @xmath105 . in this case the ensemble is said to be in a _ pure _ state , and the density operator is itself an idempotent @xmath106 , where @xmath107 . for an ensemble in a general _ mixed _ state , the length of the ensemble - average polarization vector measures the degree of alignment among the ( unit length ) polarization vectors of the individual members of the ensemble , and is called the _ polarization _ of the ensemble . in many physical situations there is a natural reference direction ; for example , in nmr computing the qubits are spin @xmath108 atomic nuclei whose intrinsic magnetic dipoles have been polarized by the application of a strong magnetic field @xcite . from a geometric perspective , however , the density operator is just the sum of a scalar and a vector , which for a pure state is related to the corresponding `` spinor '' by rotation of a fixed reference vector ( conventionally taken to be @xmath109 as above ) by @xmath105 . since the trace in the standard matrix representation is simply twice the scalar part @xmath110 of the corresponding expression in geometric algebra , the ensemble - average expectation value @xmath111 is just the component of the polarization vector along the @xmath112-th axis . unlike the _ strong _ measurements usually considered in quantum texts , where measurement of @xmath113 yields one of the random outcomes @xmath114 with probabilities @xmath115 and leaves the system in the corresponding state @xmath116 , _ weak _ measurements of ensemble - average expectation values can be made with only negligible perturbations to the ensemble as a whole @xcite this is in fact how quantum mechanical systems are usually manifest at the macroscopic level ! to see how all this relates to conventional wisdom , observe that the polarization vector of a pure state may be written in terms of the cayley - klein parameters as @xmath117 its stereographic projection from @xmath118 onto the @xmath119 plane is therefore @xmath120 multiplying by @xmath121 and simplifying the denominator using @xmath122 yields @xmath123 where @xmath124 is a square - root of @xmath46 . this is the same as the ratio @xmath125 save for the use of @xmath29 instead of @xmath126 as the imaginary unit , which explains formally why @xmath2 acts on the polarization vector in the same way that @xmath0 acts on the ratio of the cayley - klein parameters @xcite . the above interpretations apply only to single qubits ( or to ensembles consisting of noninteracting and identical qubits ) . extending them to systems of interacting and distinguishable qubits may be done in a physically significant fashion by considering the geometric algebra of _ space - time _ ( or minkowski space ) @xmath127 . this algebra , known as the _ dirac _ algebra and denoted by @xmath128 , may be defined by the generating relations among an orthonormal basis analogous to eq . ( [ eq : pauli_def ] ) : @xmath129 the corresponding geometric algebra separates into five inequivalent representations under the action of the full lorentz group @xmath130 , i.e. @xmath131 where @xmath132 with @xmath133 , for a total dimension of @xmath134 . the important point for our purposes is that the even subalgebra of the dirac algebra @xmath135 is isomorphic to the pauli algebra @xmath12 @xcite . this isomorphism may be constructed by choosing bases @xmath136 and @xmath137 , and defining an invertible linear mapping by @xmath138 these so - called _ relative spatial vectors _ @xmath139 satisfy the relations in eq . ( [ eq : pauli_def ] ) , since @xmath140 and hence generate an algebra isomorphic to @xmath12 . as bivectors in @xmath128 , however , they also generate @xmath135 , since @xmath141 ( @xmath142 , @xmath143 ) , and similarly @xmath144 thus @xmath145 ( @xmath146 ) induces an algebra isomorphism as claimed , and when the bases are understood we may identify @xmath147 . the choice of time - like vector @xmath148 in fact determines an inertial frame up to spatial rotation , in which the _ time _ @xmath149 and _ place _ @xmath150 of an event @xmath151 in that frame are given by @xmath152 ( note that upright case is used for the space - time vector @xmath153 ) . thus the invariant interval between events separated by the space - time vector @xmath151 is @xmath154 as usual , while the relative velocity between events whose space - time velocities are @xmath155 and @xmath156 is @xmath157 so that @xmath158 lies on an affine hyperplane in space - time . a great deal of physics can be done in a manifestly lorentz covariant fashion using the dirac algebra . for example , the electromagnetic field at a given point in space - time corresponds to an arbitrary bivector @xmath159 , called the _ faraday bivector _ , and the covariant form of the lorentz force equation is @xmath160 where @xmath161 is the rest mass , @xmath162 the charge and @xmath163 the space - time velocity . ( this is another example of the general rule that , in geometric algebra , the generators of motion are bivectors @xcite . ) the usual frame - dependent form is recovered by splitting the quantities in this equation by @xmath155 as above @xcite ; in particular , the faraday bivector splits into an electric and a magnetic field as @xmath164 , where @xmath165 the space - time reverse will be denoted by a tilde , e.g. in the present case @xmath166 . this is related to the spatial ( or pauli ) reverse by @xmath167 . both operations agree on the pauli - even subalgebra , but the spatial reverse _ not _ lorentz coveriant since it depends on a particular @xmath168 . returning to our previous discussion of the density operator , we observe that the space - time form of the density operator of a single qubit polarized along @xmath169 can be written as @xmath170 where @xmath171 is the polarization and @xmath155 determines the local inertial frame . it follows that the lorentz covariant form of the density operator is a time - like vector @xmath172 . under a lorentz boost @xmath173 along @xmath174 , therefore , the relativistic density operator @xmath175 transforms to @xmath176 @xmath177 this implies that in the unaccelerated frame ( with renormalization by @xmath178 ) , @xmath179 it follows that the polarization itself transforms as @xmath180 if we assume the qubit is at equilibrium with a heat bath , statistical mechanics tells us that @xmath181 where @xmath182 is the inverse temperature and @xmath183 is the energy difference between the @xmath5 and @xmath6 states @xcite . then the addition formulae for @xmath184 and @xmath185 give @xmath186 so the apparent equilibrium polarization depends on velocity . these results are not to be found in the classic treatise on relativistic thermodynamics @xcite . we will now construct a lorentz covariant _ multiparticle theory _ of qubit systems in the simplest possible way , by taking a direct sum of copies of space - time ( regarded as a vector space , rather than an algebra ) , one for each of the @xmath187 qubits , i.e. @xmath188 and considering the associated geometric algebra @xmath189 . then the even subalgebras of different particle spaces @xmath190 _ commute _ , since ( in any given bases ) @xmath191 for all @xmath142 , so that the algebra generated by the even subalgebras is isomorphic to a tensor product of these algebras , written as @xmath192 this construction of the tensor product was first used by clifford as a means of studying the tensor products of quaternion algebras @xcite ; van der waerden has in fact called it a clifford algebra of the second kind @xcite . as a means of justifying the tensor product of nonrelativistic quantum mechanics in terms of the underlying geometry of space - time , however , it is a much more recent development @xcite . a key feature of quantum mechanics , which is needed for quantum computers to be able to solve problems more efficiently than their classical counterparts , is an exponential growth in the dimension of the hilbert space of a multi - qubit system with the number of particles involved . the complex dimension of the hilbert space @xmath193 of an @xmath187-qubit system is in fact @xmath194 , and the space of operators ( linear transformations ) on @xmath193 therefore has _ dimension @xmath195 . the above construction yields a space of `` operators '' @xmath196 whose real dimension also grows exponentially , but as @xmath197 . the extra degrees of freedom are due to the presence of a different unit pseudo - scalar @xmath198 in every particle space . they can easily be removed by multiplying through by an idempotent element called the _ correlator _ : @xmath199 this commutes with everything in @xmath196 and satisfies @xmath200 for @xmath201 , so that multiplication by it homomorphically maps @xmath202 onto an ideal @xmath203 wherein all the unit pseudo - scalars have been identified , is justified by the fact that the two - sided principle ideal @xmath204 generated by @xmath205 is isomorphic to the quotient algebra @xmath206 , where @xmath207 . ] and which therefore has the correct dimension over @xmath9 . as a subalgebra , this ideal is in fact isomorphic to the algebra of @xmath208 complex matrices , and hence capable of describing all the states and transformations of ( ensembles of ) @xmath187 qubit systems . in the following , we shall generally omit @xmath205 from our expressions altogether , and use a single unit imaginary @xmath44 as in conventional quantum mechanics . on the `` even '' subalgebra @xmath209 , multiplication by the correlator turns out to be an algebra automorphism ; this algebra can thus be written as @xmath210 where the `` @xmath211 '' refers throughout to the subalgebra generated by expressions which are invariant under inversion in the origin , and @xmath212{0pt}{1.6ex}}^{\otimes\smash n}$ ] to the algebra generated over @xmath9 by the kronecker products of matrices in the group @xmath0 . this subalgebra has real dimension @xmath213 , but is mapped onto a left - ideal of dimension @xmath214 by right - multiplication with another idempotent which is given by the tensor product of those considered earlier , namely @xmath215 where @xmath216 for @xmath217 . henceforth , the term `` even subalgebra '' will refer to @xmath218{0pt}{1.6ex}}^{\otimes\smash n}$ ] ( suitably correlated ) unless otherwise stated . in terms of the usual matrix representation , right - multiplication of an element of the even subalgebra @xmath96 by @xmath219 likewise sets all but the first column to zero , so that @xmath220 transforms like a `` spinor '' in @xmath221 under left - multiplication by single particle rotations @xmath222{0pt}{1.6ex}}^{\otimes\smash n}$ ] . unlike the single particle case , however , this one column does not uniquely determine an element of the even subalgebra @xmath218{0pt}{1.6ex}}^{\otimes\smash n}\!/{{{\boldsymbol}{c}}}$ ] . what has been proposed instead @xcite is to use the fact that @xmath223 `` absorbs '' @xmath109 s to distribute copies of the latter across the correlator , converting it to what will here be called the _ directional _ correlator @xmath224 , i.e. @xmath225 where @xmath226 it can be shown that right - multiplication by @xmath224 , unlike @xmath205 , reduces the dimensionality to @xmath214 , thereby permitting the objects in this _ reduced _ even subalgebra @xmath218{0pt}{1.6ex}}^{\otimes\smash n } \!/ { { { \boldsymbol}{d}}}$ ] to be regarded as spinors , analogous to @xmath49 for a single qubit . in the corresponding left - ideal , @xmath227 serves as the unit imaginary , since @xmath228 , but is required to always operate from the _ right_. henceforth , unless otherwise mentioned , we will regard spinors @xmath218{0pt}{1.6ex}}^{\otimes\smash n } \!/ { { { \boldsymbol}{d } } } = ( { ( { { \mathcal{g}}}_3^+)}{\rule[0pt]{0pt}{1.6ex}}^{\otimes\smash { { { \boldsymbol}{c } } } ) / ( { { { \boldsymbol}{cd}}})$ ] as a left - ideal in the @xmath205-correlated even subalgebra , drop both @xmath205 and the superscripts on the @xmath44 s as above , and use @xmath224 as a short - hand for @xmath229 . in the case of two qubits , for example , the identifications are induced by @xmath224 are @xmath230 ( where the two columns differ by operation with @xmath126 ) . from this it may be seen that any `` spinor '' in @xmath231 can be written as @xmath232{0pt}{10pt } } \right . \ ! ( \alpha_0 + \beta_0 { { { \boldsymbol}{k } } } ) - { { { { \boldsymbol}{\iota}}}}{{{\boldsymbol}\sigma\!}}_{{\mathsf{y}}}^{\,2 } ( \alpha_1 + \beta_1 { { { \boldsymbol}{k } } } ) \ , - \\ & \left . { \rule[0pt]{0pt}{10pt } } { { { { \boldsymbol}{\iota}}}}{{{\boldsymbol}\sigma\!}}_{{\mathsf{y}}}^{\,1 } ( \alpha_2 + \beta_2 { { { \boldsymbol}{k } } } ) + { { { { \boldsymbol}{\iota}}}}{{{\boldsymbol}\sigma\!}}_{{\mathsf{y}}}^{\,1 } { { { { \boldsymbol}{\iota}}}}{{{\boldsymbol}\sigma\!}}_{{\mathsf{y}}}^{\,2 } ( \alpha_3 + \beta_3 { { { \boldsymbol}{k } } } ) \right ) { { { \boldsymbol}{d } } } \end{split}\ ] ] ( cf . @xcite ) . alternatively , again using eq . ( [ eq : dequiv2 ] ) , a unit norm spinor may be factorized into a product of entities in the _ correlated and reduced _ even subalgebra , namely @xmath233 , where @xmath234 thus when @xmath235 , the factor @xmath236 becomes @xmath237 so that the arguments of the exponentials involving @xmath238 and @xmath239 in the first two factors are shifted by @xmath240 while the total phase is shifted by @xmath241 . it follows that @xmath236 rotates the first two factors in the planes defined by their conjugate spinors @xmath242 , [ -s_2^ * ; s_1^*]$ ] . thus on right - multiplying by @xmath223 and expanding in the usual basis , we obtain ( up to an overall phase ) @xmath243 this is known as the _ schmidt decomposition _ @xcite . it is useful in studying the _ entanglement _ of bipartite quantum systems , which ( in conventional terms ) means that @xmath244 can not be written as a product @xmath245 for any one - particle spinors @xmath246 . in fact it is just the singular value decomposition in disguise , since ( for example ) on arranging the entries of a two - qubit spinor @xmath247 $ ] in a @xmath248 matrix , we can write @xmath249 where @xmath250 is a @xmath248 diagonal matrix containing the singular values @xmath251 and @xmath252 , @xmath253 are unitary matrices with columns @xmath254 , @xmath255 , respectively . since the entries of the dyadic products @xmath256 , @xmath257 are exactly the same as the kronecker matrix products @xmath258 , @xmath259 , the equivalence with eq . ( [ eq : schmidt ] ) follows with @xmath260 , @xmath261 , and the kronecker products of the columns of @xmath252 and @xmath253 identified with conjugate pairs of single qubit spinors whose relative phases are given by @xmath262 . clearly a two - qubit spinor is unentangled if and only if @xmath263 , which is equivalent to @xmath264 or @xmath265 . thus @xmath236 describes the entanglement of the qubits , and is accordingly called the _ tangler_. the geometric algebra approach clearly provides deeper insight into the structure of entanglement than does one based on mechanical matrix algebra . in particular , the fact that @xmath266 is even and reversion - symmetric in the dirac as well as the pauli algebra implies that it is the sum of a scalar and a four - vector in the two - particle dirac algebra @xmath267 . since lorentz transformations of the spinors cancel , this entity is in fact a lorentz invariant , and dividing out the total phase @xmath268 as @xmath269 yields the _ square _ of the tangler directly . the availability of such powerful methods of manipulating entities in the multiparticle dirac algebra promises to be useful in finding analogs of the schmidt decomposition for three or more qubits . quantum computers operate on information stored in the states of quantum systems . the systems are usually assumed to be arrays of distinguishable qubits ( two - state subsystems ) , whose basis states @xmath5 and @xmath6 correspond to the binary digits @xmath3 and @xmath4 , respectively , while the operations are usually taken to be unitary . general unitary transformations of the qubits are built up from simpler ones that affect only a few qubits at a time , which are called _ quantum logic gates_. the representation of these gates in suitable products of clifford algebras has been described in refs . the goal here will be to show how gates act upon spinors in the even subalgebra , and how they can be extended to a wider class of nonunitary quantum operations on density operators . given the isomorphism between the algebra of @xmath208 matrices over @xmath38 and @xmath203 relative to a choice of basis in each particle space , it is straightforward to interpret matrices in the former as geometric entities in the latter . a matrix @xmath270 , however , does not generally correspond to an entity @xmath271 in the _ even _ subalgebra @xmath218{0pt}{1.6ex}}^{\otimes\smash n } \!/ { { { \boldsymbol}{c}}}$ ] , so that @xmath272{0pt}{1.6ex}}^{\otimes\smash n } \!/ { { { \boldsymbol}{d}}}$ ] for a general spinor @xmath273{0pt}{1.6ex}}^{\otimes\smash n } \!/ { { { \boldsymbol}{d}}}$ ] . nevertheless , letting @xmath274 be the idempotent `` opposite '' to @xmath223 , and noting that this satisfies @xmath275 , the product of @xmath276 with @xmath223 may be written as @xmath277 where the `` hat '' on @xmath278 denotes its image under inversion in the origin ( so that @xmath279 ) , and hence @xmath280 is a projection onto the even subalgebra . because @xmath281{0pt}{1.6ex}}^{\otimes\smash n } \!/ { { { \boldsymbol}{c}}}$ ] and @xmath282 are both @xmath283-dimensional , nothing is lost in this projection ! thus we can drop the right - factor of @xmath223 as usual , and define the action of @xmath271 on @xmath284{0pt}{1.6ex}}^{\otimes\smash n } \!/ { { { \boldsymbol}{d}}}$ ] as @xmath285 more generally , the usual action of the pauli matrices on spinors corresponds to the following action of the basis vectors on the reduced even subalgebra @xcite : @xmath286 the simplest logic gate is the of a single qubit , which operates on the computational basis as follows : @xmath287 thus it might appear reasonable to represent the by @xmath288 , but when @xmath289 is applied a superposition @xmath290 , we get @xmath291 instead of @xmath292 again . for a single qubit this difference is just an overall rotation by @xmath83 about @xmath109 , but a second qubit can be affected by this phase difference between the first qubit s states . therefore the correct representation of the gate in @xmath0 is actually @xmath293 , which preserves this superposition up to an irrelevant overall phase shift : @xmath294 @xmath295 more interesting logical operations on the qubits must be able to transform the state of one _ conditional _ on that of another . the usual way in which this is done is via the or _ controlled - not _ gate . as a matrix in @xmath296 , this is represented in the computational basis by @xmath297 which makes it clear that this operation s the second qubit whenever the first is @xmath4 . the corresponding operator in geometric algebra is @xmath298 this may also be written in exponential form as @xmath299 physical implementations of this operation by e.g. nmr typically expand the exponential into a product of relatively simple commuting factors which can be performed sequentially @xcite . note that since @xmath300 is also an idempotent , @xmath301 differs from @xmath302 by a swap of the @xmath303 and @xmath169 axes for both qubits . this self - inverse operation , called the _ hadamard transform _ @xmath304 , is simply a rotation by @xmath83 about the @xmath305 axis . sandwiching @xmath301 by hadamards @xmath306 to just the second qubit gives @xmath307 so the can also be viewed as a rotated phase shift of the state @xmath308 by @xmath83 . the hadamard gate has the important feature of transforming basis states into superpositions thereof ; indeed , as an element of the even subalgebra , it actually represents the spinor of a uniform superposition directly : @xmath309 thus , by using the relations ( [ eq : pauli_rules ] ) , we can show that applying a hadamard to one of two qubits in the state @xmath308 followed by a gate to the other yields the entangled singlet state : @xmath310 @xmath311{0pt}{12pt } } ( 1 + { { { { \boldsymbol}{\iota}}}}){{{\boldsymbol}{e}}}_+^1 + ( 1 - { { { { \boldsymbol}{\iota}}}}){{{\boldsymbol}{e}}}_-^1 \ , { { { { \boldsymbol}{\iota}}}}{{{\boldsymbol}\sigma\!}}_{{\mathsf{x}}}^{\,2 } \right ) \circ \left ( ( { { { { \boldsymbol}{\iota}}}}{{{\boldsymbol}\sigma\!}}_{{\mathsf{z}}}^{\,1 } - { { { { \boldsymbol}{\iota}}}}{{{\boldsymbol}\sigma\!}}_{{\mathsf{x}}}^{\,1 } ) ( -{{{{\boldsymbol}{\iota}}}}{{{\boldsymbol}\sigma\!}}_{{\mathsf{y}}}^{\,2 } ) \right ) \\ = ~ & { { \textstyle\frac{1}{2}}}\ , \left ( { \rule[0pt]{0pt}{12pt } } \left ( ( 1 + { { { { \boldsymbol}{\iota } } } } ) \circ ( { { { { \boldsymbol}{\iota}}}}{{{\boldsymbol}\sigma\!}}_{{\mathsf{z}}}^{\,1 } ) \right ) ( -{{{{\boldsymbol}{\iota}}}}{{{\boldsymbol}\sigma\!}}_{{\mathsf{y}}}^{\,2 } ) + \left ( ( 1 - { { { { \boldsymbol}{\iota } } } } ) \circ ( -{{{{\boldsymbol}{\iota}}}}{{{\boldsymbol}\sigma\!}}_{{\mathsf{x}}}^{\,1 } ) \right ) \left ( ( { { { { \boldsymbol}{\iota}}}}{{{\boldsymbol}\sigma\!}}_{{\mathsf{x}}}^{\,2 } ) \circ ( -{{{{\boldsymbol}{\iota}}}}{{{\boldsymbol}\sigma\!}}_{{\mathsf{y}}}^{\,2 } ) \right ) \right ) \\ = ~ & { { \textstyle\frac{1}{2}}}(1 - { { { { \boldsymbol}{\iota}}}}{{{\boldsymbol}\sigma\!}}_{{\mathsf{z}}}^{\,1 } ) { { { { \boldsymbol}{\iota}}}}{{{\boldsymbol}\sigma\!}}_{{\mathsf{y}}}^{\,2 } - { { \textstyle\frac{1}{2}}}({{{{\boldsymbol}{\iota}}}}{{{\boldsymbol}\sigma\!}}_{{\mathsf{x}}}^{\,1 } - { { { { \boldsymbol}{\iota}}}}{{{\boldsymbol}\sigma\!}}_{{\mathsf{y}}}^{\,1 } ) { { { { \boldsymbol}{\iota}}}}{{{\boldsymbol}\sigma\!}}_{{\mathsf{z}}}^{\,2 } ~\stackrel{{{{\boldsymbol}{d}}}}{\longleftrightarrow}~ { { \textstyle\frac{{{{{{\boldsymbol}{\iota}}}}}}{{2 } } } } ( { { { \boldsymbol}\sigma\!}}_{{\mathsf{y}}}^{\,2 } + { { { \boldsymbol}\sigma\!}}_{{\mathsf{x}}}^{\,2 } - { { { \boldsymbol}\sigma\!}}_{{\mathsf{y}}}^{\,1 } - { { { \boldsymbol}\sigma\!}}_{{\mathsf{x}}}^{\,1 } ) \end{split}\ ] ] @xmath312 . `` quantum '' gates like @xmath304 are not , of course , found in conventional boolean logic , and are an essential component of all quantum algorithms that are more efficient than their classical counterparts @xcite . indeed , the gate together with general single qubit rotations are known to generate @xmath313 , and hence are _ universal _ for quantum logic @xcite . it turns out that unitary transformations are not the most general sort of operation that can be applied to a quantum system . most such _ quantum operations _ , however , produce a statistical outcome , and the ensemble of possible outcomes must be described by a density operator . the previous definition ( eq . ( [ eq : den_op_def ] ) ) of the density operator of an ensemble of identical and noninteracting qubits may be extended to an ensemble of multi - qubit systems as follows : @xmath314 suppressing the correlator @xmath205 as usual , @xmath92 may also be expressed in diagonal form as @xmath315 where @xmath316 corresponds to a unitary matrix @xmath317 ( in the usual @xmath109 coordinate system ) , and @xmath318 are the eigenvalues of @xmath92 . the idempotents @xmath319 are given by @xmath320{0pt}{1.9ex}}^{\,q}_{\smash{\epsilon_k^q } } \leftrightarrow { |\,\chi_k^1\cdots\chi_k^n\,\rangle}{\langle\,\chi_k^1\cdots\chi_k^n\,|}$ ] , where @xmath321 with @xmath322 equal to the @xmath162-th bit in the binary expansion of @xmath323 . it follows that @xmath324 for @xmath325 are the spinors of the ( unique , if @xmath326 ) minimal ensemble that realizes @xmath92 , which therefore describes a pure state if and only if it has rank @xmath4 as an operator . note that by eq . ( [ eq : unit_act ] ) , the density operator transforms under unitary operations as @xmath327 similarly , the ensemble - average expectation value of any observable @xmath328 is @xmath329 just as shown in eq . ( [ eq : rho1_obs ] ) for single qubit ensembles . in contrast to the case of a single qubit , however , the geometric interpretation of these observables is not straightforward . while one can certainly express @xmath92 as a finite ensemble average @xmath330 ( where the @xmath331 are probabilities with @xmath332 ) , this decomposition is highly nonunique . the minimal ensemble obtained by diagonalization , on the other hand , will generally include entangled spinors @xmath333 , for which the expectation value @xmath334 can not be expressed as a product of inner products of the factors of @xmath335 with the polarization vectors of the individual qubits ( indeed , @xmath336 itself need not be factorizable ! ) . the best one can do is to expand @xmath92 in the _ product operator _ basis consisting of all @xmath213 products of the basis vectors @xmath337 , i.e. @xmath338 where @xmath339 and @xmath340 for notational convenience . the utility of this basis is most simply demonstrated via a concrete example , namely nmr spectroscopy . here one is given a liquid sample consisting of identical molecules whose nuclear spins are chemically distinguishable , and hence constitutes an ensemble of multi - qubit systems ( see @xcite and references therein ) . the energy of interaction between the spins and an external magnetic field along @xmath169 is given by an observable called the zeeman hamiltonian , @xmath341 , where @xmath342 is the energy difference between the @xmath5 and @xmath6 states of the @xmath162-th spin in the field . in thermal equilibrium at room temperatures , the polarization of the spins relative to the strongest available fields is typically @xmath343 , and the density operator of the ensemble is essentially @xmath344 . via a suitable pulse of radio - frequency radiation , this may be rotated to @xmath345 , which evolves under the interaction with the field as @xmath346 & 2^{-n}\left ( 1 + \alpha ( \cos(\omega^1 t ) { { { \boldsymbol}\sigma\!}}_{{\mathsf{x}}}^{\,1 } - \sin(\omega^1 t ) { { { \boldsymbol}\sigma\!}}_{{\mathsf{y}}}^{\,1 } \,+\cdots \right . \\ & \left . \cdots+\ , \cos(\omega^nt ) { { { \boldsymbol}\sigma\!}}_{{\mathsf{x}}}^{\,n } - \sin(\omega^nt ) { { { \boldsymbol}\sigma\!}}_{{\mathsf{y}}}^{\,n } ) \right ) ~. \end{aligned}\ ] ] thus on measuring the total magnetization @xmath347 along the @xmath303 axis , @xmath348 ( where @xmath349 is the nuclear gyromagnetic ratio ) , one obtains the sum of the projections of the rotating magnetization vectors of the spins along the @xmath303-axis , i.e. @xmath350 whose fourier transform reveals the contribution from each spin . the way in which the factors of product operators transform like vectors under rotations accounts in large part for the computational utility of the product operator basis . of course , unless it is a natural part of the problem at hand ( as in nmr ) , one is better off not chosing a basis at all ! a _ normal quantum operation _ is a linear transformation of the density operator that may be written in operator sum form as @xcite @xmath351 where the _ kraus operators _ @xmath352 satisfy @xmath353 . the term `` normal '' here refers to the fact that such an operation preserves the scalar part of @xmath92 , since @xmath354 it is also easily seen that such quantum operations are _ positive _ , in that they preserve the positive - definiteness of @xmath92 ; in fact , these operations have a yet stronger property known as _ complete positivity _ , meaning that if the qubits to which @xmath355 applies are embedded in a larger system , then applying @xmath355 to just those qubits preserves the positive - definiteness of the larger system s density operator . that this is a nontrivial extension of positivity is shown by the two - qubit `` partial transpose '' operator @xmath356 , which carries @xmath357 but leaves all the other operator factors unchanged ; this is clearly positive on density operators not involving the second qubit , but acts on the density operator of the singlet state ( eq . ( [ eq : singlet ] ) ) as @xmath358 \equiv~ & { { \textstyle\frac{{1}}{{4 } } } } ( { { { \boldsymbol}\sigma\!}}_{{\mathsf{x}}}^{\,1 } + { { { \boldsymbol}\sigma\!}}_{{\mathsf{y}}}^{\,1 } - { { { \boldsymbol}\sigma\!}}_{{\mathsf{x}}}^{\,2 } - { { { \boldsymbol}\sigma\!}}_{{\mathsf{y}}}^{\,2 } ) { { { \boldsymbol}{e}}}_+^1 { { { \boldsymbol}{e}}}_+^2 ( { { { \boldsymbol}\sigma\!}}_{{\mathsf{x}}}^{\,1 } + { { { \boldsymbol}\sigma\!}}_{{\mathsf{y}}}^{\,1 } - { { { \boldsymbol}\sigma\!}}_{{\mathsf{x}}}^{\,2 } - { { { \boldsymbol}\sigma\!}}_{{\mathsf{y}}}^{\,2 } ) \\ = ~ & { { \textstyle\frac{{1}}{{4 } } } } ( 1 - { { { \boldsymbol}\sigma\!}}_{{\mathsf{x}}}^{\,1 } { { { \boldsymbol}\sigma\!}}_{{\mathsf{x}}}^{\,2 } - { { { \boldsymbol}\sigma\!}}_{{\mathsf{y}}}^{\,1 } { { { \boldsymbol}\sigma\!}}_{{\mathsf{y}}}^{\,2 } - { { { \boldsymbol}\sigma\!}}_{{\mathsf{z}}}^{\,1 } { { { \boldsymbol}\sigma\!}}_{{\mathsf{z}}}^{\,2 } ) \end{aligned } \\ \mapsto\quad & \omega_{{\mathsf{t}}}^1({\boldsymbol}\psi_- ) ~=~ { { \textstyle\frac{{1}}{{4 } } } } ( 1 - { { { \boldsymbol}\sigma\!}}_{{\mathsf{x}}}^{\,1 } { { { \boldsymbol}\sigma\!}}_{{\mathsf{x}}}^{\,2 } + { { { \boldsymbol}\sigma\!}}_{{\mathsf{y}}}^{\,1 } { { { \boldsymbol}\sigma\!}}_{{\mathsf{y}}}^{\,2 } - { { { \boldsymbol}\sigma\!}}_{{\mathsf{z}}}^{\,1 } { { { \boldsymbol}\sigma\!}}_{{\mathsf{z}}}^{\,2 } ) ~ , \end{split}\ ] ] which has eigenvalues @xmath359 $ ] . a quantum operation @xmath355 is called _ unital _ if it preserves the identity itself , i.e. @xmath360 , or equivalently , @xmath361 . perhaps the most important example of a normal unital operation is found in the _ _ contraction _ _ by a single qubit @xmath362 , which may be written in operator sum form as @xcite : @xmath363 this may also be expressed by dropping all terms in the product operator expansion of @xmath92 depending on @xmath162 , and multiplying the remaining terms by a factor of @xmath1 . note that , while @xmath364 is normal and unital , this factor means that the contraction itself is neither . the factor is nevertheless required if the result is to be interpreted as a density operator for the remaining qubits , since the contraction by the second qubit of the above singlet state is @xmath365 ( not @xmath108 ) . this example also illustrates an important way in which general quantum operations are realized in practice , despite the fact that the universe as a whole evolves unitarily . as shown previously , the superposition state with spinor @xmath366 is converted into the singlet state with density operator @xmath367 by letting it interact with a second qubit so as to effect the operation @xmath301 . the contraction then corresponds to `` discarding '' the second qubit ( i.e. ensuring that it does not further interact with the first and hence can be ignored ) , which yields the density operator @xmath108 of the totally mixed state for the first qubit . since the basis states @xmath368 are unaffected by @xmath301 , the net quantum operation on the first qubit corresponds to what is known in quantum communications theory as the _ phase damping channel _ @xmath369 with damping parameter @xmath370 . phase damping is also known as @xmath371 relaxation in nmr , and as _ decoherence _ in quantum information processing ; it is widely believed to be the dominant mechanism by which classical statistical mechanics arises from the underlying unitary dynamics @xcite . to illustrate the utility of geometric algebra in the study of general quantum operations , an eigenvalue characterization of normal , unital , one - bit quantum operations @xmath355 will now be derived . this characterization was originally given by fujiwara & algoet @xcite , although the derivation here parallels that more recently obtained using matrix methods king & ruskai @xcite . this derivation will regard the kraus operators @xmath372 as `` complex quaternions '' @xmath373 with @xmath374 , and consider the action of an arbitrary operation @xmath355 on the scalar and vector parts of @xmath375 separately . first , the action on @xmath4 is @xmath376 the first summation is symmetric with respect to spatial reversion and inversion , i.e. scalar , while the second ( excluding the @xmath44 ) is reversion antisymmetric but inversion symmetric , i.e. a bivector . writing @xmath377 and @xmath378 , so that @xmath379 , we may further expand the second summation as follows : @xmath380 \\ = ~ & { \textstyle\sum}_k \begin{aligned}[t ] & \left [ \left ( \alpha_k\beta_k + { { { { \boldsymbol}{\iota}}}}(\alpha_k { { { \boldsymbol}{b}}}_k - \beta_k { { { \boldsymbol}{a}}}_k ) + { { { \boldsymbol}{b}}}_k { { { \boldsymbol}{a}}}_k \right ) { \rule[0pt]{0pt}{10pt } } \right . \ , - \\ & \left . { \rule[0pt]{0pt}{10pt } } \left ( \alpha_k\beta_k + { { { { \boldsymbol}{\iota}}}}(\beta_k { { { \boldsymbol}{a}}}_k - \alpha_k { { { \boldsymbol}{b}}}_k ) + { { { \boldsymbol}{a}}}_k { { { \boldsymbol}{b}}}_k \right ) \right ] \end{aligned } \\ = ~ & { \textstyle\sum}_k \left[\ , 2 { { { { \boldsymbol}{\iota } } } } ( \alpha_k { { { \boldsymbol}{b}}}_k - \beta_k { { { \boldsymbol}{a}}}_k ) - 2 { { { \boldsymbol}{a}}}_k \wedge { { { \boldsymbol}{b}}}_k \ , \right ] \end{split}\ ] ] thus @xmath355 is unital if and only if @xmath381 ( where `` @xmath382 '' denotes the cross product ) and @xmath383 similarly , the action on @xmath384 is @xmath385 the first summation is over different dilation / rotations of @xmath68 ; the second summation ( excluding the @xmath44 ) is reversion and inversion antisymmetric , i.e. a trivector , and may be further expanded as above : @xmath386 \\ = ~ & { \textstyle\sum}_k \begin{aligned}[t ] & \left [ \left ( \alpha_k\beta_k{{{\boldsymbol}{r } } } + { { { { \boldsymbol}{\iota}}}}(\alpha_k { { { \boldsymbol}{b}}}_k { { { \boldsymbol}{r } } } - \beta_k { { { \boldsymbol}{r } } } { { { \boldsymbol}{a}}}_k ) + { { { \boldsymbol}{b}}}_k { { { \boldsymbol}{r } } } { { { \boldsymbol}{a}}}_k \right ) { \rule[0pt]{0pt}{10pt } } \right . \ , - \\ & \left . { \rule[0pt]{0pt}{10pt } } \left ( \alpha_k\beta_k{{{\boldsymbol}{r } } } + { { { { \boldsymbol}{\iota}}}}(\beta_k { { { \boldsymbol}{a}}}_k { { { \boldsymbol}{r } } } - \alpha_k { { { \boldsymbol}{r } } } { { { \boldsymbol}{b}}}_k ) + { { { \boldsymbol}{a}}}_k { { { \boldsymbol}{r } } } { { { \boldsymbol}{b}}}_k \right ) \right ] \end{aligned } \\ = ~ & { \textstyle\sum}_k \left [ \ , 2{{{{\boldsymbol}{\iota}}}}\alpha_k { { { \boldsymbol}{b}}}_k \cdot { { { \boldsymbol}{r } } } - 2{{{{\boldsymbol}{\iota}}}}\beta_k { { { \boldsymbol}{a}}}_k \cdot { { { \boldsymbol}{r } } } + 2 { { { \boldsymbol}{b}}}_k \wedge { { { \boldsymbol}{r } } } \wedge { { { \boldsymbol}{a}}}_k \ , \right ] \end{split}\ ] ] multiplying through by @xmath387 converts this to @xmath388 ~,\ ] ] which vanishes if and only if @xmath389 if @xmath355 is normal , this must be true for all @xmath68 , which is equivalent to @xmath390 a comparison with eq . ( [ eq : cond1a ] ) shows further that @xmath355 is both unital and normal if and only if @xmath391 and @xmath392 . if we regard a normal @xmath355 as an affine transformation , i.e. @xmath393 ( @xmath394 ) , we see from the derivation leading up to ( [ eq : cond1a ] ) that @xmath395 similarly , the above vectors @xmath396 are @xmath397 i.e. a sum of independent dilation / rotations of each basis vector . it follows that normal and unital quantum operations @xmath355 may be characterized by finding conditions for the linear map @xmath398 to be written as a sum of dilation / rotations of @xmath68 . to this end , we expand @xmath399 as @xmath400 with a similar expansion for @xmath401 . thus on assuming that @xmath355 is diagonal , i.e. @xmath402 , we get @xmath403 & { { { \boldsymbol}\sigma\!}}_\mu \ , { \textstyle\sum}_k ( \alpha_k^2 + \beta_k^2 - \| { { { \boldsymbol}{a}}}_k \|^2 - \| { { { \boldsymbol}{b}}}_k \|^2 ) + 2\ , { { { \boldsymbol}\sigma\!}}_\mu \times { \textstyle\sum}_k ( \alpha_k { { { \boldsymbol}{a}}}_k + \beta_k { { { \boldsymbol}{b}}}_k ) \\ & + \ , 2\ , { \textstyle\sum}_k \left ( ( { { { \boldsymbol}\sigma\!}}_\mu \cdot { { { \boldsymbol}{a}}}_k ) { { { \boldsymbol}{a}}}_k + ( { { { \boldsymbol}\sigma\!}}_\mu \cdot { { { \boldsymbol}{b}}}_k ) { { { \boldsymbol}{b}}}_k \right ) ~. \end{aligned}\ ] ] dotting both sides by @xmath113 now yields @xmath404 so we have simple expressions for the eigenvalues . now consider the vector obtained from the first line of this equation , i.e. @xmath405 @xmath406 where @xmath407 since the coefficients of the @xmath408 s are nonnegative and sum to @xmath4 by ( [ eq : cond1b ] ) , this shows that the vector @xmath409 lies within the tetrahedron @xmath410 , which is the condition on the eigenvalues found by fujiwara and algoet @xcite as well as by king and ruskai @xcite . it is also known that an arbitrary linear map @xmath355 has an operator sum representation if and only if it is completely positive @xcite , so the above can also be viewed as a characterization of complete positivity for normal and unital maps of a single qubit s density operator . finally , it is worth stressing once again that , because of the isomorphisms which exist between the pauli algebra and the even subalgebra of the dirac algebra , every step of the above derivation carries with it a natural interpretation in space - time , and is in fact even easier to carry out when the full power of the dirac algebra is used . a. barenco , c. h. bennett , r. cleve , d. p. divincenzo , n. margolus , p. shor , t. sleator , j. a. smolin , and h. weinfurter , _ elementary gates for quantum computation _ , a * 52 * ( 1995 ) , 34573467 . w. k. clifford , _ applications of grassmann s extensive algebra _ , am . j. math . * 1 * ( 1878 ) , 350358 ; see also _ mathematical papers by w. k. clifford _ , ( r. tucker , ed . ) , chelsea publ . co. , bronx ny , 1968 . c. j. l. doran , a. n. lasenby , s. f. gull , s. s. somaroo , and a. d. challinor , _ spacetime algebra and electron physics _ , in _ advances in imaging and electron physics _ ( p. hawkes , ed . ) , academic press , englewood cliffs , nj , 1996 , pp . 271386 . t. f. havel , d. g. cory , s. s. somaroo , and c .- h . tseng , _ geometric algebra methods in quantum information processing by nmr spectroscopy _ , in _ geometric algebra with applications in science and engineering _ ( e. bayro corrochano and g. sobczyk , eds . ) , birkhauser , boston , usa , 2001 . t. f. havel , s. s. somaroo , c .- h . tseng , and d. g. cory , _ principles and demonstrations of quantum information processing by nmr spectroscopy _ , applicable algebra in engineering , communications and computing ( t. beth and m. grassl , special edition eds . ) * 10 * ( 2000 ) , 339374 ( see also ` quant - ph/9812086 ` ) . t. f. havel , _ qubit logic , algebra and geometry _ , in _ automated deduction in geometry _ ( j. richter - gebert and d. wang , eds . ) , lect . notes artif . , vol . * 2061 * , springer - verlag , berlin / heidelberg , frg , 2001 .

we address the issue of variable selection in the regression model with very high ambient dimension , that is , when the number of variables is very large . the main focus is on the situation where the number of relevant variables , called intrinsic dimension , is much smaller than the ambient dimension . without assuming any parametric form of the underlying regression function , we get tight conditions making it possible to consistently estimate the set of relevant variables . these conditions relate the intrinsic dimension to the ambient dimension and to the sample size . the procedure that is provably consistent under these tight conditions is based on comparing quadratic functionals of the empirical fourier coefficients with appropriately chosen threshold values . the asymptotic analysis reveals the presence of two quite different re gimes . the first regime is when the intrinsic dimension is fixed . in this case the situation in nonparametric regression is the same as in linear regression , that is , consistent variable selection is possible if and only if is small compared to the sample size . the picture is different in the second regime , that is , when the number of relevant variables denoted by tends to infinity as . then we prove that consistent variable selection in nonparametric set - up is possible only if is small compared to . we apply these results to derive minimax separation rates for the problem of variable selection .

1106.4293

real - world data such as those obtained from neuroscience , chemometrics , data mining or sensor - rich environments are often extremely high - dimensional , severely underconstrained ( few data samples compared to the dimensionality of the data ) and interspersed with a large number of irrelevant or redundant features . furthermore , in most situations the data is contaminated by noise , making it even more difficult to retrieve useful information from the data . relevant variable selection is a compelling approach for addressing statistical issues in the scenario of high - dimensional and noisy data with small sample size . starting from mallows @xcite , akaike @xcite , schwarz @xcite who introduced , respectively , the famous criteria @xmath7 , aic and bic , the problem of variable selection was extensively studied in the statistical and machine learning literature both from the theoretical and algorithmic viewpoints . it appears , however , that the theoretical limits of performing variable selection in the context of nonparametric regression are still poorly understood , especially when the number of variables , denoted by @xmath0 and referred to as ambient dimension , is much larger than the sample size @xmath2 . the purpose of the present work is to explore this setting under the assumption that the number of relevant variables , hereafter called intrinsic dimension and denoted by @xmath8 , may grow with the sample size but remains much smaller than @xmath0 . in the important particular case of linear regression , the latter scenario was the subject of a number of recent studies . many of them rely on @xmath9-norm penalization @xcite and constitute an attractive alternative to iterative variable selection procedures @xcite and to marginal regression or correlation screening @xcite . promising results for feature selection are also obtained by conformal prediction @xcite , ( minimax ) concave penalties @xcite , bayesian approach @xcite and higher criticism @xcite . extensions to other settings including logistic regression , generalized linear model and ising model were carried out in @xcite , respectively . variable selection in the context of groups of variables with disjoint or overlapping groups was studied by @xcite . hierarchical procedures for selection of relevant variables were proposed by @xcite . it is now well understood that in the gaussian sequence model and in the high - dimensional linear regression with a gram matrix satisfying some variant of irrepresentable condition , consistent estimation of the pattern of relevant variables also called the sparsity pattern is possible under the condition @xmath10 as @xmath4 @xcite . furthermore , it is well known that if @xmath11 remains bounded from below by some positive constant when @xmath4 , then it is impossible to consistently recover the sparsity pattern @xcite . thus , a tight condition exists that describes in an exhaustive manner the interplay between the quantities @xmath8 , @xmath0 and @xmath2 that guarantees the existence of consistent estimators . the situation is very different in the case of nonlinear regression , since , to our knowledge , there is no result providing tight conditions for consistent estimation of the sparsity pattern . lafferty and wasserman @xcite and bertin and lecu ' e @xcite , in papers closely related to the present work , considered the problem of variable selection in nonparametric gaussian regression model . they proved the consistency of the proposed procedures under some assumptions that in the light of the present work turn out to be suboptimal . more precisely , lafferty and wasserman @xcite assumed the unknown regression function to be four times continuously differentiable with bounded derivatives . the algorithm they proposed , termed rodeo , is a greedy procedure performing simultaneously local bandwidth choice and variable selection . rodeo is shown to converge when the ambient dimension @xmath0 is @xmath12 while the intrinsic dimension @xmath8 does not increase with @xmath2 . on the other hand , bertin and lecu @xcite proposed a procedure based on the @xmath9-penalization of local polynomial estimators and proved its consistency when @xmath13 , but @xmath0 is allowed to be as large as @xmath6 , up to a constant . they also had a weaker assumption on the regression function merely assumed to belong to the holder class with smoothness @xmath14 . to complete the picture , let us mention that estimation and hypotheses testing problems for high - dimensional nonparametric regression under sparse additive modeling were recently addressed in @xcite . this brief review of the literature reveals that there is an important gap in consistency conditions for the linear regression and for the nonlinear one . for instance , if the intrinsic dimension @xmath8 is fixed , then the condition guaranteeing consistent estimation of the sparsity pattern is @xmath15 in linear regression , whereas it is @xmath16 in the nonparametric case . while it is undeniable that the nonparametric regression is much more complex than the linear one , it is , however , not easy to find a justification to such an important gap between two conditions . the situation is even worse in the case where @xmath17 . in fact , for the linear model with at most polynomially increasing ambient dimension @xmath18 , it is possible to estimate the sparsity pattern for intrinsic dimensions @xmath8 as large as @xmath19 , for some @xmath20 . in other words , the sparsity index can be almost on the same order as the sample size . in contrast , in nonparametric regression , there is no procedure that is proved to converge to the true sparsity pattern when both @xmath2 and @xmath8 tend to infinity , even if @xmath8 grows extremely slowly . in the present work , we fill this gap by introducing a simple variable selection procedure that selects the relevant variables by comparing some quadratic functionals of empirical fourier coefficients to prescribed significance levels . consistency of this procedure is established under some conditions on the triplet @xmath21 , and the tightness of these conditions is proved . the main take - away messages deduced from our results are the following : * when the number of relevant variables @xmath8 is fixed and the sample size @xmath2 tends to infinity , there exist positive real numbers @xmath22 and @xmath23 such that ( a ) if @xmath24 the estimator proposed in section [ sec3 ] is consistent and ( b ) no estimator of the sparsity pattern may be consistent if @xmath25 . * when the number of relevant variables @xmath8 tends to infinity with @xmath4 , then there exist real numbers @xmath26 and @xmath27 , @xmath28 such that @xmath29 , @xmath30 and ( a ) if @xmath31 the estimator proposed in section [ sec3 ] is consistent and ( b ) no estimator of the sparsity pattern may be consistent if @xmath32 . * in particular , if @xmath0 grows not faster than a polynomial in @xmath2 , then there exist positive real numbers @xmath33 and @xmath34 such that ( a ) if @xmath35 , the estimator proposed in section [ sec3 ] is consistent , and ( b ) no estimator of the sparsity pattern may be consistent if @xmath36 . in the regime of a growing intrinsic dimension @xmath17 and a moderately large ambient dimension @xmath37 , for some @xmath38 , we make a concentrated effort to get the constant @xmath39 as close as possible to the constant @xmath40 . this goal is reached for the model of gaussian white noise and , very surprisingly , it required from us to apply some tools from complex analysis , such as the jacobi @xmath41-function and the saddle point method , in order to evaluate the number of lattice points lying in a ball of an euclidean space with increasing dimension . the rest of the paper is organized as follows . the notation and assumptions necessary for stating our main results are presented in section [ sec2 ] . in section [ sec3 ] , an estimator of the set of relevant variables is introduced and its consistency is established , in the case where the data come from the gaussian white noise model . the main condition required in the consistency result involves the number of lattice points in a ball of a high - dimensional euclidean space . an asymptotic equivalent for this number is presented in section [ sec4 ] . results on impossibility of consistent estimation of the sparsity pattern are derived in section [ sec5 ] . section [ secadapt ] is devoted to exploring adaptation to the unknown parameters ( smoothness and degree of significance ) and recovering minimax rates of separation . then , in section [ sec6 ] , we show that some of our results can be extended to the model of nonparametric regression . the relations between consistency and inconsistency results are discussed in section [ sec7 ] . the technical parts of the proofs are postponed to the . we are interested in the variable selection task ( also known as model selection , feature selection , sparsity pattern estimation ) in the context of high - dimensional nonlinear regression . let @xmath42^d\to\rr$ ] denote the unknown regression function . we assume that the number of variables @xmath0 is very large , possibly much larger than the sample size @xmath2 , but only a small number of these variables contribute to the fluctuations of the regression function @xmath43 . to be more precise , we assume that for some small subset @xmath44 of the index set @xmath45 satisfying @xmath46 , there is a function @xmath47 such that @xmath48 where @xmath49 stands for the subvector of @xmath50 obtained by removing from @xmath50 all the coordinates with indices lying outside @xmath44 . in what follows , we allow @xmath0 and @xmath8 to depend on @xmath2 , but we will not always indicate this dependence in notation . note also that the genuine intrinsic dimension is @xmath51 ; @xmath8 is merely a known upper bound on the intrinsic dimension . in what follows , we use the standard notation for the vector and sequence norms : @xmath52 . let us stress right away that the primary aim of this work is to understand when it is possible to estimate the sparsity pattern @xmath44 ( with theoretical guarantees on the convergence of the estimator ) and when it is impossible . the estimator that we will define in the next sections is intended to show the possibility of consistent estimation , rather than to provide a practical procedure for recovering the sparsity pattern . therefore , the estimator will be allowed to depend on different constants appearing in conditions imposed on the regression function @xmath43 and on some characteristics of the noise . to make the consistent estimation of the set @xmath44 realizable , we impose some smoothness and identifiability assumptions on @xmath43 . in order to describe the smoothness assumption imposed on @xmath43 , let us introduce the trigonometric fourier basis , @xmath53 and @xmath54 where @xmath55 denotes the set of all @xmath56 such that the first nonzero element of @xmath57 is positive , and @xmath58 stands for the usual inner product in @xmath59 . in what follows , we use the notation @xmath60 for designing the scalar product in @xmath61^d;\rr)$ ] , that is , @xmath62^d } \sfh(\mathbf{x})\tilde\sfh(\mathbf{x } ) \,d\mathbf{x}$ ] for every @xmath63^d;\rr)$ ] . using this orthonormal fourier basis , we define @xmath64 to ease notation , we set @xmath65=\langle\sff,\varphi_\bk \rangle$ ] for all @xmath66 . in addition to the smoothness , we need also to require that the relevant variables are sufficiently relevant for making their identification possible . this is done by means of the following condition . the regression function @xmath43 belongs to @xmath67 . furthermore , for some subset @xmath68 of cardinality @xmath69 , there exists a function @xmath70 such that @xmath71 , @xmath72 , and it holds that @xmath73=\sum_{\bk : k_j\neq0 } \theta_\bk[\sff]^2\geq\kappa \qquad\forall j\in j.\ ] ] one easily checks that @xmath74=0 $ ] for every @xmath75 that does not lie in the sparsity pattern . this provides a characterization of the sparsity pattern as the set of indices of nonzero coefficients of the vector @xmath76=(q_1[\sff],\ldots , q_d[\sff])$ ] . prior to describing the procedures for estimating @xmath44 , let us comment on condition [ c1 ] . it is important to note that the identifiability assumption ( [ ident ] ) can be rewritten as @xmath77^d } ( \sff(\mathbf{x})-\int_0 ^ 1\sff(\mathbf{x } ) \,dx_j ) ^2 \,d\mathbf{x}\ge\kappa$ ] and , therefore , is not intrinsically related to the basis we have chosen . in the case of continuously differentiable and @xmath78-periodic function @xmath43 , the smoothness assumption @xmath79 as well can be rewritten without using the trigonometric basis , since @xmath80 ^ 2=(2\pi ) ^{-2}\int_{[0,1]^d}[\partial_j \sff(\mathbf{x})]^2 \,d\mathbf{x}$ ] . thus condition [ c1 ] is essentially a constraint on the function @xmath43 itself and not on its representation in the specific basis of trigonometric functions . the results of this work can be extended with minor modifications to other types of smoothness conditions imposed on @xmath43 , such as hlder continuity or besov - regularity . in these cases the trigonometric basis ( [ trig ] ) should be replaced by a basis adapted to the smoothness condition ( spline , wavelet , etc . ) . furthermore , even in the case of sobolev smoothness , one can replace the set @xmath67 corresponding to smoothness order @xmath78 by any sobolev ellipsoid of smoothness @xmath81 ; see , for instance , @xcite where the case @xmath82 is explored . roughly speaking , the role of the smoothness assumption is to reduce the statistical model with infinite - dimensional parameter @xmath43 to a finite - dimensional model having good approximation properties . any value of smoothness order @xmath81 leads to this reduction . the value @xmath83 is chosen for simplicity of exposition only . to convey the main ideas without taking care of some technical details , we start by focusing our attention on the gaussian white noise model that was proved to be asymptotically equivalent to the model of regression @xcite , as well as to other nonparametric models @xcite . thus , we assume that the available data consists of the gaussian process @xmath84^d;\rr)\}$ ] such that @xmath85=\int_{[0,1]^d } \sff ( \mathbf{x } ) \phi ( \mathbf{x } ) \,d\mathbf{x},\qquad \operatorname{cov}_f \bigl(\by(\phi),\by\bigl(\phi'\bigr)\bigr)=\frac1n\int _ { [ 0,1]^d } \phi ( \mathbf{x})\phi'(\mathbf{x } ) \,d \mathbf{x}.\ ] ] it is well known that these two properties uniquely characterize the probability distribution of a gaussian process . an alternative representation of @xmath86 is @xmath87^d,\ ] ] where @xmath88 is a @xmath0-parameter brownian sheet . note that minimax estimation and detection of the function @xmath43 in this set - up ( but without sparsity assumption ) was studied by @xcite . we intend to tackle the variable selection problem by multiple hypotheses testing ; each hypothesis concerns a group of the fourier coefficients of the observed signal and suggests that all the elements within the group are zero . the rationale behind this approach is the following simple observation : since the trigonometric basis is orthonormal and contains the constant function , @xmath89=\langle f,\varphi_\bk \rangle=0 \qquad\forall\bk \mbox { s.t . } k_j \not=0.\ ] ] this observation entails that if the intrinsic dimension @xmath90 is small as compared to @xmath0 , then the sequence of fourier coefficients is sparse . furthermore , as explained below , there is a sort of group sparsity with overlapping groups . for every @xmath91 , we denote by @xmath92 the set of all subsets @xmath93 of @xmath45 having exactly @xmath94 elements : @xmath95 . for every multi - index @xmath66 , we denote by @xmath96 the set of indices corresponding to nonzero entries of @xmath57 . to define the blocks of coefficients @xmath97 that will be tested for significance , we introduce the following notation : for every @xmath98 and for every @xmath99 , we set @xmath100= \bigl(\theta_\bk[\sff]\dvtx j \in\supp(\bk)\subset i \bigr).\ ] ] it follows from ( [ crit1 ] ) that the characterization @xmath101 \bigr\|_p=0,\ ] ] holds true for every @xmath102 $ ] . furthermore , again in view of ( [ crit1 ] ) , the maximum over @xmath93 of the norms @xmath103 \|_p$ ] is attained when @xmath104 and is equal to the maximum over all subsets @xmath93 such that @xmath105 . summarizing these arguments , we can formulate the problem of variable selection as a problem of testing @xmath0 null hypotheses @xmath106 \bigr\|_p=0 \qquad \forall i\subset\{1,\ldots , d\ } \mbox { such that } \operatorname{card}(i)\le d^*.\ ] ] if the hypothesis @xmath107 is rejected , then the @xmath75th covariate is declared as relevant . note that by virtue of assumption [ c1 ] , the alternatives can be written as @xmath108 \bigr\|_2 ^ 2\ge\kappa\qquad\mbox{for some } i\subset\{1,\ldots , d\ } \mbox { such that } \operatorname{card}(i)\le d^*.\hspace*{-35pt}\ ] ] our estimator is based on this characterization of the sparsity pattern . if we denote by @xmath109 the observable random variable @xmath110 , we have @xmath111+n^{-1/2 } \xi_\bk,\qquad \theta_\bk=\langle f,\varphi_\bk \rangle , \bk\in\zz^d,\ ] ] where @xmath112 form a countable family of independent gaussian random variables with zero mean and variance equal to one . according to this property , @xmath109 is a good estimate of @xmath65 $ ] : it is unbiased and with a mean squared error equal to @xmath113 . using the plug - in argument , this suggests to estimate @xmath114 by @xmath115 and the norm of @xmath114 by the norm of @xmath116 . however , since this amounts to estimating an infinite - dimensional vector , the error of estimation will be infinitely large . to cope with this issue , we restrict the set of indices for which @xmath117 is estimated by @xmath109 to a finite set , outside of which @xmath117 will be merely estimated by @xmath118 . such a restriction is justified by the fact that @xmath43 is assumed to be smooth : fourier coefficients corresponding to very high frequencies are very small . let us fix an integer @xmath119 , the cut - off level , and denote , for @xmath120 , @xmath121 since the alternatives @xmath122 are concerned with the 2-norm , we build our test statistic on an estimate of the norm @xmath123\|_2 $ ] . to this end , we introduce @xmath124 which is an unbiased estimator of @xmath125 . note that when @xmath126 , the quantity @xmath127 approaches @xmath123\|_2 ^ 2 $ ] . it is clear that larger values of @xmath128 lead to a smaller bias while the variance get increased . moreover , the variance of @xmath129 is proportional to the cardinality of the set @xmath130 . the latter is an increasing function of @xmath131 . therefore , if we aim at getting comparable estimation accuracies when estimating the functionals @xmath123\|_2 ^ 2 $ ] by @xmath129 for various @xmath93 s , it is reasonable to make the cut - off level @xmath128 vary with the cardinality of @xmath93 . thus , we consider a multivariate cut - off @xmath132 . for a subset @xmath93 of cardinality @xmath133 , we test significance of the vector @xmath134 $ ] by comparing its estimate @xmath135 with a prescribed threshold @xmath136 . this leads us to define an estimator of the set @xmath44 by @xmath137 where @xmath138 and @xmath139 are two vectors of tuning parameters . as already mentioned , the role of @xmath140 is to ensure that the truncated sums @xmath141 do not deviate too much from the complete sums @xmath142 . quantitatively speaking , for a given @xmath143 , we would like to choose @xmath144 s so that @xmath145 , where @xmath146 . this guarantee can be achieved due to the smoothness assumption . indeed , as proved in ( [ minor ] ) ( cf . appendix [ appa1 ] ) , it holds that @xmath147 therefore , choosing @xmath148 , for every @xmath149 , entails the inequality @xmath150 , which indicates that the relevance of variables is not affected too much by the truncation . pushing further the analogy with the hypotheses testing , we define type i error of an estimator @xmath151 of @xmath44 as the one of having @xmath152 , that is , classifying some irrelevant variables as relevant . the type ii error is then that of having @xmath153 , which amounts to classifying some relevant variables as irrelevant . as in the testing problem , handling the type i error is easier since the distribution of the test statistic is independent of @xmath43 . in fact , this is the max of a finite family of random variables drawn from translated and scaled @xmath154-distributions . using the bonferroni adjustment leads to the following control of the first kind error . [ prop0 ] let us denote by @xmath155 the cardinality of the set @xmath156 . if for some @xmath157 and for every @xmath149 , @xmath158 then the type i error @xmath159 is upper - bounded by @xmath160 , and therefore tends to 0 as @xmath161 . this proposition shows that the type i error of a variable selection procedure may be made small by choosing a sufficiently high threshold . by doing this , we run the risk to reject @xmath107 very often and to drastically underestimate the set of relevant variables . the next result establishes a necessary condition , which will be shown to be tight , ensuring that such an underestimation does not occur . [ prop2 ] let condition be satisfied with some known constants @xmath162 and @xmath163 , and let @xmath146 . for some real numbers @xmath143 and @xmath157 , set @xmath148 , @xmath149 , and define @xmath136 to be equal to the right - hand side of ( [ lambda ] ) . if the condition @xmath164 is fulfilled , then @xmath165 is consistent and satisfies the inequalities @xmath166 and @xmath167 . condition ( [ ordre2 ] ) , ensuring the consistency of the variable selection procedure @xmath151 , admits a very natural interpretation : it is possible to detect relevant variables if the degree of relevance @xmath168 is larger than a multiple of the threshold @xmath169 , the latter being chosen according to the noise level . a first observation is that this theorem provides interesting insight to the possibility of consistent recovery of the sparsity pattern @xmath44 in the context of fixed intrinsic dimension . in fact , when @xmath8 remains bounded from above when @xmath170 and @xmath171 , then we get that @xmath172 provided that=-1 @xmath173=0 although we did not find ( exactly ) this result in the statistical literature on variable selection , it can be checked that ( [ eq9a ] ) is a necessary and sufficient condition for recovering the sparsity pattern @xmath44 in linear regression with fixed sparsity @xmath8 and growing dimension @xmath0 and sample size @xmath2 . thus , in the regime of fixed or bounded @xmath8 , the sparsity pattern estimation in nonparametric regression is not more difficult than in the parametric linear regression , as far as only the consistency of estimation is considered and the precise value of the constant in ( [ eq9a ] ) is neglected . furthermore , there is a simple estimator @xmath174 of @xmath44 ( cf . equation ( 3 ) in @xcite ) , which is provably consistent under condition ( [ eq9a ] ) . this estimator can be seen as a procedure of testing hypotheses @xmath107 of form ( [ h0 ] ) with @xmath175 and , therefore , it does not really exploit the structure of the fourier coefficients of the regression function . to some extent , this is the reason why in the regime of growing intrinsic dimension @xmath17 , the estimator @xmath176 proposed by @xcite is no longer optimal . in fact , when @xmath17 , the term @xmath177 present in ( [ ordre2 ] ) tends to infinity as well . furthermore , as we show in section [ sec4 ] , this convergence takes place at an exponential rate in @xmath8 . therefore , in this asymptotic set - up it is crucial to have the right order of @xmath177 in the condition that ensures the consistency . as shown in section [ sec5 ] , this is the case for condition ( [ ordre2 ] ) . an apparent drawback of the estimator @xmath151 is the large dimensionality of tuning parameters involved in @xmath178 . however , theorem [ prop2 ] reveals that for achieving good selection power , it is sufficient to select the @xmath179-dimensional tuning parameter @xmath180 on a one - dimensional curve parameterized by @xmath181 . indeed , once the value of @xmath182 is given , theorem [ prop2 ] advocates for choosing @xmath183 \\[-8pt ] \nonumber \lambda_\ell&=&\frac { 2\sqrt{a n(\ell,\vartheta)d^*\log(2ed / d^*)}+2ad^*\log(2ed / d^*)}{n}\end{aligned}\ ] ] for every @xmath149 . as discussed in section [ ssecadapt ] , this property allows us to relax the requirement that the values @xmath184 and @xmath168 involved in [ c1 ] are known in advance . the result of the last theorem is in some sense adaptive w.r.t . the unknown sparsity . indeed , while the estimator @xmath178 involves @xmath8 , which is merely a known upper bound on the true sparsity @xmath146 and may be significantly larger than @xmath3 , it is the true sparsity @xmath3 that appears in condition ( [ ordre2 ] ) as a first argument of the quantity @xmath185 . this point is important given the exponential rate of divergence of @xmath186 when its first argument tends to infinity . on the other hand , if condition ( [ ordre2 ] ) is satisfied with @xmath187 instead of @xmath188 , then the consistent estimation of @xmath44 can be achieved by a slightly simpler procedure , @xmath189 the proof of this statement is similar to that of theorem [ prop2 ] and will be omitted . the aim of the present section is to investigate the properties of the quantity @xmath190 that is involved in the conditions ensuring the consistency of the proposed procedures . quite surprisingly , the asymptotic behavior of @xmath190 turns out to be related to the jacobi @xmath41-function . to show this , let us introduce some notation . for a positive number @xmath191 , we set @xmath192 along with @xmath193 and @xmath194 . in simple words , @xmath195 is the number of lattice points lying in the @xmath8-dimensional ball with radius @xmath196 and centered at the origin , while @xmath197 is the number of ( integer ) lattice points lying in the @xmath198-dimensional ball with radius @xmath199 and centered at the origin ; see figure [ fig01 ] for an illustration . with this notation , the quantity @xmath200 of theorem [ prop2 ] can be written as @xmath201 . by volumetric arguments , one can check that @xmath202 , where @xmath203 is the volume of the unit ball in @xmath204 . furthermore , similar bounds hold true for @xmath197 as well . unfortunately , when @xmath17 , these inequalities are not accurate enough to yield nontrivial results in the problem of variable selection we are dealing with . this is especially true for the results on impossibility of consistent estimation stated in section [ sec5 ] . in the three dimensional space ( @xmath205 ) . red points are those of @xmath206 while blue stars are those of @xmath207 . in this example , @xmath208 . ] in order to determine the asymptotic behavior of @xmath195 and @xmath197 when @xmath8 tends to infinity , we will rely on their integral representation through jacobi s @xmath209-function . recall that the latter is given by @xmath210 , which is well defined for any complex number @xmath211 belonging to the unit ball @xmath212 . to briefly explain where the relation between @xmath213 and the @xmath41-function comes from , let us denote by @xmath214 the sequence of coefficients of the power series of @xmath215 , that is , @xmath216 . one easily checks that @xmath217 , @xmath218 . thus , for every @xmath219 such that @xmath220 is integer , we have @xmath221 . as a consequence of cauchy s theorem , we get @xmath222 where the integral is taken over any circle @xmath223 with @xmath224 . exploiting this representation and applying the saddle - point method thoroughly described in @xcite , we get the following result . [ prop1 ] let @xmath225 be an integer and let @xmath226 . there is a unique solution @xmath227 in @xmath228 to the equation @xmath229 . furthermore , the function @xmath230 is increasing and @xmath231 . for @xmath28 , the following equivalences hold true : @xmath232 as @xmath8 tends to infinity . hereafter , it will be useful to note that the second part of proposition [ prop1 ] yields @xmath233 \\[-8pt ] \eqntext{\mbox{as } d^*\to\infty,}\end{aligned}\ ] ] with @xmath234 . furthermore , while the asymptotic equivalences of proposition [ prop1 ] are established for integer values of @xmath225 , relation @xmath235 holds true for any positive real number @xmath219 @xcite . in order to get an idea of how the terms @xmath227 and @xmath236 depend on @xmath219 , we depicted in figure [ fig1 ] the plots of these quantities as functions of @xmath225 . and @xmath237 . one can observe that both functions are increasing , the first one converges to @xmath78 very rapidly , while the second one seems to diverge very slowly . ] combining relation ( [ eqlog ] ) with theorem [ prop2 ] , we get the following result . [ cor2 ] let condition be satisfied with some known constants @xmath162 and @xmath163 . consider the asymptotic set - up in which both @xmath238 and @xmath239 tend to infinity as @xmath240 . assume that @xmath0 grows at a sub - exponential rate in @xmath2 , that is , @xmath241 . if @xmath242 with @xmath243 , then consistent estimation of @xmath44 is possible and can be achieved , for instance , by the estimator @xmath178 . in this section , we focus our attention on the functional class @xmath244 of all functions satisfying assumption [ c1(@xmath245 ) ] . for emphasizing that @xmath44 is the sparsity pattern of the function @xmath43 , we write @xmath246 instead of @xmath44 . we assume that @xmath247 . the goal is to provide conditions under which the consistent estimation of the sparsity support is impossible , that is , there exists a constant @xmath248 and an integer @xmath249 such that , if @xmath250 , @xmath251 where the @xmath252 is over all possible estimators of @xmath246 . to this end , we introduce a set of @xmath253 probability distributions @xmath254 on @xmath255 and use the fact that @xmath256 these measures @xmath257 will be chosen in such a way that for each @xmath258 there is a set @xmath259 of cardinality @xmath8 such that @xmath260 and all the sets @xmath261 are distinct . the measure @xmath262 is the dirac measure in @xmath118 . considering these @xmath257s as `` priors '' on @xmath255 and defining the corresponding `` posteriors '' @xmath263 by @xmath264 we can write inequality ( [ minor1 ] ) as @xmath265 where the @xmath252 is taken over all random @xmath266 taking values in @xmath267 . the latter @xmath252 will be controlled using a suitable version of the fano lemma . to state it , we denote by @xmath268 the kullback leibler divergence between two probability measures @xmath269 and @xmath270 defined on the same probability space . [ lemfano ] let @xmath271 be an integer , @xmath272 be a measurable space and let @xmath273 be probability measures on @xmath272 . let us set @xmath274 , where the @xmath252 is taken over all measurable functions @xmath275 . if for some @xmath276 , @xmath277 , then @xmath278 . we apply this lemma with @xmath279 being the set of all arrays @xmath280 such that for some @xmath281 the entries @xmath282 for every @xmath57 larger than @xmath283 in @xmath284-norm . it follows from fano s lemma that one can deduce a lower bound on @xmath285 , the quantity we are interested in , from an upper bound on the average kullback leibler divergence between @xmath286 and @xmath287 . with these tools at hand , we are in a position to state the main result on the impossibility of consistent estimation of the sparsity pattern in the case when the conditions of theorem [ prop2 ] are violated . [ thm3 ] assume that @xmath288 and @xmath289 . let @xmath290 be the largest integer satisfying @xmath291 , where the jacobi @xmath41-function @xmath292 and @xmath293 are those defined in section [ sec4 ] . if for some @xmath294 , @xmath295 then , for @xmath8 large enough , @xmath296 . if for some @xmath294 , @xmath297 then @xmath296 it is worth stressing here that condition ( [ ordre3 ] ) is the converse of condition ( [ ordre2 ] ) of theorem [ prop2 ] in the case @xmath17 , in the sense that condition ( [ ordre2 ] ) amounts to requiring that the left - hand side of ( [ ordre3 ] ) is smaller than some constant . there is , however , one difference between the quantities involved in these conditions : the term @xmath298 of ( [ ordre2 ] ) is replaced by @xmath299 in condition ( [ ordre3 ] ) . one can wonder how close @xmath300 is to @xmath182 . to give a qualitative answer to this question , we plotted in figure [ fig2 ] the curve of the mapping @xmath301 along with the bisector @xmath302 . we observe that the difference between two curves is small compared to @xmath182 . as we discuss it later , this property shows that the constants involved in the necessary condition and in the sufficient condition for consistent estimation of @xmath44 are very close , especially for large values of @xmath182 . ( blue ) and the bisector ( red ) . ] the estimator @xmath303 we have introduced in section [ sec3 ] is clearly nonadaptive : the tuning parameters @xmath304 recommended by the developed theory involve the values @xmath184 and @xmath168 , which are generally unknown . fortunately , we can take advantage of the fact that the choice of @xmath305 and @xmath306 is governed by the one - dimensional parameter @xmath181 . therefore , it is realistic to assume that a finite grid of values @xmath307 is available containing a true value of @xmath182 . the following result provides an adaptive procedure of variable selection with guaranteed control of the error . [ propadapt ] let @xmath307 and @xmath143 be given values , and set . ] @xmath308 for every @xmath309 , let us denote @xmath310 with @xmath311 and @xmath312 if the condition @xmath313 is fulfilled , then the estimator @xmath314 satisfies @xmath315 . in simple words , if the grid of possible values @xmath316 has a cardinality @xmath283 which is not too large [ i.e. , @xmath317 , then declaring a variable relevant if at least one of the procedures @xmath318 suggests its relevance provides a consistent and adaptive variable selection strategy . the proof of this statement follows immediately from proposition [ prop0 ] and theorem [ prop2 ] . indeed , applying proposition [ prop0 ] with @xmath319 yields @xmath320 , while theorem [ prop2 ] ensures that @xmath321 . since the methodology of section [ sec3 ] takes its roots in the theory of hypotheses testing , one naturally wonders what are the minimax rates of separation in the problem of variable selection . the results stated in foregoing sections allow us to answer this question in the case of sobolev smoothness 1 and alternatives separated in @xmath322-norm . the following result , the proof of which is postponed to the appendix [ appprop5 ] provides minimax rates . we assume herein that the true sparsity @xmath146 and its known upper estimate @xmath8 are such that @xmath323 is bounded from above by some constant . [ propmmx ] @xmath324there is a constant @xmath325 depending only on @xmath184 such that if @xmath326 then there exists a consistent estimator of @xmath44 . furthermore , the consistency is uniform in @xmath327 . on the other hand , there is a constant @xmath328 depending only on @xmath184 such that if @xmath329 then uniformly consistent estimation of @xmath44 is impossible . borrowing the terminology of the theory of hypotheses testing , we say that @xmath330 is the minimax rate of separation in the problem of variable selection for sobolev smoothness one . these results readily extend to sobolev smoothness of any order @xmath331 , in which case the rate of separation takes the form @xmath332 . the first term in this maximum coincides , up to the logarithmic term , with the minimax rate of separation in the problem of detection of an @xmath3-dimensional signal @xcite . note , however , that in our case this logarithmic inflation is unavoidable . it is the price to pay for not knowing in advance which @xmath3 variables are relevant . so far , we have analyzed the situation in which noisy observations of the regression function @xmath333 are available at all points @xmath334^d$ ] . let us turn now to the more realistic model of nonparametric regression , when the observed noisy values of @xmath43 are sampled at random in the unit hypercube @xmath335^d$ ] . more precisely , we assume that @xmath2 independent and identically distributed pairs of input - output variables @xmath336 , @xmath337 are observed that obey the regression model @xmath338 the input variables @xmath339 are assumed to take values in @xmath59 while the output variables @xmath340 are scalar . as usual , @xmath341 are such that @xmath342=0 $ ] , @xmath337 ; additional conditions will be imposed later . without requiring from @xmath43 to be of a special parametric form , we aim at recovering the set @xmath68 of its relevant variables . the noise magnitude @xmath343 is assumed to be known . it is clear that the estimation of @xmath44 can not be accomplished without imposing some further assumptions on @xmath43 and on the distribution @xmath344 of the input variables . roughly speaking , we will assume that @xmath43 is differentiable with a squared integrable gradient and that @xmath344 admits a density which is bounded from below . more precisely , let @xmath345 denote the density of @xmath344 w.r.t . the lebesgue measure . @xmath346 for any @xmath347^d$ ] and that @xmath348 for any @xmath334^d$ ] . the next assumptions imposed to the regression function and to the noise require their boundedness in an appropriate sense . these assumptions are needed in order to prove , by means of a concentration inequality , the closeness of the empirical coefficients to the true ones . the @xmath349^d,\rr , p_x)$ ] and @xmath61^d,\rr , p_x)$ ] norms of the function @xmath43 are bounded from above , respectively , by @xmath350 and @xmath351 , that is , @xmath352 and @xmath353\le l_2 ^ 2 $ ] . the noise variables satisfy a.e . e^{t^2/2}$ ] for all @xmath355 . we stress once again that the primary aim of this work is merely to understand when it is possible to consistently estimate the sparsity pattern . the estimator that we will define is intended to show the possibility of consistent estimation , rather than being a practical procedure for recovering the sparsity pattern . therefore , the estimator will be allowed to depend on the parameters @xmath356 , @xmath184 , @xmath168 and @xmath351 appearing in conditions [ c1][c3 ] . the estimator of the sparsity pattern @xmath44 that we are going to introduce now is based on the following simple observation : if @xmath357 , then @xmath65=0 $ ] for every @xmath57 such that @xmath358 . in contrast , if @xmath359 , then there exists @xmath66 with @xmath358 such that @xmath360|>0 $ ] . to turn this observation into an estimator of @xmath44 , we start by estimating the fourier coefficients @xmath65 $ ] by their empirical counterparts , @xmath361 then , for every @xmath362 and for any @xmath225 , we introduce the notation @xmath363 . the estimator of @xmath44 is defined by @xmath364 where @xmath128 and @xmath365 are some parameters to be defined later . the next result , the proof of which is placed in the supplementary material @xcite , provides consistency guarantees for @xmath366 . [ thm1 ] let conditions be fulfilled with some known values @xmath367 , @xmath368 and @xmath351 . assume furthermore that the design density @xmath345 and an upper estimate on the noise magnitude @xmath343 are available . set @xmath369 and @xmath370 . if the following conditions are satisfied : @xmath371 \\[-8pt ] \nonumber \frac{128({\sigma}+l_2)^2d^*n(d^*,\vartheta)\log(24\sqrt{\vartheta } d / d^*)}{n\gmin^2 } & < & \kappa,\end{aligned}\ ] ] then the estimator @xmath372 satisfies @xmath373 . if we take a look at the conditions of theorem [ thm1 ] ensuring the consistency of @xmath374 , it becomes clear that the strongest requirement is the second inequality in ( [ cond3 ] ) . roughly speaking , this condition requires that @xmath375 is bounded from above by some constant . according to results stated in section [ sec4 ] , @xmath187 diverges exponentially fast , making inequality ( [ cond3 ] ) impossible for @xmath8 larger than @xmath376 up to a multiplicative constant . it is also worth stressing that although we require the @xmath344-a.e . boundedness of @xmath43 by some constant @xmath350 , this constant is not needed for computing the estimator proposed in theorem [ thm1 ] . only constants related to some quadratic functionals of the sequence of fourier coefficients @xmath65 $ ] are involved in the tuning parameters @xmath128 and @xmath377 . this point might be important for designing practical estimators of @xmath44 , since the estimation of quadratic functionals is more realistic ( see , e.g. , @xcite ) than the estimation of @xmath378-norm . theorem [ thm1 ] can be reformulated to characterize the level of relevance @xmath168 for the relevant components of @xmath379 making their identification possible . in fact , an alternative way of stating theorem [ thm1 ] is the following : under conditions [ c1][c4 ] if @xmath182 is an arbitrary tuning parameter satisfying the first inequality in ( [ cond3 ] ) , then the estimator @xmath380with @xmath128 and @xmath377 chosen as in theorem [ thm1]satisfies @xmath381 if the smallest level of relevance @xmath168 for components @xmath382 of @xmath379 with @xmath359 is not smaller than @xmath383 . this statement can be easily deduced from the proof of theorem [ thm1 ] ; cf . the supplementary material @xcite . a natural question is now to check that the assumptions of theorem [ thm1 ] are tight in the asymptotic regimes of fixed sparsity and increasing ambient dimension , as well as increasing sparsity . we will only establish an analogue of claim ( ii ) of theorem [ thm3 ] . an attempt to prove a result similar to claim ( i ) of theorem [ thm3 ] was done in @xcite , theorem 2 . however , the result of @xcite involves a stringent assumption on the empirical gram matrix ( cf . condition ( 6 ) in @xcite ) and , unfortunately , we are unable to prove the existence of a sampling scheme for which this assumption is fulfilled . we assume that the errors @xmath384 are i.i.d . standard gaussian , and we focus our attention on the functional class @xmath255 . the following simple result shows that the conditions of theorem [ thm1 ] are tight in the case of fixed intrinsic dimension . [ prop4 ] let the design @xmath385^d$ ] be either deterministic or random . if for some positive @xmath386 , the inequality @xmath387 = 0.19em plus 0.05em minus 0.02em holds true , then there is a constant @xmath388 such that . the results proved in previous sections almost exhaustively answer the questions on the existence of consistent estimators of the sparsity pattern in the model of gaussian white noise and , to a smaller extent , in nonparametric regression . in fact as far as only rates of convergence are of interest , the result obtained in theorem [ prop2 ] is shown in section [ sec5 ] to be unimprovable . thus only the problem of finding sharp constants remains open . to make these statements more precise , let us consider the simplified set - up @xmath389 and define the following two regimes : * the regime of fixed sparsity , that is , when the sample size @xmath2 and the ambient dimension @xmath0 tend to infinity but the intrinsic dimension @xmath8 remains constant or bounded . * the regime of increasing sparsity , that is , when the intrinsic dimension @xmath8 tends to infinity along with the sample size @xmath2 and the ambient dimension @xmath0 . for simplicity , we will assume that @xmath390 for some @xmath20 . in the fixed sparsity regime , in view of theorems [ prop2 ] and [ thm1 ] , consistent estimation of the sparsity pattern can be achieved both in the gaussian white noise model and nonparametric regression as soon as @xmath391 , where @xmath392 is the constant defined by @xmath393 for the gaussian white noise model and @xmath394 for the regression model . on the other hand , by theorem [ thm3 ] and proposition [ prop4 ] , consistent estimation of the sparsity pattern is impossible if@xmath395 with @xmath396 . thus , up to multiplicative constants @xmath392 and @xmath397 ( which are clearly not sharp ) , the results of theorems [ prop2 ] and [ thm1 ] can not be improved in the regime of fixed sparsity . in the regime of increasing sparsity , the results we get in the model of gaussian white noise are much stronger than those for nonparametric regression . in the former model , taking the logarithm of both sides of inequality ( [ ordre2 ] ) and using formula ( [ eqlog ] ) for @xmath398 , we see that consistent estimation of @xmath44 is possible when , for some @xmath143 and for all @xmath2 , the following two conditions are fulfilled : @xmath399 with some constants @xmath400 and @xmath401 . on the other hand , theorem [ thm3 ] yields that there are some constants @xmath402 and @xmath403 such that it is impossible to consistently estimate @xmath44 if either one of the conditions @xmath404 is satisfied . first note that the left - hand side of the second condition in ( [ eq10 ] ) is exactly the same as the left - hand side of ( [ eq12b ] ) . if we compare now the left - hand side of the first condition in ( [ eq10 ] ) with the left - hand side of ( [ eq11 ] ) , we see that only the coefficients of @xmath8 differ . to measure the degree of difference of these two coefficients we draw in figure [ fig4 ] the plots of the functions @xmath405 and @xmath406 , with @xmath407 as is theorem [ thm3 ] . one can observe that the two curves are very close , especially for relatively large values of @xmath184 . this implies that the conditions in ( [ eq10 ] ) are tight . a simple consequence of inequalities ( [ eq10 ] ) and ( [ eq11 ] ) is that the consistent recovery of the sparsity pattern is possible under the condition @xmath408 and impossible for @xmath409 as @xmath4 , provided that @xmath410 . ( blue curve ) and @xmath406 ( red curve ) . ] still in the regime of increasing sparsity , but for nonparametric regression , we proved that consistent estimation of the sparsity pattern is possible whenever @xmath411 with some constants @xmath412 and @xmath413 . as we have already mentioned , the second condition in ( [ eq10b ] ) is tight , up to the choice of @xmath414 , in view of proposition [ prop4 ] . it is natural to expect that the first condition is tight as well , since it is in the model of gaussian white noise , which has the reputation of being simpler than the model of nonparametric regression . however , we do not have a mathematical proof of this statement . let us stress now that , all over this work , we have deliberately avoided any discussion on the computational aspects of the variable selection in nonparametric regression . the goal in this paper was to investigate the possibility of consistent recovery without paying attention to the complexity of the selection procedure . this lead to some conditions that could be considered a benchmark for assessing the properties of sparsity pattern estimators . as for the estimators proposed in section [ sec3 ] , it is worth noting that their computational complexity is not always prohibitively large . a recommended strategy is to compute the coefficients @xmath415 in a stepwise manner ; at each step @xmath416 only the coefficients @xmath415 with @xmath417 need to be computed and compared with the threshold . if some @xmath415 exceeds the threshold , then all the variables @xmath418 corresponding to nonzero coordinates of @xmath57 are considered as relevant . we can stop this computation as soon as the number of variables classified as relevant attains @xmath8 . while the worst - case complexity of this procedure is exponential , there are many functions @xmath43 for which the complexity of the procedure will be polynomial in @xmath0 . for example , this is the case for additive models in which @xmath419 for some univariate functions @xmath420 . note also that in the present study we focused exclusively on the consistency of variable selection without paying any attention to the consistency of regression function estimation . a thorough analysis of the latter problem being left to a future work , let us simply remark that in the case of fixed @xmath8 , under the conditions of theorem [ thm1 ] , it is straightforward to construct a consistent estimator of the regression function . in fact , it suffices to use a projection estimator with a properly chosen truncation parameter on the set of relevant variables . the situation is much more delicate in the case when the sparsity @xmath8 grows to infinity along with the sample size @xmath2 . presumably , condition ( [ eq10 ] ) is no longer sufficient for consistently estimating the regression function . the rationale behind this conjecture is that the minimax rate of convergence for estimating @xmath43 in our context , if we assume in addition that the set of relevant variables is known , is equal to @xmath421 . if the left - hand side of ( [ eq10 ] ) is equal to a constant and @xmath241 , then the aforementioned minimax rate does not tend to zero , making thus the estimator inconsistent . finally , we would like to mention that the selection of relevant variables is a challenging statistical task , which might be useful to perform independently of the task of regression function estimation . indeed , if we succeed in identifying relevant variables on a data - set having a small sample size , we can continue the data collection process more efficiently by recording only the values of relevant variables . this may considerably reduce the memory costs related to the data storage and the financial costs necessary for collecting new data . then , the regression function may be estimated more accurately on the base of this new ( larger ) data - set . to ease notation , we write @xmath151 instead of @xmath422 . it is clear that @xmath423 if and only if @xmath424 such that @xmath425 , where @xmath426 . for every @xmath427 , let us set @xmath428 and @xmath429 so that @xmath430 for @xmath431 , the first two terms of the last sum vanish and , therefore , we have @xmath432 where the last equality results from the fact that @xmath433 if @xmath434 . the random variable @xmath435 , being a centered sum of squares of independent standard gaussian random variables , follows a translated @xmath154-distribution . the tails of this distribution can be evaluated using the following result . [ lemmassart ] let @xmath436 be independent standard gaussian random variables . for every @xmath437 and for every vector @xmath438 , the following inequalities hold true : @xmath439 we apply this lemma to @xmath440 , for which @xmath441 and @xmath442 . setting @xmath443 and using the union bound , we get @xmath444 one checks that @xmath445 holds true for every pair of integers @xmath446 such that @xmath447 ; cf . the supplementary material @xcite for a proof . hence , for @xmath448 , we get @xmath449 . we begin with proving a stronger result that implies the claim of theorem [ prop2 ] . [ propnew ] let @xmath450 be a real number from @xmath228 . if for every @xmath359 and for @xmath146 the inequality @xmath451^{1/2}+ \biggl [ \frac{2\log(2s/\alpha)}{n } \biggr]^{1/2 } \biggr\}^2\hspace*{-35pt}\ ] ] holds true , then @xmath452 . to bound from above the probability of type ii error , we rely on the equivalence : @xmath453 if and only if @xmath454 such that@xmath455 . recall that @xmath146 . using bonferroni s inequality , we get @xmath456 \\[-8pt ] \nonumber & \le&\sum_{j\in j } \pb \bigl(\widehat q_{m_{s},j}^j\le\lambda_s \bigr ) \le s \max_{j\in j}\pb \bigl(\widehat q_{m_{s},j}^j\le \lambda_s \bigr).\end{aligned}\ ] ] by virtue of decomposition ( [ notation1 ] ) , @xmath457 one checks that @xmath458 is a drawn from @xmath154-distribution with @xmath459 degrees of freedom . therefore , using lemma [ lemmassart ] stated in previous section , we get @xmath460 . therefore , @xmath461 is upper - bounded by @xmath462 using the condition of the proposition , we get @xmath463 . combining this inequality with ( [ bonf2 ] ) , we get the result of proposition [ propnew ] . to deduce the claim of theorem [ prop2 ] from that of proposition [ propnew ] , we use the following lower bound : @xmath464 \\[-8pt ] \nonumber & \geq&\kappa- m_s^{-2}\sum_{j\in\supp(\bk)\subset j } \theta_{\bk } ^2\| \bk\|_2 ^ 2 \geq \kappa- m_s^{-2 } ls\end{aligned}\ ] ] for every @xmath359 . our choice of @xmath465 , @xmath466 , ensures that @xmath467 . finally , using a very rough bound ( which is sufficient for our purposes ) , the right - hand side in ( [ qq1 ] ) can be upper - bounded by @xmath468 if @xmath450 is chosen to be equal to @xmath469 . therefore , if @xmath470 then ( [ qq1 ] ) holds true with @xmath471 and , therefore , the type ii error has a probability less than or equal to @xmath469 . _ proof of the first assertion_. this proof can be found in @xcite ; we repeat here the arguments therein for the sake of keeping the paper self - contained . recall that @xmath195 admits an integral representation with the integrand @xmath472.\ ] ] for any @xmath473 , we define @xmath474 in such a way that @xmath475 by virtue of the cauchy schwarz inequality , it holds that @xmath476 , @xmath477 , implying that @xmath478 for all @xmath479 , that is , @xmath480 is strictly decreasing . furthermore , @xmath480 is obviously continuous with @xmath481 and @xmath482 . these properties imply the existence and the uniqueness of @xmath483 such that @xmath484 . furthermore , as the inverse of a decreasing function , the function @xmath485 is decreasing as well . we set @xmath486 so that @xmath230 is increasing . we also have @xmath487 _ proof of the second assertion_. we apply the saddle - point method to the integral representing @xmath488 ; see , for example , chapter ix in @xcite . it holds that @xmath489 \\[-8pt ] \nonumber & = & \frac{1}{2\pi i}\oint_{|z|=\zb } \bigl\{z(1-z)\bigr\}^{-1}e^{d^*\sfl_\gamma(z)}{\,dz}.\end{aligned}\ ] ] the first assertion of the proposition provided us with a real number @xmath227 such that @xmath490 and @xmath491 . the tangent to the steepest descent curve at @xmath227 is vertical . the path we choose for integration is the circle with center 0 and radius @xmath227 . as this circle and the steepest descent curve have the same tangent at @xmath227 , applying formula ( 1.8.1 ) of @xcite [ with @xmath492 since @xmath493 is real and positive ] , we get that @xmath494 when @xmath17 , as soon as the condition stands for the real part of the complex number @xmath495 . ] @xmath496\leq-\mu$ ] is satisfied for some @xmath497 and for any @xmath211 belonging to the circle @xmath498 and lying not too close to @xmath227 . to check that this is indeed the case , we remark that @xmath499=\log |\frac{\sfh(z)}{z^\gamma } |$ ] . hence , if @xmath500 with @xmath501 $ ] for some @xmath5020,\pi[$ ] , then @xmath503 therefore @xmath504\leq-\mu$ ] with @xmath505 . this completes the proof for the term @xmath195 . the term @xmath197 can be dealt in the same way . to prove ( i ) we apply lemma [ lemfano ] with @xmath506 in conjunction with a standard result , the proof of which can be found in @xcite and in the supplementary material @xcite . [ lem2 ] let @xmath507 be a subset of @xmath508 of cardinality @xmath509 and @xmath510 be a constant . define @xmath511 as a discrete measure supported on the finite set of functions @xmath512 such that @xmath513 for every @xmath514 . if we define the probability measure @xmath515 by @xmath516 , for every measurable set @xmath517 , and @xmath518 , then @xmath519 . without loss of generality , we can assume @xmath520 ( the general case can be reduced to this one by replacing @xmath184 and @xmath2 , respectively , by @xmath521 and @xmath522 ) . thus , @xmath523 . we denote the set @xmath524 by @xmath67 and choose @xmath525 as follows : @xmath262 is the dirac measure @xmath526 , @xmath527 is defined as in lemma [ lem2 ] with @xmath528 and @xmath529}^{-1/2}$ ] . the measures @xmath530 are defined similarly and correspond to the @xmath531 remaining sparsity patterns of cardinality @xmath8 . in view of inequality ( [ minor2 ] ) and lemma [ lemfano ] , it suffices to show that the measures @xmath257 satisfy @xmath532 and @xmath533 . combining lemma [ lem2 ] with @xmath534 and inequality ( [ ordre3 ] ) , we get @xmath535 . now , let us show that@xmath536 . by symmetry , this will imply that @xmath532 for every @xmath94 . since @xmath527 is supported by the set @xmath537 , it is clear that @xmath538=a^2[n_1(d^*,\gamma_l)-n_2(d^*,\gamma_l)]=1 $ ] and @xmath539&=&\sum _ { \bk \in \mathcal c_1(d^*,\gamma_l ) } k_j^2 a^2= \frac1{d^ * } \sum_{j=1}^{d^*}\sum _ { \bk\in\mathcal c_1(d^*,\gamma_l ) } k_j^2 a^2\leq a^2 \gamma_l n_1\bigl(d^ * , \gamma_l\bigr ) \\ & \le&\gamma_l \frac{n_1(d^*,\gamma_l ) } { n(d^*,\gamma_l ) } , \qquad j=1,\ldots , d^ * .\end{aligned}\ ] ] the results stated in section [ sec4 ] imply that @xmath540 . our choice of @xmath541 ensures that , for @xmath8 large enough , @xmath542 . this completes the proof of claim ( i ) . to prove ( ii ) , we still use lemma [ lemfano ] with @xmath543 and @xmath544 , where for every @xmath545 , @xmath546 is chosen as follows . let @xmath547 be all the subsets of @xmath45 containing exactly @xmath8 elements . we define @xmath546 , for @xmath548 , by its fourier coefficients @xmath549 as follows : @xmath550 obviously , all the functions @xmath546 belong to @xmath551 and , moreover , each @xmath552 has @xmath553 as sparsity pattern . one easily checks that our choice of @xmath552 implies @xmath554 . therefore , if @xmath555 , the desired inequality is satisfied . to conclude , it suffices to note that @xmath556 . in view of theorem [ prop2 ] , applied with @xmath319 and @xmath557 , the consistent [ uniformly in @xmath327 ] estimation of @xmath44 is possible if @xmath559 since @xmath323 is upper - bounded by some constant , there is a constant @xmath560 such that the left - hand side of the last display is upper - bounded by @xmath561 as proved in lemma [ lemcard0 ] below , @xmath562 . thus there is a constant @xmath563 such that @xmath564 combining these results , we see that under the conditions @xmath565 and @xmath566 consistent estimation of @xmath44 is possible . taking @xmath567 , we complete the proof of the first claim of the proposition . to prove the second assertion , we apply theorem [ thm3 ] . since it holds that @xmath568 , we deduce from theorem [ thm3 ] that there are some constants @xmath569 and @xmath570 such that if @xmath571 then consistent estimation of @xmath44 is impossible . since the @xmath3-dimensional @xmath351 ball with radius @xmath572 contains the @xmath350 ball of radius @xmath573 , @xmath574 for some constant @xmath575 . by rearranging different terms , we get the desired result . [ lemcard0 ] for every @xmath576 and @xmath577 , @xmath578 . one readily checks that if @xmath579 , then the hypercube centered at @xmath57 with side of length @xmath78 is included in the ball centered at the origin and having radius @xmath580 . therefore , @xmath581 $ ] , where @xmath582 $ ] stands for the volume of the unit ball in @xmath204 . using the well - known formula for the latter and the stirling approximation , for every @xmath583 , we get @xmath584= \frac{2\pi^{d^*/2}}{d^*\gamma(d^*/2 ) } \le0.4\frac{(4\pi e / d^*)^{d^*/2}}{\sqrt{2d^*}}$ ] . this implies that @xmath585 and the result follows . the authors would like to thank the reviewers for very useful remarks .

packet - switching wdm ring networks with a hotspot transporting unicast , multicast , and broadcast traffic are important components of high - speed metropolitan area networks . for an arbitrary multicast fanout traffic model with uniform , hotspot destination , and hotspot source packet traffic , we analyze the maximum achievable long - run average packet throughput , which we refer to as _ multicast capacity _ , of bi - directional shortest - path routed wdm rings . we identify three segments that can experience the maximum utilization , and thus , limit the multicast capacity . we characterize the segment utilization probabilities through bounds and approximations , which we verify through simulations . we discover that shortest - path routing can lead to utilization probabilities above one half for moderate to large portions of hotspot source multi- and broadcast traffic , and consequently multicast capacities of less than two simultaneous packet transmissions . we outline a one - copy routing strategy that guarantees a multicast capacity of at least two simultaneous packet transmissions for arbitrary hotspot source traffic . keywords : hotspot traffic , multicast , packet throughput , shortest path routing , spatial reuse , wavelength division multiplexing ( wdm ) .

0804.3215

0.275 in optical packet - switched ring wavelength division multiplexing ( wdm ) networks have emerged as a promising solution to alleviate the capacity shortage in the metropolitan area , which is commonly referred to as metro gap . packet - switched ring networks , such as the resilient packet ring ( rpr ) @xcite , overcome many of the shortcomings of circuit - switched ring networks , such as low provisioning flexibility for packet data traffic @xcite . in addition , the use of multiple wavelength channels in wdm ring networks , see e.g. , @xcite , overcomes a key limitation of rpr , which was originally designed for a single - wavelength channel in each ring direction . in optical packet - switched ring networks , the destination nodes typically remove ( strip ) the packets destined to them from the ring . destination stripping _ allows the destination node as well as other nodes downstream to utilize the wavelength channel for their own transmissions . with this so - called _ spatial wavelength reuse _ , multiple simultaneous transmissions can take place on any given wavelength channel . spatial wavelength reuse is maximized through shortest path routing , whereby the source node sends a packet in the ring direction that reaches the destination with the smallest hop distance , i.e. , traversing the smallest number of intermediate network nodes . multicast traffic is widely expected to account for a large portion of the metro area traffic due to multi - party communication applications , such as tele - conferences @xcite , virtual private network interconnections , interactive distance learning , distributed games , and content distribution . these multi - party applications are expected to demand substantial bandwidths due to the trend to deliver the video component of multimedia content in the high - definition television ( hdtv ) format or in video formats with even higher resolutions , e.g. , for digital cinema and tele - immersion applications . while there is at present scant quantitative information about the multicast traffic volume , there is ample anecdotal evidence of the emerging significance of this traffic type @xcite . as a result , multicasting has been identified as an important service in optical networks @xcite and has begun to attract significant attention in optical networking research as outlined in section [ lit : sec ] . metropolitan area networks consist typically of edge rings that interconnect several access networks ( e.g. , ethernet passive optical networks ) and connect to a metro core ring @xcite . the metro core ring interconnects several metro edge rings and connects to the wide area network . the node connecting a metro edge ring to the metro core ring is typically a traffic hotspot as it collects / distributes traffic destined to / originating from other metro edge rings or the wide area network . similarly , the node connecting the metro core ring to the wide area network is typically a traffic hotspot . examining the capacity of optical packet - switched ring networks for hotspot traffic is therefore very important . in this paper we examine the multicast capacity ( maximum achievable long run average multicast packet throughput ) of bidirectional wdm optical ring networks with a single hotspot for a general fanout traffic model comprising unicast , multicast , and broadcast traffic . we consider an arbitrary traffic mix composed of uniform traffic , hotspot destination traffic ( from regular nodes to the hotspot ) , and hotspot source traffic ( from the hotspot to regular nodes ) . we study the widely considered node architecture that allows nodes to transmit on all wavelength channels , but to receive only on one channel . we initially examine shortest path routing by deriving bounds and approximations for the ring segment utilization probabilities due to uniform , hotspot destination , and hotspot source packet traffic . we prove that there are three ring segments ( in a given ring direction ) that govern the maximum segment utilization probability . for the clockwise direction in a network with nodes @xmath0 and wavelengths @xmath1 ( with @xmath2 ) , whereby node 1 receives on wavelength 1 , node 2 on wavelength 2 , @xmath3 , node @xmath4 on wavelength @xmath4 , node @xmath5 on wavelength 1 , and so on , and with node @xmath6 denoting the index of the hotspot node , the three critical segments are identified as : * the segment connecting the hotspot , node @xmath6 , to node 1 on wavelength 1 , * the segment connecting node @xmath7 to node @xmath4 on wavelength @xmath4 , and * the segment connecting node @xmath8 to node @xmath6 on wavelength @xmath4 . the utilization on these three segments limits the maximum achievable multicast packet throughput . we observe from the derived utilization probability expressions that the utilizations of the first two identified segments exceed 1/2 ( and approach 1 ) for large fractions of hotspot source multi- and broadcast traffic , whereas the utilization of the third identified segment is always less than or equal to 1/2 . thus , shortest path routing achieves a long run average multicast throughput of less than two simultaneous packet transmissions ( and approaching one simultaneous packet transmission ) for large portions of hotspot source multi- and broadcast traffic . we specify one - copy routing which sends only one packet copy for hotspot source traffic , while uniform and hotspot destination packet traffic is still served using shortest path routing . one - copy routing ensures a capacity of at least two simultaneous packet transmissions for arbitrary hotspot source traffic , and at least approximately two simultaneous packet transmissions for arbitrary overall traffic . we verify the accuracy of our bounds and approximations for the segment utilization probabilities , which are exact in the limit @xmath9 , through comparisons with utilization probabilities obtained from discrete event simulations . we also quantify the gains in maximum achievable multicast throughput achieved by the one - copy routing strategy over shortest path routing through simulations . this paper is structured as follows . in the following subsection , we review related work . in section [ model : sec ] , we introduce the detailed network and traffic models and formally define the multicast capacity . in section [ gen_prop : sec ] , we establish fundamental properties of the ring segment utilization in wdm packet rings with shortest path routing . in section [ util_laneqla : sec ] , we derive bounds and approximations for the ring segment utilization due to uniform , hotspot destination , and hotspot source packet traffic on the wavelengths that the hotspot is not receiving on , i.e. , wavelengths @xmath10 in the model outlined above . in section [ util_laeqla : sec ] , we derive similar utilization probability bounds and approximations for wavelength @xmath4 that the hotspot receives on . in section [ max_segment : sec ] , we prove that the three specific segments identified above govern the maximum segment utilization and multicast capacity in the network , and discuss implications for packet routing . in section [ num : sec ] , we present numerical results obtained with the derived utilization bounds and approximations and compare with verifying simulations . we conclude in section [ concl : sec ] . there has been increasing research interest in recent years for the wide range of aspects of multicast in general mesh circuit - switched wdm networks , including lightpath design , see for instance @xcite , traffic grooming , see e.g. , @xcite , routing and wavelength assignment , see e.g. , @xcite , and connection carrying capacity @xcite . similarly , multicasting in packet - switched single - hop star wdm networks has been intensely investigated , see for instance @xcite . in contrast to these studies , we focus on packet - switched wdm ring networks in this paper . multicasting in circuit - switched wdm rings , which are fundamentally different from the packet - switched networks considered in this paper , has been extensively examined in the literature . the scheduling of connections and cost - effective design of bidirectional wdm rings was addressed , for instance in @xcite . cost - effective traffic grooming approaches in wdm rings have been studied for instance in @xcite . the routing and wavelength assignment in reconfigurable bidirectional wdm rings with wavelength converters was examined in @xcite . the wavelength assignment for multicasting in circuit - switched wdm ring networks has been studied in @xcite . for unicast traffic , the throughputs achieved by different circuit - switched and packet - switched optical ring network architectures are compared in @xcite . optical _ packet - switched _ wdm ring networks have been experimentally demonstrated , see for instance @xcite , and studied for unicast traffic , see for instance @xcite . multicasting in packet - switched wdm ring networks has received increasing interest in recent years @xcite . the photonics level issues involved in multicasting over ring wdm networks are explored in @xcite , while a node architecture suitable for multicasting is studied in @xcite . the general network architecture and mac protocol issues arising from multicasting in packet - switched wdm ring networks are addressed in @xcite . the fairness issues arising when transmitting a mix of unicast and multicast traffic in a ring wdm network are examined in @xcite . the multicast capacity of packet - switched wdm ring networks has been examined for uniform packet traffic in @xcite . in contrast , we consider non - uniform traffic with a hotspot node , as it commonly arises in metro edge rings @xcite . studies of non - uniform traffic in optical networks have generally focused on issues arising in circuit - switched optical networks , see for instance @xcite . a comparison of circuit - switching to optical burst switching network technologies , including a brief comparison for non - uniform traffic , was conducted in @xcite . the throughput characteristics of a mesh network interconnecting routers on an optical ring through fiber shortcuts for non - uniform unicast traffic were examined in @xcite . the study @xcite considered the throughput characteristics of a ring network with uniform unicast traffic , where the nodes may adjust their send probabilities in a non - uniform manner . the multicast capacity of a single - wavelength packet - switched ring with non - uniform traffic was examined in @xcite . in contrast to these works , we consider non - uniform traffic with an arbitrary fanout , which accommodates a wide range of unicast , multicast , and broadcast traffic mixes , in a wdm ring network . we let @xmath6 denote the number of network nodes , which we index sequentially by @xmath11 , in the clockwise direction and let @xmath12 denote the set of network nodes . for convenience , we label the nodes modulo @xmath6 , e.g. , node @xmath6 is also denoted by @xmath13 or @xmath14 . we consider the family of node structures where each node can transmit on any wavelength using either one or multiple tunable transmitters @xmath15 or an array of @xmath4 fixed - tuned transmitters @xmath16 , and receive on one wavelength using a single fixed - tuned receiver @xmath17 . for @xmath18 , each node has its own home channel for reception . for @xmath19 , each wavelength is shared by @xmath20 nodes , which we assume to be an integer . for @xmath21 , we let @xmath22 denote the clockwise oriented ring segment connecting node @xmath23 to node @xmath24 . analogously , we let @xmath25 denote the counter clockwise oriented ring segment connecting node @xmath24 to node @xmath23 . each ring deploys the same set of wavelength channels @xmath26 , one set on the clockwise ring and another set on the counterclockwise ring . the nodes @xmath27 with @xmath28 share the drop wavelength @xmath29 . we refer to the incoming edges of these nodes , i.e. , the edges @xmath30 and @xmath31 , as _ critical edges _ on @xmath29 . for multicast traffic , the sending node generates a copy of the multicast packet for each wavelength that is drop wavelength for at least one destination node . denote by @xmath32 the node that is the sender . we introduce the random set of destinations ( fanout set ) @xmath33 moreover , we define the set of active nodes @xmath34 as the union of the sender and all destinations , i.e. , @xmath35 . we consider a traffic model combining a portion @xmath36 of uniform traffic , a portion @xmath37 of hotspot destination traffic , and a portion @xmath38 of hotspot source traffic with @xmath39 and @xmath40 : uniform@xmath41traffic : : a given generated packet is a _ uniform traffic _ packet with probability @xmath36 . for such a packet , the sending node is chosen uniformly at random amongst all network nodes @xmath42 . once the sender @xmath32 is chosen , the number of receivers ( fanout ) @xmath43 is chosen at random according to a discrete probability distribution @xmath44 . once the fanout @xmath45 is chosen , the random set of destinations ( fanout set ) @xmath46 is chosen uniformly at random amongst all subsets of @xmath47 having cardinality @xmath45 . we denote by @xmath48 the probability measure associated with uniform traffic . hotspot@xmath41destination@xmath41traffic : : a given packet is a _ hotspot destination traffic _ packet with probability @xmath37 . for a hotspot destination traffic packet , node @xmath6 is always a destination . the sending node is chosen uniformly at random amongst the other nodes @xmath49 . once the sender @xmath32 is chosen , the fanout @xmath43 is chosen at random according to a discrete probability distribution @xmath50 . once the fanout @xmath45 is chosen , a random fanout subset @xmath51 is chosen uniformly at random amongst all subsets of @xmath52 having cardinality @xmath53 , and the fanout set is @xmath54 . we denote by @xmath55 the probability measure associated with hotspot destination traffic . hotspot@xmath41source@xmath41traffic : : a given packet is a _ hotspot source traffic _ packet with probability @xmath38 . for such a packet , the sending node is chosen to be node @xmath6 . the fanout @xmath56 is chosen at random according to a discrete prob . distribution @xmath57 . once the fanout @xmath45 is chosen , a random fanout set @xmath58 is chosen uniformly at random amongst all subsets of @xmath49 having cardinality @xmath45 . we denote by @xmath59 the probability measures associated with hotspot source traffic . while our analysis assumes that the traffic type , the source node , the fanout , and the fanout set are drawn independently at random , this independence assumption is not critical for the analysis . our results hold also for traffic patterns with correlations , as long as the long run average segment utilizations are equivalent to the utilizations with the independence assumption . for instance , our results hold for a correlated traffic model where a given source node transmits with a probability @xmath60 to exactly the same set of destinations as the previous packet sent by the node , and with probability @xmath61 to an independently randomly drawn number and set of destination nodes . we denote by @xmath62 the probability measure @xmath48 conditioned upon @xmath63 , and define @xmath64 and @xmath65 analogously . we denote the set of nodes with drop wavelength @xmath29 by @xmath66 the set of all destinations with drop wavelength @xmath29 is then @xmath67 moreover , we use the following notation : for @xmath68 we denote the probability of @xmath69 destinations on wavelength @xmath29 by @xmath70 , @xmath71 , and @xmath72 for uniform , hotspot destination , and hotspot source traffic , respectively . since the fanout set is chosen uniformly at random among all subsets of @xmath47 having cardinality @xmath45 , these usage - probabilities can be expressed by @xmath73 , @xmath74 , and @xmath75 . depending on whether the sender is on the drop - wavelength or not , we obtain slightly different expressions . as will become evident shortly , it suffices to focus on the case where the sender is on the considered drop wavelength @xmath29 , i.e. , @xmath76 , since the relevant probabilities are estimated through comparisons with transformations ( enlarged , reduced or right / left - shifted ring introduced in appendix [ appendixa ] ) that put the sender in @xmath77 . through elementary combinatorial considerations we obtain the following probability distributions : for uniform traffic , the probability for having @xmath78 destinations on wavelength @xmath29 is @xmath79 for hotspot destination traffic , we obtain for wavelengths @xmath80 and @xmath81 @xmath82 as well as for wavelength @xmath4 homing the hotspot and @xmath83 @xmath84 finally , for hotspot source traffic , we obtain for @xmath80 and @xmath78 @xmath85 and for @xmath86 and @xmath87 @xmath88 for a given wavelength @xmath29 , we denote by @xmath89 the probability measure @xmath48 conditioned upon @xmath90 , and define @xmath91 and @xmath92 analogously . whenever it is clear which wavelength @xmath29 is considered we omit the subscript @xmath29 and write @xmath93 , or @xmath94 . we introduce the set of active nodes @xmath95 on a given drop wavelength @xmath29 as @xmath96 we order the nodes in this set in increasing order of their node indices , i.e. , @xmath97 and consider the `` gaps '' @xmath98 between successive nodes in the set @xmath99 we have denoted here again by @xmath100 the random number of destinations with drop wavelength @xmath29 . for shortest path routing , i.e. , to maximize spatial wavelength reuse , we determine the largest of these gaps . since there may be a tie among the largest gaps ( in which case one of the largest gaps is chosen uniformly at random ) , we denote the selected largest gap as `` @xmath101 '' " ( for `` chosen largest gap '' ) . suppose the @xmath101 is between nodes @xmath102 and @xmath103 . with shortest path routing , the packet is then sent from the sender @xmath32 to node @xmath102 , and from the sender @xmath32 to node @xmath103 in the opposite direction . thus , the largest gap is not traversed by the packet transmission . note that by symmetry , @xmath104 , and @xmath105 . more generally , for reasons of symmetry , it suffices to compute the utilization probabilities for the clockwise oriented edges . for @xmath106 , we abbreviate @xmath107 it will be convenient to call node @xmath6 also node @xmath13 . we let @xmath108 , be a random variable denoting the first node bordering the chosen largest gap on wavelength @xmath29 , when this gap is considered clockwise . the utilization probability for the clockwise segment @xmath109 on wavelength @xmath29 is given by @xmath110 our primary performance metric is the maximum packet throughout ( stability limit ) . more specifically , we define the ( effective ) multicast capacity @xmath111 as the maximum number of packets ( with a given traffic pattern ) that can be sent simultaneously in the long run , and note that @xmath111 is given as the reciprocal of the largest ring segment utilization probability , i.e. , @xmath112 first , we prove a general recursion formula for shortest path routing . [ prop31 ] let @xmath113 be a fixed wavelength . for all nodes @xmath114 , @xmath115 there are two complementary events leading to @xmath116 : ( a ) the packet traverses ( on wavelength @xmath29 ) both the clockwise segment @xmath117 and the preceding clockwise segment @xmath118 , i.e. , the sender is a node @xmath119 , and ( b ) node @xmath109 is the sender ( @xmath120 ) and transmits the packet in the clockwise direction , so that it traverses segment @xmath121 following node @xmath109 ( in the clockwise direction ) . formally , @xmath122 next , note that the event that the clockwise segment @xmath118 is traversed can be decomposed into two complementary events , namely ( a ) segments @xmath118 and @xmath117 are traversed , and ( b ) segment @xmath118 is traversed , but not segment @xmath117 , i.e. , @xmath123 similarly , we can decompose the event of node @xmath109 being the sender as @xmath124 hence , we can express @xmath125 as @xmath126 now , note that there are two complementary events that result in the clg to start at node @xmath109 , such that clockwise segment @xmath127 is inside the clg : @xmath128 node @xmath109 is the last destination node reached by the clockwise transmission , i.e. , segment @xmath109 is used , but segment @xmath127 is not used , and @xmath129 node @xmath109 is the sender and transmits only a packet copy in the counter clockwise direction . hence , @xmath130 therefore , we obtain the general recursion @xmath131 we introduce the left ( counter clockwise ) shift and the right ( clockwise ) shift of node @xmath109 to be @xmath132 and @xmath133 given by @xmath134 the counter clockwise shift maps a node @xmath109 not homed on @xmath29 onto the nearest node in the counter clockwise direction that is homed on @xmath29 . similarly , the clockwise shift maps a node @xmath109 not homed on @xmath29 onto the closest node in the clockwise direction that is homed on @xmath29 . for the traffic on wavelength @xmath29 , we obtain by repeated application of proposition [ prop31 ] @xmath135 note that the clg on @xmath29 can only start @xmath128 at the source node , irrespective of whether it is on @xmath29 , or @xmath129 at a destination node on @xmath29 . consider a given node @xmath109 that is not on @xmath29 , then the nodes in @xmath136 are not on @xmath29 . ( if node @xmath109 is on @xmath29 , i.e. , @xmath137 , then trivially the set @xmath138 is empty and @xmath139 . ) hence , the clg on @xmath29 can only start at a node in @xmath140 if that node is the source node , i.e. , @xmath141 next , note that the event that a node in @xmath138 is the source node can be decomposed into the two complementary events @xmath128 a node in @xmath142 is the source node and the clg on @xmath29 starts at that node , and @xmath129 a node in @xmath142 is the source node and the clg does not start at that node . hence , @xmath143 inserting ( [ glambdaequals ] ) and ( [ pscomplement ] ) in ( [ pnlambda1 ] ) we obtain @xmath144 which directly leads to [ cor critical non critical ] the usage of non - critical segments is smaller than the usage of critical segments , more precisely for @xmath145 : @xmath146 to compare the expected length of the largest gap on a wavelength in the wdm ring with the expected length of the largest gap in the single wavelength ring , we introduce the enlarged and reduced ring in appendix [ appendixa ] . in brief , in the enlarged ring , an extra node is added on the considered wavelength between the @xmath29-neighbors of the source node . this enlargement results in @xmath147 a set of @xmath148 nodes homed on the considered wavelength , and @xmath149 an enlarged set of active nodes @xmath150 containing the original destination nodes plus the added extra node ( which in a sense represents the source node on the considered wavelength ) for a total of @xmath151 active nodes . the expected length of the largest gap on this enlarged wavelength ring with @xmath152 active nodes among @xmath148 nodes homed on the wavelength ( a ) is equivalent to @xmath4 times the expected length of the largest gap on a single wavelength ring with @xmath153 destination nodes and one source node among @xmath6 nodes homed on the ring , and ( b ) provides an upper bound on the expected length of the largest gap on the original wavelength ring ( before the enlargement ) . in the reduced ring , the left- and right - shifted source node are merged into one node on the considered wavelength , resulting @xmath147 in a set of @xmath154 nodes homed on the considered wavelength , and @xmath149 a set @xmath155 of @xmath156 , or @xmath152 active nodes . the expected length of the largest gap decreases with increasing number of active nodes , hence we consider the case with @xmath152 active nodes for a lower bound . the expected length of the largest gap on the reduced wavelength ring with @xmath152 active nodes among @xmath157 nodes homed on the wavelength ( a ) is equivalent to @xmath4 times the expected length of the largest gap on a single wavelength ring with @xmath153 destination nodes and one source node among @xmath6 nodes homed on the ring , and ( b ) provides a lower bound on the expected length of the largest gap on the original wavelength ring ( before the reduction ) . from these two constructions , which are formally provided in appendix [ appendixa ] , we directly obtain : [ prop33 ] given that the cardinality of @xmath158 is @xmath69 , the expected length of the clg on wavelength @xmath29 is bounded by : @xmath159 where @xmath160 denotes the expected length of the clg for a single wavelength ring with @xmath6 nodes , when the active set is chosen uniformly at random from all subsets of @xmath161 with cardinality @xmath162 . the expected length of the largest gap @xmath163 @xcite is given for @xmath164 , by @xmath165 , where @xmath166 denotes the distribution of the length of the largest gap . let @xmath167 denote the probability that an arbitrary gap has @xmath168 hops . then the distribution @xmath169 may be computed using the recursion @xmath170 together with the initialization @xmath171 and @xmath172 , where @xmath173 denotes the kronecker delta . whereby , @xmath171 means a ring with only one active node has only one gap of length @xmath6 , hence the largest gap has length @xmath6 with probability one . similarly , @xmath172 means a ring with all nodes active ( broadcast case ) has @xmath6 gaps with length one , hence the largest gap has length @xmath174 with probability one . this initialization directly implies @xmath175 as well as @xmath176 . obviously , we have to set @xmath177 for @xmath178 . in the setting of uniform traffic , one has for all @xmath179 and @xmath180 , for reasons of symmetry : @xmath181 for @xmath179 , the difference between critical and non - critical edges , corresponding to corollary [ cor critical non critical ] , can be estimated by @xmath182 with shortest path routing , on average @xmath183 segments are traversed on @xmath29 to serve a uniform traffic packet . equivalently , we obtain the expected number of traversed segments by summing the utilization probabilities of the individual segments , i.e. , as @xmath184 , which , due to symmetry , equals @xmath185 . hence , @xmath186 and @xmath187 expressing @xmath188 using corollary [ cor critical non critical ] , we obtain @xmath189 solving for @xmath190 yields @xmath191 hence , the inequalities ( [ estimating the difference ] ) lead to @xmath192 employing the bounds for @xmath193 from proposition [ prop33 ] gives @xmath194 the only difference to uniform traffic is that @xmath6 can not be a sender , since it is already a destination , i.e. , @xmath195 using @xmath196 , we obtain @xmath197 due to the factor @xmath198 , the second term is negligible in the context of large networks . since node @xmath6 is the sender ( and given that there is at least one destination node on @xmath29 ) , it sends a packet copy over segment @xmath199 on wavelength @xmath29 if the clg on @xmath29 starts at a node with index @xmath109 or higher . hence , the usage probability of a segment can be computed as @xmath200 for @xmath106 . we notice immediately that @xmath201 is monotone decreasing in @xmath109 . moreover , for all @xmath202 , equation ( [ critical non critical ] ) simplifies to @xmath203 since the sender is node @xmath204 and consequently @xmath205 for the considered @xmath206 . since @xmath201 is monotone decreasing in @xmath109 , the maximally used critical segment on wavelength @xmath29 is @xmath207 . with node @xmath6 being the sender , the clg on @xmath29 can only start at the source node @xmath204 , or at a destination node homed on @xmath29 . if the clg does not start at @xmath204 , the segment @xmath207 leading to the first node homed on @xmath29 , namely node @xmath29 , is utilized . hence , @xmath208 observe that @xmath209 which is exploited in section [ summarylaneqla : sec ] . enlarging the ring leads to @xmath210 since the gaps bordering node @xmath13 are enlarged whereas the lengths of all other gaps are unchanged . a right shifting of @xmath32 yields the following lower bound : @xmath211 thus , @xmath212 for @xmath80 we obtain from ( [ segment_util : eqn ] ) and ( [ hsdestsegutil ] ) @xmath214 using corollary [ cor critical non critical ] for @xmath215 and ( [ gamma usage for ungleich ] ) for @xmath94 yields @xmath216 i.e. , the segment number @xmath29 experiences the maximum utilization on wavelength @xmath29 . moreover , inequality ( [ usage halb leer halb voll ] ) yields @xmath217 i.e. , the first segment on wavelength 1 , experiences the maximum utilization among all segments on all wavelengths @xmath213 . from ( [ p n lambda ] ) in conjunction with ( [ pellalphabounds ] ) and ( [ qellgammalambdabounds ] ) we obtain @xmath218 and @xmath219 we obtain an approximation of the segment utilization by considering the behavior of these bounds for large @xmath220 . large @xmath221 imply @xmath222 as well as @xmath223 , and @xmath224 . intuitively , this last relation means that the expected length of the largest gap on a ring network with @xmath69 destination nodes among @xmath154 nodes is approximately equal to the largest gap when there are @xmath69 destination nodes among @xmath225 nodes . with these considerations we can simplify the bounds given above and obtain the approximation ( valid for large @xmath221 ) : @xmath226 for uniform traffic this case , of course , does not differ from the case @xmath80 . since @xmath6 is a destination node , by symmetry it is reached by a clockwise transmission with probability one half , i.e. , @xmath228 for hotspot destination traffic , node @xmath6 can not be the sender , i.e. , @xmath229 . hence , by proposition [ prop31 ] : @xmath230 moreover , we have from corollary [ cor critical non critical ] with @xmath231 and @xmath86 : @xmath232 to estimate @xmath233 , we introduce , as before , the left- resp . right - shift of @xmath32 , given by @xmath234 left and right shifting of @xmath32 leads to the following bounds for the probability @xmath235 , which are proven in appendix [ appendixbb ] . [ prop in beta ] for hotspot destination traffic , conditioning on the cardinality of @xmath236 to be @xmath69 , the probability that the clg starts at node @xmath13 is bounded by : @xmath237 inserting the bounds from proposition [ prop in beta ] and noting that @xmath238 leads to @xmath239 and @xmath240 since we know that @xmath6 is the sender and has drop wavelength @xmath4 , we have a symmetric setting on @xmath236 and can directly apply the results of the single wavelength setting @xcite . in particular , we obtain from section 3.1.3 in @xcite for @xmath241 @xmath242 and @xmath243 inserting the bounds derived in the preceding sections in ( [ segment_util : eqn ] ) , we obtain @xmath244 and @xmath245 whereby @xmath246 is given by setting @xmath86 in ( [ mu_laell : eqn ] ) . moreover , @xmath247 and @xmath248 considering again these bounds for large @xmath221 , we obtain the approximations : @xmath249 as well as @xmath250 with ( [ max lambda neq lambda ] ) and a detailed consideration of wavelength @xmath86 , we prove in appendix [ appendixc ] the main theoretical result : [ mainthm ] the maximum segment utilization probability is @xmath251 it thus remains to compute the three probabilities on the right hand side . we have no exact result in the most general setting ( it would be possible to give recursive formulae , but these would be prohibitively complex ) . however , we have given upper and lower bounds and approximations in sections [ summarylaneqla : sec ] and [ summarylaeqla : sec ] , which match rather well in most situations , as demonstrated in the next section , and have the same asymptotics when @xmath252 while @xmath4 remains fixed . toward assessing the considered shortest - path routing strategy , we directly observe , that @xmath253 is always less or equal to @xmath254 . on the other hand , the first two usage probabilities will , for @xmath38 large enough , become larger than @xmath254 , especially for hotspot source traffic with moderate to large fanouts . hence , shortest - path routing will result in a multicast capacity of less than two for large portions of hotspot source multi- and broadcast traffic , which may arise in content distribution , such as for ip tv . the intuitive explanation for the high utilization of the segments @xmath255 and @xmath256 with shortest - path routing for multi- and broadcast hotspot source traffic is a follows . consider the transmission of a given hotspot source traffic packet with destinations on wavelength @xmath4 homing the hotspot . if the packet has a single destination uniformly distributed among the other @xmath157 nodes homed on wavelength @xmath4 , then the clg is adjacent and to the left ( i.e. , in the counter clockwise sense ) of the hotspot with probability one half . hence , with probability one half a packet copy is sent in the clockwise direction , utilizing the segment @xmath256 . with an increasing number of uniformly distributed destination nodes on wavelength @xmath4 , it becomes less likely that the clg is adjacent and to the left of the hotspot , resulting in increased utilization of segment @xmath256 . in the extreme case of a broadcast destined from the hotspot to all other @xmath157 nodes homed on @xmath4 , the clg is adjacent and to the left of the hotspot with probability @xmath257 , i.e. , segment @xmath256 is utilized with probability @xmath258 . with probability @xmath259 the clg is not adjacent to the hotspot , resulting in two packet copy transmissions , i.e. , a packet copy is sent in each ring direction . for wavelength 1 , the situation is subtly different due to the rotational offset of the nodes homed on wavelength 1 from the hotspot . that is , node 1 has a hop distance of 1 from the hotspot ( in the clockwise direction ) , whereas the highest indexed node on wavelength 1 , namely node @xmath260 has a hop distance of @xmath7 from the hotspot ( in the counter clockwise direction ) . as for wavelength @xmath4 , for a given packet with a single uniformly distributed destination on wavelength 1 , the clg is adjacent and to the left of the hotspot with probability one half , and the packet consequently utilizes segment @xmath255 with probability one half . with increasing number of destinations , the probability of the clg being adjacent and to the left of the hotspot decreases , and the utilization of segment @xmath255 increases , similar to the case for wavelength @xmath4 . for a broadcast destined to all @xmath221 nodes on wavelength 1 , the situation is different from wavelength @xmath4 , in that the clg is never adjacent to the hotspot , i.e. , the hotspot always sends two packet copies , one in each ring direction . to overcome the high utilization of the segments @xmath255 and @xmath256 due to hotpot source multi- and broadcast traffic , we propose _ one - copy ( oc ) routing _ : with one - copy routing , uniform traffic and hotspot destination traffic are still served using shortest path routing . hotspot source traffic is served using the following counter - based policy . we define the counter @xmath261 to denote the number of nodes homed on @xmath29 that would need to be traversed to reach all destinations on @xmath29 with one packet transmission in the clockwise direction ( whereby the final reached destination node counts as a traversed node ) . if @xmath262 , then one packet copy is sent in the clockwise direction to reach all destinations . if @xmath263 , then one packet copy is sent in the counter clockwise direction to reach all destinations . ties , i.e. , @xmath264 , are served in either clockwise or counter clockwise direction with probability one half . for hotspot source traffic with arbitrary traffic fanout , this counter - based one - copy routing ensures a maximum utilization of one half on any ring segment . note that the counter - based policy considers only the nodes homed on the considered wavelength @xmath29 to ensure that the rotational offset between the wavelength @xmath4 homing the hotspot and the considered wavelength @xmath29 does not affect the routing decisions . we propose the following strategy for switching between shortest path ( sp ) and one - copy ( oc ) routing . shortest path routing is employed if both ( [ sim 1 ] ) and ( [ sim 2 ] ) are less than one half . if ( [ sim 1 ] ) or ( [ sim 2 ] ) exceeds one half , then one - copy routing is used . for the practical implementation of this switching strategy , the hotspot can periodically estimate the current traffic parameters , i.e. , the traffic portions @xmath36 , @xmath37 , and @xmath38 as well as the corresponding fanout distributions @xmath265 , @xmath266 , and @xmath267 , for instance , through a combination of traffic measurements and historic traffic patterns , similar to @xcite . from these traffic parameter estimates , the hotspot can then evaluate ( [ sim 1 ] ) and ( [ sim 2 ] ) . to obtain a more refined criterion for switching between shortest path routing and one - copy routing we proceed as follows . we characterize the maximum segment utilization with shortest path routing more explicitly by inserting ( [ sim 1 ] ) , ( [ sim 2 ] ) , and ( [ sim 3 ] ) in ( [ mainthm : eqn ] ) to obtain : @xmath268 whereby we noted that the definition of @xmath70 in ( [ mu_laell : eqn ] ) directly implies that @xmath70 is independent of @xmath29 . clearly , the hotspot source traffic does not influence the maximum segment utilization as long as @xmath269 and @xmath270 thus , if @xmath271 , then all traffic is served using shortest path routing . we next note that theorem [ mainthm ] does not hold for the one - copy routing strategy . we therefore bound the maximum segment utilization probability with one - copy routing by observing that ( [ pellalphabounds ] ) together with proposition [ cor critical non critical ] and ( [ estimating the difference ] ) implies that asymptotically for all @xmath272 @xmath273 hence , @xmath274 is asymptotically constant . moreover , similar as in the single wavelength case @xcite , we have @xmath275 therefore , the maximum segment utilization with one - copy routing is ( approximately ) bounded by @xmath276 comparing ( [ oc_util : eqn ] ) with ( [ mainthm : eqn ] ) we observe that the maximum segment utilization with one - copy routing is smaller than with shortest path routing if the following threshold conditions hold : * if @xmath277 , then set @xmath278 otherwise set @xmath279 . * if @xmath280 , then set @xmath281 otherwise set @xmath282 . if @xmath283 , then one - copy routing is employed . for @xmath38 values between @xmath284 and @xmath285 , the hotspot could numerically evaluate the maximum segment utilization probability of shortest path routing with the derived approximations . the hotspot could also obtain the segment utilization probabilities with one - copy routing through discrete event simulations to determine whether shortest path routing or one - copy routing of the hotspot traffic is preferable for a given set of traffic parameter estimates . in this section we present numerical results obtained from the derived bounds and approximations of the utilization probabilities as well as verifying simulations . we initially simulate individual , stochastically independent packets generated according to the traffic model of section [ model : sec ] and routed according to the shortest path routing policy . we determine estimates of the utilization probabilities of the three segments @xmath255 , @xmath256 , and @xmath286 and denote these probabilities by @xmath287 , @xmath288 , and @xmath289 . each simulation is run until the 99% confidence intervals of the utilization probability estimates are less than 1% of the corresponding sample means . we consider a networks with @xmath290 wavelength channels in each ring direction . we examine the accuracy of the derived bounds and approximations by plotting the segment utilization probabilities as a function of the number of network nodes @xmath291 and comparing with the corresponding simulation results . [ cols="^,^,^ " , ] @xmath285 + + @xmath292 & 0.122 & 0.283 + @xmath293 & 0.126 & 0.302 + @xmath294 & 0.972 & @xmath295 + + @xmath292 & 0.0017 & 0.028 + @xmath293 & 0.025 & 0.073 + @xmath294 & 0.212 & 0.456 + we observe from fig . [ num5](a ) that for hotspot source traffic with large fanout , sp routing achieves significantly smaller maximum segment utilizations than oc routing for @xmath38 values up to a cross - over point , which lies between @xmath284 and @xmath285 . similarly , we observe from fig . [ num5](b ) that for small @xmath38 , sp routing achieves significantly smaller maximum segment utilizations than oc routing for hotspot destination traffic with small fanout . for example , for unicast hotspot destination traffic ( i.e. , @xmath296 ) , for @xmath297 , sp routing gives a multicast capacity of @xmath298 compared to @xmath299 with oc routing . by switching from sp routing to oc routing when the fraction of hotspot source traffic @xmath38 exceeds 0.31 , the smaller maximum utilization probability , i.e. , higher multicast capacity can be achieved across the range of fractions of hotspot source traffic @xmath38 . we have analytically characterized the segment utilization probabilities in a bi - directional wdm packet ring network with a single hotspot . we have considered arbitrary mixes of unicast , multicast , and broadcast traffic in combination with an arbitrary mix of uniform , hotspot destination , and hotspot source traffic . for shortest - path routing , we found that there are three segments that can attain the maximum utilization , which in turn limits the maximum achievable long - run average multicast packet throughput ( multicast capacity ) . through verifying simulations , we found that our bounds and approximations of the segment utilization probabilities , which are exact in the limit for many nodes in a network with a fixed number of wavelength channels , are fairly accurate for networks with on the order of ten nodes receiving on a wavelength . importantly , we observed from our segment utilization analysis that shortest - path routing does _ not _ maximize the achievable multicast packet throughput when there is a significant portion of multi- or broadcast traffic emanating from the hotspot , as arises with multimedia distribution , such as ip tv networks . we proposed a one - copy routing strategy with an achievable long run average multicast packet throughout of about two simultaneous packet transmissions for such distribution scenarios . this study focused on the maximum achievable multicast packet throughput , but did not consider packet delay . a thorough study of the packet delay in wdm ring networks with a hotspot transporting multicast traffic is an important direction for future research . in this appendix , we first define the enlarging and reducing of the set of " @xmath29-active nodes " @xmath302 . suppose that @xmath90 . depending on the setting , and with @xmath77 denoting the set of nodes homed on a given wavelength @xmath29 , the set @xmath158 is chosen uniformly at random among * all subsets of @xmath77 ( uniform traffic and for @xmath80 also hotspot destination and source traffic ) , or * all subsets of @xmath77 that contain @xmath6 ( hotspot destination traffic for @xmath227 since @xmath6 is always a destination for hotspot destination traffic ) , or * all subsets of @xmath77 that do not contain @xmath6 ( hotspot source traffic for @xmath227 since @xmath6 is always the source for hotspot source traffic ) . assuming @xmath303 , we define : enlarged@xmath41ring : : we enlarge the set @xmath77 by injecting an extra node homed on @xmath29 between @xmath304 and @xmath305 ( and correspondingly @xmath7 nodes homed on the other wavelengths ) . after a re - numeration starting with @xmath13 at the new node ( which is accordingly homed on wavelength @xmath4 after the re - numeration ) , we obtain @xmath306 . we define the enlarged set @xmath307 to equal the renumbered set @xmath158 united with the new node . this procedure leads to a random set of active nodes @xmath308 that is uniformly distributed among all subsets of @xmath309 with cardinality @xmath310 containing node @xmath13 . note that the largest gap of the enlarged set is larger or equal to the largest gap of @xmath95 . reduced@xmath41ring : : we transform the set @xmath77 by merging the nodes @xmath304 and @xmath305 to a single active node ( eliminating the @xmath311 nodes inbetween ) . after re - numeration starting with @xmath13 at this merged node , we obtain an active set @xmath312 on @xmath313 . + depending on the cardinality of @xmath314 the new active set @xmath312 has @xmath152 , @xmath69 , or @xmath315 elements . more specifically , if neither the left- nor the right - shifted source node was a destination node , then @xmath316 . if either the left- or the right - shifted source node was a destination node , then @xmath317 . if both the left- and right - shifted source node were destination nodes , then @xmath318 . in each of these cases @xmath312 is uniformly distributed among all subsets of @xmath319 with cardinality @xmath320 that contains node @xmath13 . + observe that in all cases , the largest gap of @xmath312 is smaller or equal to the largest gap of @xmath95 . we also define the following transformations : left ( counter clockwise ) shifting : : since @xmath32 is uniformly distributed on @xmath321 , the set @xmath322 is a random subset of @xmath77 . we can think of @xmath323 as being chosen uniformly at random among all subsets of @xmath77 having cardinality @xmath324 and subject to the same conditions as @xmath158 . + notice that @xmath325 if @xmath326 and @xmath327 otherwise . right ( clockwise ) shifting : : analogously we define @xmath328 this is a random set chosen uniformly at random among all subsets of @xmath77 having cardinality @xmath329 and subject to the same conditions as @xmath158 . conditioned on @xmath330 , we obtain @xmath331 hence , we only have to consider the case @xmath332 . we will not explicitly write down this condition . consider the right shifting and denote by @xmath333 the starting point of the chosen largest gap of @xmath334 . since @xmath335 is the only fixed active node , the first gap , i.e. , @xmath336 , is the only one that never shrinks , while the last gap , i.e. , @xmath337 , is the only one that never grows . therefore , @xmath338 for reasons of symmetry , we have @xmath339 and @xmath340 the remaining probabilities can be computed as @xmath341 , leading to the desired upper bound . analogously , the left shifting yields a lower bound , namely @xmath342 again for reasons of symmetry , we obtain @xmath343 and @xmath344 finally , we have , of course , @xmath345 and @xmath346 due to equation ( [ max lambda neq lambda ] ) , we only have to prove the case of drop wavelength @xmath4 . corollary [ cor critical non critical ] tell us that it suffices to consider the critical segments . let @xmath347 with @xmath348 be a critical segment for @xmath4 . analogously to the proof of the domination principle in @xcite , we reduce the domination principle for hotspot destination traffic to the statement @xmath349 and for hotspot source traffic to : @xmath350 in the @xmath38 ( hotspot source traffic ) setting , we know that @xmath6 is the sender , and thus @xmath351 hence , we do not need to consider the nodes on the other drop wavelengths and the proof is exactly the same as in the single wavelength case @xcite , see also figure c.2 . we will now use the same strategy for the more complicated proof in the @xmath37 ( hotspot destination traffic ) setting . let @xmath352 denote the number of active nodes finding themselves between the nodes @xmath6 and @xmath109 ( clockwise ) , i.e. , @xmath353 for @xmath354 we denote @xmath355 for the probability measure @xmath94 conditioned on @xmath356 we denote again @xmath347 for @xmath357 . we will show that @xmath358 in case that @xmath330 we can again use the proof of the one wavelength scenario . this is also true if @xmath359 , since we do not claim anything about these nodes . hence , we only have to investigate the case @xmath360 . from now on we assume this to be the case . we decompose the left hand side into two parts , @xmath361 for the first summand of ( [ two parts ] ) , we proceed similarly to the case of a single wavelength , namely @xmath362 we obtain @xmath363 this probability can be computed precisely @xmath364 we now use the fact that , conditionally on @xmath365 , @xmath366 and @xmath367 hence , we obtain with @xmath368 that @xmath369 for the second part of ( [ two parts ] ) , we obtain @xmath370 we have @xmath371 now , we use that @xmath372 for @xmath373 . hence , we obtain @xmath374 note that , conditioned on @xmath365 , we have @xmath375 and @xmath376 summarizing , we obtain , using @xmath377 , that @xmath378 it remains to show that @xmath379 this can be shown by @xmath380 for the last inequality , we used that for @xmath381 and @xmath382@xmath383 and , for reasons of symmetry , @xmath384 the last step we need is a comparison of ( [ nearly finish ] ) and ( [ looks good ] ) . the only difference arises , when both of the events , @xmath385 and @xmath386 , take place . then , @xmath387 this occurs with probability @xmath388 and explains the additional factor in the decomposition ( [ nearly finish ] ) . we are grateful to martin herzog , formerly of emt , inrs , and ravi seshachala of arizona state university for assistance with the numerical and simulation evaluations . s. spadaro , j. sol - pareta , d. careglio , k. wajda , and a. szymaski , `` positioning of the rpr standard in contemporary operator environments , '' _ ieee network _ , vol . 18 , no . 2 , pp . 3540 , march / april 2004 . j. fransson , m. johansson , m. roughan , l. andrew , and m. a. summerfield , `` design of a medium access control protocol for a wdma / tdma photonic ring network , '' in _ proc . , ieee globecom _ , vol . 1 , nov . 1998 , pp . 307312 . m. a. marsan , a. bianco , e. leonardi , m. meo , and f. neri , `` mac protocols and fairness control in wdm multirings with tunable transmitters and fixed receivers , '' _ ieee journal of lightwave technology _ , vol . 14 , no . 6 , pp . 12301244 , june 1996 . m. j. spencer and m. a. summerfield , `` wrap : a medium access control protocol for wavelength routed passive optical networks , '' _ ieee / osa journal of lightwave technology _ , vol . 18 , no . 12 , pp . 16571676 , dec . 2000 . i. m. white , e. s .- t . hsueh , k. v. shrikhande , m. s. rogge , and l. g. kazovsky , `` demonstration and system analysis of the hornet , '' _ ieee / osa journal of lightwave technology _ , vol . 21 , no . 11 , 24892498 , nov . 2003 . g. feng , c. k. siew , and t .- s . p. yum , `` architectural design and bandwidth demand analysis for multiparty videoconferencing on sonet / atm rings , '' _ ieee journal on selected areas in communications _ 20 , no . 8 , pp . 15801588 , oct . 2002 . n. singhal , l. sahasrabuddhe , and b. mukherjee , `` optimal multicasting of multiple light - trees of different bandwidth granularities in a wdm mesh network with sparse splitting capabilities , '' _ ieee / acm transactions on networking _ , vol . 14 , no . 5 , pp . 11041117 , oct . 2006 . s. sankaranarayanan and s. subramaniam , `` comprehensive performance modeling and analysis of multicasting in optical networks , '' _ ieee journal on selected areas in communications _ , 21 , no . 11 , 13991413 , nov . j. wang , x. qi , and b. chen , `` wavelength assignment for multicast in all - optical wdm networks with splitting constraints , '' _ ieee / acm transactions on networking _ , vol . 14 , no . 1 , 169182 , feb . 2006 . hsu , t .- l . liu , and n .- f . huang , `` multicast traffic scheduling in single - hop wdm networks with arbitrary tuning latencies , '' _ ieee transactions on communications _ , vol . 52 , no . 17471757 , oct . 2004 . k. naik , d. wei , d. krizanc , and s .- y . kuo , `` a reservation - based multicast protocol for wdm optical star networks , '' _ ieee journal on selected areas in communications _ , vol . 22 , no . 9 , pp . 16701680 , nov . 2004 . j. wang , w. cho , v. r. vemuri , and b. mukherjee , `` improved approaches for cost - effective traffic grooming in wdm ring networks : ilp formulations and single - hop and multihop connections , '' _ ieee / osa journal of lightwave technology _ , vol . 19 , no . 11 , pp . 16451653 , nov . chen and e. modiano , `` efficient routing and wavelength assignment for reconfigurable wdm ring networks with wavelength converters , '' _ ieee / acm transactions on networking _ , vol . 13 , no . 1 , 173186 , feb . 2005 . i. ferrel , a. mettler , e. miller , and r. libeskind - hadas , `` virtual topologies for multicasting with multiple originators in wdm networks , '' _ ieee / acm transactions on networking _ , vol . 14 , no . 1 , pp . 183190 , feb . 2006 . r. libeskind - hadas , j. r. hartline , k. dresner , e. drucker , and k. ray , `` multicast virtual topologies in wdm paths and rings with splitting loss , '' in _ proc . conf . on computer communications and networks _ , oct . 2002 , pp . 318321 . j. wang , b. chen , and r. uma , `` dynamic wavelength assignment for multicast in all - optical wdm networks to maximize the network capacity , '' _ ieee journal on selected areas in communications _ , 21 , no . 8 , 12741284 , oct . 2003 . a. carena , v. d. feo , j. m. finochietto , r. gaudino , f. neri , c. piglione , and p. poggiolini , `` ringo : an experimental wdm optical packet network for metro applications , '' _ ieee journal on selected areas in communications _ 22 , no . 8 , pp . 15611571 , oct . m. a. marsan , a. bianco , e. leonardi , m. meo , and f. neri , `` on the capacity of mac protocols for all optical wdm multi rings with tunable transmitters and fixed receivers , '' in _ proc . , ieee infocom _ , . 3 , mar . 1996 , pp . 12061216 . m. boroditsky , c. f. lam , s. l. woodward , n. j. frigo , and m. d. feuer , `` power management for enhanced system performance of passive optical rings , '' _ iee proceedings optoelectronics _ 150 , no . 3 , pp . 229234 , june 2003 . e. shimada , s. fujiwara , k. okazaki , and i. sasase , `` a transmission system for multicast traffic with preallocation scheme in wdm ring networks , '' in _ proc . ieee pacific rim conf . on comm . , computers and signal proc . 2003 , pp . 490493 . a. a. m. saleh and j. m. simmons , `` architectural principles of optical regional and metropolitan access networks , '' _ ieee / osa journal of lightwave technology _ , vol . 17 , no . 12 , pp . 24312448 , dec . 1999 . j. wang , w. cho , v. r. vemuri , and b. mukherjee , `` improved approaches for cost - effective traffic grooming in wdm ring networks : ilp formulations and single - hop and multihop connections , '' _ ieee / osa journal of lightwave technology _ , vol . 19 , no . 11 , pp . 16451653 , nov . 2001 . x. zhang and c. qiao , `` an effective and comprehensive approach for traffic grooming and wavelength assignment in sonet / wdm rings , '' _ ieee / acm transactions on networking _ , vol . 8 , no . 5 , pp . 608617 , oct . 2000 . a. zapata , m. duser , j. spencer , p. bayvel , i. demiguel , d. breuer , n. hanik , and a. gladisch , `` next - generation 100-gigabit metro ethernet ( 100 gbme ) using multiwavelength optical rings , '' _ ieee / osa journal of lightwave technology _ , vol . 22 , no . 11 , pp . 24202434 , nov . i. rubin and h. hua , `` synthesis and throughput behaviour of wdm meshed - ring networks under nonuniform traffic loading , '' _ ieee / osa journal of lightwave technology _ , vol . 15 , no . 8 , pp . 15131521 , m. chaitou , g. hebuterne , and h. castel , `` modelling multi - channel optical slotted rings with fixed transmitter and fixed receivers , '' in _ proc . , advanced industrial conference on telecommunications / service assurance with partial and intermittent resources conference/ e - learning on telecommunications workshop _ , july 2005 , pp . 115120 . m. an der heiden , m. sortais , m. scheutzow , m. reisslein , p. seeling , m. herzog , and m. maier , `` multicast capacity of optical packet ring for hotspot traffic , '' _ ieee / osa journal of lightwave technology _ , vol . 25 , no . 9 , 26382652 , sept . 2007 . a. bianco , j. m. finochietto , g. giarratana , f. neri , and c. piglione , `` measurement - based reconfiguration in optical ring metro networks , '' _ ieee / osa journal of lightwave technology _ , vol . 23 , no . 10 , 31563166 , oct . h. elbiaze and o. cherkaoui , `` exploiting self - similar traffic analysis in network resource control : the ip over wdm networks case , '' in _ proc . , ieee int . on autonomic and autonomous systems and int . conf . on networking and services ( icas - icns ) _ , oct . 2005 , pp . 6571 . a. elwalid , d. mitra , i. saniee , and i. widjaja , `` routing and protection in gmpls networks : from shortest paths to optimized designs , '' _ ieee / osa journal of lightwave technology _ , vol . 21 , no . 11 , 28282838 , nov . 2003 . e. oki , k. shiomoto , s. okamoto , w. imajuku , and n. yamanaka , `` heuristic multi - layer optimum topology design scheme based on traffic measurement for ip + photonic networks , '' in _ proc . , optical fiber communication conference and exhibit ( ofc ) _ , mar . 2002 , pp . 104105 .

according to a recent proposal [ s. takayama _ et al . _ , appl . phys . lett . * 87 * , 061107 ( 2005 ) ] , the triangular lattice of triangular air holes may allow to achieve a complete photonic band gap in two - dimensional photonic crystal slabs . in this work we present a systematic theoretical study of this photonic lattice in a high - index membrane , and a comparison with the conventional triangular lattice of circular holes , by means of the guided - mode expansion method whose detailed formulation is described here . photonic mode dispersion below and above the light line , gap maps , and intrinsic diffraction losses of quasi - guided modes are calculated for the periodic lattice as well as for line- and point - defects defined therein . the main results are summarized as follows : ( i ) the triangular lattice of triangular holes does indeed have a complete photonic band gap for the fundamental guided mode , but the useful region is generally limited by the presence of second - order waveguide modes ; ( ii ) the lattice may support the usual photonic band gap for even modes ( quasi - te polarization ) and several band gaps for odd modes ( quasi - tm polarization ) , which could be tuned in order to achieve doubly - resonant frequency conversion between an even mode at the fundamental frequency and an odd mode at the second - harmonic frequency ; ( iii ) diffraction losses of quasi - guided modes in the triangular lattices with circular and triangular holes , and in line - defect waveguides or point - defect cavities based on these geometries , are comparable . the results point to the interest of the triangular lattice of triangular holes for nonlinear optics , and show the usefulness of the guided - mode expansion method for calculating photonic band dispersion and diffraction losses , especially for higher - lying photonic modes .

0706.0395

photonic crystals ( phc ) embedded in planar waveguides , also called phc slabs , are at the heart of current research on photonic crystals@xcite because of the possibility of confining light in all spatial directions combined with the advantage of a lithographic definition of the pattern . the propagation of light in these systems can be controlled by a two - dimensional ( 2d ) photonic lattice in the waveguide plane , and by total internal reflection in the perpendicular direction . waveguides with strong refractive index contrast ( like the suspended membrane , or air bridge ) support truly guided modes lying below the light line dispersion in the cladding materials.@xcite however , most of the photonic modes ( or all of them , in the case of waveguides with weak refractive index contrast ) lie above the light line and are only quasi - guided , as they are subject to intrinsic losses due to out - of - plane diffraction.@xcite the physical properties of phc slabs can be significantly different from those of the corresponding 2d system , for several reasons . first , the 2d photonic modes are subject to confinement in the vertical waveguide and the resulting blue shift is strongly polarization - dependent . second , the presence of second- and higher - order waveguide modes can produce a complicated pattern of photonic bands , especially for higher - lying states . third , diffraction losses of quasi - guided modes are an inherent feature of phc slabs which is absent in the ideal 2d case . for all of these reasons , some basic and well - known properties of 2d photonic crystals can not be easily translated to phc slabs . for example , it is well known that the triangular lattice of circular air holes in 2d supports a complete photonic band gap for all propagation directions and light polarizations at sufficiently large air fractions.@xcite nevertheless , the same lattice realized in a high - index membrane does not possess a complete band gap , as the odd modes with respect to a horizontal mirror plane ( often called quasi - tm modes ) are subject to a strong and non - uniform blue shift which eliminates the gap.@xcite indeed , most applications of phc slabs employ the even modes ( often called quasi - te ) , which do possess a band gap for all propagation directions . it was recently suggested@xcite that the triangular lattice of triangular air holes in a high - index membrane gives rise to a complete photonic gap for both even and odd modes . the physical mechanism is the reduction of symmetry of the basis in the unit cell , as compared to the hexagonal symmetry of the 2d lattice , giving rise to a splitting between the first and the second odd bands at the k point of the brillouin zone : when realized in a high - index membrane , this gap overlaps the usual gap between the first and the second even bands , giving rise to a complete gap for all directions and polarizations . the experimental results reported in ref . support the existence of a complete photonic gap . a main purpose of this paper is to perform a systematic study of the triangular lattice of triangular holes , as compared to the triangular lattice of circular holes , both being realized in a high - index dielectric membrane . we calculate the photonic band dispersion , gap maps , and intrinsic losses of quasi - guided modes for the 2d lattice . we also treat line - defect waveguides obtained by removing a full row of holes , and point cavities consisting of three missing holes . in addition to a determination of photonic gaps for even and odd modes as a function of membrane thickness and air fraction , we compare diffraction losses for the lattices with conventional ( circular ) and reduced symmetry ( triangular ) holes : this comparison is important in order to assess the possible usefulness of the reduced - symmetry lattice . among the results , we find interesting prospects of the triangular lattice of triangular holes for nonlinear optics , as it may allow to achieve doubly - resonant second - harmonic generation ( shg ) with an even ( quasi - te ) fundamental wave and an odd ( quasi - tm ) harmonic wave when line - defect waveguides or photonic cavities are introduced . the calculations reported in this paper are performed with an approach which we name guided - mode expansion ( gme ) method . maxwell equations are treated by expanding the magnetic field into the basis of guided modes of an effective homogeneous waveguide , and by solving the resulting eigenvalue equation numerically . intrinsic losses of quasi - guided modes are obtained by calculating the coupling to leaky modes of the effective waveguide within perturbation theory ( i.e. , the photonic analog of fermi golden rule for quantum - mechanical problems ) . the gme method , although being an approximate one ( since the basis of guided modes is not a complete basis set ) , has been applied to a variety of photonic lattices and has proven to be useful especially for obtaining quasi - guided modes and their diffraction losses.@xcite it has also been successfully employed for the interpretation of optical experiments on phc slabs.@xcite another purpose of this paper is to provide a detailed description of the gme method , together with convergence tests and exemplifying applications . the rest of this work is organized as follows . in section ii we outline the gme method and discuss a few convergence tests . section iii contains the results for the 2d triangular lattice of triangular holes in a high - index membrane , namely photonic dispersion , gap maps and intrinsic losses . in sec . iv we present a few results for line - defect waveguides and point cavities in the triangular lattice of triangular holes . section v contains a discussion of the results and of prospective applications of the investigated lattice . technical details , which are needed by the reader in order to implement the gme method , are given in the appendices . as phc slabs are intermediate between 2d photonic crystals and dielectric slab waveguides , it is reasonable to describe photonic modes in these systems starting from slab waveguide modes and introducing the effect of a dielectric modulation in the core and cladding layers . this is the central idea of the guided - mode expansion method , in which phc slab modes are expanded in the basis of guided modes of an effective homogeneous waveguide and coupling to radiative modes is taken into account by perturbation theory . in this section we describe the formalism for calculating mode dispersion and intrinsic losses , for the general case of an asymmetric phc slab , and discuss convergence of the method . the system we are considering is shown in fig . [ fig : scheme]a . it consists of a phc slab made of three layers 1,2,3 , each of which is homogeneous in the vertical ( @xmath0 ) direction and is patterned with a 2d photonic lattice in the @xmath1 plane . the core layer has a thickness @xmath2 , while the lower and upper claddings are taken to be semi - infinite.@xcite patterning of each layer is characterized by the same 2d bravais lattice , while the basis in the unit cell can be different - a typical situation is that the core layer 2 is patterned , while layers 1 and 3 are not ( i.e. , they can be described by the same bravais lattice as layer 2 , but with a vanishing air fraction ) . writing @xmath3 , the dielectric constant @xmath4 is piecewise constant in the @xmath0 direction and takes the form @xmath5 in each layer @xmath6 . we assume a magnetic permeability @xmath7 . must have the same 2d bravais lattice for the three layers in ( a ) , but it can have different bases in the unit cell . the change in grey scales in going from ( a ) to ( b ) suggests that the effective waveguide is characterized by an average dielectric constant in each layer.,scaledwidth=60.0% ] as is well known , maxwell equations for the electric and magnetic fields * e * , * h * with harmonic time dependence can be transformed into the second - order equation for the magnetic field @xmath8 = \frac{\omega^2}{c^2 } { \bf h } , \label{eq : basic}\ ] ] with the condition @xmath9 . by expanding the magnetic field in an orthonormal set of basis states as @xmath10 the orthonormality condition being expressed by @xmath11 eq . ( [ eq : basic ] ) becomes equivalent to a linear eigenvalue problem @xmath12 where the matrix @xmath13 is given by @xmath14 this formulation of the electromagnetic problem bears strong analogies to the quantum - mechanical treatment of electrons , with the hermitian matrix @xmath13 playing the role of a quantum hamiltonian . once the magnetic field of a photonic mode is known , the electric field can be obtained from @xmath15 and the electric field eigenmodes are orthonormal according to @xmath16 as is well known@xcite and easily verified . in order to define an appropriate basis set @xmath17 for the expansion ( [ eq : expansion ] ) , we introduce an effective slab waveguide made of three homogeneous layers with dielectric constants @xmath18 , @xmath19 , @xmath20 , as illustrated in fig . [ fig : scheme]b . we take @xmath21 as the spatial average of @xmath22 in each layer : @xmath23 where the integral extends over a unit cell of area @xmath24 . this choice for the average dielectric constants is convenient , but by no means unique - numerical tests and other issues related to the choice of @xmath21 are discussed in subsec . [ sec : convergence ] . in any case , we assume that the average dielectric constants fulfill the inequalities @xmath25 in order for the effective slab to support a set of guided modes . their explicit form is well known@xcite and it will be summarized here in order to specify our notation , as well as for the paper to be self - contained . let us denote by @xmath26 the 2d wavevector in the @xmath1 plane with modulus @xmath27 and unit vector @xmath28 , by @xmath29 a unit vector perpendicular to both * g * and @xmath30 , and by @xmath31 the frequency of a guided mode which satisfies @xmath32 . moreover we define the following quantities : @xmath33 which represent the real ( imaginary ) parts of the wavevector in the core ( upper / lower cladding ) , respectively . the guided mode frequencies are found by solving the following implicit equations ( with the suffix @xmath27 being understood for simplicity ) : @xmath34 for transverse electric ( te ) polarization , and @xmath35 for transverse magnetic ( tm ) polarization . the guided modes at a given wavevector * g * are labelled by the index @xmath36 and the eigenfrequencies are denoted by @xmath37 . explicit forms for the electric and magnetic fields of the guided modes are given in appendix [ app : guided - modes ] , as they are needed for the calculation of the matrix elements ( [ eq : matrix - elements ] ) . the guided modes of the effective waveguide in fig . [ fig : scheme]b depend on a wavevector * g * which can take any value in the 2d plane . however , photonic modes in the phc slab of fig . [ fig : scheme]a have the form dictated by bloch - floquet theorem and they depend on a wavevector * k * which is usually restricted to the first brillouin zone of the 2d lattice : indeed , the effect of the dielectric modulation @xmath38 is to fold the guided modes of the effective waveguide to the first brillouin zone and to produce photonic bands and band gaps . we therefore write @xmath39 , where the bloch vector * k * lies in the first brillouin zone and * g * is a reciprocal lattice vector . as the basis states for the expansion ( [ eq : expansion ] ) of the magnetic field , we choose the guided modes @xmath40 of the effective waveguide given in appendix [ app : guided - modes ] . the guided - mode expansion of the magnetic field therefore reads@xcite @xmath41 the basis set is orthonormal according to eq . ( [ eq : orthonorm ] ) , but not complete , since the radiative modes of the effective waveguide are not included in the basis set . this approximation will be partially lifted in sec . [ sec : losses ] , where the effect of radiative modes in determining diffraction losses will be taken into account . with this choice for the basis set , the general index @xmath42 can be written as @xmath43 . the matrix elements @xmath13 of eq . ( [ eq : matrix - elements ] ) can be calculated in a straightforward way from the field profiles of the guided modes : the resulting analytic expressions are somewhat lengthy and are given in appendix [ app : guided - elements ] . they depend on the inverse dielectric matrices @xmath44 in the various layers . like in usual 2d plane wave calculations , a numerically convenient approach to calculate these matrix elements is to define the dielectric matrix @xmath45 and to find @xmath46 by numerical matrix inversion as @xmath47 : this procedure is known to have much better convergence properties as a function of the number of plane waves,@xcite since the truncation rules for fourier series in the presence of discontinuous functions are better represented.@xcite if the photonic lattices in layers 13 have a center of inversion ( the same for all layers ) , the dielectric matrices @xmath48 , @xmath46 are symmetric in @xmath49 and @xmath13 turns out to be a real , symmetric matrix when the phases of the fields are chosen as in appendix [ app : guided - modes ] . in the most general case of a photonic lattice without a center of inversion ( like the triangular lattice with triangular holes investigated in this work ) , the matrix @xmath13 is complex hermitian . line and point defects can be treated by introducing a supercell in one or two directions , respectively , like in 2d plane - wave calculations . the formalism presented here refers to the most general case of an asymmetric phc slab . if the structure is symmetric under reflection by a horizontal mirror plane @xmath50 [ i.e. , if the upper and lower claddings have the same dielectric pattern and are made of the same dielectric materials , @xmath51 , the basis states as well as the photonic eigenmodes can be classified as even or odd with respect to the mirror plane . we denote by @xmath52 the corresponding specular reflection operator and we shall refer to even ( odd ) states with respect to this operator as @xmath53 ( @xmath54 ) modes . the simplest way to exploit this mirror symmetry is to solve the eigenvalue problem ( [ eq : linear ] ) separately for @xmath53 and @xmath54 states by keeping only the appropriate solutions of eqs . ( [ eq : tepol]),([eq : tmpol ] ) , as detailed in appendix [ app : symmetry ] . there we also discuss the issue of polarization mixing in the framework of the gme method : in general , a photonic mode of a phc slab is a superposition of te - polarized and tm - polarized basis states and all six components of the electric and magnetic fields are usually non - vanishing . another useful symmetry property holds when the bloch vector * k * lies along special symmetry directions , for which mirror reflection @xmath55 with respect to a vertical plane ( @xmath56 ) is a symmetry operation of the phc slab : in this case the eigenmodes can be classified as even or odd with respect to this mirror symmetry , i.e. , they have @xmath57 or @xmath58 respectively . vertical mirror symmetry can be exploited by transforming the matrix @xmath13 with a unitary transformation which decouples the blocks corresponding to @xmath59 states . when a photonic mode in the phc slab falls above the cladding light line ( or light lines , if the waveguide is asymmetric ) , it is coupled to leaky modes of the slab by the dielectric modulation and it becomes quasi - guided , i.e. , it is subject to intrinsic losses due to scattering out of the plane . the losses can be represented by an imaginary part of the frequency @xmath60 , which is related to the q - factor of a resonance by @xmath61 . within the gme method , the imaginary part of frequency can be calculated by time - dependent perturbation theory , like in fermi golden rule for quantum mechanics . the imaginary part of the squared frequency of a phc mode with bloch vector * k * , whose frequency lies above the cladding light lines ( or at least above one of them ) , is given by @xmath62 where the matrix element between a guided and a leaky phc slab mode is given by @xmath63 and @xmath64 is the 1d photonic density of states ( dos ) at fixed in - plane wave vector for radiation states that are outgoing in medium @xmath6 : @xmath65 notice the sum over reciprocal lattice vectors and polarizations in eq . ( [ eq : fermi ] ) , as all diffraction processes contribute to @xmath66 . equations ( [ eq : fermi])([eq : dos ] ) generalize the expressions given in ref . to the case of an asymmetric phc slab and to situations in which processes with all @xmath67 contribute to diffraction losses . once @xmath66 is found , the imaginary part of frequency is easily obtained as @xmath68 . , @xmath69 have to be made.,scaledwidth=60.0% ] for a given wavevector * g * and polarization @xmath70 of the radiative modes , there are two scattering channels , corresponding to states with an outgoing component in the lower cladding ( medium 1 ) or in the upper cladding ( medium 3).@xcite the radiation modes are schematically shown in fig . [ fig : scatt - states ] . the state which is outgoing in medium 1 corresponds to a photonic dos with @xmath71 in eq . ( [ eq : dos ] ) , while the outgoing state in medium 3 corresponds to a dos with @xmath72 . radiation states are normalized according to eq . ( [ eq : orthonorm ] ) , and only the field components in the cladding regions are relevant in determining the normalization . @xcite since the field profile of a scattering state tends to a plane - wave form in the far field , the photonic density of states ( [ eq : dos ] ) is appropriate whatever the explicit field profile of radiative phc slab modes . in the present gme method , we evaluate the matrix element by approximating the radiation modes of the phc slab with those of the effective waveguide . this approximation is consistent with the treatment of the previous subsection , as the set of guided+radiation modes of the effective waveguide is orthonormal according to eq . ( [ eq : orthonorm ] ) , and will be shown later to yield accurate results for diffraction losses when compared to more exact methods . in appendix [ app : rad - modes ] we give the explicit form of radiation states of the effective waveguide which are outgoing in either the lower cladding ( fig . [ fig : scatt - states]a ) or in the upper cladding ( fig . [ fig : scatt - states]b ) , for both te and tm polarizations . using expansion ( [ eq : gme - expansion ] ) for the magnetic field of a phc slab mode , the matrix element ( [ eq : k - rad ] ) becomes @xmath73 where @xmath74)\cdot ( \mbox{\boldmath$\nabla$}\times\textbf{h}_{\textbf{k}+\textbf{g } ' , \lambda , j}^{\mathrm{rad}}({\bf r } ) ) \ , \mathrm{d}\textbf{r } \label{eq : guided - rad - gme}\ ] ] are matrix elements between guided and radiation modes of the effective waveguide , which can be calculated analytically . in appendix [ app : rad - elements ] we collect the matrix elements ( [ eq : guided - rad - gme ] ) , for all possible combinations of polarizations . those expressions are implemented together with eqs . ( [ eq : fermi])([eq : guided - rad - gme ] ) , in order to evaluate the imaginary part of frequency and hence the intrinsic diffraction losses of quasi - guided modes . the calculation of losses is computationally more demanding than the photonic dispersion alone , as the eigenvectors of the linear problem ( [ eq : linear ] ) are also needed . the formalism for intrinsic diffraction losses presented in this subsection and in the appendix refers to the general case of an asymmetric phc slab . if the slab is symmetric , and if the phc slab modes are found as @xmath75 states as outlined in appendix [ app : symmetry ] , it is possible to define and use symmetric or antisymmetric outgoing states instead of the scattering states illustrated in fig . [ fig : scatt - states ] . if this is done , @xmath53 ( @xmath54 ) quasi - guided modes are coupled only to symmetric ( antisymmetric ) radiative states . from a computational point of view there is no special advantage in using symmetry - adapted radiative states , as the computing time needed to implement the perturbative formula eq . ( [ eq : fermi ] ) is usually small . thus even in the case of a symmetric phc slab it is convenient to use the general loss formalism , and simply obtain @xmath53 or @xmath54 phc slab states by selecting the basis states of the effective waveguide , as explained in appendix [ app : symmetry ] . in this subsection we discuss the convergence properties of the gme as a function of numerical parameters and give a brief overview of previously published results and comparison with other approaches . truncation parameters in the gme method are the choice and number @xmath76 of plane waves and the number @xmath77 of guided modes of the effective waveguide kept in the expansion . the dimension of the eigenvalue problem ( [ eq : linear ] ) is @xmath78 , or about half of this value when vertical parity @xmath55 is used . the choice of plane waves can be made in analogy to usual 2d plane - wave expansion : the important point is how to calculate the inverse dielectric matrix @xmath46 properly , as already noticed in subsec . [ sec : dispersion ] . in the present implementation we take an isotropic cutoff for reciprocal lattice vectors and adopt the common procedure of inverting the dielectric matrix numerically:@xcite while this is adequate for many purposes , improved implementations of the plane - wave expansion @xcite could also be introduced in the gme method in order to speed up convergence . in the expansion , for @xmath53 modes in the triangular lattice of circular air holes defined in a high - index suspended membrane . the frequencies in ( a ) are calculated at the symmetry points @xmath79,k , m , while the imaginary parts in ( b ) are calculated at @xmath80 along the @xmath79-k direction . parameters are : @xmath81 , @xmath82 , @xmath83 . the effective core dielectric constant is fixed to @xmath84 , as given by eq . ( [ eq : average ] ) . , scaledwidth=38.0% ] the optimal @xmath77 depends on waveguide parameters and on the frequency region considered : typically , for phc slabs defined in a thin high - index membrane , @xmath85 in each parity sector ( i.e. , for both @xmath53 and @xmath54 modes ) yields very accurate results for the lower - lying modes . as an example , in fig . [ fig : conv_nal ] we show the real and imaginary parts of the frequency for a few photonic modes as a function of the number of guided modes @xmath77 , for @xmath53 modes in the triangular lattice of circular air holes . structure parameters are similar to those used later in the paper . the number of plane waves is truncated to @xmath86 , which is sufficient for convergence in the present lattice . it can be seen that both the mode energies and the imaginary parts are stable for @xmath87 , while a good approximation is obtained already for @xmath88 or even @xmath89 . this is very convenient for the purpose of design and search of parameters , as a photonic structure in a high - index contrast phc slab can be optimized within a fully 3d approach , with the same computing time of a conventional 2d plane - wave expansion . the precise calculation ( e.g. , with @xmath90 ) can be performed _ a posteriori _ once the structure parameters are optimized . modes in the triangular lattice of circular air holes defined in a high - index suspended membrane . the frequencies in ( a ) are calculated at the symmetry points @xmath79,k , m , while the imaginary parts in ( b ) are calculated at @xmath80 along the @xmath79-k direction . parameters are : @xmath81 , @xmath82 , @xmath83 . the number of guided modes in the expansion is fixed to @xmath85 . the vertical bar indicates the average dielectric constant , eq . ( [ eq : average ] ) , usually employed in the gme method ( in the present case , @xmath84 ) . , scaledwidth=38.0% ] it is also important to discuss the choice of the dielectric constant of the effective waveguide , which is usually taken to be the spatial average ( [ eq : average ] ) in each layer . while this choice is a natural one , it might be asked whether the results depend on the assumed value of @xmath21 . in fig . [ fig : conv_eps ] we show the real and imaginary parts of frequency for a few modes , with the same parameters of the previous figure , as a function of the effective dielectric constant in the core region . it can be seen that the results are rather insensitive to the assumed value of @xmath19 , except when it becomes smaller than about five . all curves are flat around the value @xmath91 which corresponds to the spatial average . this satisfactory behavior arises because the gme method employs the set of _ all _ guided modes of the effective waveguide ( although in practice they are truncated to a finite number for numerical convenience ) , thereby compensating for different choices of the effective waveguide used to define the basis set . thus the choice of the dielectric constant of the effective waveguide as the spatial average of @xmath92 is justified , at least for the low air fractions that are commonly employed . although one might consider extending the method by introducing an anisotropic dielectric tensor for each layer of the effective waveguide , fig . [ fig : conv_eps ] strongly suggests that the results of the gme method would not be changed by such a complicated extension which is therefore unnecessary . the approximations of the gme method are : ( i ) for photonic dispersion , the shift of guided and quasi - guided modes due to second - order coupling to leaky modes of the effective waveguide is neglected ; ( ii ) for diffraction losses , the density of radiative modes of the phc slab is approximated with that of the effective waveguide . the effect of these approximations can be judged by comparing the results with those of other methods that are known to be exact within numerical accuracy . a full analysis of this issue is outside the scope of this paper , however comparisons between the results of gme and those obtained by other methods have already been published and are shortly summarized here . the dispersion of quasi - guided modes was found to be in very good agreement with the resonance positions in reflectance or transmittance spectra found from a scattering - matrix treatment.@xcite in fact , the complex frequencies of photonic modes can be obtained from the poles of a scattering matrix.@xcite the intrinsic propagation losses of line - defect waveguides in the triangular lattice obtained with different methods ( gme,@xcite finite - difference time - domain or fdtd@xcite and from the poles of a scattering matrix@xcite ) were compared for the same waveguide parameters and were found to agree very well with each other : a detailed comparison is presented in ref . . a comparative study of gme and fourier modal methods applied to cavity modes in one - dimensional phc slabs has been reported in ref . , and very good agreement was found for the photonic dispersion as well as the intrinsic losses . finally , the q - factors of cavity modes in l1 , l2 and l3 nanocavities ( one , two or three missing holes in the triangular lattice ) as a function of nearby hole displacement are found to be almost identical when calculated with either the gme method@xcite or the poles of a scattering matrix.@xcite agreement with experiments performed on phc slabs made of silicon - on - insulator,@xcite si membranes,@xcite gaas / algaas,@xcite and silicon nitride@xcite is also very satisfactory . the set of these comparisons indicates that the gme is a reliable method for both photonic mode dispersion and intrinsic losses , whose main advantages are computational efficiency and ease of application to various kinds of vertical waveguides and of photonic lattices , both periodic in 2d and containing line- and point defects . and triangle side @xmath93 ; the unit cell in real space is indicated . the lattice is patterned in a high index suspended membrane of thickness @xmath2 . the brillouin zone with main symmetry directions in reciprocal space is also drawn.,scaledwidth=40.0% ] we report in this section a systematic study of the triangular lattice of triangular air holes patterned on a free - standing high - index slab of thickness @xmath2 , as schematically shown in fig . [ fig : scheme - tria ] . the relevant dimensions of this lattice are the triangular basis side , @xmath94 , and the slab thickness . throughout this section we assume ( unless otherwise specified ) that the lattice is patterned in a material with dielectric constant @xmath83 , which is appropriate for silicon at the usual telecom wavelength @xmath95 @xmath42 m but also to other materials with similar dielectric constants ( e.g. , gaas ) . for the calculations shown in the present section , we employed up to 109 plane waves and 4 guided modes per parity sector in the basis set of the gme method . all the calculations have been done on an ordinary pentium computer with low computational effort . the complex fourier transform of the dielectric constant for the triangular lattice of triangular holes is given in appendix [ app : four - trans ] . notice that the lattice is invariant under rotations by @xmath96 and is not centrosymmetric , but since @xmath97 by time - reversal symmetry , the same symmetry directions of the usual triangular lattice of circular air holes do apply . ) and ( b ) odd ( @xmath54 ) modes of the phc slab with a triangular lattice of triangular holes , parameters as in ref . : @xmath98 , @xmath99 , @xmath100 . light dispersion in air is represented by dotted lines.,scaledwidth=40.0% ] the first realization of such a lattice has been reported in ref . , with the purpose of measuring a complete photonic band gap in a waveguide - embedded 2d phc thanks to a reduction of symmetry for the basis in the unit cell of the triangular lattice , as already pointed out in the introduction . the importance of finding a complete photonic band gap in a waveguide - embedded photonic structure is related to the practical realization of linear waveguides or cavities with reduced losses or higher q - factors , as scattering into quasi - guided modes of opposite parity would be avoided . in fig . [ fig : takayama ] we present the photonic band dispersion calculated with the gme method for structure parameters as in ref . . even ( te - like ) and odd ( tm - like ) modes are reported in figs . [ fig : takayama]a and b , respectively . it is worth noticing that in the absence of second - order waveguide modes , the fundamental gaps of both parities would overlap in the dimensionless frequency range 0.31 - 0.32 , as it can be seen from the figure . however , the presence of a second - order mode with cut - off around @xmath101 for both polarizations , as shown in fig . [ fig : takayama ] , prevents the opening of a true photonic band gap . such second - order modes ( which do not appear in a 2d treatment with an effective index ) are unavoidable in the present structure , due to the relatively large slab thickness , @xmath102 . we believe that our result is not in contrast with the experimental findings reported in ref . , in which a strong reduction of the transmission intensity was measured along the @xmath79k and @xmath79 m directions . since the dispersion of the second - order waveguide mode is very close to the air light line in the relevant frequency region , such modes are barely coupled to the external beam focused on the sample side and also they are subject to large disorder - induced losses ; thus , it is likely that transmission into such modes is attenuated in a few lattice periods , which would explain the apparent photonic band gap observed experimentally . on the other hand , when line- or point - defects are introduced , such second - order modes are necessarily present within the gap , thus limiting the usefulness of this design for applications . in the following , we will consider a photonic band gap for the waveguide - embedded structure only when no higher - order modes are present within the apparent gaps formed by the fundamental modes , i.e. , when the phc slab is truly single - mode . with the aim of optimizing the design proposed by takayama _ _ , we performed a systematic study of the present photonic lattice at varying structure parameters . in fig . [ fig : gapmaps ] we show the photonic band dispersion and gap maps for the triangular lattice of triangular air holes with slab thickness @xmath103 ( figs . [ fig : gapmaps]a and d ) , @xmath82 ( figs . [ fig : gapmaps]b and e ) , and for the reference 2d system of infinite thickness ( figs . [ fig : gapmaps]c and f ) , respectively . as is well known for more conventional lattices , @xcite the blue shift of photonic modes with respect to the ideal 2d system due to the dielectric confinement is strongly polarization - dependent ; in particular , it is more pronounced for odd ( tm - like ) modes . this physical behavior is seen by comparing figs . [ fig : gapmaps]a - c , in which the dispersion of photonic modes is calculated for the same 2d photonic lattice in which triangular air holes have side @xmath104 . for @xmath103 the second - order cut - off is rather high in energy , thus the system can be considered as single - mode below the light line . on increasing the slab thickness , the second - order cut - off red shifts . for @xmath82 , both even and odd modes open their fundamental band gaps , but they do not overlap to give a complete photonic band gap for all symmetry directions . in order to span a wider range of structure parameters , important information can be inferred by the gap width as a function of the holes side . as it can be seen in figs . [ fig : gapmaps]d - f , the gap maps depend strongly on the waveguide thickness and differ substantially from those of the 2d system . for @xmath103 , the fundamental odd gap is closed for all values of @xmath94 , due to the larger blue shift of band edges at the k - point than at the m - point . even if both fundamental band gaps are present for @xmath82 , no complete gap opens due to the stronger blue shift of odd modes . for @xmath105 , the presence of higher - order modes prevents the opening of complete gaps as well . the gap map for the ideal 2d system presents a great variety of photonic gaps for both polarizations , and also some complete gaps at high frequencies . it is worth noting that the fundamental te and tm band gaps do not overlap in frequency , unlike in the ordinary triangular lattice of circular air holes.@xcite furthermore , the fundamental te gap turns out to be narrower than the usual gap for circular holes . even if no truly complete band gaps appear to be present for this waveguide - embedded lattice , it is interesting to investigate the formation of resonant gaps at @xmath106 and @xmath107 , to be exploited in non - linear optical applications like second - harmonic generation ( shg ) . while doubly - resonant shg has been widely studied in one - dimensional systems like multilayers and microcavities,@xcite no 2d photonic lattice giving such an interesting result has been reported in the literature . in particular , it can be seen from fig . [ fig : gapmaps]e that many odd gaps are present for a slab thickness @xmath82 . in this case , it is easy to find a condition for which the fundamental @xmath53 gap at pump frequency @xmath106 is resonant with a @xmath54 gap at @xmath107 . one of the possibilities is indeed shown in fig . [ fig : shg ] . it is substantially the same result reported in fig . [ fig : gapmaps]b , but here the odd modes ( fig . [ fig : shg]b ) are plotted in a doubled frequency range with respect to the even modes ( fig . [ fig : shg]a ) . as it can be seen , there is a good overlap in frequency between the large band gap of fig . [ fig : shg]a and the rather narrow but robust gap of fig . [ fig : shg]b . this could be interesting , e.g. , for applications to non - linear optical converters based on gaas , in which the pump s - polarized field would be converted into a p - polarized second - harmonic field.@xcite in this respect , the usefulness of the present lattice would be enhanced with _ ad hoc _ fabricated line- or point - defects . it should be reminded that these results have been obtained by using the same dielectric constant for both pump and second - harmonic modes , but it is easy to extend the present calculations to non - linear materials with optical constants of known frequency dispersion . one of the issues when dealing with the triangular lattice of triangular holes would be given by intrinsic losses . in order to show the general interest of the triangular hole - based lattice , and also to show an application of the gme method , we compare intrinsic losses of the present system to the ones of an ordinary circular hole - based phc slab . in this case , we restrict our discussion to @xmath53 modes along the @xmath79k symmetry direction . intrinsic losses are quantified by calculating the imaginary part of frequency , as detailed in the previous section . in fig . [ fig : bulk - loss]a and c we show the photonic dispersion and intrinsic losses for the first few bands of an ordinary phc slab with a triangular lattice of circular holes . it can be noticed that the modes along the chosen symmetry direction are separated according to their mirror symmetry with respect to the vertical plane , which can be either even ( @xmath108 ) or odd ( @xmath109).@xcite on the right hand side , in fig . [ fig : bulk - loss]b and d , we plot dispersion and losses for the case of triangular holes . in this case no symmetry operation @xmath55 holds , so the modes are mixed . the choice of @xmath81 for the circular holes and @xmath104 for triangular ones , respectively , allows to obtain roughly the same air fraction @xmath110 for the two lattices . this can be also verified by the roughly equal second - order cut - offs . the dispersion is only slightly modified when changing the hole shape from circular to triangular , and in particular it appears that a general blue shift occurs for the modes in fig . [ fig : bulk - loss]b with respect to the ones in [ fig : bulk - loss]a . for what concerns intrinsic losses , we can see that they remain almost of the same order on going from circular to triangular holes , at least for the lower frequency bands . this means that ( in the absence of disorder ) no particular differences are introduced by changing the basis of the lattice , apart from the loss of symmetry @xmath55 . as the band labelled with 1 is almost unchanged on going from the circular to the triangular hole shape , also the intrinsic losses are quite similar between the two lattices . for higher lying bands , in particular for the band labelled with 2 , such similarity becomes less evident on approaching the light line , owing to the blue shift of modes for the triangular hole - based lattice and to the mixing with other modes which does not occur in the conventional lattice . in particular , owing to the lack of @xmath55 symmetry , there is an anticrossing between mode 2 and the second - order waveguide mode occurring very close to the light line , which correspondingly affects the imaginary part of frequency ( see figs . [ fig : bulk - loss]b - d ) . the mode labelled with 3 has larger losses for triangular holes than for circular ones , but it is quite high in energy . modes at pump and ( b ) @xmath54 modes at second - harmonic frequencies , respectively , showing resonant band gaps at @xmath106 and @xmath107 useful for non - linear applications . parameters are as in fig . [ fig : gapmaps]b : @xmath104 , @xmath82 , @xmath111 . light dispersion in air is represented by dotted lines.,scaledwidth=40.0% ] as a conclusion to this section , we have shown that the triangular lattice of triangular air holes does indeed have a complete photonic band gap between the fundamental waveguide modes , but the gap is actually eliminated by the presence of second - order modes . the lattice can be designed for nonlinear optical applications in order to have a gap for even modes at the pump frequency which is resonant with a gap for odd modes at the sh frequency . intrinsic losses are of the same order as in the circular hole - based triangular lattice . the gme method is useful for a comparison of the two lattices and for the design of structures with desired characteristics . we show in the present section that line- and point - defects can be introduced into the triangular - hole lattice without any loss of performance with respect to the circular - hole one . it is well known that removing a row of holes in a 2d photonic lattice introduces a linear defect supporting guided modes within the photonic band gap . in a triangular lattice such waveguide is usually called w1 , and its channel width ( namely the distance between the holes surrounding the waveguide ) is @xmath112 . these guided modes exploit the usual index confinement along the vertical direction added to the gap confinement in the waveguide plane , and they are studied with great interest by the scientific community owing to prospective applications in photonic integrated circuits . modes in the triangular lattice of circular air holes with @xmath81 , @xmath82 , @xmath83 , along the @xmath79-k symmetry direction ; modes are classified according to their @xmath113 parity . ( b ) dispersion and ( d ) intrinsic losses for the same lattice of triangular air holes with @xmath104 , @xmath82 , @xmath83 . light dispersion in air is given by dotted lines.,scaledwidth=40.0% ] modes in w1 waveguides in triangular lattice of circular air holes with @xmath81 , @xmath82 , @xmath83 ; modes are classified according to their @xmath113 parity . ( b ) dispersion and ( d ) intrinsic losses for the same waveguide in the triangular lattice of triangular air holes with @xmath104 , @xmath82 , @xmath83 . light dispersion in air is given by dotted lines . the shaded regions represent the bulk modes of the respective lattices projected onto the line - defect brillouin zone.,scaledwidth=40.0% ] in fig . [ fig : w1-loss ] we report a comparison between the w1 waveguide realized in the triangular lattice with circular ( figs . [ fig : w1-loss]a - c ) and triangular holes ( figs . [ fig : w1-loss]b - d ) , respectively . for the circular hole - lattice the guided modes are classified according to their mirror symmetry with respect to the vertical plane @xmath114 . for such calculations , we used supercells along the @xmath79 m direction of widths ranging from @xmath115 to @xmath116 , with 181 to 323 plane waves and 4 guided modes in the basis set for the expansion . then , an average was taken to get final results for losses , in order to smooth out finite supercell effects.@xcite as it is seen from the figure , the dispersion of the guided modes is very similar between the two systems , apart in the region of mixing . in fact , as @xmath55 is a symmetry operation for the system in fig . [ fig : w1-loss]a , the two guided modes are completely decoupled , while in fig . [ fig : w1-loss]b they mix producing a consistent anti - crossing and a mini - gap . looking at the imaginary part of frequency , we see by comparing figs . [ fig : w1-loss]c and d that they are almost identical away from the mixing region , which means that w1 waveguides in triangular lattice of triangular holes may behave as standard w1 waveguides.@xcite such result is quite surprising , as the asymmetry of such a waveguide could be expected to lead to higher scattering of propagating light states . on the contrary , the reduction of symmetry does not increase losses in the _ perfect _ photonic lattice - of course , the role of fabrication imperfections may affect differently the two structures in _ real _ lattices . the final goal of this paper is to exploit the potential of the triangular lattice of triangular holes for non - linear optical applications . in this respect , the ultimate system for frequency conversion at the nanoscale is a photonic cavity with double resonance for pump and harmonic waves . doubly - resonant microcavities for second - harmonic generation have been studied in the context of fabry - perot cavities.@xcite according to a recent proposal , double resonance in photonic crystal cavities may allow to achieve strong coupling between single photons.@xcite it has been shown that full 3d confinement of light with a very high figure of merit ( quality factor , q ) can be achieved by properly designed point - defects in 2d phc slabs.@xcite in particular , removing three holes along the @xmath79k direction ( l3 cavity ) and optimizing the positions of the holes surrounding the cavity allows to achieve q - factors of the order of @xmath117.@xcite we compare here the l3 cavity structure made in a triangular lattice of triangular air holes to the usual l3 cavity with circular holes and the results are shown in fig . [ fig : cavity ] . for such calculations , a square supercell of dimensions @xmath118 has been taken . up to 1551 plane waves and two guided modes were used in the gme . as it can be seen from fig . [ fig : cavity]a , many dispersionless @xmath53 defect modes appear in the photonic band gap , due to the tight confinement provided by the cavity . we can evaluate the vertical q - factor of such modes ( i.e. , the one determined by intrinsic out - of - plane losses ) by the definition @xmath119 $ ] , where both @xmath106 and @xmath60 have been averaged over the folded brillouin zone in order to reduce the effects of a finite supercell size . very good agreement with measured q - factors and defect mode energies has been already found with this gme - based procedure.@xcite the q - factor of the un - optimized l3 cavity in the lattice of circular holes is theoretically estimated to be @xmath120 . when the same cavity is realized in the triangular lattice of triangular holes ( fig . [ fig : cavity]b ) , we see that the fundamental mode is almost unchanged , and its q - factor is comparable to the one of fig . [ fig : cavity]a . the slight blue shift of the defect mode and of the lower band edge in fig . [ fig : cavity]b with respect to fig . [ fig : cavity]a is due to the slightly different air fractions of the two lattices . as we have already shown in fig . [ fig : shg ] , we can design a triangular lattice with triangular holes having a doubly - resonant band gap . thus , we plot in fig . [ fig : cavity]c the dispersion of @xmath54 modes within the doubled frequency range with respect to fig . [ fig : cavity]b , in analogy with fig . [ fig : shg ] . many defect modes appear within the @xmath107 band gap , and the lowest mode has a q - factor of almost @xmath121 , which is a very promising value considering that the structure has not been optimized in any way . there is , in fact , much room for improving the q - factors at both @xmath106 and @xmath107 by means of geometry optimization of the nearby holes . the present result make us confident that high - q , doubly - resonant nanocavities may be realized in phc slabs made of non - linear materials , bringing new exciting results in integrated optics at nanoscale level . the triangular lattice of triangular holes in a high - index dielectric slab has been thoroughly investigated by means of the guided - mode expansion method , whose detailed formulation and examples of applications have been given in the paper . the investigated lattice does have a complete band gap for both polarizations when the fundamental waveguide mode is considered , consistently with the results of ref . , but the gap is actually eliminated by the presence of second - order waveguide modes . the intrinsic losses of quasi - guided modes in the triangular lattice with triangular and circular holes , and in line - defect waveguides as well as nanocavities defined therein , are comparable . an interesting application of the triangular lattice of triangular holes for nonlinear optics follows from the fact that a gap for @xmath53 ( te - like ) modes at the pump frequency can be designed to be in resonance with a gap for @xmath54 ( tm - like ) modes at the harmonic frequency : thus , resonant second - harmonic generation exploiting band - edge effects may be realized . promising results are also found for point cavities : the l3 cavity in the triangular lattice of triangular holes supports a @xmath53 cavity mode at @xmath106 which can be resonant with a @xmath54 cavity mode at @xmath107 . there is much room for improving the design of doubly - resonant , high - q nanocavities by geometry optimization . these results are interesting in view of application of the triangular lattice of triangular holes for nanoscale nonlinear optical processes . the authors are grateful to m. agio , h. benisty , d. cassagne , ph . lalanne , m. le vassor dyerville , m. liscidini , and c. sauvan for useful conversations and suggestions . they also would like to acknowledge k. hennessy and w.t.m . irvine for providing them with the preprint of ref . prior to its publication , thereby motivating the search for doubly resonant phc cavities . this work was supported by miur through cofin project `` silicon - based photonic crystals for the control of light propagation and emission '' and firb project `` miniaturized electron and photon systems '' , as well as by cnr - infm through pra photonic . the frequencies of the guided modes of the effective waveguide are found from the implicit equations ( [ eq : tepol ] ) for te polarization , and ( [ eq : tmpol ] ) for tm polarization . the guided modes are labelled by the wavevector * g * and the mode index @xmath122,@xcite which can be combined into a single index @xmath123 . the mode profiles for te polarization can be written as follows ( with the time dependence @xmath124 being understood ) , with reference to the geometry of fig . [ fig : scheme ] : @xmath125 @xmath126 in the above expressions , @xmath127 is a normalization surface , which disappears from the final results , while the magnetic field is obtained from the electric field thorugh the maxwell equation @xmath128 . application of standard transfer - matrix theory leads to the following relations between coefficients ( with the suffix @xmath42 being understood for simplicity ) : @xmath129\nonumber\\ \label{eq : tecoeff3}\end{aligned}\ ] ] and to the well known implicit equation ( [ eq : tepol ] ) of the main text . the normalization integral ( [ eq : orthonorm ] ) implies the relation @xmath130=1\end{aligned}\ ] ] which , together with eqs . ( [ eq : tecoeff1])([eq : tecoeff3 ] ) , determines all coefficients @xmath131 and @xmath132 . the guided modes for tm polarization are characterized by the following field profiles : @xmath133 @xmath134 , & |z|<\frac{d}{2 } \\ \frac{1}{\bar{\epsilon}_1}d_{1\mu}(-\chi_{1\mu}\hat{g}+ig\hat{z } ) e^ { \chi_{1\mu}(z+d/2 ) } , & z<-\frac{d}{2}\\ \end{array } \right.\ ] ] where the electric field follows from the magnetic field profile through eq . ( [ eq : efield ] ) . the appropriate relations between coefficients can be derived from ( [ eq : tecoeff1])-([eq : tecoeff3 ] ) by the replacements @xmath135 , @xmath136 and @xmath137 , @xmath138 ( except in the trigonometric functions ) , @xmath139 . the normalization integral of the magnetic field gives the condition @xmath140=1 \ , .\ ] ] we calculate the matrix elements @xmath13 of eq . ( [ eq : matrix - elements ] ) between guided modes of the effective waveguide from the field profiles given in appendix [ app : guided - modes ] for te and tm polarizations . we adopt the following notations : @xmath141 and @xmath142 , where @xmath143 is the bloch - floquet vector , @xmath144 and @xmath145 are reciprocal lattice vectors and @xmath122 , @xmath146 are the indices of guided modes of the effective waveguide at wavevectors @xmath147 , @xmath148 respectively . there are four kinds of matrix elements : te - te , tm - tm , te - tm and tm - te . the @xmath0 integral in eq . ( [ eq : matrix - elements ] ) can be broken into three terms over the regions 1,2,3 where the dielectric constant does not depend on @xmath0 , and can be expressed in terms of the following integrals : @xmath149 the @xmath1 integrals in eq . ( [ eq : matrix - elements ] ) yield the inverse dielectric matrices in the three layers , @xmath46 , defined in eq . ( [ eq : eta ] ) . the matrix elements are then obtained as follows : @xmath150 \right\}\,,\nonumber\\\end{aligned}\ ] ] @xmath151,\end{aligned}\ ] ] @xmath152 \right\},\end{aligned}\ ] ] @xmath153\right\}.\end{aligned}\ ] ] if the phc slab is invariant under reflection through a mirror plane bisecting the slab ( which we take as @xmath50 ) , the effective waveguide has the same reflection symmetry and its eigenmodes can be classified as even or odd with respect to the reflection operator @xmath52 defined in the main text . the spatial symmetries of the nonvanishing field components are given in table [ tab : fields ] , assuming a wavevector along the @xmath154 direction . the guided mode frequencies are found from the following implicit equations : @xmath155 notice that the first - order te and tm modes of the effective waveguide are described by eqs . ( [ eq : te - even ] ) and ( [ eq : tm - odd ] ) , respectively . by keeping only @xmath53 ( @xmath54 ) basis states in the guided - mode expansion , photonic eigenmodes which are even ( odd ) with respect to horizontal mirror symmetry are obtained , and the computational effort in solving the eigenvalue equation ( [ eq : linear ] ) is reduced . .spatial symmetry of the electric and magnetic field components of an eigenmode of a symmetric waveguide with respect to the mirror symmetry operation @xmath52 . the wavevector is assumed to lie along the @xmath154 direction . vanishing components are denoted by `` * '' . [ cols="^,^,^,^,^,^,^",options="header " , ] [ tab : fields ] in general , a photonic eigenmode in a phc slab is a linear combination of te- and tm - polarized basis states and all six components of the electric and magnetic field are nonvanishing . nevertheless , low - lying photonic modes are often dominated by the lowest - order guided mode of the effective waveguide . in this situation , @xmath53 states are dominated by the te guided modes described by eq . ( [ eq : te - even ] ) and can be called `` quasi - te '' , while @xmath54 states are dominated by the tm guided modes described by eq . ( [ eq : tm - odd ] ) and can be called `` quasi - tm '' . this widespread terminology is useful and appropriate for large wavevectors , when the dominant field components are the spatially even ones : making reference to table [ tab : fields ] , and since the dielectric modulation @xmath156 couples guided modes with all wavevector directions in the 2d plane , the dominant components are @xmath157 ( for @xmath53 or te - like modes ) or @xmath158 ( for @xmath54 or tm - like modes ) . however , the terminology becomes inadequate when the phc slab is multimode , as it always happens at sufficiently high frequencies . if the waveguide is not symmetric , any phc slab mode is a linear combination of basis states arising from all equations ( [ eq : te - even])([eq : tm - odd ] ) and even the lowest - order te / tm guided modes of the effective waveguide are mixed . in this appendix we give the radiation modes of the effective waveguide which are outgoing in either the lower or the upper cladding ( see fig . [ fig : scatt - states ] ) . each of these radiation modes is labelled by the wavevector * g * in the 2d plane , the frequency @xmath106 , and the polarization . in the following the quantum numbers @xmath159 will be understood in order to simplify the notations . instead of distinguishing between real and imaginary components of the wavevectors in the three regions as in eq . ( [ eq : chi - and - q ] ) , we define three complex wavevectors as @xmath160 while the wavevector @xmath161 in the core region is always real , the wavevectors @xmath162 in the cladding regions can be either real or purely imaginary according to the values of the dielectric constants and of @xmath159 . if the effective waveguide is symmetric ( @xmath163 ) all radiation modes have real @xmath164 , while if @xmath165 there is a frequency region @xmath166 in which @xmath167 is real but @xmath168 is imaginary : i.e. , a quasi - guided mode in this region of @xmath169-@xmath106 space is diffracted to the lower ( but not to the upper ) cladding . this phenomenon is automatically taken into account by the kronecker @xmath170-function in the dos formula ( [ eq : dos ] ) , as the photonic dos vanishes when @xmath171 is imaginary . therefore we shall write down the electric and magnetic field components of radiative waveguide modes assuming that all @xmath172 s are real - if one of them is purely imaginary , that radiation mode does not carry an energy flux and it does not contribute to scattering loss . with this proviso , the radiation modes of the effective waveguide for te polarization in the three regions @xmath173 can be written as follows ( setting @xmath174 , @xmath175 , @xmath176 and with the time dependence @xmath177 being understood ) : @xmath178,\ ] ] @xmath179\,.\end{aligned}\ ] ] for the scattering state which is outgoing in the lower cladding ( fig . [ fig : scatt - states]a ) , @xmath180 and the normalization condition ( [ eq : orthonorm ] ) determines the coefficient of the outgoing component@xcite as @xmath181 . all other coefficients are then found from a standard transfer - matrix calculation , which yields @xmath182= \frac{1}{2q_2}\left [ \begin{array}{cc } ( q_2+q_1)e^{iq_2d/2 } & ( q_2-q_1)e^{iq_2d/2}\\ ( q_2-q_1)e^{-iq_2d/2 } & ( q_2+q_1)e^{-iq_2d/2}\\ \end{array } \right ] \left [ \begin{array}{cc } w_1 \\ x_1 \\ \end{array } \right ] \label{eq : transfer - matrix-1}\ ] ] and @xmath183= \frac{1}{2q_3}\left [ \begin{array}{cc } ( q_3+q_2)e^{iq_2d/2 } & ( q_3-q_2)e^{-iq_2d/2}\\ ( q_3-q_2)e^{iq_2d/2 } & ( q_3+q_2)e^{-iq_2d/2}\\ \end{array } \right ] \left [ \begin{array}{cc } w_2 \\ x_2 \\ \end{array } \right ] . \label{eq : transfer - matrix-2}\ ] ] on the other hand , for the state outgoing in the upper cladding ( fig . [ fig : scatt - states]b ) , @xmath184 and the proper normalization condition@xcite is @xmath185 . the other coefficients can then be found from the inverse relations of ( [ eq : transfer - matrix-1 ] ) , ( [ eq : transfer - matrix-2 ] ) . for what concerns tm - polarized radiation modes , the field profiles are given by @xmath186,\ ] ] @xmath187.\end{aligned}\ ] ] for the scattering state outgoing in the lower cladding ( fig . [ fig : scatt - states]a ) , @xmath188 and the normalization condition ( [ eq : orthonorm ] ) determines the coefficient of the outgoing component as @xmath189 . for the state outgoing in the upper cladding ( fig . [ fig : scatt - states]b ) , @xmath190 and the proper normalization condition is @xmath191 . all other coefficients can be found from transfer - matrix theory , the relevant expressions being obtained from ( [ eq : transfer - matrix-1 ] ) , ( [ eq : transfer - matrix-2 ] ) by the replacements @xmath192 , @xmath193 and @xmath194 ( except in the exponential functions ) . we calculate the matrix elements ( [ eq : guided - rad - gme ] ) between guided and radiation modes of the effective waveguide , to be used in formula ( [ eq : k - rad - gme ] ) for the loss calculation . the expressions are similar to those of appendix [ app : guided - elements ] , but care must be taken to distinguish between the different quantum numbers . guided modes are labelled by the wavevector @xmath195 and by the index @xmath122 , which are combined as before into a single index @xmath42 . radiation modes are labelled by the wavevector @xmath196 , the frequency @xmath106 , the polarization @xmath70 , and an additional index which specifies whether the mode is outgoing in medium 1 or 3 : all these quantum numbers will be grouped into a single index @xmath197 . like for the matrix elements between guided modes , there are four possible combinations of polarizations . the @xmath0 integral can be broken into three terms over the regions 1,2,3 and it can be expressed in terms of the following integrals : @xmath198 the matrix elements between guided and radiative modes are found as follows : @xmath199 \right\}\ , \nonumber\\\end{aligned}\ ] ] @xmath200\nonumber\\ & & + { \eta}_3(\textbf{g},\textbf{g } ' ) c_{3\mu}^ * [ ( gg'-i\chi_{3\mu}q_{3r}\hat{g}\cdot\hat{g}')y_{3r}i_{3-}\nonumber\\ & & + ( gg'+i\chi_{3\mu}q_{3r}\hat{g}\cdot\hat{g}')z_{3r}i_{3+}]\nonumber\\ & & + { \eta}_2(\textbf{g},\textbf{g } ' ) [ ( c_{2\mu}^*y_{2r}+d_{2\mu}^*z_{2r})\nonumber\\ & & \times(gg'+q_{\mu}q_{2r}\hat{g}\cdot\hat{g } ' ) i_{2-}\nonumber\\ & & + ( c_{2\mu}^*z_{2r}+d_{2\mu}^*y_{2r } ) ( gg'-q_{\mu}q_{2r}\hat{g}\cdot\hat{g}')i_{2+}]\ , , \end{aligned}\ ] ] @xmath201 \right\ } \,,\end{aligned}\ ] ] @xmath202 \right\ } \ , .\end{aligned}\ ] ] we give here the complex fourier transform of the dielectric constant in the unit cell for the triangular lattice of triangular holes , to be used for the calculations shown in secs . [ sec : trianglatt ] and [ sec : defects ] . let us refer to the geometry shown in fig . [ fig : scheme - tria ] , the side of the triangle and the lattice constant being denoted by @xmath93 and @xmath203 , respectively . the triangle is made of a medium of dielectric constant @xmath204 ( in the case of an air hole , @xmath205 ) embedded in a background with dielectric constant @xmath206 . the origin of coordinates in the unit cell is taken to be at the center of the triangle . the filling factor of the triangle is with @xmath214 , @xmath215 . in this work , the triangles are oriented in such a way that one of the triangle sides is along the @xmath216-axis as in fig . [ fig : scheme - tria ] ( i.e. , along the @xmath79-k direction of the triangular lattice ) : if other orientations have to be considered , the simplest way is to apply a 2d rotation matrix to the vector * g * before calculating the fourier transform with the above formulas . d. gerace , m. galli , d. bajoni , g. guizzetti , l.c . andreani , f. riboli , m. melchiorri , n. daldosso , l. pavesi , g. pucker , s. cabrini , l. businaro , and e. di fabrizio , appl . phys . lett . * 87 * , 211116 ( 2005 ) . in order to derive the formulas of this section , it is in fact necessary to consider a finite thickness @xmath228 of the cladding layers and to take the limit @xmath229 of the resulting expressions . these details will be omitted for simplicity . notice that the index @xmath122 takes into account both the order and the polarization of the guided mode . for example , if two te and two tm guided modes are kept in the expansion , there are four distinct values of @xmath122 . this can be shown formally by introducing a normalization box of width @xmath228 in the @xmath0 direction . when @xmath231 , the contributions of the cladding regions to the normalization integral extend over a very large thickness and the core region becomes irrelevant . on the other hand , the contributions of core and claddings to the matrix element ( [ eq : guided - rad - gme ] ) are of the same order in @xmath232 ( they are both proportional to @xmath233 ) because the integrals in the cladding regions are exponentially damped by the field of the guided modes . since the fictitious width @xmath228 disappears from the final result for @xmath60 , it has been set equal to unity in eq . ( [ eq : dos ] ) and in appendix [ app : rad - modes ] for the sake of simplicity . notice that in materials with the zincblende structure , the symmetry of the nonlinear susceptibility tensor allows for shg with a te - polarized pump and a tm - polarized harmonic in a ( 001)-oriented slab .

the cosmic microwave background ( cmb ) radiation lensing is a promising tool to study the physics of early universe . in this work we probe the imprints of deviations from isotropy and scale invariance of primordial curvature perturbation power spectrum on cmb lensing potential and convergence . specifically , we consider a scale - dependent hemispherical asymmetry in primordial power spectrum . we show that the cmb lensing potential and convergence and also the cross - correlation of the cmb lensing and late time galaxy convergence can probe the amplitude and the scale dependence of the dipole modulation . as another example , we consider a primordial power spectrum with local feature . we show that the cmb lensing and the cross - correlation of the cmb lensing and galaxy lensing can probe the amplitude and the shape of the local feature . we show that the cross correlation of cmb lensing convergence and galaxy lensing is capable to probe the effects of local features in power spectrum on smaller scales than the cmb lensing . finally we showed that the current data can constrain the amplitude and moment dependence of dipole asymmetry .

1511.05534

the standard model of cosmology known as @xmath0cdm is established as a cornerstone of modern cosmology . the cosmic microwave background ( cmb ) radiation @xcite and large scale structure ( lss ) surveys @xcite indicate that almost all observations can be described by a 6-parameter @xmath0cdm model with an initial condition of almost gaussian , isotropic , scale invariant and adiabatic perturbations which can be sourced during inflation @xcite . recent observations of planck collaboration shows almost no deviation from standard picture of single field slow roll inflation @xcite . on the other hand the bicep2/keck - planck results shows that there is no detection of gravitational waves @xcite . despite the success of the standard model , the physics of early universe , the natures of dark matter and dark energy are still unknown . the large scale structure surveys open a new horizon to test the cosmological models in different redshifts and sub - cmb scales . one of the recent developments of the field is the detection of the lensed cmb map with unprecedented accuracy @xcite . the cmb lensing is caused by the gravitational effect of structures in the line of sight of the photons from the last scattering surface @xcite . each ray of cmb is deflected due to gravitational lensing effect and accordingly we observe the temperature of each point on the lensed cmb map , @xmath1 , as a map of the temperature in a nearby point @xmath2 ( note that in this work a quantity accompanied with the symbol @xmath3 represents the corresponding lensed quantity ) . the lensing potential encapsulates all information on the deflection of light bundle due to lensing and it is related to the integrated potential of the matter distribution in all redshift up to the surface of last scattering . the calculation of the two point statistics in lensed cmb maps can be a probe of the matter distribution ( matter power spectrum ) in the universe . accordingly , the cmb lensing is a unique observational tool to constrain the cosmological models , specially the evolution of the growth of the structures @xcite . the cmb lensing can not only change the temperature map of cmb but also it can change the polarization maps by being a new source of b - mode polarization @xcite . in addition , it introduces non - gaussianities @xcite . therefore , in order to study the primordial universe , we should subtract the effects of cmb lensing on cmb map correctly in order to avoid the degeneracies @xcite . having said that , the cmb lensing in its own can be used as a unique observational tool to constrain the physics of early universe @xcite . it can probe the curvature perturbations power spectrum and capture any deviation from standard assumptions . in this work we use the cmb lensing as a probe of early universe physics , specially to study the effects of deviation from scale invariant and isotropic initial conditions . this is motivated from the fact that there are indications of anomalies in cmb data ( like power deficit , cold spot , alignment of quadrupole and octupole @xcite ) . in addition , the planck data indicates the existence of hemispherical asymmetry in primordial power spectrum @xcite , first seen in wmap data @xcite , and also verified with the local variance estimator @xcite . if these anomalies are not the statistical flukes , then it is worth to investigate their effects on cmb- and lss - related observations in order to understand their origins . we propose that the cmb lensing and also cmb lensing cross - correlation with cosmic shear can be used as a unique probe for detection of these anomalies and their scale dependence . the structure of the work is as follows . in sec . [ sec : theory ] we review the theoretical background , which is needed for cmb lensing , cosmic shear and their cross - correlation . in sec . [ sec : deviation ] we study the effects of deviation from isotropy , specifically the dipole modulation in power spectrum , and the deviation from simple scale - invariance via a localized feature in power spectrum . the conclusions and discussions are presented in sec . [ sec - conclusion ] . also in app.([app1 ] ) and app . ( [ app2 ] ) we probe the parameter space of dipole modulation and gaussian feature . in this work we use the cosmological parameters of the base @xmath4 model from planck 2015 @xcite results : @xmath5 , @xmath6 , @xmath7 , @xmath8 , @xmath9 and @xmath10 . in this section we review the theoretical background for the mechanism of lensing . in first subsection we focus on cmb lensing . the cross - correlation of cmb lensing with cosmic shear is the subject of the study in the second sub - section . the geometry of the universe is described by the perturbed frw metric in `` conformal newtonian '' gauge ds^2=a^2 ( ) , where @xmath11 is the conformal time , @xmath12 is the comoving distance and @xmath13 , @xmath14 are the bardeen potentials , the scalar degrees of freedoms for perturbations . we assume that @xmath15 , which is the case if the cosmic fluid does not have anisotropic stress and general relativity ( gr ) is the correct classical theory of gravity . the light source from a distant objects in universe is distracted by the intervening structures before the line of sight . a crucial parameter that shows the amount of convergence / deconvergence of light bundles is the lensing potential @xmath16 defined as @xcite : ( ) = -2_0^_*d ( ) ( , ) , where the integration parameter @xmath12 is the comoving distance , @xmath17 is the comoving distance to the source and @xmath18 represents the direction of the observed cmb light . by the knowledge of the lensing potential , we can calculate the deflection angle as @xmath19 . the corresponding lensed temperature @xmath20 is related to the un - lensed temperature @xmath21 via the effect of deflection as : @xmath22 . note that @xmath23 is the angular gradient , consequently the gravitational potential , the lensing potential and the deflection angle are dimensionless quantities . now we can relate the two point function of the lensing potential to the power spectrum of gravitational potential in fourier space as [ eq : lenscor ] ( ) ^*()=4_0^_*d_0^_*d ( ) ( ) e^i**k.**e^-i**k. * * ( * k*,())^*(*k*,( ) ) , where @xmath24 is the fourier transform of gravitational potential . we can relate the gravitational potential @xmath13 to the primordial curvature perturbation @xmath25 via transfer function @xmath26 as : [ eq : transfer ] ( * k*,)=(*k*,)(*k * ) . correspondingly , the lensing potential will be related to the primordial curvature perturbation @xmath27 as follows ( ) ^*()=4_0^_*d_0^_*d()()e^i**k.(-)**(k,())(k,())p_(k ) [ 116 ] in which the primordial power spectrum @xmath27 is defined via [ eq : power ] ( * k*)^*(*k* ) = ( 2 ) ^3p_(k)^3(*k-*k ) . * * using the definition of the flat wave and the orthogonality of spherical harmonics , the cross - correlation of lensing potential becomes @xmath28 where @xmath29 is the dimensionless curvature power spectrum with its amplitude and scale dependence being constrained by the cmb data @xcite . the lensing potential can be expanded in terms of spherical harmonics as well @xmath30 for a statistically isotropic field , the angular power spectrum is defined as @xmath31 where @xmath32 is the angular potential of the lensing . now considering the fact that the integral over @xmath12 and @xmath33 is symmetric and interchangeable from eq . ( [ 120 ] ) we obtain c^_l=16_(k ) ( _ 0^_*d()(k,)j_l(k ) ) ^2 [ eq : clpsi ] . eq.([eq : clpsi ] ) shows that the angular power spectrum of cmb lensing depends on the primordial power spectrum @xmath34 . we emphasis that in obtaining eq.([eq : clpsi ] ) we have assumed that the appropriate transfer function @xmath26 used in this equation is not modified compared to the base @xmath0cdm universe . this is justified because we assume that anomalies of our interest , such as dipole asymmetry or local feature , happen during inflation and universe follows its traditional history after inflation . accordingly , at least to leading order , all the effects of the primordial universe are imprinted in primordial power and they can be investigated in cmb lensing . in cmb lensing studies we can interchangeably express the lensing power in terms of the late time matter power spectrum and to test the predictions of the models for matter distribution of late time universe . in this light , cmb lensing is a probe of early universe through eq.([eq : clpsi ] ) and also the distribution of the matter in late time . in below we discuss the theoretical background of the relation between the cmb lensing potential power and the matter power spectrum . defining the gravitational potential power spectrum @xmath35 via [ eq : powerpot ] ( * k*,)^*(*k*, ) = ( 2)^3p_(k,,)^3(*k*-*k* ) . the angular power spectrum of lensing becomes c_l^=k^2dk_0^_*d_0^_*dp_(k;,)j_l(k)j_l(k ) ( ) ( ) . now using the poisson equation we can relate the power of gravitational potential to the matter power as follows p_(k,)= p_(k , z ) , where @xmath36 is the density parameter of matter in present time , @xmath37 is the matter power spectrum which can be expressed in terms of growth function @xmath38 and the matter transfer function @xmath39 via p_m(k , z ) = a t^2(k ) d^2(z ) k^n_s , where @xmath40 is the spectral index . in this work we use the transfer function of eisenstein and hu @xcite for transfer function which captures the baryon acoustic oscillations . in the high moment approximation @xmath41 , we can use the limber formula @xmath42 to study the lensing potential in small angular scales , yielding c_l^ 9_0^(1+z)^2()^2p_m(k , z ) . now using the fact that in small angle limit the wavenumber becomes @xmath43 and also by exchanging the variable of integral from comoving distance to redshift results in l^4c_l^=_0^z_*dzw^(z)p_m(,z ) , where @xmath44 is the kernel of lensing potential obtained as [ eq : wphi ] w^(z)=9 ^ 2 ( ) ^2_m^2(1+z)^2 , in which @xmath45 . we can also find the angular power spectrum of convergence @xmath46 , which is defined as follows ^2 _ ( ) . accordingly , the angular power spectrum @xmath47 is obtained to be [ eq : powerkcmb ] c_l^_c_c=_0^z_*(1+z)^2h_0 ^ 3_m^2 ^ 2 ( ) ^2p_m(k , z ) . worth to mention that observationally both the lensing and convergence power spectrum are used to constrain the cosmological models . in the next sections we study the deviation from isotropic and scale invariant initial conditions for primordial power spectrum . -1.5 cm versus redshift are plotted where @xmath48 represent each of the kernels . the black solid line represents the cmb lensing potential kernel ( @xmath49 ) introduced in eq.([eq : wphi ] ) . the red long dashed line shows the cmb lensing convergence kernel ( @xmath50 ) introduce in eq.([eq : wkcmb ] ) . the blue dashed line shows the galaxy lensing convergence kernel ( @xmath51 ) introduced in eq.([eq : wkg ] ) . for @xmath52 we used the galaxy distribution from hand et al . @xcite.,title="fig : " ] on cosmological scales besides the cmb lensing there is another lensing effect known as the cosmic shear . this effect is the distortion of the light of the source galaxies in weak gravitational lensing limit by the surrounding matter distribution @xcite . this effect has been seen in different weak lensing surveys @xcite . in order to study the weak lensing effect of structures in late time , we define the convergence as the weighted integral of matter density contrast in line of sight as follows ( ) = _ 0^dz w(z)((z),z ) . here @xmath53 is the weighting function which is related to the cosmological model and to the distribution of the sources via [ eq : wkg ] w^_g(z)=_m h_0 ^ 2(z)_z^dz , where @xmath54 represents the redshift of the source and @xmath55 is the distribution of sources in redshift space . in the case of the cmb lensing , the source is localized at @xmath56 and we obtain eq . ( [ eq : powerkcmb ] ) with the kernel of cmb lensing @xmath57 defined as below [ eq : wkcmb ] w^_c=_m h_0 ^ 2(z ) . for the cosmic shear , we need to know the distribution of the source galaxies . beside the cmb lensing and cosmic shear angular power spectrum , we can plot the cross correlation of the cmb - cosmic shear as : c_l^_c_g=_0^dzw^_c(z)w^_g(z)p(k , z ) . in fig . ( [ kernels ] ) , we plot the kernels of the cmb lensing potential ( @xmath49 ) introduced in eq . ( [ eq : wphi ] ) , cmb lensing convergence ( @xmath50 ) introduced in eq . ( [ eq : wkcmb ] ) and the galaxy lensing convergence ( @xmath51 ) introduced in eq.([eq : wkg ] ) as a function of the redshift . an important point to indicate is that the kernels in the standard @xmath0cdm model are just functions of the redshift . the figure shows the effectiveness of each redshift bin on the overall signal . also one must be cautious that the galaxy lensing kernel is very sensitive to the selection function @xmath58 , which shows the number density of the observed galaxies in each redshift bin . this distribution is very well affected by the characteristics of the galaxy survey . for @xmath52 we used the galaxy distribution from hand et al . @xcite . in the next section we will discuss the deviation from standard primordial conditions and study its effect on cmb and galaxy lensing . as mentioned before the primordial fluctuations are nearly gaussian , nearly scale invariant and isotropic . in this section we propose a number of phenomenological models for primordial power spectrum , which deviate from these standard assumptions . our goal is to study the imprints of these deviations on the cmb lensing . in grishchuk and zeldovich @xcite there is a theoretical motivation to study the effect of the dipole asymmetry in primordial universe and later its implications on observable universe such as on cmb . there were indications of dipole asymmetry in wmap data @xcite . the subsequent planck observations also seem to support the existence of dipole asymmetry @xcite , also see @xcite for follow up works on cmb analysis in the presence of dipole asymmetry . this has caused significant interests on theoretical and observational implications of hemispherical asymmetry . the simplest realization of hemispherical asymmetry is in the form of dipole modulation t()= ( ) , in which @xmath59 is the isotropic temperature fluctuation , @xmath60 is the amplitude of dipole , @xmath18 is the direction of the observation and @xmath61 is the preferred direction in sky . the planck team found @xmath62 in galactic coordinate with the amplitude @xmath63 . in addition , the planck data indicates a non - trivial scale dependence for dipole asymmetry , i.e. @xmath64 , such that the amplitude of dipole falls off rapidly for angular scale @xmath65 . there is no convincing theoretical understanding of hemispherical asymmetry . one promising approach is the idea of long mode modulations @xcite . in this picture a long mode @xmath66 , which is much bigger than the hubble radius , causes the asymmetry by modulating the background inflationary parameters such as the inflaton field or its velocity or by modulating the surface of end of inflation . however , it is not easy for this idea to work in simple models of inflation such as in single field models . indeed it is shown in @xcite that the amplitude of dipole modulation is related to the amplitude of local - type non - gaussianity @xmath67 . in single field models of inflation with small ( actually zero @xmath67 ) there is no chance to generate dipole asymmetry with large enough amplitude . therefore , one has to look for beyond simple slow roll models of inflation employing ideas such as multiple fields models like curvaton scenarios , iso - curvature perturbations , domain walls , non - vacuum inflationary initial conditions etc . for a list of various theoretical works on these directions see @xcite . if the power asymmetry is not a statistical fluke it indicates towards a non - trivial early universe physics . now using the fact that @xmath68 the primordial power spectrum is obtained to be [ dipole ] _ = ^iso(k ) . in this section we would like to study the effect of this dipole modulation on cmb lensing . however , before that we extend the above dipole modulation to more general modulation considering azimuthally symmetric in which @xmath69,\end{aligned}\ ] ] where @xmath70 is the angle between the patch in the sky and the preference direction and @xmath71 is the coefficient which determines the amplitude of anisotropy and in general can be a function of wavenumber . the specific case of @xmath72 gives the dipole modulation . in order to obtain the angular power spectrum of cmb lensing potential , we use eqs . ( [ 120 ] ) and ( [ 123 ] ) : @xmath73 j_l(k\chi ) j_{l'}(k\chi ' ) y^*_{lm}(\widehat{n } ) y_{l'm'}(\widehat{n'})\delta_{ll ' } \delta_{mm ' } \label{148}\end{aligned}\ ] ] the anisotropic angular power spectrum @xmath74 is defined as below @xmath75 the isotropic part is the same as eq.([130 ] ) . for the anisotropic contribution , we use an approximation to simplify the equation . suppose we perform the statistics in a patch in the sky which has the angle @xmath70 relative to the preference direction . we assume the patch is large enough to produce large enough statistics and small enough to be considered isotropic in the sky . this approximation is valid as long as the variation in asymmetry power spectrum across the sky is small . considering the anisotropic part of eq . ( [ 148 ] ) we obtain @xmath76 using statistical isotropy in each patch @xmath77 . now by expanding @xmath78 in spherical harmonics space with respect to @xmath70 @xmath79 and rewriting eq . ( [ 154 ] ) in terms of the new coefficient of expansion , we obtain @xmath80 for different patches the coefficients @xmath81 are different , so @xmath81 is a function of @xmath70 . as it is obvious from eq . ( [ cllm ] ) , the asymmetry in lensing power spectrum is induced from the asymmetric nature of the primordial curvature perturbation power spectrum in eq . ( [ pr - general ] ) . having presented the general form of lensing power spectrum for arbitrary shape of asymmetry , in the next sub - section we study the special case of dipole modulation . in the special case of dipole asymmetry in the primordial power spectrum we have @xmath82 = \mathcal { p_r}^{iso}(k ) \big[1 + 2 a(k ) \sqrt{\frac{\pi}{3 } } y_{10}(\theta ) \big].\end{aligned}\ ] ] note that we have allowed for the scale dependence of dipole amplitude @xmath83 . now considering the fact that the dipole modulation is a specific sample of general anisotropy in real space , from eq . ( [ cllm ] ) we obtain @xmath84 where @xmath85 is replaced by @xmath83 . one important issue to consider is the scale dependence of dipole asymmetry . as mentioned before the cmb data shows that the dipole is effective only on large scales ( small moments ) . in order to quantify this scale dependence we consider the amplitude of the dipole modulation @xmath83 to have the following scaling form [ eq : adnd ] a(k)= a_d ( ) ^n_d-1 , where @xmath60 is a constant , @xmath86 represents the scales at which the anisotropy is damped and @xmath87 is the spectral index of dipole power spectrum . in this work we set the value of the pivot wavenumber of dipole modulation to @xmath88 . this value is translated to the moment of @xmath89 , where we assume is the mean value of moment which cmb dipole modulation dies off . -1.5 cm , @xmath90 and @xmath91 . the data points are from planck lensing map reconstruction.,title="fig : " ] one can compare @xmath92 in different directions in order to probe the amplitude of dipole modulation . it is also possible to test the scale dependence of @xmath93 . taking into account the scale dependent anisotropy in camb code one can estimate the lensing potential power spectrum for different patches in the sky and check the consistency of model with observations . we consider the observable quantity introduced below in order to check the prediction of anisotropic model on cmb lensing potential [ cl - ns ] c_l^(n - s)/|c_l= where @xmath94 and @xmath95 are the powers in northern and southern hemispheres of the sky defined with respect to the asymmetric direction of dipole and @xmath96 is the isotropic power spectrum . note that @xmath97 is an observable quantity . by comparing between the northern and southern hemispheres of cmb sky , one can obtain this quantity from observation . in appendix [ app1 ] we investigate that how the @xmath98 will depend on the parameters of dipole modulation , like dipole amplitude , the spectral index of dipole modulation and also the effect of the observing patch . + on the other hand , theoretically , we can obtain @xmath97 from the camb code and we can examine the consistency of the model . in fig . ( [ fig : clphi - di - data - norm ] ) we plot the full sky angular power spectrum of cmb lensing potential for the @xmath0cdm case and the cases with the maximum power modulation ( northern hemisphere ) and minimum power modulation ( southern hemisphere ) with @xmath99 , @xmath90 and @xmath91 . fig . ( [ fig : clphi - di - data - norm ] ) shows that the maximum deviation from standard case is in low multipole ( @xmath100 ) , however the analysis of large patches to detect the dipole modulation is difficult due to limitation from cosmic variance . beside the cmb lensing potential , we can investigate the effect of the dipole modulation on the cmb lensing convergence as well . for this purpose , and similar to @xmath97 defined in eq . ( [ cl - ns ] ) , we define @xmath101 as the difference of the convergence angular power spectrum between the northern and southern hemispheres . in appendix [ app1 ] , we discuss the dependence of @xmath101 on the spectral index of the dipole modulation . -1.5 cm : for @xmath102 ( black solid curve ) , @xmath90 ( red long dashed curve ) and @xmath103 ( blue dashed curve ) with @xmath99 and @xmath91 for all curves.,title="fig : " ] in fig . ( [ fig : kcmbkcmb - dipole - nd ] ) , we present the effects of @xmath87 on @xmath101 , it is obvious that the behavior of the cmb convergence power spectrum is very similar to the lensing power , accordingly the effect of change in amplitude and the dependence of the power to @xmath70 is very similar to figs . ( [ clpsiamp ] ) and ( [ clphidegree ] ) . however the cmb convergence can be used as a complementary observation for lensing potential power to probe modulation and its scale dependence . -1.5 cm , @xmath90 and @xmath91 . the data points are from act project @xcite.,title="fig : " ] in fig . ( [ fig : kcmbkcmb - dipole - data - log ] ) we plot @xmath47 for the northern and southern hemispheres with @xmath90 , @xmath104 and @xmath99 . we compare the results with the data from atacama cosmology telescope ( act ) @xcite . the important point is that the most significant deviation occurs in low multipoles while the convergence data from act can span the high multipoles . this is because the telescope has a field of view of @xmath105 which scan the cmb in three frequencies in a seasonal way . however , the small field of view of the telescope will help to map the cmb lensing in different patches with different angles with respect to the dipole direction . accordingly , our proposal for detecting the signal of dipole modulation is to study the cmb lensing convergence in each patch of observational sky separately without combining all the data . -1.5 cm , @xmath90 and @xmath91 . the data points are from act project @xcite.,title="fig : " ] in the search for dipole asymmetry in late time lensing data , we can also use the cosmic shear observation and use the cross - correlation of the convergence of galaxy lensing with cmb lensing . the cross - correlation of the cmb lensing and galaxy lensing permits us to probe the signal of modulation for larger angular moments as well . for this purpose , and similar to the case of lensing potential @xmath97 , we define @xmath106 as a measure of the difference of the cross angular power spectrum of cmb lensing convergence and galaxy lensing convergence between two hemispheres . the physics of cross - correlation and its relation to the dipole asymmetry parameters is studied in app . [ app1 ] . in fig . ( [ fig : kcmbkgal - dipole - data - norm ] ) we compare the signal of cross - correlation of cmb lensing and galaxy lensing @xmath107 . the curve descriptions are the same as in fig . ( [ fig : kcmbkcmb - dipole - data - log ] ) . the data points are from the joint analysis of the act data for cmb lensing and cfhtlens data for galaxy lensing @xcite . an important point to indicate is that the cross angular power spectrum of cmb lensing convergence and galaxy lensing convergence depends on the kernel of the galaxies , accordingly each galaxy sample can change the moment dependence of the signal due to its specific kernel . this is important because future lss surveys will provide us with new catalogs of galaxy lensing in different redshifts and accordingly different kernels . as a final word in this subsection we use the act convergence data and cross correlation of cmb lensing and cosmic shear data @xcite to put constrains on the parameters of dipole modulation . we use only the act data for cmb lensing as this project maps a patch of sky and not the whole cmb sky . the planck sky takes the average of all angles with respect to the dipole direction . the act data is taken from a strip - survey of galactic coordinate of right ascension @xmath108 and declination @xmath109 @xcite . we convert this area in galactic coordinate and by approximation we assume that the whole patch ( center of the patch as a representative ) has a cosine angle of @xmath110 with dipole direction @xmath111 . the lensing amplitude is fixed with the data for each sample and then we put the constrain in @xmath60 and @xmath87 defined in eq.([eq : adnd ] ) . in fig . ( [ fig : rev - counter ] ) we plot the @xmath112 and @xmath113 confidence levels . -1.5 cm and @xmath113 confidence levels of the amplitude of dipole @xmath60 and the spectral index @xmath87 are plotted using the act convergence data and act - cfthlens cross correlation of cmb convergence , galaxy convergence . , title="fig : " ] the figure shows that the data are consistent with null detection of dipole asymmetry . however we should keep in mind that the data points are very sparse with large error - bars . future cmb lensing - cosmic shear data can tighten the constraint on the dipole parameters . also fig.([fig : rev - counter ] ) shows that recent data put an upper threshold on the amplitude of dipole modulation with a specific spectral @xmath87 . one way to check the deviation from the standard initial conditions for primordial perturbations is to study the behavior of primordial power in different wavenumbers . scale invariance is a key property of primordial fluctuations , which is observed from the scales of cmb down to the scale of group of galaxies at @xmath114 . however , there is no direct test of scale invariance or its violation on smaller scales . in this sub - section we study the effect of a strong violation of scale invariance in primordial power spectrum on small scales @xmath115 which is now out of the access of large scale structure surveys . as a toy model , we consider a model of primordial power spectrum with localized feature at a specific wavenumber @xmath116 ( the subscript @xmath117 stands for feature ) in the following form @xmath118 \ , , \ ] ] where @xmath119 is the standard power spectrum of @xmath0cdm model with isotropic , scale invariance and gaussian initial conditions . in these parameterizations @xmath120 measures the amplitude of the local feature at the wavenumber @xmath116 . our goal is to see how this local feature affect the observation of the cmb - lensing . for a relevant study concerning the effects of local feature on the abundance of large scale structures see @xcite . here let us pause to provide the theoretical motivation for generating local feature and its naturalness in inflationary dynamics . in its simplest realization , inflation is driven by a single field slowly rolling over a flat potential . in this simple picture , the primordial power spectrum is very nearly scale - invariant in all observable scales . however , there is no fundamental reason that inflationary dynamics should follow this simple picture . indeed , it is quite reasonable that there are many fields during inflation which can affect the dynamics of inflaton field sometime during its evolution . in particular , if some of the fields trigger instabilities in field space , then this instability may imprint itself as a local violation of slow roll conditions on inflaton dynamics . this local violation of slow roll conditions can be captured effectively as a local glitch in power spectrum as presented in eq . ( [ eq : powerbump ] ) . for a realization of these kinds of inflationary scenarios see @xcite . of course the local feature represented by eq . ( [ eq : powerbump ] ) is oversimplified and a real primordial feature may have a more non - trivial shape than this highly localized shape . having that said , we use eq . ( [ eq : powerbump ] ) as a toy model which can help us to perform the analysis analytically . as we shall see , in our numerical analysis and plots we extend this artificial feature into a more realistic feature in which the bump has a gaussian width . inserting the primordial power spectrum given in eq . ( [ eq : powerbump ] ) in the angular power spectrum of lensing potential defined in eq.([eq : clpsi ] ) , we can calculate the modified power spectrum of cmb - lensing potential . keeping in mind that the power spectrum does not depend on the direction of wavenumber @xmath121 we can separate the integral over the comoving part and the wavenumber part as follows @xmath122 now we can calculate the modification induced from the local feature @xmath123 . this modification is a function of @xmath124 , @xmath116 and also the standard model parameters defined as below c^(f)_l(a_f , k_f)c_l^-|c_l^= ( k_f ) _ 0^ _ * d ( ) t(k_f ; ) j_l(k_f ) ^2 , where @xmath125 is the angular power spectrum of lensing potential in the standard case , i.e. in the absence of local feature . -1.5 cm cdm ( blue dashed line ) and for a gaussian feature with @xmath126 , @xmath127 and @xmath128 . the data point is taken from lensing reconstruction of planck data.,title="fig : " ] -1.5 cm for @xmath129 ( solid black curve ) , @xmath127 ( red long dashed curve ) and @xmath130 ( blue dashed curve ) with @xmath131 and @xmath128 in all curves . , title="fig : " ] -1.5 cm cdm ( blue dashed curve ) and a gaussian feature with @xmath132 , @xmath127 and @xmath128 ( black solid curve ) . the data point is taken from act data . , title="fig : " ] as mentioned before , the local feature presented in eq . ( [ eq : powerbump ] ) is over - simplified mainly to allow us to perform the above integrals analytically . however , in order to study more realistic cases , we can change the dirac delta - function in eq . ( [ eq : powerbump ] ) by a gaussian function defined as [ eq : gauss ] _ = e^-(k - k_f)^2/2 ^ 2_f , where @xmath133 is the width of the gaussian function of the feature and @xmath116 is the wavenumber where the feature is centered . we use the above realistic feature in our numerical plots . in fig . ( [ clphi - data - log ] ) we have plotted the angular power spectrum of cmb lensing potential in log scale . the data points are from the planck collaboration . the prediction of @xmath0cdm model with standard initial conditions and the gaussian feature with @xmath128 , @xmath126 and @xmath127 are jointly plotted . generally a feature with positive amplitude increases the amplitude of power . the dependence of the cmb lensing potential to the parameters of power spectrum with a gaussian feature is discussed and plotted in detail in appendix ( [ app2 ] ) . now we can also investigate the effect of the feature on the angular power spectrum of lensing convergence . in fig . ( [ clkckc - f - kf ] ) the fractional change in angular power spectrum of lensing convergence , @xmath134 , is plotted for different values of characteristic wavenumber @xmath116 . the dependence of this fractional change in convergence power spectrum to the amplitude and the width of the feature is the same as in the cmb lensing potential . the important point to mention here is that the general behavior of the lensing convergence signal is the same as lensing potential . in fig . ( [ clkckc - data - log ] ) we plot the cmb lensing convergence power spectrum @xmath135 for the gaussian feature with @xmath127 , @xmath136 and @xmath137 . the convergence data points are from act collaboration@xcite . as we discussed in the dipole modulation case study , the cross correlation of the cmb lensing and galaxy lensing improves the observational opportunity to test the deviations from the isotropic and scale invariant power spectrum for smaller scales ( higher moments ) . accordingly we study the effect of feature on the cross correlation power @xmath138 . -1.5 cm cdm ( blue dashed curve ) and for the case of gaussian feature with @xmath136 , @xmath127 and @xmath128 ( black solid curve ) . the data point is taken from act collaboration @xcite , title="fig : " ] in fig . ( [ clkckg - data - norm ] ) , we plot the angular power spectrum of the cross correlation of cmb lensing and galaxy lensing for both the standard case and the feature model with @xmath136 , @xmath127 and @xmath128 . the data points are taken from the act data . a very important point to indicate is that the convergence data has a higher resolution and it spans the moments up to @xmath139 . accordingly , the lensing convergence is a more prominent observation to test the deviation from scale invariance by the excess in amplitude on smaller scales . cosmological observations strongly support the standard model of cosmology with the initial conditions which are nearly isotropic , gaussian , adiabatic and scale invariant . however it is crucial to test the properties of initial conditions with different observations and on different scales . in this work we have proposed that the cmb lensing can be a prominent new observational tool to investigate the physics of initial conditions . the cmb lensing can be a unique way to test these deviations . this is because the map of last scattering photons is distorted by the structures between the observer and cmb while these late time structures carry rich information about the physics of early universe encoded in their statistical properties . in addition , we also suggested that the cross - correlation of cmb and galaxy lensing can also be used as a probe of initial conditions . in the case of the cross - correlation , each galaxy sample probes a different range of wavenumbers due to its corresponding kernel which is related to the distribution of the galaxies . there are indications of hemispherical asymmetry in cmb observations . we have proposed that both the cmb lensing potential and cmb lensing convergence can be used to probe the amplitude of dipole modulation . an important feature is that the amplitude of the cmb lensing potential is sensitive to both the amplitude of dipole and also to the orientation with respect to the direction of dipole modulation . the future cmb observations are planned to scan the last scattering surface with high resolution in small patches . accordingly they can be used to address the predictions of this study . in addition the cmb lensing can probe the scale dependence of the dipole modulation . the planck data suggests a non - trivial scale - dependence for dipole amplitude in which the amplitude of dipole falls off rapidly for @xmath140 . we showed that the cmb lensing potential ( convergence ) can be used to constrain the scale dependence of dipole modulation . in this work we considered the simple scaling relation @xmath141 for dipole amplitude and have plotted the signal of cmb lensing as a function of the spectral index of dipole modulation . we have used the cmb lensing convergence autocorrelation and the cosmic shear convergence cross correlation with cmb to constrain the amplitude of @xmath142 and the spectral index of dipole modulation . another example of deviation from the standard primordial initial conditions which can be tested via cmb lensing is the deviation from scale invariance . the observational data from cmb and lss show that up to scales of groups of galaxies the primordial power spectrum is scale invariant . however , there is no direct evidence that this scale invariance continues on smaller scales . in the second part of this work we have studied the question of deviation from scale invariance in the form of local feature in primordial power spectrum . in order to model this feature we have assumed a gaussian - type bump in the primordial power on small scales @xmath143 . we studied the effect of the amplitude of the bump , the location of feature in momentum space @xmath116 and the effect of the variance of gaussian bump on the cmb lensing potential and convergence . we also studied the effect of feature on the cross correlation of cmb lensing convergence and galaxy lensing . we showed that this cross correlation has the ability to probe the signals in higher moments ( small scales ) in examining the scale dependence of primordial power spectrum . there are other questions which one can study in this direction . for example , one can study the question of statistical anisotropy in primordial power spectrum . this can arise for example in models of inflation containing background vector fields and gauge fields , for a review see @xcite . because of the background vector field , the rotational symmetry is broken . the primordial power spectrum has the form [ si ] ( k)= , in which @xmath144 represents the preferred direction in the sky and @xmath145 measures the amplitude of quadrupole anisotropy . the date from planck observations yield the constrain @xmath146 @xcite . note that the anisotropy in eq . ( [ si ] ) has a different nature than the dipole asymmetry in eq . ( [ dipole ] ) . statistical anisotropy given in eq . ( [ si ] ) is defined in fourier space and measures anisotropy for each point in the sky while dipole asymmetry given in eq . ( [ dipole ] ) holds in real space and only distinguishes two opposite hemispheres and not the individual points . it will be interesting to use the cmb lensing potential and cmb lensing convergence to probe the statistical anisotropy of the primordial perturbations . we would like to thank anthony lewis , anthony challinor and yashar akrami for insightful comments . we also thank the anonymous referee for his / her detailed and insightful comments and suggestions which helped to improve the contents and the presentation of the manuscript . in this appendix we investigate the parameter space of the dipole asymmetry effect on the lensing potential correlation function and also on the cross correlation of the cmb convergence and cosmic shear lensing . the idea is to use the observable @xmath147 which is introduced in sec . ( [ sec : deviation ] ) as a probe to pin down the physics of dipole asymmetry . -1.5 cm for different values of @xmath70 : @xmath148 ( black solid curve ) , @xmath149 ( red long dashed curve ) and @xmath150 ( blue dashed curve ) with @xmath99 and @xmath91 for all curves.,title="fig:",width=238,height=188 ] -1.5 cm for different values of @xmath70 : @xmath148 ( black solid curve ) , @xmath149 ( red long dashed curve ) and @xmath150 ( blue dashed curve ) with @xmath99 and @xmath91 for all curves.,title="fig:",width=238,height=188 ] -1.5 cm for different values of @xmath70 : @xmath148 ( black solid curve ) , @xmath149 ( red long dashed curve ) and @xmath150 ( blue dashed curve ) with @xmath99 and @xmath91 for all curves.,title="fig:",width=238,height=188 ] in fig . ( [ clpsiamp ] ) @xmath98 , as defined in eq . ( [ cl - ns ] ) , is plotted versus the moment @xmath151 for different values of dipole amplitude : @xmath152 ( black solid curve ) , @xmath99 ( red long dashed curve ) and @xmath153 ( blue dashed curve ) . in all cases the spectral index of dipole modulation is set to @xmath90 and the pivot wavenumber of dipole modulation is set to @xmath91 . the figure shows that probing the cmb lensing in two opposite hemispheres can constrain the amplitude of the dipole modulation . in fig . ( [ clphiindex ] ) , @xmath98 is plotted for different values of @xmath87 : @xmath102 ( black solid curve ) , @xmath90 ( red long dashed curve ) , @xmath103 ( blue dashed curve ) . in all cases @xmath99 and @xmath91 . the figure shows that a smaller spectral index introduces more negative slope for the power and accordingly the difference in dipole modulation in low and high momenta are more enhanced . the fixed point where all powers have the same magnitude represents the pivot wavenumber . as described before the pivot number is chosen to have the pivot moment of @xmath154 . accordingly we assert that the cmb lensing is a promising tool to constrain the scale dependence of the dipole modulation . measuring the high and low moments of lensing potential can be used as criteria to constrain the spectral index of the modulation . in fig . ( [ clphidegree ] ) , @xmath98 is plotted for different values of @xmath70 : @xmath148 ( black solid curve ) , @xmath149 ( red long dashed curve ) and @xmath150 ( blue dashed curve ) in which in all plots @xmath99 and @xmath155 . it is obvious that the maximum difference can be obtained in direction of dipole . -1.4 cm for different values of @xmath70 : @xmath148 ( black solid curve ) , @xmath149 ( red long dashed curve ) and @xmath150 ( blue dashed curve ) with @xmath99 and @xmath155 for all curves.,title="fig:",width=245,height=188 ] -1.4 cm for different values of @xmath70 : @xmath148 ( black solid curve ) , @xmath149 ( red long dashed curve ) and @xmath150 ( blue dashed curve ) with @xmath99 and @xmath155 for all curves.,title="fig:",width=245,height=188 ] -0.7 cm for different values of @xmath70 : @xmath148 ( black solid curve ) , @xmath149 ( red long dashed curve ) and @xmath150 ( blue dashed curve ) with @xmath99 and @xmath155 for all curves.,title="fig:",width=245,height=188 ] as we discussed extensively in sec . ( [ sec : deviation ] ) , another promising tool to study the deviation from standard primordial perturbation is the cross correlation of cmb convergence with cosmic shear data . a crucial point to indicate here is that each cross correlation signal is affected by the distribution of the galaxies , where it can probe the dipole asymmetry in different moments . in figs . ( [ fig : kcmbkgal - dipole - amp ] ) , ( [ fig : kcmbkgal - dipole - nd ] ) and ( [ fig : kcmbkgal - dipole - degree ] ) we study respectively the effect of @xmath60 , @xmath87 and @xmath70 on @xmath106 . the curve descriptions respectively are the same as in figs . ( [ clpsiamp ] ) , ( [ clphiindex ] ) and ( [ clphidegree ] ) . in this appendix we probe the parameter space of a gaussian feature in primordial power spectrum with cmb lensing and cmb convergence - galaxy convergence cross correlation data . the idea is the same as raised in app . ( [ app1 ] ) where we plot the difference in powers of the lensing and convergence for the cases with the feature compared with the standard one . in fig . ( [ clphi - f - amp ] ) , the fractional change in lensing potential due to feature , @xmath158 , is plotted . the feature has gaussian shape as in eq . ( [ eq : gauss ] ) with @xmath159 and @xmath137 held fixed . plots are for different values of the amplitude , @xmath131 ( blue dotted line ) , @xmath160 ( red long dashed line ) and @xmath161 ( green solid line ) . fig . ( [ clphi - f - amp ] ) shows that for features with large amplitudes the cmb lensing can probe the deviation from the standard case . in fig . ( [ clphi - f - kf ] ) we have plotted @xmath158 for the feature with @xmath131 and @xmath137 for different wavenumbers . an important fact is that by going to smaller scales ( larger wavenumbers ) , the signal of deviation from scale invariance appears in higher moments @xmath151 . accordingly , if we want to probe the deviation of the initial power spectrum from scale invariance in smaller and smaller scales , we have to reconstruct the cmb lensing maps in higher moments with corresponding resolution . the highest moment of cmb lensing potential reconstruction is now in the moment of @xmath162 . in fig . ( [ clphi - f - sigmaf ] ) , we have plotted the same quantity as in previous figures for different value of @xmath133 . the smaller values of variance lead to features in the angular power spectrum which have higher amplitudes and are more localized . as mentioned we can also use the cmb convergence - galaxy convergence correlation data which in this case the signal is sourced by the distribution of galaxies as well . in fig . ( [ clkckg - f - amp ] ) we plot @xmath163 for different values of amplitude with @xmath127 and @xmath128 . it is worth to mention that the peak is shifted to smaller moments . in fig . ( [ clkckg - f - kf ] ) and fig . ( [ clkckg - f - sigmaf ] ) we study respectively the effects of change in @xmath116 and @xmath133 on @xmath163 . m. tegmark _ et al . _ [ sdss collaboration ] , `` cosmological parameters from sdss and wmap , '' phys . d * 69 * , 103501 ( 2004 ) [ astro - ph/0310723 ] . d. j. eisenstein _ et al . _ [ sdss collaboration ] , `` detection of the baryon acoustic peak in the large - scale correlation function of sdss luminous red galaxies , '' astrophys . j. * 633 * , 560 ( 2005 ) [ astro - ph/0501171 ] . m. tegmark _ et al . _ [ sdss collaboration ] , `` cosmological constraints from the sdss luminous red galaxies , '' phys . d * 74 * , 123507 ( 2006 ) [ astro - ph/0608632 ] . a. h. guth , `` the inflationary universe : a possible solution to the horizon and flatness problems , '' phys . d * 23 * , 347 ( 1981 ) . a. d. linde , `` a new inflationary universe scenario : a possible solution of the horizon , flatness , homogeneity , isotropy and primordial monopole problems , '' phys . b * 108 * , 389 ( 1982 ) . a. albrecht and p. j. steinhardt , phys . lett . * 48 * , 1220 ( 1982 ) . k. sato , mon . not . * 195 * , 467 ( 1981 ) . a. a. starobinsky , phys . b * 91 * , 99 ( 1980 ) . p. a. r. ade _ et al . _ [ planck collaboration ] , `` planck 2015 . constraints on inflation , '' arxiv:1502.02114 [ astro-ph.co ] . p. a. r. ade _ et al . _ [ bicep2 and planck collaborations ] , `` a joint analysis of bicep2/keck array and planck data , '' [ arxiv:1502.00612 [ astro-ph.co ] ] . k. array _ et al . _ [ bicep2 s collaboration ] , `` bicep2 / keck array vi : improved constraints on cosmology and foregrounds when adding 95 ghz data from keck array , '' arxiv:1510.09217 [ astro-ph.co ] . s. das , t. louis , m. r. nolta , g. e. addison , e. s. battistelli , j. r. bond , e. calabrese and d. c. m. j. devlin _ et al . _ , `` the atacama cosmology telescope : temperature and gravitational lensing power spectrum measurements from three seasons of data , '' jcap * 1404 * , 014 ( 2014 ) [ arxiv:1301.1037 [ astro-ph.co ] ] . r. keisler , c. l. reichardt , k. a. aird , b. a. benson , l. e. bleem , j. e. carlstrom , c. l. chang and h. m. cho _ et al . _ , `` a measurement of the damping tail of the cosmic microwave background power spectrum with the south pole telescope , '' astrophys . j. * 743 * , 28 ( 2011 ) [ arxiv:1105.3182 [ astro-ph.co ] ] . p. a. r. ade _ et al . _ [ planck collaboration ] , `` planck 2013 results . . gravitational lensing by large - scale structure , '' astron . astrophys . * 571 * ( 2014 ) a17 [ arxiv:1303.5077 [ astro-ph.co ] ] . p. a. r. ade _ et al . _ [ planck collaboration ] , `` planck 2015 results . gravitational lensing , '' arxiv:1502.01591 [ astro-ph.co ] . a. lewis and a. challinor , `` weak gravitational lensing of the cmb , '' phys . rept . * 429 * , 1 ( 2006 ) [ astro - ph/0601594 ] . p. a. r. ade _ et al . _ [ xxx collaboration ] , `` planck 2015 results . dark energy and modified gravity , '' arxiv:1502.01590 [ astro-ph.co ] . d. hanson , k. m. smith , a. challinor and m. liguori , `` cmb lensing and primordial non - gaussianity , '' phys . d * 80 * , 083004 ( 2009 ) [ arxiv:0905.4732 [ astro-ph.co ] ] . l. amendola , g. ballesteros and v. pettorino , `` effects of modified gravity on b - mode polarization , '' phys . d * 90 * , no . 4 , 043009 ( 2014 ) [ arxiv:1405.7004 [ astro-ph.co ] ] . t. okamoto and w. hu , `` cmb lensing reconstruction on the full sky , '' phys . d * 67 * , 083002 ( 2003 ) [ astro - ph/0301031 ] . r. pearson , a. lewis and d. regan , `` cmb lensing and primordial squeezed non - gaussianity , '' jcap * 1203 * , 011 ( 2012 ) [ arxiv:1201.1010 [ astro-ph.co ] ] . p. a. r. ade _ et al . _ [ planck collaboration ] , `` planck 2013 results isotropy and statistics of the cmb , '' astron . astrophys . * 571 * , a23 ( 2014 ) [ arxiv:1303.5083 [ astro-ph.co ] ] . p. a. r. ade _ et al . _ [ planck collaboration ] , `` planck 2015 results . xvi . isotropy and statistics of the cmb , '' arxiv:1506.07135 [ astro-ph.co ] . h. k. eriksen , f. k. hansen , a. j. banday , k. m. gorski and p. b. lilje , `` asymmetries in the cosmic microwave background anisotropy field , '' astrophys . j. * 605 * , 14 ( 2004 ) [ erratum - ibid . * 609 * , 1198 ( 2004 ) ] [ astro - ph/0307507 ] . h. k. eriksen , g. huey , r. saha , f. k. hansen , j. dick , a. j. banday , k. m. gorski and p. jain _ et al . _ , `` a re - analysis of the three - year wmap temperature power spectrum and likelihood , '' astrophys . j. * 656 * , 641 ( 2007 ) [ astro - ph/0606088 ] . h. k. eriksen , a. j. banday , k. m. gorski , f. k. hansen and p. b. lilje , `` hemispherical power asymmetry in the three - year wilkinson microwave anisotropy probe sky maps , '' astrophys . j. * 660 * , l81 ( 2007 ) [ astro - ph/0701089 ] . y. akrami , y. fantaye , a. shafieloo , h. k. eriksen , f. k. hansen , a. j. banday and k. m. grski , `` power asymmetry in wmap and planck temperature sky maps as measured by a local variance estimator , '' astrophys . j. * 784 * , l42 ( 2014 ) [ arxiv:1402.0870 [ astro-ph.co ] ] . d. j. eisenstein and w. hu , `` baryonic features in the matter transfer function , '' astrophys . j. * 496 * , 605 ( 1998 ) [ astro - ph/9709112 ] . m. bartelmann and p. schneider , `` weak gravitational lensing , '' phys . * 340 * , 291 ( 2001 ) [ astro - ph/9912508 ] . f. bernardeau , `` gravitational lenses , '' astro - ph/9901117 . d. m. wittman , j. a. tyson , d. kirkman , i. dellantonio and g. bernstein , `` detection of weak gravitational lensing distortions of distant galaxies by cosmic dark matter at large scales , '' nature * 405 * , 143 ( 2000 ) [ astro - ph/0003014 ] . l. fu _ et al . _ , `` very weak lensing in the cfhtls wide : cosmology from cosmic shear in the linear regime , '' astron . astrophys . * 479 * ( 2008 ) 9 [ arxiv:0712.0884 [ astro - ph ] ] . n. hand _ et al . _ , `` first measurement of the cross - correlation of cmb lensing and galaxy lensing , '' phys . d * 91 * , no . 6 , 062001 ( 2015 ) [ arxiv:1311.6200 [ astro-ph.co ] ] . s. aiola , b. wang , a. kosowsky , t. kahniashvili and h. firouzjahi , `` microwave background correlations from dipole anisotropy modulation , '' phys . d * 92 * , 063008 ( 2015 ) [ arxiv:1506.04405 [ astro-ph.co ] ] . s. mukherjee , p. k. aluri , s. das , s. shaikh and t. souradeep , arxiv:1510.00154 [ astro-ph.co ] . s. mukherjee and t. souradeep , arxiv:1509.06736 [ astro-ph.co ] . s. adhikari , mon . not . soc . * 446 * , no . 4 , 4232 ( 2015 ) [ arxiv:1408.5396 [ astro-ph.co ] ] . a. l. erickcek , m. kamionkowski and s. m. carroll , phys . d * 78 * , 123520 ( 2008 ) [ arxiv:0806.0377 [ astro - ph ] ] ; a. l. erickcek , s. m. carroll and m. kamionkowski , phys . d * 78 * , 083012 ( 2008 ) [ arxiv:0808.1570 [ astro - ph ] ] ; a. l. erickcek , c. m. hirata and m. kamionkowski , phys . d * 80 * , 083507 ( 2009 ) [ arxiv:0907.0705 [ astro-ph.co ] ] . l. dai , d. jeong , m. kamionkowski and j. chluba , phys . d * 87 * , 123005 ( 2013 ) [ arxiv:1303.6949 [ astro-ph.co ] ] . m. h. namjoo , s. baghram and h. firouzjahi , phys . d * 88 * , 083527 ( 2013 ) [ arxiv:1305.0813 [ astro-ph.co ] ] . @xcite a. a. abolhasani , s. baghram , h. firouzjahi and m. h. namjoo , phys . d * 89 * , no . 6 , 063511 ( 2014 ) [ arxiv:1306.6932 [ astro-ph.co ] ] . m. h. namjoo , a. a. abolhasani , s. baghram and h. firouzjahi , `` cmb hemispherical asymmetry : long mode modulation and non - gaussianity , '' jcap * 1408 * , 002 ( 2014 ) [ arxiv:1405.7317 [ astro-ph.co ] ] . d. h. lyth , jcap * 1308 * , 007 ( 2013 ) [ arxiv:1304.1270 [ astro-ph.co ] ] . j. f. donoghue , k. dutta and a. ross , phys . d * 80 * , 023526 ( 2009 ) [ astro - ph/0703455 [ astro - ph ] ] ; l. wang and a. mazumdar , phys . d * 88 * , 023512 ( 2013 ) [ arxiv:1304.6399 [ astro-ph.co ] ] ; a. mazumdar and l. wang , jcap * 1310 * , 049 ( 2013 ) [ arxiv:1306.5736 [ astro-ph.co ] ] ; s. jazayeri , y. akrami , h. firouzjahi , a. r. solomon and y. wang , jcap * 1411 * , 044 ( 2014 ) [ arxiv:1408.3057 [ astro-ph.co ] ] . m. h. namjoo , a. a. abolhasani , h. assadullahi , s. baghram , h. firouzjahi and d. wands , jcap * 1505 * , no . 05 , 015 ( 2015 ) [ arxiv:1411.5312 [ astro-ph.co ] ] . h. assadullahi , h. firouzjahi , m. h. namjoo and d. wands , jcap * 1504 * , no . 04 , 017 ( 2015 ) [ arxiv:1410.8036 [ astro-ph.co ] ] . h. firouzjahi , j. o. gong and m. h. namjoo , jcap * 1411 * , no . 11 , 037 ( 2014 ) [ arxiv:1405.0159 [ astro-ph.co ] ] . a. r. liddle and m. corts , phys . lett . * 111 * , 111302 ( 2013 ) [ arxiv:1306.5698 [ astro-ph.co ] ] ; t. kobayashi , m. corts and a. r. liddle , jcap * 1505 * , no . 05 , 029 ( 2015 ) [ arxiv:1501.05864 [ astro-ph.co ] ] . g. damico , r. gobbetti , m. kleban and m. schillo , jcap * 1311 * , 013 ( 2013 ) [ arxiv:1306.6872 [ astro-ph.co ] ] ; z. -g . liu , z. -k . guo and y. -s . piao , phys . d * 88 * , 063539 ( 2013 ) [ arxiv:1304.6527 [ astro-ph.co ] ] ; z. -g . liu , z. -k . guo and y. -s . piao , arxiv:1311.1599 [ astro-ph.co ] ; y. -f . cai , w. zhao and y. zhang , arxiv:1307.4090 [ astro-ph.co ] ; z. chang , x. li and s. wang , arxiv:1307.4542 [ astro-ph.co ] ; k. kohri , c. -m . lin and t. matsuda , arxiv:1308.5790 [ hep - ph ] ; c. t. byrnes and e. r. m. tarrant , jcap * 1507 * , 007 ( 2015 ) [ arxiv:1502.07339 [ astro-ph.co ] ] . et al . _ , `` the atacama cosmology telescope : a measurement of the cosmic microwave background power spectrum at 148 and 218 ghz from the 2008 southern survey , '' astrophys . j. * 729 * , 62 ( 2011 ) doi:10.1088/0004 - 637x/729/1/62 [ arxiv:1009.0847 [ astro-ph.co ] ] . a. a. abolhasani , h. firouzjahi , s. khosravi and m. sasaki , jcap * 1211 * , 012 ( 2012 ) [ arxiv:1204.3722 [ astro-ph.co ] ] . r. emami , arxiv:1511.01683 [ astro-ph.co ] . p. a. r. ade _ et al . _ [ planck collaboration ] , arxiv:1303.5082 [ astro-ph.co ] .

we investigate the association between the radio `` plateau '' states and the large superluminal flares in grs 1915 + 105 and propose a qualitative scenario to explain this association . to investigate the properties of the source during a superluminal flare , we present gmrt observations during a radio flare which turned out to be a pre - plateau flare as shown by the contemporaneous ryle telescope observations . a major superluminal ejection was observed at the end of this `` plateau '' state ( dhawan et al . 2003 ) , associated with highly variable x - ray emission showing x - ray soft dips . this episode , thus has all the three types of radio emission : a pre - plateau flare , a `` plateau '' state and superluminal jets . we analyze all the available rxte - pca data during this episode and show that : ( 1 ) the pre - flare `` plateau '' state consists of a three - component x - ray spectra which includes a multicolor disk - blackbody , a comptonized component and a power - law and ( 2 ) the compton cloud , which is responsible for the comptonizing component , is ejected away during the x - ray soft dips . we investigate all the available monitoring data on this source and identify several candidate superluminal flare events and analyze the contemporaneous rxte pointed observations . we detect a strong correlation between the average x - ray flux during the `` plateau '' state and the total energy emitted in radio during the subsequent radio flare . we find that the sequence of events is similar for all large radio flares with a fast rise and exponential decay morphology . based on these results , we propose a qualitative scenario in which the separating ejecta during the superluminal flares are observed due to the interaction of the matter blob ejected during the x - ray soft dips , with the steady jet already established during the `` plateau '' state . this picture can explain all types of radio emission observed from this source in terms of its x - ray emission characteristics .

astro-ph0308096

the black - hole candidate grs 1915 + 105 is one of the most studied x - ray sources in our galaxy due to its remarkably rich phenomenological diversity across the entire electromagnetic spectrum except the visible band . it was first detected in 1992 with the watch instrument on - board the granat satellite ( castro - tirado , brandt , & lund 1992 ) . in contrast with other transient x - ray sources , it has never been switched off since detection . its optical counter - part has not been detected due to a very high line - of - sight absorption ( @xmath026 magnitude , chaty et al . 1996 ) . in the absence of optical observations , the binary parameters as well as the nature of the companion were unknown till 2001 . finally nine years after its detection , greiner et al . ( 2001 ) , by near - infrared spectroscopic observations with vlt , established it as a low mass x - ray binary ( lmxb ) system with a k - m iii type companion star of mass 1.2@xmath10.2 m@xmath2 . the black hole mass was determined to be 14@xmath14 m@xmath2 which is the largest mass known so far for any black hole candidate in an x - ray binary . grs 1915 + 105 achieved the status of micro - quasar much earlier in 1994 when mirabel & rodriguez ( 1994 ) detected episodes of superluminal ejections from this source , first time in any galactic source . observations of many such episodes of superluminal ejections during the past eight years make grs 1915 + 105 distinct among other micro - quasars in our galaxy . apart from huge superluminal radio flares , grs 1915 + 105 also shows various other types of radio emission like radio oscillations ( pooley & fender 1997 ) and extended periods of steady , flat spectrum radio emission known as radio `` plateau '' states ( fender et al . these radio `` plateau '' states are particularly interesting due to their association with the large radio flares . it is found that the `` plateau '' states are almost always followed by large radio flares with steep spectrum ( hannikainen et al . 1998 ; fender et al . 1999 ; klein - wolt et al . 2002 ) , though so far there is no satisfactory explanation for this observed association . the true glory of this source came into notice after the launch of nasa s _ rossi x - ray timing explorer ( rxte ) _ in 1996 . pointed observations with rxte - pca showed extremely rich and exciting morphology in the x - ray intensity patterns ( greiner , morgan , & remillard 1996 ; morgan , remillard , & greiner 1997 ; muno , morgan , & remillard 1999 ) . similar diversity in the x - ray emission from this source was also observed with the _ indian x - ray astronomy experiment ( ixae ) _ ( paul et al . 1997 , 1998 ; yadav et al . belloni et al . ( 2000 ) carried out a detailed analysis of all the rxte pointed observations in 1996 - 97 and classified the complex diversity of the x - ray emission exhibited by this source into 12 classes on the basis of light curves and hardness ratios ( one more variability class was identified later on - see klein - wolt et al . 2002 ; naik , rao & chakrabarti 2002 ) . they also put an important step forward in understanding the complexity of this source by showing that all the variability classes occur because of repeated transitions of the source in three basic states . they showed that the transitions between these three states , named as states a , b , and c , can be very fast i.e. within a few seconds . the states a and b are characterized by a soft spectrum with the total luminosity being low and high , respectively . the state c is characterized by a hard spectrum and the presence of strong quasi periodic oscillations ( qpo ) in the power density spectrum . this source is probably the best example for a strong connection between the accretion disk , manifested in the x - ray band , and the jet , manifested in the radio / ir band , in a black hole system ( pooley & fender 1997 ; eikenberry et al . 1998 ; 2000 ; mirabel et al . 1998 ) . study of the correlated behavior of the source in the x - ray and the radio / ir wave bands plays a very important role in the endeavor of understanding this source and many attempts have been made so far to relate the x - ray emission characteristics with the episodes of various types of radio emission . muno et al . ( 2001 ) studied representative observations in detail and showed that the properties of the qpo , like frequency , phase - lag , coherence etc . are correlated with the radio emission . rau & greiner ( 2003 ) showed that the properties of accretion disk are not correlated with the radio emission . naik & rao ( 2000 ) showed that the radio emission is high only during three variability classes : @xmath3 , @xmath4 and a sub - class of @xmath5 ( @xmath6/@xmath7 or @xmath8 . ) into four sub - classes @xmath9 , @xmath10 , @xmath11 and @xmath12 based on their order of occurrence . out of these four sub - classes only @xmath6 and @xmath7 show high radio emission . however , later more such occurrences of class @xmath5 have been observed during which the radio emission is high . thus , instead of numbered sub - classes , class @xmath5 can be divided into two generic sub - classes radio - quiet ( @xmath13 ) and radio - loud ( @xmath8 ) . the sub - class @xmath8 corresponds to the radio `` plateau '' states . ] they suggested that the radio emission is associated with the soft x - ray dips ( short periods of state a ) observed in the classes @xmath3 and @xmath4 , apart from the sub - class @xmath8 . naik et al . ( 2001 ) detected similar dips in the x - ray light curve during the presence of a large radio flare indicating their association with large radio flares . klein - wolt et al . ( 2002 ) studied all simultaneous observations of the source by rxte and ryle telescope and showed that the radio emission is associated with only state c and not with states a or b. they also found an one - to - one relation between series of long ( @xmath14 100 s ) , well separated state c intervals and radio oscillation events . they suggested a scenario , according to which the long uninterrupted state c intervals are associated with continuous jets which give `` plateau '' radio emission , whereas if the state c intervals are well separated by other states then the radio oscillation events are observed . a different picture , however , emerges when the association of the radio emission with the different sub - classes of @xmath5 ( state c ) are examined in detail ( see discussion in section [ scenario ] for more details on this ) . the sub - classes of class @xmath5 can be divided into two different types of hard states : radio - loud hard state ( i.e. @xmath15 ) and radio - quiet hard state ( i.e. @xmath16 ) ( vadawale et al . 2001b ; trudolyubov 2001 ) . by a detailed spectral study of rxte observations of the source belonging to the two types of hard states , vadawale et al . ( 2001b ) showed that the wide band x - ray spectrum of state @xmath16 consists of two components : a multicolor disk - blackbody and a comptonized component whereas the wide band x - ray spectrum of state @xmath15 consists of three components : a multicolor disk - blackbody , a comptonized component and a power - low . they further showed that the additional power - law component can be modeled as due to synchrotron radiation in x - rays , coming from the base of a continuous jet which is present only during the @xmath15 states . this three - component picture gets further support from other results such as : only the comptonized component is responsible for the low - frequency qpo ( rao et al . 2000 ) and both the qpo as well as the comptonized component are absent during the soft dips ( short state a periods ) observed in class @xmath4 ( vadawale et al . 2001a ) . it should be noted that the radio - loud hard state also occurs as a part of other variability classes like @xmath4 and @xmath3 . hence we will refer it generically as @xmath15 , whereas the particular sub - class of class @xmath5 during which the radio emission is high , will be referred as @xmath8 . in this paper we explore the association of the radio `` plateau '' states with large superluminal radio flares and try to envisage a scenario to explain this association as well as the various types of radio emission observed from this source . since there are not many detailed studies of such episodes of superluminal jet emissions , in the first part of the paper we present a detailed study of a complete sequence of flare morphology described , sequentially , by a small flare , `` plateau '' state , a disturbed accretion disk and a superluminal jet . we report the observations of the source with _ giant meter - wave radio telescope ( gmrt ) _ at 1.28 ghz when the source exhibits a flare which turned out to be a pre - plateau flare . we also present the results of a detailed analysis of all the pointed rxte observations during this flare cycle which confirm the three - component x - ray spectral description of the pre - flare `` plateau '' state . in order to confirm the three - component description during all occurrences of the `` plateau '' state , we further carry out similar analysis of representative pointed rxte observations prior to all huge radio flares identified in the gbi monitoring data on this source . finally we summarize the results and propose a phenomenological model for different types of radio emission from this source . radio observations of the micro - quasar grs 1915 + 105 , at 1.28 ghz with a band pass of 16 mhz , were carried out with gmrt ( swarup et al . 1991 ) on 2001 june 18 , 22 , 23 , 27 , 28 , 29 , 30 and july 1 . the flux density scale was set by observing the primary calibrator 3c286 / 3c48 . a phase calibrator was observed before and after a 45 minutes scan on grs 1915 + 105 . the integration time during the source observations was 32 s. the data reduction and analysis were done by using the standard astronomical image processing system ( aips ) package . the observations on 2001 june 18 , 22 , 23 , and 27 were of @xmath0 30 - 60 minutes , whereas on june 28 , 29 , 30 and july 1 , the observations were of @xmath0 8 - 10 hours . light curves , with 32 s integration time , were extracted from each of the individual observations of the source and are shown in figure [ gmrt ] . it can be seen from the figure that the source was quite steady on june 18 with a very low flux density ( @xmath010 mjy ) . the source was found to be bright in radio on june 22 with a flux density of @xmath0 50 mjy , along with a sharp dip with a decrease in the flux density by a factor of @xmath0 2 . on june 23 the source was steady at about 30 mjy . it was very weak ( @xmath0 5 mjy ) on june 27 . from june 28 onwards the observations were of long durations ( 8 10 hours ) and hence a wide range of flux variabilities could be studied . on june 28 the source was found to be very weak at the beginning of the observations ( utc 17 hour ) , but the radio flux density gradually increased to @xmath0 50 mjy followed by a sharp decrease to @xmath0 5 mjy at the end of observations ( utc 24 hour ) . on the other hand , on june 29 , the source was bright in radio with flux density @xmath0 70 mjy at the beginning which decayed down to 10 mjy towards the end . the source was relatively weak and steady on june 30 and july 1 . it can be seen that the radio emission is highly variable over time scales of a few tens of minutes , similar to that observed during mini flares ( or `` baby jets '' ) . analyzing many individual `` clean '' flares from the same data , ishwara - chandra et al . ( 2002 ) have tried to model these flares as due to adiabatically expanding synchrotron emitting clouds . to corroborate the results from the gmrt observations , we obtained radio monitoring data of grs 1915 + 105 from 2001 june 15 to july 25 observed with the ryle telescope at 15.2 ghz ( pooley & fender 1997 ) . radio light curves at 15.2 ghz are shown in figure [ ryle ] . a strong radio flare was observed on june 21 which was narrowly missed by gmrt . radio flares with low peak flux density were also observed with both gmrt and ryle telescopes on june 28 and 29 . similar flux densities at 1.4 ghz and 15.2 ghz suggest the flat nature of the spectrum during the flares . it can be seen , from figure [ ryle ] , that the source went into a radio - loud `` plateau '' state after the gmrt observations which is followed by a radio flare on july 16 , observed with the _ ratan _ telescope ( dhawan et al . the flux density of this flare was much higher at lower frequencies ( as observed by ratan ) compared to that at 15 ghz ( as observed by ryle ) suggesting a steep spectrum during the flare . this source was also observed by vlba on july 16 , 17 and 18 at 2.25 , 8.3 and 15 ghz . vlba observations clearly showed an ejecta well separated from the core ( dhawan et al . 2003 ) . these observations suggest that this superluminal flare episode followed a typical sequence of events observed during the past episodes of superluminal flares . it started with a small flare , which was observed by gmrt , followed by a long radio `` plateau '' state , followed by superluminal ejections . it is thought that the different types of radio emission are results of changes in the accretion flow which are directly reflected in x - rays . thus to understand the sequence of these events it is necessary to make a detailed analysis of the x - ray observations during this time period . to investigate the x - ray properties of the source during and before this 2001 july 16 flaring episode , we analyzed all the 20 available rxte pointed observations of grs 1915@xmath17105 from 2001 june 30 to 2001 july 16 . standard procedures for data reduction , response matrix generation , and background estimation were followed using the software package heasoft 5.2 . for spectral analysis we used standard-2 mode data from proportional counter array ( pca ) and archive mode data from hexte cluster-0 . we added the count rate spectrum of all layers of individual proportional counter units ( pcus ) and all available pcus together . analyzing the spectra of crab with the new response matrices , it is found that 0.5% systematic error is required to achieve an acceptable fit and the value of the power - law photon index is also found to be close to the canonical value of 2.1 . systematic error of 0.5% is , therefore , added to the pca spectra of all the observations used in the present work . x - ray light curves with 1 s time resolution ( standard-1 mode ) are generated for all the 20 observations . it is found that during all the rxte / pca observations before 2001 july 16 , the source was in a hard state ( i.e. the source state is @xmath15 whereas the class is @xmath8 ) . light curves of all the rxte / pca observations on july 16 and the last two rxte / pca observations before july 16 , along with the x - ray ( rxte / asm ) and radio ( ryle telescope ) monitoring light curves , are shown in figure [ jul_lc ] . it can be seen from figure [ jul_lc ] that the last three pointed rxte / pca observations on july 16 clearly show evidence for a disturbed accretion disc . the first three pointed observations on july 16 ( when the radio flare is already on ) show enhanced x - ray emission , which could be either an enhanced @xmath8 class or the steady part of the x - ray emission during a disturbed accretion disc . because of the gaps between the pointed observations , it is not possible to precisely identify the time of occurrence of the state change from the x - ray light curves . we carried out a detailed timing and spectral analysis of all the 20 rxte observations to examine the x - ray properties so that the time of occurrence of the state change could be better understood . for timing analysis , single - bit mode data , available in the energy ranges of 3.6 5.7 kev and 5.7 14.8 kev , are used . power density spectra ( pds ) of 256 bin light curves are generated and co - added for every 16 seconds to achieve better signal to noise ratio . results of timing and spectral analysis are given in table 1 , along with the log of observations . the parameters for one representative observation per day ( before the radio flare ) and all observations carried out on july 16 ( when the radio flare commenced ) are given in the table . quasi periodic oscillations ( qpos ) in 1 3 hz frequency range are seen in the pds of all the rxte observations made before july 16 , whereas the pds of all observations on july 16 show qpos at @xmath14 4 hz ( see table [ jul]a ) . along with the pds , phase - lag spectra are also generated by multiplying the imaginary part of the coefficients of the power density spectra in the above two energy ranges ( reig et al . individual phase - lag spectra for each 16 s interval were also co - added . phase - lag spectra of all the observations are shown in figure [ phase ] . the top panel of figure [ phase ] shows the phase - lag spectra of all observations made before july 16 ( `` plateau '' state ) and the bottom panel shows the phase - lag spectra of all observations made on july 16 ( including those showing flare / dips , however , these phase - lag spectra are calculated only for the steady portion of the light curve ) . it can be seen that there is a clear difference in the overall shape of the phase - lag spectra . in order to quantify the shape of the phase - lag spectra of the observations before and on july 16 , we applied ks test which shows that these two distributions are different with 99.99% confidence level significance . the reason for this difference in the shape of the phase - lag spectra is not clear , however , it is possible that the ejected matter during the x - ray disturbance ( see next section ) might be playing some role here as suggested by pottschmidt et al . however , the difference clearly shows that the source state before july 16 is different from the state during the steady observations on july 16 and the disturbance in x - rays has already occurred before the first observation on july 16 . xspec package ( ver . 11.2 ) is used for the spectral analysis . a simultaneous fit to the pca ( in the energy range of 3 - 35 kev ) and hexte ( in the energy range of 20 - 150 kev ) data was carried out , keeping the relative normalization as a free parameter . a fit with any continuum model shows strong residuals near 6.4 kev and any fit without a gaussian line at this energy is not acceptable . such residuals , though with a low amplitude , are also seen in the crab spectra obtained with pca . it is not clear what fraction of the residuals is really due to the source and what fraction is due to an instrumental artifact . the line feature at 6.4 kev , therefore , will not be discussed further in this paper . it is found that a model consisting of three components : a disk - blackbody , a power - law and a comptonized component ( comptt in xspec , see titarchuk 1994 ) is required for a statistically and physically acceptable fit to the source spectra ( see vadawale et al . 2001b for a comparison of the various models ) . the best - fit parameters along with the continuum flux of individual spectral components are given in table [ jul ] . it can be seen from table [ jul ] that the equivalent hydrogen column density is very high compared to the values reported in the literature ( e.g. muno et al . 1999 ; vilhu et al . spectral fitting with this three - component model requires n@xmath18 to be within 12 @xmath19 15 @xmath20 with 99% confidence level . the derived inner - disk radii ( r@xmath21 ) are also found to be high during these observations , compared to the reported values ( e.g. rau & greiner 2003 ) . however , these apparently high values of r@xmath21 are physically plausible i.e. 15 25 r@xmath22 for a 14 m@xmath23 black hole . the inner disk temperature ( kt@xmath21 ) and the plasma temperature ( kt@xmath24 ) for all the observations are @xmath0 0.5 kev and 711 kev respectively and are acceptable . it should be noted that on june 30 ( mjd 52090 ) , when the radio emission was relatively low ( the `` plateau '' started on july 1 ) , the spectral and timing parameters are different from the rest of the observations with the value of n@xmath18 being close to the reported value , which gives confidence for the applicability of this three - component model to the @xmath15 state of this source . the presence of short duration and irregular flares and dips in the x - ray light curves of rxte observations on july 16 makes spectral fitting extremely difficult during the individual flare / dip events . to examine the properties of the source during these flares and dips , we follow the ratio method used by vadawale et al . ( 2001a ) . figure [ jul16_rat ] shows the standard-1 light curve in the 260 kev energy band and the dynamic pds along with the ratios of the standard-1 light curve to the light curves in 28 , 815 , 1523 and 2360 kev energy ranges . the second panel of the figure shows that the low frequency 47 hz qpo , present in the steady part of the light curve , is absent during the dips . the third panel of the figure shows that during the dips , the decrease in the count - rate in 2 - 8 kev range is less than that in the total count - rate . inverted shapes of dips in the fourth and the fifth panels show that the counts in the energy band of 8 - 23 kev are preferentially decreased compared to the decrease in the total count rate ( see figures [ beta ] and [ theta ] for a clear picture where the dips are of longer duration ) . preferential decrease in the count rate in the 8 - 23 kev energy band strongly supports the inference drawn by vadawale et al . ( 2001a ) that the comptonized component , which is dominant in this energy band , is absent during these dips . the absence of qpo during these dips ( figure [ jul16_rat ] , second panel ) , and the observation that the comptonized component is responsible for the qpo in the hard ( @xmath15 ) states ( rao et al . 2000 ) , further strengthen the suggestion that the comptonized component vanishes during the dips . our detailed analysis around the 2001 july radio flare ( during which a separate ejecta was detected by vlbi , exhibiting superluminal ejection ) shows that before the flare , the source was in @xmath15 state consisting of three spectral components : a multi - color disk blackbody , a power - law , and a comptonized component . during the radio flare , the source showed a dipping behavior during which the comptonized component of the x - ray spectra appears to be vanishing . a similar behavior of the source was observed during the superluminal ejection episode of 1999 june ( dhawan et al . 2003 ) and 1997 october ( fender et al . the source was in @xmath15 state before the flare ( rao et al . 2000 ; vadawale et al . 2001b ) and pointed observation during the flare showed soft dips ( state a period ) during which the comptonized component vanishes . this leads to an interesting suggestion that the preceding @xmath15 state and the presence of soft dips ( state a periods ) are the characteristics of all superluminal ejection episodes . to verify this suggestion we carried out a similar analysis for all possible superluminal radio flares observed during the period of 1996 december to 2000 april when gbi radio monitoring data are available . apart from the superluminal episodes in 1997 october and 1999 june , there is one more occasion , 1998 may , during which a separated ejecta is reported ( dhawan et al . 2000 ) . however , a complete gbi monitoring radio light curve from 1996 december to 2000 april shows many other large flares and it might be possible that some such ejection episodes are missed . therefore , we selected all large flares ( @xmath14 350 mjy at 2.25 ghz ) from the gbi monitoring data . it is found that there are 9 flares satisfying this criteria , the characteristic properties of which are given in table [ radio_flares ] . the gbi and asm monitoring radio light curves of the source during all these 9 selected radio flares are shown in figure [ all_flares ] along with the complete monitoring light curves from gbi , asm and batse . all the large radio flares , during which separate ejecta are reported so far , show a fast rise and exponential decay ( fred ) with decay time constant of the order of 3 days . it can be seen , from table [ radio_flares ] and figure [ all_flares ] , that the characteristic properties of 3 other radio flares ( 1998 april 13 , 1998 june 3 and 1999 december 23 ) are found to be similar to those 3 flares during which superluminal ejection is reported . hence we assume that all flares with the fred morphology are superluminal flare episodes and we give the integrated flare energy ( in units of jy.day ) in table 2 . the gbi light curve also shows that the source was in radio `` plateau '' state prior to the all six ` assumed ' superluminal flares whereas the asm light curves during these periods show that the source was in low and steady x - ray state . the properties of the three radio flares without fred such as shape of the radio flares , variability in asm light curve prior to the flares , are different from those seen during the superluminal radio flares . a closer look at table [ radio_flares ] reveals an interesting fact that the average asm count rate and the gbi flux during the pre - flare `` plateau '' state show a strong correlation suggesting a strong coupling between the x - ray and radio emission . in figure [ corr ] ( left panel ) we show a plot of gbi flux against the asm count rate along with a best - fit straight line . we derive a correlation coefficient of 0.96 ( 98% confidence level ) . this could be similar to the correlation between the x - ray and the radio emissions found in the galactic black hole candidate sources ( see eg . gallo , fender & pooley 2003 ; choudhury et al . 2003 ) . we also find that the average pre - flare asm count rate and the total flux integrated over the duration of the flare are also strongly correlated . in figure [ corr ] ( right panel ) we show this correlation ( with a correlation coefficient of 0.98 ) which is significant at 99% confidence level . the correlation between the pre - flare asm count rate and the integrated flare flux is particularly interesting , given the temporal and spatial separation between the two types of emission . it shows that the pre - flare `` plateau '' state plays a crucial role in the formation of the large superluminal radio flares . to quantify the timing and the spectral properties of the pre - flare states of all the six selected flares , we have done a detailed analysis of the rxte pointed observations during these steady states . as the properties of the source are rather stable during these states , we analyzed two rxte observations immediately before each flare . we also analyzed the rxte observations on 1997 october 30 and 1999 june 8 , which are the only pointed observations of the source during the rising phase of superluminal flares . we adopted a similar analysis procedure described in the previous section and fitted the same three - component model to the source spectra of all the observations . the log of observations and the best - fit spectral parameters are given in table [ all_year ] . the results of analysis of the two pointed observations during the onset of superluminal flares are shown in figure [ beta ] and [ theta ] , which are similar to figure [ jul16_rat ] . we find that x - ray spectrum during all the pointed rxte observations immediately before the selected flares can be fitted by our model consisting of three components : a disk - blackbody , a comptonized component and a power - law . this confirms our three - component description of the @xmath15 state . both the available observations close to but after the onset of the radio flare shows dipping behavior as shown in figures [ beta ] and [ theta ] . it can be seen that , though the morphology of the x - ray light curves ( classes according to belloni et al . 2000 ) is different in all the pointed observations during the superluminal flares ( figures [ jul16_rat ] , [ beta ] and [ theta ] ) , the properties of the source during the dips are the same . the absence of qpos during the dips in x - ray light curves ( panel b of the above three figures ) and the preferential decrease in 8 23 kev count rate ( as seen from the inverted shape of the ratios in panels d and e of these figures ) confirm the absence of the comptonized component during the dips . we have verified , for a large number of observations containing similar soft dips ( state a periods ) , that the characteristics of all the soft dips are similar . thus it seems that the preferential decrease in @xmath0 8 25 kev range , suggesting the absence of the comptonized component , is rather a general feature of all the soft dips ( short state a periods ) seen in this source . this leads to an interpretation that the central compton cloud is ejected away during all the soft dips . similar conclusion of mass ejection during dips of type a is drawn by yadav ( 2001 ) on the basis of their spectral study during the class @xmath3 . we have presented a detailed study of all the candidate superluminal ejection episodes of grs 1915 + 105 , identified from the gbi , asm , and batse monitoring observations . it should be noted here that there are at least two other episodes of large flares , first one during 2000 july - august , reported by fender et al . ( 2002a ) which follows the `` plateau '' state large flare sequence and the other during 2001 march reported in fender et al . ( 2002b ) for which the superluminal ejection was observed directly . we have verified that the results presented here are consistent with the rxte pointed observations during 2000 july - august 2000 flare episode . for the 2001 march flare episode , unfortunately , both rxte - pca and rxte - asm coverage are poor and hence the pre - flare state can not be established , but we believe that it must be @xmath15 . results presented in the previous sections confirm the suggestions that ( i ) the wide band x - ray spectrum during the @xmath15 state has three spectral components : a multicolor disk - blackbody , a comptonized component and a power - law ( ii ) the soft dips ( short periods of state a ) are present in the x - ray light curve observed during the onset of superluminal flares , ( iii ) the comptonized component is absent during the soft dips . these results lead to an interesting suggestion that the compton cloud is ejected away during the soft dips . formation of a central hot region can be explained under the two component accretion flow ( tcaf ) paradigm as the hot , quasi - spherical , post shock region near the black hole ( chakrabarti & titarchuk 1995 , chakrabarti 2001 and references there in ) which is responsible for the presence of the comptonized component ( rao et al . ejection of the central compton cloud can be explained if magnetic field is included in the tcaf framework ( nandi et al . the pre - flare @xmath15 states are known to have a steady jet and the sudden ejection of matter in the previously established jet raises an interesting possibility for the origin of the superluminal flares in terms of the internal shock model . apparent superluminal motion of blobs during the large radio flares of galactic micro - quasars is one of the most spectacular phenomena of high energy astrophysics . grs 1915 + 105 is the only source which shows repeated episodes of large radio flares exhibiting superluminal motion . the apparent superluminal motion can be understood as geometric effect due to very high intrinsic velocity ( i.e. close to @xmath25 ) of the moving blob at some angle with the line of sight . so far attempts have been made to understand the superluminal motions based on two different ideas : ( 1 ) matter blob moving at very high velocity and ( 2 ) shocks moving through a relatively slowly moving medium . atoyan & aharonian ( 1999 ) showed that it is possible to reproduce the observed radio light curves of the separating blobs in terms of ejection and evolution of discrete plasmons , however , the instantaneous power requirement is very high in this model . further they found that a single population of relativistic particles accelerated at the time of ejection of the plasmon can not explain the observations and a continuous replenishment of relativistic particles is necessary . the continuous acceleration of particles comes naturally in the alternative model , the internal shock model proposed by kaiser , sunyaev & spruit ( 2000 ) . they show that if the moving blobs are the traveling shock front arising due to differential velocity of the moving continuous jet material then it is possible to reproduce the observed light curves . the total energy requirement is the same as the plasmon model , however , the rate at which the energy supplied from the central source is much lower than that in the plasmon model . the presence of the underlying continuous jet and the differential velocity of the moving matter are assumed _ a priori _ in the internal shock model and no attempt has been made to explain them . however , it can be seen that , in the context of the present work , these requirements are met quite naturally if the sudden ejection of the compton cloud ( i.e. the state a dip ) occurs after a prolonged period of @xmath15 state . the velocity of the ejected matter blob can be higher than the velocity of the matter in the underlying continuous jet , and hence as the ejected matter blob passes through the jet it creates an internal shock at the leading front where particles can be accelerated . gallo et al . ( 2003 ) showed that , in other gbhcs , the velocity of the matter in steady jets is lower than that during transient events . this gives further confidence to the picture of the ejected mass driving through slower steady flow causing the internal shock . this moving internal shock can then be observed as separating blobs in the radio images . this suggestion provides a rather complete picture for the different types of radio emission from this source as described in the next section . the results obtained from our analysis as well as the various results on radio flares reported in the literature , can be summarized as follows . * all superluminal radio flares , during which a separated ejecta is reported , as well as candidate superluminal flares for which there is no report of observation , are preceded by the `` radio - loud hard ( @xmath15 ) state '' . * the source spectrum during @xmath15 state consists of three components : a multicolor disk - blackbody , a comptonized component and a power - law . large values of the derived inner disk radius suggest a geometry in which there is a hot spherical compton cloud ( responsible for the comptonized component ) surrounded by a thin disk ( responsible for the disk - blackbody component ) . * whenever there is a pointed x - ray observation during the onset of a superluminal radio flare ( 3 out of 7 cases ) , the light curve shows soft dips ( small state a episodes ) . during such dips , the comptonized component is absent in the source spectrum . * high radio emission is observed from the source only during the x - ray variability classes having soft dips ( short state a periods ) apart from the radio `` plateau '' periods . * all superluminal flares show a characteristic fast rise and exponential decay profile in their radio light curve , which can not be explained by the blob - ejection mechanism . such profiles can be explained by the internal shock model of kaiser et al.(1999 ) . according to this model , the motion of the matter blob through the already present matter is seen as the moving ejecta during high spatial resolution radio observations . * the radio `` plateau '' ( @xmath15 ) state is known to have steady jets which eject a continuous stream of matter . thus if a blob - ejection occurs after a `` plateau '' state , it will satisfy the requirement of the internal shock model and gives rise to a superluminal radio flare . * the total integrated flux during the superluminal flares is highly correlated with the observed x - ray flux during the preceding @xmath15 states . these points suggest a scenario for different types of radio emission observed from this source . * the @xmath15 states have continuous jets and prolonged episodes of such states give rise to steady high radio emission with flat spectrum which arises from the continuous jet . this is a fairly robust statement as the compact jets have actually been observed during the episodes of @xmath15 states ( dhawan et al . * the ejection of the compton cloud during a soft ( state a ) dip gives rise to a small radio flare . repeated occurrence of such dips gives rise to radio oscillation events . it should be noted that the amplitude of the isolated flares or the radio oscillation events will depend on various aspects like the amount of mass outflow during the immediately previous @xmath15 state , the amount of mass ejected during the dips etc . hence , it might not be possible to always observe small radio flares or oscillation when such dips are present in x - ray , though , in such cases the average radio emission should be high . however , this means that whenever radio oscillations are observed , the soft dips must be present in the x - ray light curves . * if the occurrence of the soft ( state a ) dip is preceded by a prolonged period of @xmath15 state having a continuous jet then it satisfies the requirements of the internal shock model and gives rise to the observed superluminal motion of blobs as seen in high resolution radio images . a schematic picture of this scenario is shown in figure [ schematic ] . according to this scenario , any short time variation in the radio flux is caused by the occurrence of the soft ( state a ) dips during which the compton cloud is ejected . this is slightly different from the inference drawn by klein - wolt et al . ( 2002 ) that only state c episodes are related to the radio emission . they find that it is necessary to have the state c episodes well - separated in order to have radio oscillation events . however , there are x - ray variability classes like @xmath26 which show long , well separated periods of state c , but the average radio flux during class @xmath26 is much lower compared to the classes @xmath3 , @xmath4 which also have well separated long state c periods ( naik & rao 2000 ) . this can be explained in our scenario according to which , the state c periods in class @xmath26 are actually state @xmath16 periods whereas the state c periods in classes @xmath3 , @xmath4 are actually state @xmath15 periods . further , it can be seen that such well - separated state c episodes , when the radio oscillations are observed , are almost always accompanied by state a dips . thus the one - to - one relation between the state c episodes and the radio oscillation events found by klein - wolt et al . ( 2002 ) is valid in our scenario also . it should be noted that figure 3 g of klein - wolt et al . ( 2002 ) does not show any oscillations in the presence of class @xmath4 . according to our scenario it is possible not to have radio oscillations in the presence of state a dips , i.e. because of close ejection events etc . , but not vise - versa . as a counter example , figure 10 of dhawan et al . ( 2000 ) show clear oscillations during the presence of class @xmath4 . this picture is further supported by the pointed radio observation near the onset of the superluminal flare in 1997 october by fender et . al ( 1999 ) , which shows core oscillations as well as superluminal jets on 1997 october 30 and 31 . on both these days the source was in class @xmath3 which shows a type dips after c state ( yadav 2001 ) . the hypothesis of mass ejection during the soft dips is also supported by the calculation of time delay between emissions at different frequencies as presented by ishwara - chandra et al . they show that the ir emission follows the mass ejection with almost zero time delay . it is known that ir flares start at the time of spike ( in beta class ) which marks the beginning of a type dips ( eikenberry et al . 1998 ) . there are some important cross - checks for this scenario . first of all , it predicts that the separated ejecta should not be observed for any radio flare other than those having fred profile and are preceded by @xmath15 states . also whenever radio oscillations are observed , the soft ( state a ) dips must be present in the x - ray light curve . a large data base of radio and x - ray observation of grs 1915 + 105 has been collected so far and all the reported observations satisfies these criteria . however , the most important aspect of this scenario , to be tested with future observations , is the three - component description of the wide band x - ray spectra during the @xmath15 state . this can be verified , for example , by high sensitivity observations above 50 kev , where the differences in the variation of the rms power of the qpo with energy between @xmath15 and @xmath16 state will show the existence of additional component which is not participating in the qpo . it should be noted that according to this picture , the low luminosity and soft spectrum of state a is due to the absence of the central compton cloud . however this applies only to the short duration state a period appearing as soft dips . there exists a variability class @xmath27 during which the source appears to be in state a for much longer duration ( @xmath0days ) which can not be explained in the present picture ( see , however , vadawale et al . 2003 for an attempt to understand the class @xmath27 as due to obscuration of the compton cloud ) . this mechanism suggests that whenever there is a @xmath15 state , the next soft x - ray dip events will give rise to a superluminal flare . thus it provides a very important handle in predicting the occurrence of a flare which can help to plan new coordinated observation of large flares . such observations can lead to further understanding of very important questions like exact mechanism of ejection , cause of ejection , acceleration of the matter etc . it can be seen that a continuous hard x - ray and radio monitoring is very important in distinguishing @xmath16 state and @xmath15 states . unfortunately , after cgro - batse there is no hard x - ray monitor currently available making an important observation handle on the occurrence of @xmath15 state unavailable . the solar x - ray spectrometer ( soxs ) , which is a small phoswich type open hard x - ray detector ( malkar et al . 2003 ) , will also perform daily monitoring of a few bright sources and hence might be useful in predicting large superluminal flares from this source . in this paper we have tried to explain the observed association between the radio `` plateau '' or the @xmath15 states and the superluminal flares in grs 1915 + 105 . all @xmath15 states have very similar x - ray spectral properties and show a three - component spectra . we have further shown that the soft x - ray dips are observed near the onset of radio flares . during such dips the comptonized component is absent , which we propose as due to the ejection of the central compton cloud . this ejected matter is then responsible for the low frequency emission which can be observed as an isolated flare or regular oscillations ( if the dips are repeating ) . such ejections might give rise to superluminal flares if these dip events are preceded by a @xmath15 state according to the internal shock model for the superluminal flares . two definite predictions of this scenario are that ( 1 ) whenever radio oscillations are observed , the soft dips will be present in the x - ray light curves and ( 2 ) if @xmath15 is not preceded by any radio flare it will not have the characteristic fred shape and will not show separated ejecta . it also suggests that in the future whenever @xmath15 is observed , there are very good chances to catch a superluminal flare in ` live action ' and well coordinated multi wavelength observations of such an event will help understanding of the fundamental questions underlying such events . we thank the members of the rxte , batse , and gbi teams for making the data publicly available . this research has made extensive use of data obtained through the high energy astrophysics science archive research center ( heasarc ) online service , provided by nasa / goddard space flight center . the green bank interferometer ( gbi ) is a facility of the national science foundation operated by the nrao in support of nasa high energy astrophysics programs . atoyan , a. m. & aharonian , f. a. 1999 , mnras , 302 , 253 belloni , t. , klein - wolt , m. , mendez , m. , et al . 2000 , a&a , 355 , 271 castro - tirado , a.j . , brandt , s. , & lund , n. 1992 , iau circ . chakrabarti , s. k. 2001 , apsss , 276 191 chakrabarti , s. k. & titarchuk , l. g. 1995 , apj , 455 , 623 chaty , s. , mirabel , i. f. , duc , p. a. , wink , j. e. & rodriguez , l. f. 1996 , a&a , 310 , 825 dhawan v. et al . 2003 , in preparation dhawan , v. , mirabel , i. f. & rodriguez , l. f. 2000 , apj , 543 , 373 eikenberry , s. s. , matthews , k. , morgan , e. h. , remillard , r. a. & nelson , r. w. 1998 , apj , 494 , l61 eikenberry , s. s. , matthews , k. , muno , m. et al . 2000 , apj , 532 , l33 fender , r. p. , rayner , d. , mccormick , d. g. et al . 2002b , mnras , 336 , 39 fender , r. p. , rayner , d. , trushkin , s. a. et al . 2002a , mnras , 330 , 212 fender , r. p. , garrington , s. t. , mckay , d. j. et al . 1999 , mnras , 304 , 865 gallo , e. , fender , r. p. & pooley , g. g. 2003 , accepted in mnras , astro - ph/0305231 greiner , j. , morgan , e. h. & remillard , r. a. 1996 , apj , 473 , l107 greiner , j. , cuby , j. g. , & mccaughrean , m. j. 2001 , nature , 414 , 522 hannikainen , d. , hunstead , r. & campbell - wilson , d. 1998 , new astronomy reviews , 42 , 601 ishwara - chandra , c. h. , yadav , j. s. , & pramesh rao , a. 2002 , a&a , 388 , l33 kaiser , c. r. , sunyaev , r. & spruit , h. c. 2000 , a&a , 356 , 975 klein - wolt , m. , fender , r. p. , pooley , g. g. , et al . 2002 , mnras , 331 , 745 malkar , j. p. , janawlekar , s. b. , kothare , a. t. et al . , 2003 , basi , to appear mirabel , i. f. , & rodriguez , l. f. 1994 , nature , 371 , 46 mirabel , i. f. , dhawan , v. , chaty , s. et al . 1998 , a&a , 330 , l9 morgan , e. h. , remillard , r. a. & greiner , j. 1997 , apj , 482 , 1010 muno , m. p. , morgan , e. h. & remillard , r. a. 1999 , apj , 527 , 321 muno , m. p. , remillard , r. a. , morgan , e. h. , et al . 2001 , apj , 556 , 515 naik , s. , rao , a. r. & chakrabarti , s. k. 2002 , japa , 23 , 213 naik , s. , agrawal , p. c. , rao , a. r. , et al . 2001 , apj , 546 , 1075 naik , s. , & rao , a. r. 2000 , a&a , 362 , 691 nandi , a. , chakrabarti , s. k. , vadawale , s. v. , & rao , a. r. 2001 , a&a , 380 , 245 paul , b. , agrawal , p. c. , rao , a. r. et al . 1997 , a&a , 320 , l37 paul , b. , agrawal , p. c. , rao , a. r. et al . 1998 , a&as , 128 , 145 pooley , g. g. & fender , r. p. 1997 , mnras , 292 , 925 pottschmidt , k. , wilms , j. , nowak , m. a. et al . 2000 , a&a , 357 , l17 rao , a. r. , naik , s. , vadawale , s. v. & chakrabarti , s. k. 2000 , a&a , 360 , l25 rau , a. , & greiner , j. 2003 , a&a , 397 , 711 reig , p. , belloni , t. , van der klis , m. 2000 , apj , 541 , 883 swarup , g. , ananthakrishnan , s. , kapahi , v. k. , et al . 1991 , curr . sci , 60 , 95 titarchuk , lev 1994 , apj , 434 , 570 trudolyubov , s. p. 2001 , apj , 558 , 276 vadawale , s. v. , rao , a. r. & naik , s. 2003 , proc . of the 4th microquasar workshop , eds . durouchoux , y. fuchs and j. rodriguez , ( center for space physics , kolkata ) , 338 vadawale , s. v. , rao , a. r. , & chakrabarti , s. k. 2001b , a&a , 372 , 793 vadawale , s. v. , rao , a. r. , nandi , a. , & chakrabarti , s. k. 2001a , a&a , 370 , l17 vilhu , o. , poutanen , j. , nikula , p. & nevalainen , j. 2001 , apj , 553 , l51 yadav , j. s. , rao , a. r. , agrawal , p. c , et al . 1999 , apj , 517 , 935 yadav , j. s. 2001 , apj , 548 , 876 cccccccc + obsid@xmath28 & date & start & @xmath29 & + & & time ( ut ) & ( hz ) & f@xmath30 & f@xmath31 & f@xmath32 & f@xmath33 + a-65 - 00 & june 30 & 21:35:28 & 3.124 & 2.44 & 0.27 & 1.38 & 0.77 + a-66 - 00 & july 06 & 21:18:24 & 1.901 & 1.95 & 0.20 & 0.95 & 0.71 + a-66 - 01 & july 07 & 00:41:36 & 1.908 & 1.92 & 0.22 & 1.01 & 0.60 + b-04 - 00 & july 09 & 01:59:28 & 2.202 & 2.07 & 0.23 & 1.09 & 0.64 + b-04 - 05 & july 10 & 05:52:16 & 2.738 & 2.33 & 0.29 & 0.71 & 1.23 + b-04 - 08 & july 11 & 01:41:36 & 2.434 & 2.09 & 0.24 & 0.75 & 1.01 + + c-01 - 02 & july 16 & 00:51:28 & 4.341 & 3.65 & 0.46 & 0.52 & 2.59 + c-01 - 01 & july 16 & 02:30:40 & 3.998 & 3.25 & 0.50 & 1.52 & 1.07 + c-01 - 00 & july 16 & 08:10:56 & 4.533 & 3.70 & 0.63 & 2.28 & 0.61 + c-01 - 03 & july 16 & 11:20:48 & 4.194 & 3.07 & 0.53 & 1.42 & 0.95 + c-01 - 04 & july 16 & 18:03:28 & 4.874 & 3.90 & 0.68 & 1.47 & 1.55 + c-01 - 05 & july 16 & 21:21:20 & 4.705 & 3.72 & 0.62 & 1.38 & 1.55 + c-01 - 06 & july 16 & 23:02:40 & 4.895 & 3.89 & 0.67 & 1.04 & 1.99 + + + obsid@xmath28 & n@xmath18 & kt@xmath21 & kt@xmath34 & @xmath35 & @xmath36 & r@xmath21 @xmath37 & @xmath38(dof ) + & ( 10@xmath39 @xmath40 ) & ( kev ) & ( kev ) & & & ( km ) & + a-65 - 00 & 1.65@xmath41 & 1.06@xmath42 & 7.95@xmath43 & 6.38@xmath44 & 2.17@xmath45 & 34.79 & 1.98(87 ) + a-66 - 00 & 13.69@xmath46 & 0.53@xmath45 & 7.97@xmath47 & 6.88@xmath48 & 2.27@xmath49 & 558.98 & 0.97(87 ) + a-66 - 01 & 14.22@xmath50 & 0.54@xmath45 & 7.32@xmath51 & 7.26@xmath52 & 2.17@xmath53 & 567.59 & 0.98(81 ) + b-04 - 00 & 14.53@xmath54 & 0.52@xmath45 & 9.77@xmath55 & 6.01@xmath56 & 2.35@xmath57 & 721.83 & 1.40(87 ) + b-04 - 05 & 14.73@xmath58 & 0.52@xmath45 & 6.73@xmath59 & 8.08@xmath60 & 2.52@xmath61 & 453.99 & 0.93(87 ) + b-04 - 08 & 13.10@xmath62 & 0.54@xmath45 & 7.02@xmath63 & 7.83@xmath64 & 2.46@xmath65 & 575.62 & 0.98(87 ) + + c-01 - 02 & 14.46@xmath66 & 0.53@xmath45 & 7.88@xmath67 & 8.09@xmath68 & 2.83@xmath69 & 914.69 & 1.04(87 ) + c-01 - 01 & 13.47@xmath70 & 0.56@xmath45 & 8.85@xmath71 & 5.31@xmath72 & 2.52@xmath73 & 728.57 & 1.23(87 ) + c-01 - 00 & 14.46@xmath74 & 0.56@xmath45 & 10.86@xmath75 & 4.15@xmath57 & 2.31@xmath76 & 847.52 & 0.99(87 ) + c-01 - 03 & 13.90@xmath77 & 0.57@xmath45 & 9.21@xmath78 & 5.16@xmath79 & 2.53@xmath80 & 724.04 & 1.16(87 ) + c-01 - 04 & 14.41@xmath81 & 0.57@xmath45 & 8.35@xmath82 & 5.18@xmath83 & 2.58@xmath84 & 830.45 & 0.93(87 ) + c-01 - 05 & 12.97@xmath85 & 0.59@xmath45 & 8.23@xmath86 & 5.39@xmath87 & 2.56@xmath88 & 605.30 & 0.68(87 ) + c-01 - 06 & 14.39@xmath89 & 0.56@xmath45 & 7.12@xmath90 & 6.29@xmath91 & 2.71@xmath92 & 835.85 & 1.02(87 ) + + + + mjd & date & peak flux & decay time & integrated@xmath93 & duration & ave . gbi & ave . asm & asm & no . of + & & ( mjy ) & ( day ) & flux ( jy.day ) & ( day ) & flux ( mjy ) & count s@xmath94 & variability ( % ) & pca obs . + 50751 & 1997 oct 31 & 550 & 3.16 & 1.65@xmath95 & 19 & 50.3 & 35.9 & 5.44 & 5 + 50916 & 1998 apr 13 & 920 & 3.01 & 2.34 & 6 & 90.1 & 48.2 & 2.70 & 4 + 50933 & 1998 apr 30 & 580 & 3.82 & 2.14 & 6 & 91.4 & 44.9 & 3.31 & 2 + 50967 & 1998 jun 03 & 710 & 2.13 & 1.37 & 12 & 54.4 & 36.5 & 5.10 & 12 + 51004 & 1998 jul 10 & 480 & no fred & & 7 & 112.7 & 43.6 & 18.82 & + 51337 & 1999 jun 08 & 490 & 2.68 & 1.46@xmath95 & 5 & 43.3 & 35.2 & 6.05 & 7 + 51499 & 1999 nov 17 & 510 & no fred & & 8 & 82.8 & 69.2 & 16.60 & + 51535 & 1999 dec 23 & 510 & 3.86@xmath96&1.67 & 10 & 52.3 & 38.9 & 3.96 & 7 + 51577 & 2000 feb 03 & 420 & no fred & & 12 & 69.8 & 30.0 & 11.50 & 4 + + + + + cccccccc + obsid@xmath28 & date & start & @xmath29 & + & & time ( ut ) & ( hz ) & f@xmath30 & f@xmath31 & f@xmath32 & f@xmath33 + p-51 - 00 & 1997 oct 22 & 07:04:48 & 1.401 & 2.012 & 0.175 & 0.437 & 1.301 + p-52 - 000 & 1997 oct 25 & 06:30:40 & 1.882 & 2.042 & 0.245 & 0.412 & 1.258 + q-09 - 01 & 1998 apr 10 & 00:02:24 & 2.175 & 2.439 & 0.313 & 0.826 & 1.134 + q-10 - 00 & 1998 apr 11 & 09:28:32 & 1.739 & 2.423 & 0.291 & 0.729 & 1.247 + r-15 - 00 & 1998 apr 22 & 21:16:00 & 1.458 & 2.443 & 0.265 & 1.029 & 0.999 + r-16 - 00 & 1998 apr 28 & 16:16:48 & 1.399 & 2.335 & 0.288 & 1.044 & 0.828 + r-20 - 00 & 1998 may 24 & 19:27:28 & 0.692 & 2.131 & 0.165 & 0.876 & 0.993 + r-21 - 00 & 1998 may 31 & 19:25:20 & 1.763 & 2.144 & 0.252 & 0.846 & 0.912 + s-16 - 02 & 1999 jun 03 & 03:02:24 & 1.759 & 1.978 & 0.229 & 0.727 & 0.919 + s-17 - 01 & 1999 jun 07 & 02:52:32 & 2.874 & 2.295 & 0.313 & 0.664 & 1.182 + s-42 - 00 & 1999 dec 13 & 11:56:32 & 1.435 & 2.137 & 0.245 & 0.680 & 1.065 + s-43 - 00 & 1999 dec 21 & 11:32:16 & 2.121 & 2.231 & 0.253 & 0.710 & 1.158 + + obsid@xmath28 & n@xmath18 & kt@xmath21 & kt@xmath34 & @xmath35 & @xmath36 & r@xmath21 @xmath37 & @xmath38(dof ) + & ( 10@xmath39 @xmath40 ) & ( kev ) & ( kev ) & & & ( km ) & + p-51 - 00 & 13.27@xmath97 & 0.48@xmath45 & 4.33@xmath98 & 11.90@xmath99 & 2.75@xmath100 & 787.22 & 1.34(94 ) + p-52 - 000 & 14.68@xmath101 & 0.48@xmath45 & 4.13@xmath102 & 11.49@xmath103 & 2.76@xmath104 & 1000.98 & 1.07(94 ) + q-09 - 01 & 12.39@xmath105 & 0.51@xmath45 & 3.92@xmath106 & 9.00@xmath107 & 2.79@xmath108 & 755.98 & 1.05(94 ) + q-10 - 00 & 14.95@xmath109 & 0.46@xmath45 & 4.04@xmath110 & 9.65@xmath111 & 2.97@xmath112 & 1392.43 & 0.83(94 ) + r-15 - 00 & 11.12@xmath113 & 0.52@xmath45 & 4.20@xmath114 & 9.07@xmath115 & 2.72@xmath116 & 607.12 & 1.25(94 ) + r-16 - 00 & 13.55@xmath117 & 0.50@xmath45 & 4.18@xmath118 & 8.75@xmath119 & 2.50@xmath120 & 831.36 & 1.43(94 ) + r-20 - 00 & 14.23@xmath121 & 0.50@xmath45 & 5.20@xmath122 & 8.66@xmath123 & 2.61@xmath124 & 691.93 & 1.07(94 ) + r-21 - 00 & 12.99@xmath125 & 0.51@xmath45 & 4.96@xmath126 & 8.46@xmath127 & 2.69@xmath128 & 741.65 & 0.93(94 ) + s-16 - 02 & 12.43@xmath129 & 0.51@xmath45 & 5.98@xmath130 & 7.68@xmath131 & 2.52@xmath132 & 660.93 & 1.16(88 ) + s-17 - 01 & 13.12@xmath133 & 0.50@xmath45 & 5.58@xmath134 & 7.90@xmath135 & 2.75@xmath136 & 895.49 & 0.93(87 ) + s-42 - 00 & 12.94@xmath137 & 0.51@xmath138 & 4.29@xmath139 & 10.93@xmath140 & 2.69@xmath141 & 710.50 & 1.00(88 ) + s-43 - 00 & 13.61@xmath142 & 0.48@xmath138 & 5.25@xmath143 & 8.29@xmath144 & 2.82@xmath145 & 981.34 & 0.81(88 ) + + + +

we outline the basic questions that are being studied in the theory of entanglement . following a brief review of some of the main achievements of entanglement theory for finite - dimensional quantum systems such as qubits , we will consider entanglement in infinite - dimensional systems . asking for a theory of entanglement in such systems under experimentally feasible operations leads to the development of the theory of entanglement of gaussian states . results of this theory are presented and the tools that have been developed for it are applied to a number of problems .

quant-ph0312071

when john bell , then a theoretical physicist at cern , published his now famous theorem in 1964 , it was at first hardly noticed by the scientific community . only within the subsequent decade the fundamental character of its content was truly appreciated @xcite : in a sense , it fleshed out the notion of entangled quantum systems being potentially correlated in a way that classical systems can not . entanglement can indeed be viewed as the furthest and most radical departure of quantum mechanics from classical physics . in recent years , yet another shift in the way quantum correlations are perceived has taken place : entanglement has been recognized as a resource that may be exploited to perform completely novel information processing tasks or enhance the efficiency of known tasks . as for any resource , a detailed understanding of its properties and the resulting possibilities and limitations for its manipulation are an important pre - condition for the full exploitation of its potential . quite naturally , the systematic investigation of quantum entanglement in particular is a major goal of theoretical research in quantum information science @xcite . this program of research has been initiated with the investigation of the entanglement properties of bi - partite quantum systems prepared in pure quantum states @xcite . today , most characteristics of pure state entanglement are in fact well understood @xcite . however , our understanding of mixed state entanglement is much less complete . most proven results are restricted to situations where the constituent parts are qubits , quantum two - level systems , which is due to a whole range of special mathematical properties only satisfied by qubit systems @xcite . in contrast , the study of bi - partite entanglement between higher - dimensional systems is much more involved . indeed , new types of entanglement emerge such as bound entanglement @xcite , and one encounters a rich and complex proliferation of types of entanglement with growing dimension of its constituent parts . considering the complex structure that is emerging even in the case of low - dimensional quantum systems , one may be tempted to come to the conclusion that to say anything meaningful about the entanglement of infinite - dimensional quantum systems such as field modes of light or harmonic oscillators is an enterprise doomed to failure . this is , however , not so : it has become clear in recent years that generic statements on the entanglement of such quantum systems can indeed be made . most importantly , for a large class of states many exact results on entanglement theory can be found , even in cases where there is no finite - dimensional counterpart . a brief description of the development of this theory will be the main subject of this article . it should be noted that in this article , a certain emphasis is put on the results that the research groups at imperial college and in potsdam have contributed to this area @xcite . this article will be structured in the following way . after a short introduction to some key notions of entanglement that are valid for any dimensions we will then move on to outline the key questions that any theory of entanglement will aim to answer when it addresses the resource character of quantum entanglement . some major results and definitions from the theory of entanglement for finite dimensional systems will be presented . from there we will move on to prepare the ground for the study of the entanglement properties of infinite - dimensional systems by introducing some basic notions that are helpful for the description of states and dynamics . then we will exhibit some of the problems that occur when one attempts to study entanglement of infinite - dimensional systems without any restrictions . this will serve as a motivation to restrict entanglement theory to so - called gaussian states . again , basic results are presented ; we will then ask the experimentally motivated question of the manipulation of gaussian states by gaussian operations . we will outline results that have been obtained on this question , including a no - go theorem concerning entanglement distillation of gaussian states and gaussian operations , and results on general entanglement manipulation for gaussian states . in the conclusions we will outline further developments and applications of the theory of gaussian entanglement in other areas of quantum information science . the power of entanglement manifests itself particularly clearly in quantum communication tasks that have to be performed over large distances . such tasks will typically require that , say , two parties involved in the communication protocol establish a shared entangled state of two quantum particles between them . this may for example be accomplished by one party preparing two entangled particles locally followed by the transmission of one of these particles through a physical channel such as an optical fiber to the other party . for large distances this task is complicated by the unavoidable presence of unwanted interactions with the environment causing decoherence and dissipation that will tend to destroy entanglement . the two distant parties that wish to create a shared entangled state between them will therefore generally end up with a partially entangled mixed state . consequently , in order to be able to implement an quantum communication protocol , they will need to be able to repair the partially entanglement state . this aim will be hampered by the fact that the parties are spatially separated , as this implies that they can only implement general quantum operations in their respective local laboratories and coordinate their respective actions by classical communication . global quantum operations affecting particles in separate laboratories are not available to them . the set of local operations and classical communication is usually abbreviated locc and forms the basis for much of the study of quantum entanglement , though more general classes of operations can also be relevant @xcite . the very notion of entanglement is actually tied to the set of locc : a state is called disentangled if it can be prepared using locc only . indeed , the most general two - party state that can be generated from a product state employing locc is a mixture of product states of the form @xmath0 where the tensor product refers to the parties @xmath1 and @xmath2 . such a state , usually called a separable state , may exhibit correlations , but they are not of genuinely quantum mechanical nature as , again , this state may be created using locc @xcite . a separable state allows for a description in terms of a local classical model . any state that can not be cast into the form eq.([separable ] ) ( or appropriately approximated by such a state for infinite dimensional systems ) will be called entangled . the development of a theory of entanglement , or in other words the study of what can and can not be achieved under locc , is a key concern in the field . such a theory will generally aim to answer the following three central questions : while the definition of separability is easy to formulate , it is very difficult to decide in practice whether a given state is separable or not . following eq.([separable ] ) , in order to show that a state is separable , it appears that one has to construct explicitly a decomposition of the state into tensor products . this is a very difficult and potentially lengthy task especially for high dimensional systems . for low dimensional systems , however , the separability question can be decided in a different and more efficient way using the theory of positive but not completely positive maps . in fact , a simple necessary and sufficient criterion for the separability of a quantum state can be based on the properties of the transposition and its application on a single sub - system . clearly , the transposition is a positive map in the sense that it maps any positive operator onto a positive operator , i.e. , if @xmath3 is positive then so is @xmath4 . the same then applies when the transposition is applied to one sub - system , say system @xmath2 , of a separable state , because @xmath5 is again a valid state . however , when we apply this so - called partial transposition to an inseparable state , then there is no guarantee that the result is again a positive operator , i.e. , a physical state . indeed , when applying the partial transposition to the state with state vector @xmath6 ) , we find @xmath7 , which is evidently not positive any more . this observation gave rise to the conjecture that a state is inseparable exactly if the partial transpose is not a positive operator . it has indeed been proven that this conjecture is correct for systems composed of two qubits or a qubit and a qutrit @xcite . it should be noted that the partial transposition on qubits corresponds to time reversal . this observation already suggests the appropriate application of the partial transposition criterion for infinite dimensional systems . in systems with larger dimension that six the situation is fairly different , and the question of separability is far more complicated : one encounters for example inseparable states whose partial transposition is nevertheless positive again a phenomenon commonly referred to as bound entanglement @xcite . while there is a systematic strategy for deciding the separability of a state @xcite , it is known that the problem has a computational complexity that lies in np - hard @xcite . a natural next question would be : given a quantum state of which one knows that it is entangled , how can it be manipulated by locc ? one may envision , for example , the situation that an experimental procedure generates a particular state @xmath3 , but that a state @xmath8 different from @xmath3 is actually required . is it possible to achieve the transformation @xmath9 employing locc only ? in general , this is a extraordinarily difficult question to answer , but some results have been achieved for finite dimensional systems . most progress has been made for pure quantum states for which the basic structures are known . there are two regimes which may fruitfully be considered . firstly , one may study the situation where a single copy of a quantum state is held by two parties ( the multiple copy case is a special case of this setting ) . in this case , the mathematical structure underlying the manipulation of quantum states by locc is provided by the theory of majorization , and both necessary and sufficient conditions for the interconversion between two states are known , together with protocols that achieve the task @xcite . let us write the initial and final state vectors as @xmath10 and @xmath11 in their schmidt - basis , @xmath12 where @xmath13 denotes the dimension of each of the quantum systems . we can take the schmidt coefficients to be given in decreasing order , i.e. , @xmath14 and @xmath15 . the question of the interconvertability between the states can then be decided from the knowledge of the real schmidt coefficients only . one finds that a locc transformation that converts @xmath16 to @xmath17 with unit probability exists if and only if the @xmath18 are majorized @xcite by @xmath19 , that is , exactly if for all @xmath20 @xmath21 various refinements of this result have been found that provide the largest success probabilities for the interconversion between two states by locc together with the optimal protocol where a deterministic interconversion is not possible @xcite . these results allow in principle to decide any question concerning the locc interconversion of pure state by employing techniques from linear programming @xcite . although the basic mathematical structure is well understood , surprising and not yet fully understood effects such as entanglement catalysis are still possible @xcite . secondly , further to the study of the locc transformation of single copies , a fruitful area is the study of the asymptotic regime ie the limit of a very large number of identical quantum states . while this may be viewed as a limiting case of the above theorem , it has indeed predated it . it turns out that in this case , a single number , the entropy of entanglement , will determine what is possible and what is not . the entropy of entanglement for a pure state with state vector @xmath22 is defined as @xmath23 where @xmath24 denotes the von - neumann entropy defined as @xmath25 $ ] and @xmath26 denotes the partial trace over subsystem b. with this quantity , we find that in the asymptotic limit of a large number of identically prepared systems , the transformation from @xmath27 copies of @xmath28 under locc to @xmath29 copies of @xmath30 will be possible with fidelity that approaches unity in the limit of @xmath27 approaching infinity . for mixed states no comparable set of theorems is available . indeed , even very simple questions concerning the interconvertability of states can not be provided with necessary and sufficient conditions for their availability . the quest for such conditions together with the questions of the best efficiencies for procedures of salient interest that are available has led to the development of the theory of the quantification of entanglement . an entanglement measure grasps this notion of quantifying entanglement : it is a mathematical quantity that captures the essential properties that we would associate with the amount of entanglement and that is ideally related to some operational procedure . any function @xmath31 mapping the state space on positive real numbers that deserves the name entanglement measure should satisfy the subsequent list of requirements : @xcite 1 . @xmath32 vanishes if the state @xmath3 is separable . 2 . @xmath31 is invariant under local unitary transformations , i.e. , local basis changes . 3 . @xmath31 does not increase on average under locc , i.e. , @xmath33 where in a locc the state @xmath34 with label @xmath35 is obtained with probability @xmath36 . 4 . for pure state @xmath37 the measure reduces to the entropy of entanglement @xmath38 a function @xmath31 mapping state space on positive numbers that satisfies the first three conditions is called an entanglement monotone while an entanglement measure also satisfies condition ( 4 ) . often , convexity of @xmath31 or asymptotic continuity are taken as additional reasonable properties ( see , e.g. , ref . in fact , conditions ( 13 ) together with asymptotic continuity uniquely specify the entanglement measure to be the entropy of entanglement on pure states ( 4 ) , up to a real number that is merely a scaling of the quantity @xcite . these conditions alone , yet , do not uniquely specify a measure of entanglement on mixed states . the extremal measures @xcite giving bounds for all others are the distillable entanglement @xcite , quantifying the degree to which maximally entangled states can be extracted under locc from a given supply of identically prepared states , and the entanglement cost @xcite , quantifying the resources in terms of maximally entangled states necessary in a preparation procedure of a state under locc . the entanglement cost is in fact nothing but the asymptotic version of the entanglement of formation @xcite . the relative entropy of entanglement quantifies to which extent a given state can be distinguished from a separable state @xcite . in its asymptotic version @xcite it provides moreover a tight upper bound for distillable entanglement . practically most straightforward to use is probably the negativity @xcite , which is known to be an entanglement monotone @xcite , and whose logarithm ( up to a constant ) has an interpretation as an asymptotic entanglement cost @xcite . the very recently developed squashed entanglement @xcite is a convex entanglement monotone that is moreover additive , a property that it shares with the relative entropy of entanglement with reversed arguments @xcite , but it remains finite on pure states . the same set of questions namely , characterization , manipulation and quantification will have to be addressed in the development of a theory of entanglement . in the remainder of this article we will outline what has been achieved in this respect in the infinite dimensional setting . all the above results concern bi - partite entanglement for systems comprising finite dimensional sub - systems such as qubits or qutrits . more recently , however , considerable effort has been directed towards the study of infinite dimensional quantum systems such as the photon number degree of freedom of light modes @xcite or nano - mechanical harmonic oscillators @xcite , or the state of cold atomic gases @xcite . at first sight , one would expect the theory of entanglement to become extraordinarily complicated due to the fact that now the hilbert space is no longer finite - dimensional . indeed , without imposing certain restrictions on the set of states under consideration , one even loses such elementary properties as the continuity of entanglement and its measures @xcite and questions such as distillability become difficult to answer @xcite . in this subsection we will demonstrate some of the counterintuitive properties @xcite of entanglement measures in the infinite dimensional setting and use them to motivate possible restrictions on the set of states that one wishes to consider . to clarify this issue , let us again consider the joint system to be bi - partite , consisting of parts labeled @xmath1 and @xmath2 , each of which having a finite number of degrees of freedom . both @xmath39 and @xmath40 of the joint hilbert space @xmath41 are taken to be infinite - dimensional , as in case of two modes of light . we will consider trace - norm continuity , where the trace norm @xmath42 is defined as @xmath43= \mbox{tr}[(a^\dagger a)^{1/2}]$ ] for trace - class operators @xmath1 @xcite . it is now quite straightforward to see that the entropy of entanglement ( [ entropyofentanglement ] ) is no longer trace - norm continuous . the following sequence of states exemplifies this property of the entropy of entanglement : the entropy of entanglement may be very different from zero or even infinite for states that are arbitrarily close to pure product states . in this example , let @xmath44 with @xmath45 as a sequence of pure states @xmath46 defined by @xmath47 and @xmath48 . here , @xmath49 and @xmath50 are orthonormal basis that are dense in @xmath51 and @xmath52 , respectively . the sequence of states @xmath53 in fact converges to @xmath54 in trace - norm , i.e. , @xmath55 while @xmath56 . obviously , @xmath31 is not continuous around the state @xmath54 . in fact , the states with infinite von - neumann entropy is trace - norm dense in state space @xcite . as can readily be verified , the set of pure states with infinite entropy of entanglement is also dense in the set of all pure states , and hence , the entropy of entanglement is almost everywhere infinite . that is to say , the assignment of a value which represents the degree of entanglement to an infinite - dimensional quantum system is not entirely unambiguous . this apparent counterintuitive pathology can nevertheless be tamed to a significant degree . the key observation here is concerned with the mean energy of the states . let @xmath57{0.3ex}{0.1ex } } } } + { { { \sf 1 \hspace{-0.3ex } \rule{0.1ex}{1.52ex } \rule[-.02ex]{0.3ex}{0.1ex } } } } \otimes h_b$ ] be the hamiltonian of the bi - partite quantum system . it may appear to some extent unusual in the field of quantum information theory to refer to the hamiltonians of the actual physical systems carrying the quantum states . we do not require much from @xmath58 at all , however , only that for any finite temperature @xmath59 we have @xmath60<\infty\ ] ] for all @xmath61 where @xmath62 . this is a very natural demand on the spectrum on @xmath58 . it merely means that there can be no limiting points in the spectrum of @xmath58 : what it physically implies is that the gibbs state , the state of the canonical ensemble , exists . systems such as the photon number degree of freedom of field modes of light , have this property . what are the implications for the discontinuity of the entropy of entanglement ? we observe with @xmath63 that in the above example the mean energy of the sequence of states diverges , @xmath64 & = & \lim_{k\rightarrow\infty}\sum_{n=1}^{k } \frac{1}{\log(n)^{2}}= \infty.\end{aligned}\ ] ] this divergence of the mean energy is actually generic for sequences of states with divergent entropy of entanglement @xcite , and is hence not an accident . this observation suggests how to tame the unwieldy infinities . after all , the energy that can be invested in the preparation of a state by means of physical devices is in all instances limited . this does not mean that the resulting states have a finite - dimensional carrier . it only means that their _ mean _ energy must be bounded . in fact , it turns out that for the set of states with mean energy which is bounded from above by a positive number @xmath65 , @xmath66 \leq { e_{max}}\},\ ] ] the entropy of entanglement regains its ( trace - norm ) continuity @xcite . here , @xmath67 denotes the state space , i.e. , the set of all density matrices , which is the set of all positive normalized trace - class operators . the set eq . ( [ energy ] ) is technically speaking nowhere dense in state space as @xmath58 is an unbounded operator , but it is a reasonable subset of the state space : it simply reflects the practically natural requirement that the mean energy is bounded from above . moreover , restricting the mean energy of the states under consideration recovers a number of asymptotic continuity properties of the entropy of entanglement . this strengthens the interpretation as the entanglement cost and the distillable entanglement also in the infinite - dimensional setting @xcite . similar conclusions can be reached for the entanglement of formation and the relative entropy of entanglement @xcite . in a nutshell , one may say that on subsets of state space corresponding to bounded mean energy , entanglement measures often regain trace - norm continuity . nevertheless , one has to keep in mind that the actual minimization problems involved in the evaluation of meaningful entanglement measures are often not feasible in the infinite - dimensional setting . this may again be taken as a very discouraging observation . however , in many situations of practical interest , where the semantics of the term practical may range from meaningful in mathematical physics to preparable in a quantum optical experiment , it turns out that one does not encounter all possible states from state space . instead , a most relevant class of quantum states can be described in very simple terms using ( small ) finite - dimensional matrices only , without the need of a description that is overburdened with the technicalities of infinite - dimensional hilbert spaces : this is the set of gaussian quantum states . this set of states that is of utmost practical importance will be dealt with in most of the remainder of the present article . gaussian quantum states play a key role in several fields of theoretical physics . in quantum optics , for example , they are often encountered as states of field modes of light , for reasons that will be elaborated on below . ground states of systems with canonical coordinates ( position and momentum ) where the hamiltonian is quadratic in the positions and momenta are also gaussian states @xcite . the term has been coined because the defining property is that the characteristic function associated with the state is a gaussian in phase space . for such gaussian states the theory of entanglement as well as a framework of how these states may be manipulated is well developed . typical questions in a theory of entanglement , e.g. , concerning the separability of given states , or the interconvertability of pairs of states under local operations can often be answered in full . this is true even in cases where the finite dimensional equivalent , whenever such an equivalent can be formulated , remains unsolved . one key reason for these successes is the fact that gaussian states are completely specified by their first and second moments so that questions concerning properties of gaussian states can be translated into properties of ( comparatively small ) finite - dimensional matrices . therefore the instruments of matrix theory @xcite , so useful in the study of entanglement for finite - dimensional systems such as qubits and qutrits , once more become a useful tool . the systems that will subsequently be discussed are quantum systems with @xmath13 canonical degrees of freedom . these could represent @xmath13 harmonic oscillators , or @xmath13 field modes of light . the canonical commutation relations ( ccr ) between the @xmath68 canonical self - adjoint operators corresponding to position and momentum of such a system with @xmath13 degrees of freedom , may be written in a particularly convenient form employing the row vector , @xmath69 in terms of the familiar creation and annihilation operators of the modes choosing @xmath70 , @xmath71 and @xmath72 can be expressed as @xmath73 then the canonical commutation relations ( ccr ) take the form @xmath74=i \sigma_{j , k } \ ; , \ ] ] where the skew - symmetric block diagonal real @xmath75-matrix @xmath76 given by @xmath77,\ ] ] is the so - called symplectic matrix . the phase space , isomorphic to @xmath78 , then becomes what is known as a symplectic vector space , equipped with the scalar product corresponding to this symplectic matrix . instead of referring to states , i.e. , density operators , one may equivalently refer to functions that are defined on phase space . there are many common choices of such functions in phase space , such as the wigner function , the @xmath79-function or the @xmath80-function , to name just a few @xcite . each of them is favorable in a particular physical context . for later purposes it is most convenient to introduce the characteristic function , which is the fourier transform of the wigner function . using the weyl operator @xmath81 for @xmath82 , we define the ( wigner- ) characteristic function as @xmath83.\ ] ] in quantum optics , the weyl operator is typically referred to as phase space displacement operator or glauber operator , but with a different convention concerning its arguments . for a single mode , let the complex number @xmath84 be defined as @xmath85 , then the phase space displacement operator @xmath86 of quantum optics @xcite is most commonly taken as @xmath87 . each characteristic function is uniquely associated with a state , and they are related with each other via a fourier - weyl relation . the state @xmath3 can be obtained from its characteristic function according to @xmath88 in turn , the wigner function as commonly used in quantum optics is related to the characteristic function via a fourier transform , i.e. , @xmath89 gaussian states are , as mentioned before , defined through their property that the characteristic function is a gaussian function in phase space @xcite , i.e. , @xmath90 where @xmath91 is a @xmath75-matrix and @xmath92 is a vector . as a consequence , a gaussian characteristic function can be characterized via its first and second moments only , such that a gaussian state of @xmath13 modes requires only @xmath93 real parameters for its full description , which is polynomial rather than exponential in @xmath13 . the first moments form a vector , the displacement vector @xmath94 , where @xmath95,\ ] ] @xmath96 . they are the expectation values of the canonical coordinates , and are linked to the above @xmath97 by @xmath98 . they can be made zero by means of a translation in phase space of individual oscillators . as a consequence the first moments do not carry any information about the entanglement properties of the state . the second moments are embodied in the real symmetric @xmath75 covariance matrix @xmath99 defined as @xmath100 \ , .\end{aligned}\ ] ] with this convention , the covariance matrix of the @xmath13-mode vacuum is simply @xmath101{0.3ex}{0.1ex } } } _ { 2n}$ ] . again , the link to the above matrix @xmath91 is @xmath102 . clearly , not any real symmetric @xmath75-matrix can be a legitimate covariance of a quantum state : states must respect the heisenberg uncertainty relation . in terms of the second moments the latter can be phrased in compact form as the matrix inequality @xmath103 in turn , for any real symmetric matrix @xmath99 satisfying the uncertainty principle ( [ res ] ) there exists a gaussian state the second moments of which are nothing but @xmath99 . so eq.([res ] ) implies the only restriction on legitimate covariance matrices of gaussian quantum states . this observation has quite significant implications concerning the question of separability of two - mode gaussian states shared by two parties : it has earlier been pointed out that partial transposition is a positive , but not completely positive map . hence , partial transposition must map separable states onto separable states , but there exist states for which the partial transpose is no longer positive . in fact , for bipartite qubit systems the positivity of the partial transpose is a necessary and sufficient criterion for separability . a necessary condition for separability of gaussian states can be formulated immediately , once it is understood how partial transposition is reflected on the level of covariance matrices . indeed , it has been pointed out earlier that the partial transposition on qubits is in fact time - reversal . time reversal in a system with canonical degrees of freedom is characterized by the transformation that leaves the positions invariant but reverses all momenta , @xmath104 let us now consider a system made up of @xmath68 oscillators , where @xmath13 are held by each party . applying the time reversal operation to the @xmath13 oscillators held by one of the parties , the covariance matrix will be transformed to a real symmetric matrix @xmath105 given by @xmath106 , with @xmath107 .\ ] ] the matrix @xmath108 is the matrix collecting the second moments of the partial transpose @xmath109 of @xmath3 . as the positivity of @xmath109 is equivalent to @xmath108 satisfying the heisenberg uncertainty principle , one may now merely check whether @xmath110 if @xmath111 is not positive , then the state @xmath3 must in fact be entangled . for two - mode systems , one mode held by each party , the criterion ( [ crit ] ) is both necessary and sufficient for separability of the gaussian state . one direction of the proof has been exemplified above , the converse has been proven in refs . hence , the two - mode gaussian state with covariance matrix @xmath99 is separable exactly if the covariance matrix @xmath112 obtained via partial time - reversal in one sub - system also represents a valid physical state , i.e. , if @xmath113 . it is remarkable that the above theorem can be extended considerably . in fact , for a bi - partite system the partial time - reversal also provides a necessary and sufficient criterion for the distillability of a bi - partite gaussian state of two party holds one mode and the other party holds @xmath27 modes , positivity of the partial transpose is still a necessary and sufficient criterion for separability @xcite . this result together with the discovery of bound entanglement in gaussian systems rely on an alternative criterion for the separability of gaussian states . in fact , a gaussian state represented by the covariance matrix @xmath99 is separable if and only if there exist covariance matrices @xmath114 and @xmath115 such that @xmath116 this formulation is a very useful tool to prove the validity of statements , but does not serve as a practical criterion for separability , as it is generally not straightforward to find the matrices @xmath114 and @xmath115 ( but note that this may be checked in form of a semi - definite program ) . in general , given two parties , one holding @xmath117 and the other @xmath27 oscillators , both the question of separability @xcite and distillability can be decided efficiently , although the questions of distillability requires a more sophisticated approach @xcite . these results are very encouraging but they come with one caveat . while the states have been restricted to be gaussian , no such restriction has been applied to the set of allowed operations . the result concerning the distillability for example only holds when arbitrary operations in infinite - dimensional systems are permitted , even practically very complicated ones such as the distinction between @xmath118 and @xmath119 photons in field modes of light . in the light of this observation , one should not forget that much of the significance of the gaussian states stems from the fact that the gaussian operations are so important . this is by no means a tautology : the gaussian operations are those operations that change the state , but preserve the gaussian character of any gaussian input state . this set of operations is singled out from both a theoretical perspective , but most significantly also from the perspective of practical experimental implementation . a most relevant question is then what tasks can be accomplished with gaussian states , but not under _ all _ physically available operations , but under gaussian operations only . this observation motivates the approach described in the next section , namely the investigation of what can and can not be achieved when one restricts attention to gaussian states and the so - called gaussian operations . before we can begin the development of the theory of entanglement of gaussian states under gaussian operations , we need to characterize both gaussian states ( already done in the earlier sections ) and gaussian operations . the most straightforward definition of gaussian operations simply states that an operation is gaussian exactly if it will map every gaussian input state onto a gaussian output state . the remainder of this section will present more practically useful characterizations of gaussian operations and in particular it will show that gaussian operations correspond exactly to those operations that can be implemented by means of optical elements such as beam splitters , phase shifts and squeezers together with homodyne measurements @xcite all operations that are in principle experimentally accessible with present technology @xcite . let us start with gaussian unitary transformations . the most general real linear transformation s which implements the mapping @xmath120 will have to preserve the canonical commutation relations @xmath121=i\sigma_{jk}{{\sf 1 \hspace{-0.3ex } \rule{0.1ex}{1.52ex } \rule[-.02ex]{0.3ex}{0.1ex } } } \,.$ ] this is the case exactly if the linear map @xmath24 satisfies @xmath122 the set of real @xmath75 matrices @xmath24 obeying this condition form the so - called real symplectic group @xmath123 @xcite , whose elements are called symplectic or canonical transformations . to any symplectic transformation @xmath24 also @xmath124 are symplectic . the inverse of @xmath24 is given by @xmath125 and the determinant of every symplectic matrix is @xmath126 = 1 $ ] @xcite . every symplectic transformation corresponds to a unitary operation acting on the corresponding state space . as a consequence of the stone - von neumann theorem , given a real symplectic transformation @xmath24 there exists a unique unitary transformation @xmath127 acting on the state space such that the weyl operators satisfy @xmath128 for all @xmath129 . on the level of covariance matrices @xmath99 of an @xmath13-mode system a symplectic transformation @xmath24 is reflected by a congruence @xmath130 it is instructive to consider the generators of the gaussian unitary transformations . these infinitesimal generators can be found by applying the condition that the canonical commutation relations ( on the level of weyl operators ) are preserved to an infinitesimal gaussian unitary transformation . we consider @xmath131{0.3ex}{0.1ex } } } - i\epsilon g + o(\epsilon^2)$ ] , for a real @xmath132 and arrive at the conclusion that the generators @xmath133 have the form @xmath134 , which is a hermitian quadratic expression in the canonical operators ( and thus in the annihilation / creation operators ) . as a consequence , every hamiltonian @xmath133 can in quantum optical settings be implemented using beam - splitters , phase plates and mirrors only . conversely , every array of passive optical elements is described by such a hamiltonian . a particularly important subset of all symplectic transformations is formed by those @xmath135 that are moreover orthogonal , i.e. , @xmath136 . @xmath137 is again a group , and a compact one . the elements of @xmath137 correspond to the passive transformations @xcite . in optical systems , these particular gaussian unitary operations are those that can be realized using beam splitters and phase shifts , but excluding squeezers . such a passive transformation does not change the eigenvalues of the covariance matrix and in an optical systems does not change the total photon number . hence , it can never transform a state that is not squeezed into a squeezed one . a state is called squeezed , if one of the eigenvalues of the covariance matrix @xmath99 is smaller than unity @xcite . in more physical terms this means that using passive transformations only the state can be transformed into a state such that the uncertainty with position is smaller than the respective uncertainty of the vacuum state . any @xmath138 can be decomposed into @xmath139 l,\ ] ] with @xmath140 and @xmath141 . this is the euler decomposition of symplectic transformations . physically , this means that any gaussian unitary transformation can be realized as a ( i ) passive transformation , a ( ii ) single - mode squeezing operation on each of the @xmath13 modes , and a ( iii ) subsequent second passive transformation . noisy channels , such as those corresponding to optical fibers in optical systems , can only be modeled as an irreversible quantum operation . a gaussian channel @xcite is such an irreversible quantum operation , @xmath142 . it is reflected by a transformation of the covariance matrix according to @xmath143 where @xmath144 are @xmath75-matrices , and @xmath133 is moreover symmetric . the complete positivity of the quantum operation is reflected by the condition @xmath145 such channels emerge naturally if a system is coupled to degrees of freedom that can not be controlled . often , this interaction is of gaussian nature , as the hamiltonian is a polynomial of second degree in the canonical coordinates @xcite . as the environment degrees of freedom can not be accessed , one considers a partial trace with respect to them . optical fibers manifest themselves for example as gaussian channels of this type for field modes @xcite . then , the overall dilation is reflected by a gaussian completely positive map with a transformation rule for the second moments according to eq.([channel ] ) . note that there are no restrictions on the matrix @xmath1 : this means that if one aims to implement an ( impossible ) operation with a transformation of the form @xmath146 , eq . ( [ channel ] ) together with ( [ cp ] ) show the fundamental limitations imposed by the ccr on this operation . this is relevant in the context of , for example , quantum amplifiers , attenuators , or phase conjugating mirrors approximately realising a time reversal operation . also , quantum cloning @xcite and secret sharing @xcite have been formulated as a gaussian channel . in turn , one may ask , whether it is true that any channel of this form can be realized in three steps : ( i ) appending an additional system prepared in a gaussian state ( ancillae ) , ( ii ) a gaussian unitary operation including all modes , and ( iii ) a partial trace with respect to the environment ( discarding the ancillae ) ? the structure of such a dilation has been depicted in fig . [ figure4 ] . clearly , any such operation is reflected by a transformation of the form eq . ( [ channel ] ) . the converse is not quite true , and one might have to add classical gaussian noise , by displacing the state in phase space with a classical gaussian distribution . this is reflected by the addition of a positive matrix to the covariance matrix . furthermore , one might after all perform selective measurements which preserve the gaussian character of the state . let us assume that @xmath13 modes of light are entangled with a distinguished @xmath147st mode . if one projects the mode with label @xmath147 onto the vacuum state with state vector @xmath148 , then the resulting state of the remaining @xmath13 modes is still a gaussian state . but what are its second moments ? if the original covariance matrix is of the form @xmath149,\ ] ] where @xmath1 is an @xmath75-matrix and @xmath2 is a @xmath150 matrix corresponding to the distinguished mode , then the resulting covariance matrix can be determined by a gaussian integration @xcite @xmath151{0.3ex}{0.1ex } } } } _ { 2n})^{-1 } c^t,\ ] ] which is often referred to as the schur complement of the matrix @xmath152{0.3ex}{0.1ex } } } } _ { 2n } \end{array } \right].\ ] ] note that this transformation is not of the form as specified in eq . ( [ channel ] ) . now , it follows how an ideal homodyne detection is reflected on the level of second moments . the resulting covariance matrix does not depend on the outcome of the measurement @xcite . it is again a schur complement of a matrix . indeed , with @xmath153 we have @xmath154 where the inverse is replaced by the moore - penrose pseudo - inverse , as the matrix is apparently no longer invertible . eq.([homo ] ) gives the transformation law for the covariance matrix of @xmath13 modes , after the mode with label @xmath147 has undergone an ideal homodyne detection . in fact , any gaussian operation can be thought of as a concatenation from ingredients of this above list @xcite , i.e. , any physical operation mapping all gaussians onto gaussians can be thought of as a sequence of gaussian channels ( including gaussian unitary operations ) , together with homodyne detection . this notion can be stated most clearly in terms of the isomorphism between positive operators and completely positive maps @xcite . for gaussian operations this isomorphism @xcite leads to the following form of a mapping of the general gaussian operation , on the level of second moments , @xmath155 where we have denoted @xmath156 , \ ] ] and @xmath157 , with @xmath158 being defined as in eq.([f ] ) with appropriate size . all gaussian completely positive maps on @xmath13 modes are characterized by a @xmath68-mode covariance matrix @xmath159 and a displacement vector @xmath160 @xcite . equipped with the transformation laws under these operations we may now turn to assessing what state transformations can be done in the gaussian setting . in the last section we have presented useful characterizations of gaussian operations . these characterizations form the essential tools for the study of the gaussian local manipulation of gaussian states and play important roles in many proofs in this area . we will now proceed with an exposition of known results in the area that have been discovered in the past few years . it should be noted however , that we will not proceed along the historical lines of development . while the program of the study of gaussian states under gaussian operations has originated in ref . @xcite for mixed gaussian states , we will begin with an exposition of the state of knowledge for pure states . as can be expected , the most complete understanding of the behaviour of gaussian states under gaussian operations has been reached for pure states . indeed , necessary and sufficient conditions for the possible state transformations under local gaussian operations ( glocc ) can be given . local means here that the applied gaussian quantum operations are applied locally , accompanied by classical communication . this result crucially depends on a normal form that can be obtained for any bipartite system of continuous variables @xcite . for any pure gaussian state with covariance matrix @xmath99 of an @xmath161-mode system , there exist local symplectic transformations @xmath162 such that @xmath163\nonumber\end{aligned}\ ] ] with @xmath164 . in other words , by means of local unitary gaussian operations the pure gaussian state described by @xmath99 can be transformed into a tensor product of two - mode squeezed states @xmath165 characterised by squeezing parameters @xmath166 ( see fig.[figure3 ] ) . = as a consequence the non - local properties of a pure gaussian state of @xmath167 modes can be characterised by a vector @xmath168 of two - mode squeezing parameters , which we will always assume to be given in descending order . this vector will play a role closely analogous to that of the schmidt - coefficients in the majorization criterion for pure state entanglement transformations under general locc operations . the criterion for pure - state transformations under glocc can now be stated in a very simple manner in terms of this vector of two - mode squeezing parameters . we write @xmath169 iff @xmath170 for all @xmath171 , in descending order . then , the criterion for the possibility of transforming the initial to the final state can be expressed in an extraordinarily simple manner : we have that @xmath172 = it should be noted that the criterion for pure state transformations under gaussian local operation has a simpler structure than the criterion that determines which state transformations are possible under general locc , not restricted to gaussian transformations . as already stated in earlier sections , we have that @xmath173)\prec \lambda(\mbox{tr}_{a}[\rho']).\ ] ] where @xmath174 denotes the majorisation relation eq.([majorization ] ) . in other words , a pure state @xmath3 can be transformed into another pure state @xmath8 if and only if the reduction to one part of the joint system is initially more mixed than finally in the sense of majorisation theory . this criterion for finite - dimensional systems can be immediately carried over to the infinite - dimensional case , when appropriate care is taken when extending the definition of locc convertibility : one then writes @xmath175 under locc if there exists a sequence of locc , such that the images of the maps approximate @xmath3 in trace - norm . it is then instructive to compare the restrictions imposed by the majorisation criterion with the constraints under gaussian locc . to start with , under locc , catalysis of entanglement manipulation is possible @xcite : there exist pure states of finite - dimensional bi - partite quantum systems @xmath3 , @xmath8 and @xmath176 such that @xmath177 the state @xmath176 serves as a `` catalyst '' , as the entanglement inherent in the state is not consumed in the course of the transformation . such an effect is not possible under glocc and pure gaussian states , as collective operations are never more powerful than operations on pairs of quantum systems at a time . this notably refers to entanglement transformations only , as one can think of procedures estimating certain properties of gaussian states where collective operations are superior to product operations . it should moreover be noted that the theorem governing interconvertability under glocc also covers the stochastic interconversion of pure gaussian under glocc , as it has been pointed out in the previous section that any gaussian transformation can be lifted to a trace - preserving operation . this is again in contrast to the finite - dimensional case @xcite . it is also an immediate but remarkable consequence of the above criterion that under gaussian local operations with classical communication , one can not `` concentrate '' two - mode squeezing of pure gaussian states . in particular , @xmath178 for any @xmath179 and any @xmath180 , where @xmath181 denotes a two - mode squeezed state with squeezing parameter @xmath182 . this is very much in contrast to the situation when general operations are allowed for : for any @xmath182 there exists a @xmath180 such that @xmath183 as has been shown in ref . @xcite . that is to say , this criterion already points towards the impossibility of distilling gaussian states with gaussian operations . for mixed states , both the number of parameters characterizing the state increases and the structure of state transformations becomes far more involved . to start with , a simple normal form such as the schmidt decomposition or the vector of two - mode squeezing parameters @xmath184 is not available . the simplest normal form that can be obtained for two harmonic oscillators is given by the simon normal form for the covariance matrix @xcite . if we are given a gaussian state @xmath3 of a two - mode system with covariance matrix @xmath99 , then there exist symplectic transformations @xmath185 such that @xmath186.\ ] ] further to the enlarged number of parameters characterizing the state , also the amount of classical communication in general gaussian local operations can not be bounded anymore as opposed to the pure state case where one round of one - way communication is sufficient . this latter fact has also been made use of in the proof of the key statement of ref . @xcite . nevertheless , under certain restrictions on the available protocols and with the help of the simon normal form it is possible to obtain some statements concerning the interconvertability of mixed states and one can find necessary and sufficient criteria for a transformation to be possible . one such restriction is that to local gaussian channels as have been discussed above : they can be realized by adding an ancilla system , followed by a joint unitary transformation and the subsequent discarding of the ancilla ( see fig . [ figure4 ] for illustration ) . for such local gaussian channels , one may write @xmath187 for gaussian two - mode states @xmath3 and @xmath8 , if there exists local gaussian channels @xmath188 and @xmath189 such that @xmath190 ( log stands for local gaussian channel ) . in fact , both necessary and sufficient conditions for the interconvertability under log can then be proven . for details we refer the reader to the literature , in particular ref.@xcite and for generalizations for example to a three - party version of the above statement @xcite . the availability of distillation protocols will be of crucial importance in the infinite - dimensional setting when we wish to implement long - distance quantum communication based on continuous variables . in fact , such distillation protocols must be performed at the beginning of any procedure that relies on the availability of highly entangled approximately pure shared entangled states . noise due to unwanted coupling to an environment is never entirely avoidable , and protocols have to be devised that effectively reverse this process . in the finite dimensional setting , distillation protocols are known , and using the polarization degrees of freedom of light , they can be implemented optically ( although they require photon counters in an iterative procedure that distinguish with great efficiency different numbers of photons ) @xcite . at the time it seemed fairly natural to expect that such distillation schemes can also be constructed in the case of gaussian states and gaussian operations . let us hence for a moment assume that pure gaussian two - mode squeezed states have been transmitted through lossy optical systems such as optical fibers . the resulting states are still gaussian , provided that the noisy channel was a gaussian channel , as is a good assumption in case of optical fibers . in order to distill the entanglement in an iterative procedure , one would take two pairs of identically prepared systems @xmath191 and would feed them into one step of the procedure : the output could then be taken as the input of the next step . the states are identical if they underwent the same losses . the corresponding modes will from now on be denoted as @xmath192 , @xmath193 , @xmath194 , and @xmath195 . a feasible iterative distillation protocol preserving the gaussian character of an input state would look as follows ( see fig . [ dgauss ] ) : * application of a local gaussian unitary operation , i.e. , a map of the form @xmath196 note that the @xmath197 denotes the tensor product between different copies , while the @xmath198 denotes the tensor product between the two parties . as has been discussed before , this set includes the passive transformations embodying those operations that can be implemented in optical systems using beam splitters and phase shifts . it however also includes squeezing operations . for the two - mode case , the set of local unitary transformation is a 20-dimensional manifold . note that it is not required that both parties implement the same operation . * a homodyne measurement on the modes @xmath193 and @xmath195 . * finally , the parties communicate classically about the outcome of the measurement , and perform postprocessing of the states of modes @xmath192 and @xmath194 with unitary gaussian operations . = it was expected that gaussian operations would be sufficient to provide for such an entanglement purification protocol . it was however proven in ref . @xcite that this is not the case . later this proof was extended to demonstrate that entanglement distillation of gaussian state with gaussian operations is not possible in general , allowing in particular for a @xmath68-mode input @xcite . this is after all quite an astonishing result : no matter how the gaussian local operation with classical operation is chosen , the degree of entanglement can not be increased . whatever operation is implemented , it will result in a loss rather than in an increase of entanglement . the optimal procedure is simple : not to even try to distill with gaussian operations alone . this remarkable result also implies that quantum error correction is not feasible if one is restricted to gaussian states and gaussian operations . hence , it becomes clear that is necessary to add some non - gaussian resource to the repertoire to ensure that interesting quantum information processing tasks can be carried out in the presence of noise . the statement has been based on the quantification on the degree of entanglement in terms of the logarithmic negativity that has been mentioned before , which is defined as @xmath199 for a state @xmath3 , where @xmath42 again denotes the trace norm , and @xmath200 is the partial transpose of @xmath3 . for gaussian states it has an interpretation in terms of a certain asymptotic entanglement cost @xcite . in the meantime , it should be noted , the entanglement of formation is also available for the case of symmetric two - mode gaussian states , i.e. , states for which the two reductions of the covariance matrix are identical up to local symplectic transformations @xcite . for pure ( and for symmetric mixed ) gaussian states the logarithmic negativity is related to the degree of squeezing in a monotone way ( see , e.g. , ref . these considerations , hence , imply again that with gaussian operations alone two identically prepared two - mode squeezed states can not be transformed into a single two - mode squeezed state with a higher degree of squeezing , as has already been pointed out in eq . ( [ maincrit ] ) . this difficulty sketched in this section the impossibility of distilling gaussian states with gaussian operations can however be overcome . it has been known for quite some time that the unlimited availability of arbitrary non - gaussian operations essentially requiring quantum computation allow for example for entanglement distillation of gaussian states @xcite to finite - dimensional singlets . this is an important result from the perspective of the theory of entanglement . yet , not only from a practical point of view it would be crucially important to find solutions that make merely use of minimal extensions of the set of operations into the non - gaussian regime , including only those additional operations that are to a high extent feasible in quantum optical settings . in a nutshell , it turns out that it _ is _ possible indeed to distill continuous - variable entanglement yielding highly entangled approximately gaussian states , and this can be done in optical systems using passive optical elements and photon detectors only . such procedures have been introduced and investigated in detail in refs . the key idea is to leave the gaussian setting in a first non - gaussian step which involves a single measurement only ( corresponding to the positive outcome of a photon detector ) , and then apply gaussian operations only . the philosophy of the protocol is sketched in the following : some source provides a two - mode squeezed state which is transmitted through fibers whose imperfections turn the states into symmetric noisy but still gaussian states . at that stage a single non - gaussian operation is carried out which takes the states out of the gaussian state space and may perhaps also increase entanglement . subsequent to that only gaussian operations are applied in an iterative scheme which serve two aims . firstly , they further increase the amount of entanglement and secondly , they make the states progressively more gaussian . asymptotically the states may then have a higher degree of entanglement than the original supply and at the same time approach gaussian states as closely as possible . = surprisingly indeed , such a scheme exists using as the only non - gaussian resource a photo - detector that can distinguish between the presence and absence of photons but is not able to count the precise number of photons ( yes - no detector ) @xcite . the first non - gaussian step takes two identical copies of two - mode squeezed states , and measurements are performed on two of the four modes . the outcome is accepted as successful in case that both detectors click ( i.e. , register at least one photon ) , corresponding to the kraus operators @xmath201{0.3ex}{0.1ex } } } -|0\rangle\langle0|$ ] . the reflectivity and transmittivities of the beam splitter @xmath202 are not @xmath203 , but have to be tuned appropriately . the following iterative gaussian procedure takes two pairs @xmath204 of non - gaussian states @xmath3 as input which are then mixed at a @xmath203 beam splitter , where the transformation induced by the beam splitters are given by unitaries @xmath205 with @xmath206 . two of the output modes are then fed into a photon detector , each associated with kraus operators @xmath207{0.3ex}{0.1ex } } } -|0\rangle\langle 0|,\ ] ] where @xmath148 denotes the state vector associated with the vacuum state . the state is kept in case of the vacuum outcome of both local detectors . the unnormalized final state after one step is hence given by @xmath208 the resulting two - mode states then form the basis of the next step . it is an iterative protocol , and it is event - ready , in the sense that one has a classical signal at hand which indicates whether the procedure was successful or not . no further post - processing has to be performed . generic weak convergence to two - mode gaussians can indeed be proven , and it can be demonstrated that the procedure often leads to highly entangled and squeezed states . remarkably indeed , the scheme turns out to be fairly robust with respect to detector imperfections @xcite . it is beyond the scope of this paper to present details of this procedure , however , and the reader is referred to the literature here . the previous considerations were concerned with the local manipulation of gaussian states . this is a meaningful approach if one has a distributed system , where the necessity of local operations is simply dictated by the set - up . but how does one prepare the entangled states in the first place ? experimentally , there are several ways to do it , and again , it would be beyond the scope of this paper to present these possibilities ( see , e.g. , ref.@xcite ) . instead , to exemplify the formalism developed before , we concentrate on the optimal creation of entanglement of gaussian states using passive transformations only @xcite . so in this subsection , the operations are allowed to be non - local with respect to the modes that have to be entangled . passive operations are , again in optical systems singled out as they correspond to those operations that can be realized with beam splitters and phase shifts . if simply the vacuum is fed into the additional input modes , then one even faces no practical problems of mode matching in optical systems . to start with , if one intends to entangle @xmath13 gaussian input modes with passive transformations , according to @xmath209 the input states have to be squeezed . this is immediately obvious : any state that is entangled must be squeezed , since otherwise , the covariance matrix @xmath210 of the @xmath13-mode output would satisfy @xmath211{0.3ex}{0.1ex } } } } _ { 2n}\oplus { { \sf 1 \hspace{-0.3ex } \rule{0.1ex}{1.52ex } \rule[-.02ex]{0.3ex}{0.1ex } } } _ { 2n},\ ] ] which in turn according to ( [ generalcrit ] ) implies separability . but what is the optimal operation , and , first and foremost , what is the optimal degree of entanglement that can be achieved in any two - mode output ? this question can actually be completely solved , when the degree of entanglement is quantified in terms of the negativity @xcite . let @xmath99 be the covariance matrix of the @xmath13-mode input state , gaussian but not necessarily pure , and @xmath212 be the resulting covariance matrix after application of the passive transformations . then , the maximum amount of entanglement obtained for any two - mode sub - system is given by @xmath213 where @xmath214 and @xmath215 are the two smallest eigenvalues of @xmath99 . this general solution of the problem establishes a generically valid link between the squeezing of the initial state to the degree of entanglement that can be potentially unlocked with passive transformations . the proof of this statement as found in ref . @xcite makes extensive use of the isomorphism between the groups @xmath137 and @xmath216 . the optimal entangling transformation can moreover be constructively found , and in the practically important case of two input modes , a formula can be given for the appropriate choices of transmittivities , reflectivities , and phases of the optical elements . in this article we have briefly reviewed some elements of the theory of entanglement both in finite and infinite dimensional systems . we have formulated the basic questions that any theory of entanglement aims to answer when we consider entanglement as a resource for quantum information processing protocols . these key questions are those of characterization , ( local ) manipulation , and quantification of the entanglement resource . we then pointed out that entanglement theory in finite dimensional systems such as qubits generally requires the availability of arbitrary local operations . statements concerning the feasibility or efficiency of a particular local state transformation may require the use of operations that are extraordinarily difficult to realize in practice . this motivated the development of a theory of entanglement under experimentally available operations and a particularly relevant example of such a theory can be found in infinite dimensional systems that are equivalent to harmonic oscillators . examples of such systems are the photonic degree of freedom of a light mode , nano - mechanical oscillators and , approximately , cold atomic gases . the restricted set of experimentally available operations that is being considered in this theory is that of gaussian operations which allow for the creation and manipulation of gaussian states . this restricted set of operations is of particular importance in quantum optics where it can be shown that gaussian operations are in one - to - one correspondence to the set of operations that can be implemented by simple optical tools such as phase plates , beam - splitters , and squeezer together with the addition of vacuum modes and homodyne detection . in the remainder of the article we then outlined the recent development of this theory and stated some of its key results . in particular it became clear that processes such as quantum error correction , entanglement distillation and efficient quantum computation are not possible under this restricted set of operations but require some resources that lie outside the gaussian regime . this suggested to search for protocols that employ non - gaussian resources in a minimal way . indeed , such extensions are possible in the sense that the expensive non - gaussian operations are used only in an initial step of the protocol and are then followed by purely gaussian operations . finally it should be mentioned that the recent development of the theory of entanglement for gaussian states has also provided us with many novel and useful techniques for the analytical study of entanglement in harmonic systems . this permits both , to revisit old questions such as for example concerning the connection between entanglement and the dynamical appearance of classical properties in quantum brownian motion @xcite , but also newly arising questions concerning the study of entanglement properties of interacting quantum systems both in the static case @xcite and in their dynamical behaviour @xcite . in these problems , the theory of gaussian entangled states often allows for the exact analytical solution of many questions and therefore provides an ideal playground for the exploration of quantum entanglement . discussions and collaborations on the subject presented here with k. audenaert , d. browne , k.banaszek , s. bose , j.i . cirac , g. giedke , ch . silberhorn , m.lewenstein , m.m . wolf , s. scheel , i.a . walmsley , and r.f.werner are gratefully acknowledged , as well as constructive remarks on the manuscript by j. anders . the research described here has been partly supported by the a .- v .- humboldt foundation , the esf program `` quantum information theory and quantum computation '' , epsrc , the european union via projects equip , quprodis , and quiprocone and a royal society leverhulme trust senior research fellowship . note that in the infinite setting different norms do not necessarily imply the same topologies . however , the choice of trace norm is quite natural as two states that are close in the trace norm will yield close expectation values for bounded observables or , in other words , the more difficult it is to operationally distinguish the states with generalized measurements in the non - asymptotic regime .

we consider photoproduction in ion ion ultraperipheral collisions ( upcs ) at the lhc and rhic in the coherent and incoherent quasielastic channels with and without accompanying forward neutron emission and analyze the role of nuclear gluon shadowing at small , , in these processes . we find that despite the good agreement between large nuclear gluon shadowing and the alice data in the coherent channel , in the incoherent channel , the leading twist approximation predicts the amount of nuclear suppression which is by approximately a factor of exceeds that seen in the data . we hypothesize that part of the discrepancy can be accounted for by the incoherent inelastic process of photoproduction with nucleon dissociation . to separate the high - photon - energy and low - photon - energy contributions to the cross section , we consider ion ion upcs accompanied by neutron emission due to electromagnetic excitation of one or both colliding nuclei . we describe the corresponding phenix data and make predictions for the lhc kinematics . in addition , in the incoherent quasielastic case , we show that the separation between the low - photon - energy and high - photon - energy contributions can be efficiently performed by measuring the correlation between the directions of and the emitted neutrons .

1312.6486

recently coherent and incoherent photoproduction of @xmath0 in ultraperipheral collisions ( upcs ) of nuclei was measured by the alice collaboration at the lhc @xcite . in the coherent channel , a large reduction of the coherent cross section approximately by a factor of three as compared to the impulse approximation has been reported . such a magnitude of the suppression was found to be in the reasonable agreement with the expectations of the approaches predicting significant nuclear gluon shadowing at @xmath5 ( @xmath1 is the fraction of the nucleus momentum carried by gluons ) , notably with predictions of the leading twist approach to nuclear shadowing @xcite and with the results of the eps09 global qcd fit to nuclear parton distributions @xcite . thus , charmonium photoproduction on nuclear targets is a useful tool to study nuclear gluon shadowing at small @xmath1 . the aim of this paper is twofold . first , we extend application of the leading twist approach to nuclear shadowing @xcite to incoherent quasielastic photoproduction of @xmath0 on nuclei and show that the suppression of both coherent and incoherent @xmath0 photoproduction in ion ion upcs can be described in the same framework . a comparison of the resulting theoretical prediction for the cross section of incoherent @xmath0 photoproduction in pb - pb upcs at the lhc to the alice data , which is also characterized to correspond to an incoherent quasielastic process @xcite , shows that the expected suppression due to nuclear shadowing is larger than that seen in the data . we argue that this does not only place additional constraints on models of nuclear shadowing down to @xmath6 but also indicates that additional processes can contribute to the alice data . in particular , on top of incoherent @xmath0 photoproduction on nuclei resulting from the target nucleus excitation , the @xmath7 process driven by the @xmath8 nucleon dissociation ( @xmath9 denotes products of the nucleon dissociation ) accompanied by the nucleus breakup into the @xmath10 system consisting of nucleus debris or nucleons also leads to the inelastic final state . the calculation of the @xmath7 contribution is rather involved reflecting different mechanisms of the elementary reaction at small and large @xmath11 considered in @xcite and will be addressed in a separate publication . second , we discuss specifics of and make predictions for coherent and incoherent charmonium production in nucleus nucleus upcs accompanied by forward neutron emission which can be studied at the lhc with the alice , cms and atlas detectors equipped by zero degree calorimeters ( zdc ) . the following channels can be studied : ( i ) one of the nuclei emits at least one neutron while its partner does not ( 0nxn ) ; ( ii ) both nuclei emit neutrons in opposite directions ( xnxn ) , ( iii ) neither of the nuclei emits neutrons ( 0n0n ) . we show that selection of a specific channel can strongly influence the ratio of the cross sections of incoherent to coherent @xmath0 photoproduction in pb - pb upcs at the lhc . in particular , we argue that the study of incoherent production of charmonium in ion ion upcs with the nucleus breakup allows one to separate the low - photon - energy and high - photon - energy contributions to nuclear @xmath0 photoproduction and , hence , to provide additional information on the dynamics of nuclear shadowing of the gluon distribution in nuclei which is complementary to that obtained from coherent onium production . this paper is organized as follows . in sect . [ sec : shadowing ] , we discuss the suppression of the coherent and incoherent nuclear @xmath0 photoproduction cross sections due to nuclear shadowing . we briefly recapitulate main results of the vector meson dominance and the color dipole models for these processes and present the derivation of the coherent @xmath12 and the incoherent @xmath13 cross sections in the leading twist approximation . in sect . [ sec : upcs ] , using the results of sect . [ sec : shadowing ] , we make predictions for the coherent and incoherent cross section of @xmath0 photoproduction in ion ion upcs without and with neutron emission and analyze the obtained results . section [ sec : conclusions ] presents a brief summary of the obtained results . the earliest model for production of vector mesons off nuclei is the vector meson dominance model based on hadronic degrees of freedom @xcite . in the high - energy optical limit of the glauber model , the standard expression for the cross section of coherent @xmath0 photoproduction on a nuclear target reads ( see , e.g. , @xcite ) : @xmath14 ^ 2 \phi_a(t_{\rm min } ) \ , , \label{sigglc}\ ] ] where we assumed that the multiple interactions leading to the nuclear shadowing effect do not distort the shape of the transverse momentum distribution of the vector meson . in eq . ( [ sigglc ] ) , @xmath15 , where @xmath16 is the nucleus form factor ( its normalization is @xmath17 ) whose square takes into account that the nucleus remains intact in the coherent process , @xmath18 is the four - momentum transfer squared and @xmath19 is its minimal value ( @xmath20 and @xmath21 are the masses of @xmath0 and the nucleon , respectively ) ; @xmath22 is the photon nucleus center of mass energy per nucleon ; @xmath23 and @xmath24 are the total @xmath0nucleus and @xmath0nucleon cross sections , respectively . note that in the first and second terms in eq . ( [ sigglc ] ) , the dependence on @xmath25 has been safely neglected compared to the @xmath26 term . in the optical limit of the glauber model , the @xmath23 cross section is : @xmath27 \ , , \label{gltsig}\ ] ] where @xmath28 is the nuclear width function ; @xmath29 is the nuclear density . equation ( [ gltsig ] ) describes successive multiple interactions of @xmath0 with target nucleons , whose destructive interference results in the nuclear attenuation ( shadowing ) @xcite of the @xmath0nucleus cross section , @xmath30 . the main issue with eqs . ( [ sigglc ] ) and ( [ gltsig ] ) is what value of the elementary @xmath24 cross section to use . it has been well known for a long time that if one tries to determine @xmath24 using the vector meson dominance model and the data on the elementary @xmath31 process , the obtained value of @xmath24 is small , namely , @xmath32 mb and @xmath33 mb , see , e.g. , @xcite . as a result , the effect of nuclear shadowing in the @xmath34 cross section predicted using eqs . ( [ sigglc ] ) and ( [ gltsig ] ) turns out to be small in the small-@xmath1 region , which contradicts the alice data . also , the smallness of @xmath24 serves as an indication that in the strong interaction , @xmath0 reveals properties of a small - size dipole built from a heavy quark - antiquark pair . the simple space time picture of heavy onium production in the vector meson dominance model is superseded by the one in high - energy qcd , where the process of charmonium photoproduction involves three stages : ( i ) the photon conversion into a @xmath35 component ( dipole ) long before the target , ( ii ) the interaction of the dipole with the target , and ( iii ) the conversion of the @xmath35 component into the final state vector meson . this space time picture is properly taken into account / realized in the framework of the qcd dipole approximation @xcite , where the large mass of the @xmath36-quark , @xmath37 , sets the hard scale in diffractive charmonium photoproduction , @xmath38 , and thus , leads to the dominance of the dipoles of the small transverse size @xmath39 , @xmath40 . the corresponding @xmath41 dipole target cross section reads @xcite : @xmath42 where @xmath43 is the running strong coupling constant ; @xmath44 is the gluon distribution in the target @xmath45 , which depends on the momentum fraction @xmath46 ( @xmath20 is the mass of @xmath0 ) and the hard scale squared @xmath47 . in general , in the dipole approach , the @xmath0 photoproduction cross section involves dipoles of different transverse sizes @xmath39 , which corresponds to different scales @xmath48 in eq . ( [ eq : sigma_dipole ] ) . however , using the leading logarithmic approximation and making simplifying assumptions about the gluon transverse momentum @xmath49 ( @xmath50 ) and the @xmath0 wave function ( assuming that the transverse momenta of @xmath36-quarks , @xmath51 , are small ( @xmath52 ) , while the longitudinal ones are equally shared between the @xmath36-quark and antiquark ) , it was shown that the imaginary part of the @xmath0 photoproduction amplitude involving the dipole cross section of eq . ( [ eq : sigma_dipole ] ) is expressed through the gluon density of the target at the scale of @xmath53 gev@xmath54 and one obtains @xcite : @xmath55 ^ 2 \ , , \label{eq : sigma_ryskin}\ ] ] where @xmath56 ; @xmath57 is the width of the @xmath0 electronic decay and @xmath58 is the fine - structure constant . the factors of @xmath59 and @xmath60 correct eq . ( [ eq : sigma_ryskin ] ) for the real part and the skewness of the @xmath61 scattering amplitude , respectively ; the factor of @xmath62 absorbs all effects not included in the used approximation ( @xmath63 in the non - relativistic limit for the @xmath0 wave function ) . in a more general case @xcite , ( i.e. , beyond the @xmath52 limit for the @xmath0 wave function ) , there exists a theoretical uncertainty in the value of @xmath64 in eq . ( [ eq : sigma_ryskin ] ) which means that one could use a reasonable range of values , e.g. , @xmath65 gev@xmath54 . for example , the suitable value of @xmath64 can be determined phenomenologically @xcite comparing predictions of eq . ( [ eq : sigma_ryskin ] ) for the proton target with the data . working in the framework of the color dipole model , one can calculate the @xmath66 scattering amplitude by summing multiple rescatterings of a dipole of the fixed size @xmath39 on the target nucleons essentially using eq . ( [ gltsig ] ) and then integrating over @xmath39 with the weight given by the photon and @xmath0 wave functions @xcite , see fig . [ fig : series_jpsi_eikonal ] . since , on average , @xmath0 photoproduction is dominated by small transverse size dipoles and the corresponding dipole nucleon cross section ( [ eq : sigma_dipole ] ) is small , the resulting nuclear shadowing is also small @xcite , which contradicts the alice data , see the discussion in @xcite . in this respect , the situation is similar to the vmd case considered above . in other words , in the dipole formalism , when only @xmath41-dipoles are included , nuclear shadowing in the @xmath66 scattering amplitude is a higher twist effect @xcite . at the same time , the dipole approach describes reasonably well the rhic upc data corresponding to lower energies ( larger values of @xmath1 ) , where the effect of nuclear shadowing is rather small , see fig . [ phenix ] and its discussion in sect . [ subsec : neutron_emission ] . in contrast to the dipole formalism , one can use the leading twist framework of qcd factorization theorems , which enables one to apply eq . ( [ eq : sigma_ryskin ] ) directly to nuclear targets . applying eq . ( [ eq : sigma_ryskin ] ) to the nucleus and proton targets , we obtain the following expression for the @xmath18-integrated cross section of coherent @xmath0 photoproduction on nuclei at high energy : @xmath67 ^ 2 \phi_a(t_{\rm min } ) \ , , \label{eq : sigma_pqcd}\ ] ] where @xmath68 and @xmath69 are the gluon distributions in a nucleus and the free proton , respectively ; @xmath70 is the perturbative qcd ( pqcd ) cross section on the proton calculated using eq . ( [ eq : sigma_ryskin ] ) . thus , the nuclear modification ( suppression ) of @xmath71 is given by the factor of @xmath72 < 1 $ ] quantifying the amount of nuclear gluon shadowing at small @xmath1 . the leading twist theory of nuclear shadowing @xcite is based on the space time picture of the strong interaction at high energies , the generalization of the gribov glauber theory of nuclear shadowing in soft hadron nucleus scattering @xcite to hard processes with nuclei , and the qcd collinear factorization theorems for the total and diffractive cross sections of deep inelastic scattering ( dis ) . the approach allows one to make predictions for the leading twist shadowing correction to nuclear parton distributions ( npdfs ) , structure functions and cross sections , which are given as a series in the number of simultaneous interactions with the target nucleons ( the multiple scattering series ) . the structure of each term in the series is unambiguously given by the gribov glauber theory supplemented by abramovsky gribov kancheli ( agk ) cutting rules @xcite and the qcd factorization theorems . in the graphic form , the multiple scattering series for the @xmath66 scattering amplitude in the leading twist theory of nuclear shadowing is shown in fig . [ fig : series_jpsi_coh ] , where graph @xmath73 is the impulse approximation , graph @xmath74 corresponds to double scattering ( the simultaneous interaction of the probe with two nucleons of the target ) , and graph @xmath36 corresponds to the interaction with three nucleons of the target . the multiple scattering series of fig . [ fig : series_jpsi_coh ] can be summed as follows . the gribov result on the inelastic shadowing correction in hadron nucleus scattering can be conveniently implemented using the formalism of cross section fluctuations @xcite . in this approach , the interaction of a high - energy projectile with a nucleus is a two - step process . first , long before the target , the projectile fluctuates into different configurations interacting with a hadronic target with different cross sections @xmath75 characterized by the distribution over cross sections @xmath76 . second , these fluctuations interact with the nucleus . the corresponding cross section is calculated separately for each fluctuation ( for individual @xmath75 ) using the glauber method and then averaged with @xmath76 , for details and references , see @xcite . in particular , for the @xmath66 scattering amplitude , we obtain : @xmath77 \nonumber\\ & = & \varkappa \int d^2 \vec b \left[\frac{\langle \sigma \rangle \ , t_a(\vec b)}{2}-\frac{\langle \sigma^2 \rangle \ , t_a^2(\vec b)}{2 ^ 2\ , 2!}+ \frac{\langle \sigma^3 \rangle \ , t_a^3(\vec b)}{2 ^ 3\ , 3!}- \dots \right ] \nonumber\\ & = & \varkappa a \frac{\langle \sigma \rangle}{2 } \left[1 -\frac{2}{a } \int d^2 \vec b \left(\frac{\langle \sigma^2 \rangle}{\langle \sigma \rangle } \frac{t_a^2(\vec b)}{2 ^ 2\ , 2!}- \frac{\langle \sigma^2 \rangle}{\langle \sigma \rangle}\frac{\langle \sigma^3 \rangle}{\langle \sigma^2 \rangle } \frac{t_a^3(\vec b)}{2 ^ 3\ , 3!}+\dots \right ) \right ] \ , , \label{sfac}\end{aligned}\ ] ] where @xmath78 . the factor of @xmath79 contains the factors associated with the overlap of the photon and @xmath0 wave functions ; its value is determined by the elementary @xmath80 cross section : @xmath81 . the first term in eq . ( [ sfac ] ) describes photoproduction of @xmath0 on a single nucleon and , hence , is proportional to the number of nucleons @xmath82 ; it is the impulse approximation corresponding to graph @xmath73 in fig . [ fig : series_jpsi_coh ] . the second term in eq . ( [ sfac ] ) corresponds to the simultaneous interaction of the hard probe with two nucleons of the target nucleus and gives the leading contribution to the shadowing correction ; this term corresponds to graph @xmath74 in fig . [ fig : series_jpsi_coh ] . according to the gribov glauber theory of nuclear shadowing supplemented by the collinear factorization theorem for hard diffraction in deep inelastic scattering ( dis ) @xcite , this contribution is unambiguously expressed in terms of elementary diffraction , notably , in terms of the diffractive gluon distribution of the proton @xmath83 @xcite . the corresponding interaction cross section is @xmath84 : @xmath85 where @xmath86 gev@xmath87 is the slope of the @xmath18 dependence of the diffractive cross section ; @xmath88 is the ratio of the real to the imaginary parts of the diffractive scattering amplitude ; the diffractive parton distribution @xmath89 depends on the two light - cone fractions : @xmath90 is the nucleon momentum fraction carried by the diffractive exchange presented by a zigzag line in fig . [ fig : series_jpsi_coh ] ( @xmath91 is the invariant mass of the intermediate diffractive state ) and @xmath92 is the diffractive exchange ( `` pomeron '' ) momentum fraction carried by the active parton . the structure of the interaction with three and more nucleons of the target ( graph @xmath36 in fig . [ fig : series_jpsi_coh ] and higher terms that we do not show ) presents extension of that of graph @xmath74 : in the interaction with @xmath93 nucleons of the target , the hard probe diffractively scatters off two nucleons of the target and the produced diffractive state rescatters on the remaining @xmath94 nucleons , which leads to its attenuation ( absorption ) . in particular , the third term in eq . ( [ sfac ] ) corresponds to the simultaneous interaction of the hard probe with three nucleons of the target ; its contribution corresponds to graph @xmath36 in fig . [ fig : series_jpsi_coh ] . this contribution can not in general be expressed only in terms of diffractive distributions of the proton and needs to be modeled . since the cross section of hard diffraction in @xmath95 dis exhibits the @xmath22 dependence typical for soft processes , it appears plausible to model the rescattering cross section responsible for the interaction with @xmath96 nucleons ( the solid circle in graph @xmath36 in fig . [ fig : series_jpsi_coh ] ) using the formalism of cross section fluctuations . exactly this was assumed in eq . ( [ sfac ] ) ; the corresponding effective cross section is @xmath97 where we suppressed the @xmath1 and @xmath64 dependence of @xmath98 for brevity . in practice , the @xmath98 cross section is calculated using the distribution @xmath76 modeled using the dipole formalism or @xmath99 of the pion , see details in @xcite . this is reasonable at @xmath100 few gev@xmath54 , where soft physics dominates . for larger values of @xmath64 ( e.g. , in the case of @xmath101 photoproduction ) , one can use eq . ( [ eq : sigma3 ] ) only as input at the low initial scale for the subsequent dokshitzer gribov lipatov altarelli parisi ( dglap ) evolution to the desired value of @xmath64 . for the interaction with @xmath102 nucleons ( not shown in fig . [ fig : series_jpsi_coh ] ) , we assume that the effect of cross section fluctuations is the same as for the @xmath103 term , i.e. , @xmath104 for @xmath105 . with this input , the multiple scattering series in eq . ( [ sfac ] ) can be summed and the result presented in the following compact form : @xmath106 \nonumber\\ & = & a \frac{\varkappa \langle \sigma \rangle}{2 } \left[1-\frac{\sigma_2}{\sigma_3}+ \frac{\sigma_2}{\sigma_3 } \frac{\sigma^a_3}{a \sigma_3}\right ] \ , , \label{sfac_2}\end{aligned}\ ] ] where @xmath107 is the total hadron nucleus cross section in the case when the total hadron nucleon cross section is @xmath98 . note that in eq . ( [ sfac_2 ] ) we did not take into account the small real part of the soft scattering amplitude corresponding to the @xmath98 cross section , whose numerical effect is small . expressing the @xmath108 differential cross section in terms of the amplitude of eq . ( [ sfac_2 ] ) , we obtain : @xmath109 ^ 2 \phi_a(t_{\rm min } ) \ , . \label{eq : sigma_pqcd_2}\end{aligned}\ ] ] note that the expression in the square brackets in nothing but the nuclear gluon shadowing ratio @xmath72 $ ] , i.e. , eqs . ( [ eq : sigma_pqcd_2 ] ) and ( [ eq : sigma_pqcd ] ) are consistent with each other . it is important to note that unlike the case of the color dipole formalism , the shadowing correction in eq . ( [ eq : sigma_pqcd_2 ] ) ( and also in @xmath110 ) is a leading twist quantity determined by the elementary hard diffraction in lepton proton dis . in the low nuclear density limit , when the interaction with @xmath105 nucleons can be neglected , the shadowing correction is driven by the leading twist @xmath111 cross section . at the low values of @xmath1 , the @xmath105 terms also become important ; their contributions are also leading twist quantities , which can be summed using the @xmath98 cross section . it is convenient to characterize the suppression of @xmath71 due to nuclear gluon shadowing in terms of the @xmath112 ratio : @xmath113^{1/2 } = \left[1-\frac{\sigma_2}{\sigma_3}+ \frac{\sigma_2}{\sigma_3 } \frac{\sigma^a_3}{a \sigma_3 } \right ] = r(x,\mu^2 ) \ , , \label{lorat}\ ] ] where @xmath114 is the @xmath108 cross section in the impulse approximation ( ia ) neglecting all nuclear effects except for coherence : @xmath115 theoretical predictions for the suppression factor of @xmath116 of eq . ( [ lorat ] ) calculated using the leading twist theory of nuclear shadowing @xcite and the eps09 global qcd fits @xcite agree well with the nuclear suppression factor obtained in the recent analysis of the alice data on coherent @xmath0 photoproduction in pb - pb upcs at @xmath117 and @xmath118 @xcite ( see also fig . [ cirapalice ] ) . similarly to the case of coherent @xmath0 photoproduction on nuclei considered in sect . [ subsec : coherent_case ] , the vmd model , the color dipole formalism and the leading twist approximation can be used to calculate the cross section of incoherent nuclear @xmath0 photoproduction , @xmath119 , where @xmath120 denotes the final nuclear state containing products of the nuclear disintegration ( @xmath121 ) . using completeness of the @xmath120 states , in the high - energy optical limit of the glauber model , the vmd model gives @xcite : @xmath122 where @xmath123 is the inelastic @xmath0nucleon cross section ; @xmath124 is the slope of the @xmath18 dependence of the @xmath31 scattering amplitude . note that eq . ( [ eq : sigma_incoh_vmd ] ) is valid at not too small @xmath125 . equation ( [ eq : sigma_incoh_vmd ] ) has the straightforward and well - known interpretation : the probability of incoherent ( quasielastic ) photoproduction of a vector meson on a nucleus is given by the product of the probability of elastic scattering on a single nucleon times the probability for the produced vector meson to survive the passage through the nucleus on its way out . since @xmath126 is small , the nuclear suppression of @xmath127 due to nuclear shadowing is also small . in the color dipole formalism , the coherent and incoherent nuclear @xmath0 photoproduction cross sections can be calculated on the same footing . the only difference is the order of averaging over the dipole sizes @xmath39 : in the coherent case , one first averages the @xmath66 amplitude over @xmath39 and then squares the result , while in the incoherent case , one first squares the appropriate scattering amplitude and then averages the result over @xmath39 , see , e.g. , ref . @xcite . since the relevant dipole cross section is small , similarly to the coherent case considered above , nuclear suppression of incoherent nuclear @xmath0 photoproduction cross sections is small @xcite . at small @xmath1 typical for the lhc kinematics , @xmath128 and below , the dipole formalism predictions are subject to rather significant theoretical uncertainties due to the choice of the model for the dipole cross section and for the @xmath0 wave function . nevertheless , one can still make the observation that the shadowing suppression of the incoherent cross section of @xmath0 photoproduction on nuclei appears to be larger than that for the coherent case . at the same time , in the rhic kinematics corresponding to much larger values of @xmath1 , @xmath129 , where the effect of nuclear shadowing is small , both the dipole framework and the leading twist approach provide the good description of the phenix upc data @xcite . note also that at @xmath130 , numerous versions of the dipole model correspond to the similar color dipole cross section because the models have been fitted to the same data , see , e.g. , refs . @xcite . in the leading twist formalism , the cross section of incoherent @xmath0 photoproduction on nuclei can be readily calculated using the input employed in the coherent case ( [ eq : sigma_pqcd_2 ] ) . generalizing the standard expression for the incoherent ( quasielastic ) nuclear cross section @xcite to include cross section fluctuations , we obtain in the high - energy limit : @xmath131 note that in eq . ( [ eq : sigma_incoh_qcd ] ) , averaging over cross section fluctuations should be performed at the amplitude level which explains presence of two integrals over @xmath75 and @xmath132 . note also that in eq . ( [ eq : sigma_incoh_qcd ] ) we assumed that the slopes of the @xmath18 dependence of the soft scattering amplitudes corresponding to the @xmath111 and @xmath98 cross sections are equal ; the numerical effect associated with the inequality of the slopes is negligibly small . recalling that @xmath81 and eqs . ( [ eq : sigma2 ] ) and ( [ eq : sigma3 ] ) , after some algebra , eq . ( [ eq : sigma_incoh_qcd ] ) can be written in the following compact form in terms of the @xmath111 and @xmath98 cross sections : @xmath133 \nonumber\\ & \approx & \frac{d\sigma_{\gamma p \to j/\psi p}^{\rm pqcd}(w_{\gamma p})}{dt } \int d^2 \vec{b}\ , t_a(b ) \left[1-\frac{\sigma_2}{\sigma_3}+\frac{\sigma_2}{\sigma_3}e^{-\sigma_3/2 t_a(b ) } \right]^2 \ , , \label{eq : sigma_incoh_2}\end{aligned}\ ] ] where @xmath134 . in the last line of eq . ( [ eq : sigma_incoh_2 ] ) we used @xmath135 ; we checked that this approximation works with high accuracy at the level of a few percent . the physical interpretation of eq . ( [ eq : sigma_incoh_2 ] ) is similar to that of eqs . ( [ sfac_2 ] ) and ( [ lorat ] ) : the nuclear suppression factor in the square brackets arises from multiple interactions of the produced diffractive state with nucleons of the target , which are driven by the @xmath111 and @xmath98 cross sections . a comparison of eqs . ( [ eq : sigma_incoh_2 ] ) and ( [ eq : sigma_pqcd_2 ] ) shows that in the leading twist approximation ( lta ) , nuclear suppression in both coherent and incoherent photoproduction is determined by the same quantities : @xmath136 and @xmath137 [ see eqs . ( [ lorat ] ) and ( [ eq : sigma_incoh_3 ] ) ] . the @xmath136 cross section is a model - independent quantity whose magnitude and @xmath1 dependence are fixed by the experimentally measured diffractive parton distributions , inclusive gluon distributions and dglap evolution equations , see eq . ( [ eq : sigma2 ] ) . the @xmath137 cross section is a model - dependent quantity of the lta approach , whose value is constrained using the formalism of cross section fluctuations . in general , @xmath138 [ see eq . ( [ eq : sigma3 ] ) ] ; the lower limit on the value of @xmath137 , @xmath139 , corresponds to the upper limit on the predicted nuclear shadowing . equation ( [ eq : sigma_incoh_2 ] ) defines the shadowing suppression factor for incoherent nuclear @xmath0 photoproduction , @xmath140 : @xmath141 ^ 2 \ , . \label{eq : sigma_incoh_3}\end{aligned}\ ] ] note that eqs . ( [ eq : sigma_incoh_2 ] ) and ( [ eq : sigma_incoh_3 ] ) are valid at not too small @xmath125 . figure [ fig : s_incoh_fluct ] presents our predictions for @xmath142 as a function of @xmath46 for @xmath143pb at @xmath144 gev@xmath54 . the shaded band represents the theoretical uncertainty of our predictions associated with the range of possible values of the @xmath98 cross section @xcite . one should note that since both suppression factors of @xmath116 ( [ lorat ] ) and @xmath142 ( [ eq : sigma_incoh_3 ] ) are determined by the essentially soft physics , we expect them to be numerically of a a similar magnitude , with @xmath142 being somewhat smaller than @xmath145 . indeed , at @xmath146 corresponding to @xmath147 for pb - pb upcs at @xmath148 gev , we obtain @xmath149 from fig . [ fig : s_incoh_fluct ] and @xmath150 from fig . 3 of ref . @xcite . a high energy nucleus nucleus ultraperipheral collision takes place when the colliding ions pass each other at the distance @xmath151 in the transverse plane ( impact parameter ) exceeding the sum of the nucleus radii , @xmath152 , where @xmath153 is the nuclear radius ( the upc physics is reviewed in @xcite ) . in this case , the strong interaction between the nuclei is suppressed and they interact electromagnetically via emission of quasi - real photons . thus , nucleus nucleus upcs offer a possibility to probe very high energy photon nucleus scattering and , in particular , photoproduction of @xmath0 on nuclei . the corresponding cross section has the following form : @xmath154 where @xmath155 is the photon flux ; @xmath156 is the @xmath0 rapidity , where @xmath157 is the photon energy and @xmath20 is the mass of @xmath0 ; @xmath158 is the nuclear @xmath0 photoproduction cross section ( see sect . [ sec : shadowing ] ) . note that eq . ( [ csupc ] ) includes both the case of coherent scattering without the nuclear breakup ( @xmath159 ) and the case of incoherent ( quasielastic ) scattering when the final nucleus dissociates ( @xmath121 ) . the photon flux at large impact parameters @xmath160 emitted by a fast - moving nucleus , @xmath161 , can be approximated very well by the following simple semi - classical expression for the flux of equivalent photons produced by a point - like particle with the electric charge @xmath162 : @xmath163 \ , , \label{plflux}\ ] ] where @xmath58 is the fine - structure constant ; @xmath164 and @xmath165 are modified bessel functions ; @xmath166 , where @xmath167 is the nucleus lorentz factor . the strong interactions between the colliding nuclei is suppressed by the requirement that @xmath168 . experimentally this corresponds to the selection of events with only two leptons from the @xmath0 decay and otherwise no charged particles in the whole rapidity range covered by the detector . the presence of two terms in eq . ( [ csupc ] ) reflects the fact that each nucleus can radiate a photon as well as can serve as a target . in the case of symmetric upcs ( e.g. , in the case of pb - pb upcs at the lhc and au - au upcs at rhic ) , at a fixed value of the @xmath0 rapidity @xmath169 , eq . ( [ csupc ] ) contains two contributions : one corresponding to the interaction of high - energy photons with a nucleus and another corresponding to the interaction of low - energy photons with a nucleus . in the coherent case , the separation of these overlapping contributions is not an easy problem since a priory one can not say in the interaction with which of the two nuclei @xmath0 was produced . as a result , the photoproduction cross section @xmath170 can be unambiguously extracted from the measured @xmath171 cross section only in the following two cases : ( i ) at @xmath172 , where the two photon energies are equal and , hence , both terms in eq . ( [ csupc ] ) contribute equally , and ( ii ) in the rapidity range where one of the contributions strongly dominates . exactly this situation is realized in the alice experiment @xcite : the @xmath171 cross section was measured ( i ) for @xmath173 , which allowed one to extract @xmath174 at @xmath175 gev corresponding to the gluon momentum fraction of @xmath118 ( @xmath22 is the @xmath176nucleus center - of - mass energy per nucleon ) and ( ii ) for @xmath177 , where the low - energy photon contribution dominates ( more than @xmath178 ) , which probes the nuclear gluon density at @xmath117 . as we already explained in the introduction , the nuclear suppression factor for coherent nuclear @xmath0 photoproduction determined from the corresponding upc cross section measured by the alice collaboration @xcite compares favorably with the theoretical models predicting large nuclear gluon shadowing , notably , with the leading twist approximation ( lta ) @xcite and with the eps09 @xcite result . this is illustrated in fig . [ cirapalice ] , where the alice data on the coherent @xmath4 cross section at the central and forward values of the rapidity @xmath179 are compared to the lta predictions combined with the cteq6l1 gluon parameterization @xcite at @xmath144 gev@xmath54 . one can see from from fig . [ cirapalice ] that the theoretical calculations , which are made using eqs . ( [ eq : sigma_pqcd_2 ] ) and ( [ csupc ] ) , describe the data well . in the same figure , the lta predictions for the incoherent @xmath180 cross section made using eqs . ( [ eq : sigma_incoh_2 ] ) and ( [ csupc ] ) are compared to the alice data point at @xmath181 @xcite . one can see from the comparison that the lta predicts the magnitude of suppression due to nuclear gluon shadowing exceeding the one seen in the data by approximately a factor of @xmath3 . the shaded bands in fig . [ cirapalice ] represent the dominant theoretical uncertainty of the lta predictions associated with the uncertainty in the value of the @xmath98 cross section , which in turn results in the uncertainty of the lta predictions for nuclear parton distributions @xcite . the uncertainty associated with the value of the hard scale @xmath64 , which was studied in @xcite , is much smaller and has been safely neglected . note that in our calculations , we consider quasielastic scattering and do not take into account the incoherent contribution associated with the nucleon dissociation @xmath182 @xcite . we explained in the introduction that this process could potentially contribute to the inelastic final state and , thus , affect the alice extraction of the incoherent @xmath180 cross section @xcite due to the fact that the alice detector does not cover the full range of the rapidity @xmath183 . while the calculation of the @xmath7 contribution requires a a separate publication , one can still make several qualitative observations . first , this contribution is expected to have approximately the same @xmath82 dependence as that in eq . ( [ eq : sigma_incoh_2 ] ) ( it is proportional to @xmath82 in the impulse approximation ) . second , the magnitude of this contribution is expected to be sizable : @xmath184 at @xmath185 and increases with an increase of @xmath11 when the @xmath186 cross section becomes progressively more important and eventually exceeds that of the elastic @xmath187 process ; @xmath188 for the @xmath18-integrated cross sections and for @xmath189 gev ( @xmath190 is the invariant mass of the proton - dissociative system @xmath9 ) @xcite . it would be desirable to perform an additional analysis of the alice data @xcite by assuming that the @xmath191 and @xmath182 contributions to incoherent nuclear @xmath0 photoproduction have different slopes of the @xmath18 dependence , which would enable one to experimentally estimate the contribution of the nucleon dissociation and , thus , will enable a direct comparison of the data with predictions of eq . ( [ eq : sigma_incoh_2 ] ) . in addition , it is likely that due to the interaction of the system @xmath9 with the nucleus , nucleon dissociation will lead to a larger number of neutrons originating from the nucleus dissociation . finally , the study of neutron production in the quasielastic @xmath192 process at @xmath193 gev@xmath54 , where the @xmath182 contribution dominates , may be interesting for understanding of the formation time in diffraction . besides alice , the atlas and cms detectors at the lhc are capable to measure upc production of @xmath0 in the @xmath194 range of rapidity . while for central rapidities , the interpretation of the corresponding measurements is unambiguous , it is difficult to disentangle the high - photon - energy and low - photon - energy contributions to @xmath4 for non - central values of @xmath183 and , thus , to access the small-@xmath1 region that we are interested in . in particular , according to the estimates of @xcite , for @xmath195 in pb - pb upcs at 2.76 tev , the @xmath4 rapidity distribution in the coherent case is exceedingly dominated ( by the factor of four ) by the low - photon - energy contribution corresponding to @xmath196 of the probed nuclear gluons . the high - photon - energy contribution is suppressed by the much lower photon flux and the larger nuclear gluon shadowing . hence , it is rather difficult to extract the high energy nuclear coherent @xmath0 photoproduction cross section from the upc data and , hence , to probe the nuclear gluon distribution around @xmath197 . the method to overcome this difficulty was suggested in @xcite . it is based on the observation of @xcite that coherent photoproduction of vector mesons in heavy ion upcs can be accompanied by additional photon exchanges which lead to electromagnetic excitation of one or both nuclei with the subsequent neutron emission . these neutrons will have the energy close to that of the colliding beams and can be detected by zero degree calorimeters placed at large distances on both sides of the detectors . with the additional requirement to have in the final state only two muons from the @xmath0 decay ( in addition to the neutrons ) and the large rapidity gap , i.e. , by requiring the absence of any other charged particle in the whole range of @xmath183 covered by the detector system , the strong interaction of the colliding nuclei should be suppressed . to calculate the cross section of photoproduction of vector mesons in upcs accompanied by the additional electromagnetic excitation of the colliding nuclei followed by their subsequent neutron emission , we use the model developed in @xcite . this approach is justified by the success of the calculations of @xcite describing very well the alice data on electromagnetic dissociation in pb - pb upcs @xcite . the model is based on the assumption that an additional photon exchange does not destroy coherence of the photoproduction process but influences the impact parameter of the ultraperipheral collision . the latter is taken into account by the modification of the flux of the photons participating in photoproduction : @xmath198 where the impact parameter dependent factor of @xmath199 takes into account different channels of the nuclear decay by the neutron emission ( @xmath200 0n0n , 0nxn , xnxn , @xmath201 ) . in particular , the 0n0n - channel corresponds to the selection of events without additional electromagnetic dissociation with the nucleus neutron decay ; the 0nxn - channel corresponds to one - side excitation with the nucleus neutron decay of only one of the colliding nuclei ; the xnxn - channel corresponds to mutual electromagnetic dissociation with both excited nuclei decaying by neutron emission . to obtain a rough estimate of the size of the effect , each additional photon exchange in pb - pb upcs leads to the suppression of the cross section by the factor of @xmath202 . based on the assumption that an additional photon exchange influences only the flux of photons , ( [ flux ] ) , one can try to separate the low - energy ( @xmath203 ) and high - energy ( @xmath204 ) @xmath0 photoproduction contributions in the rapidity spectra measured in heavy ion upcs . to this end , it is necessary @xcite to measure rapidity distributions of vector meson production in any two channels , e.g. , 0nxn and xnxn , and using the calculated photon fluxes for these channels , to solve simple equations at a fixed value of the rapidity @xmath183 . for the coherent case , one then obtains : @xmath205 since the photon fluxes @xmath206 and @xmath207 can be calculated with good accuracy , measurements of different channels will allow one to study nuclear gluon shadowing in a wide range of @xmath1 . according to our calculations presented and discussed below , the relative contributions of low - photon - energy and high - photon - energy @xmath0 photoproduction in these channels are strongly different : while in the 0nxn - channel they are almost equal , the high - energy photoproduction dominates in the xnxn - channel . production of forward neutrons in quasielastic incoherent photoproduction of @xmath0 in heavy ion upcs with the nuclear breakup has been considered in @xcite . since in this case the momentum transfer in elastic @xmath0 photoproduction on the nucleon can be as large as @xmath208 gev@xmath54 , this target nucleon escaping from the nucleus participates in additional quasielastic rescattering . the average excitation energy of a heavy nucleus in the one - nucleon removal process is about @xmath209 mev , which is much higher than the separation energy of @xmath210 mev of one neutron . it was shown in @xcite that in incoherent @xmath0 photoproduction in heavy ion upcs , the residual nucleus will decay emitting one or more neutrons with the probability of about @xmath211 . therefore , imposing the constraint that no neutrons are emitted , i.e. , considering the 0n0n - channel , one can almost completely ( at the level of @xmath212% ) suppress the incoherent contribution . as we mentioned in the end of sect . [ subsec : upc_coherent ] , the contribution of nucleon dissociation becomes important / dominant with an increase of the transverse momentum of @xmath0 . this process should lead to at least as many neutrons as the quasielastic process . therefore , the @xmath191 and @xmath182 contributions should be either separated experimentally or the latter should be included in the theoretical calculation of the @xmath192 cross section . the procedure for the extraction of the high - photon - energy contribution that we discuss below involves the use of the different @xmath213 dependences of the @xmath191 and @xmath182 cross sections , which allows one to separate their contributions . we also note in passing that the study of the neutron multiplicity at @xmath214 gev , where the process of nucleon dissociation dominates , would produce for the first time information about the space time formation of hadrons in the diffractive processes like @xmath182 . the recent phenix data on @xmath0 photoproduction in ultraperipheral au - au collisions at 200 gev accompanied by neutrons detected in both zdcs @xcite gives an opportunity to check main assumptions about the sources of forward neutrons in coherent and incoherent @xmath0 photoproduction in heavy ion upcs at high energies . indeed , since the nuclear gluon shadowing in these processes is small at rhic energies ( the shadowing effect leads to an approximately 20% reduction at the cross section level ) , the coherent and incoherent photoproduction cross sections can be calculated using either the dipole model or the leading twist approximation ; at @xmath130 corresponding to the rhic kinematics , the dipole model and lta predictions largely converge . in this work , we use a simple version of the dipole model which employs the standard glauber expressions [ eqs . ( [ sigglc ] ) and ( [ eq : sigma_incoh_vmd ] ) ] with the elementary @xmath215 cross section approximated by the phenomenological golec - biernat wusthoff dipole cross section @xcite . the resulting predictions are labeled `` gbw+glauber '' in fig . [ phenix ] . it is worth noting here that for the discussed kinematics , the results for the dipole nucleon cross section obtained in different dipole models are rather close since they are well constrained by the dis data for these energies @xcite . the theoretical uncertainty is much smaller than the phenix experimental errors and , hence , it is not shown in fig . [ phenix ] . note also that in the discussed model , the nuclear shadowing effect is driven by the @xmath216 dipole cross section and , hence , shadowing is suppressed ( it is a higher twist effect ) for the dipoles of such a small size , see the discussion above . selection of events with two - side neutron detection means that in the xnxn - channel , coherent production should take into account the mutual electromagnetic dissociation that requires at least two additional photon exchanges . in contrast , according to the predictions of @xcite , incoherent production is associated with excitation and neutron decay of only the target nucleus . hence , in this case , detection of the xnxn - channel requires only one photon exchange to excite the nucleus , which serves as a source of the photon flux in the process of incoherent production . figure [ phenix ] presents a comparison of the phenix results for the xnxn - channel with our theoretical calculations using the simple dipole described above . despite large experimental errors , it seems that the agreement between our calculations and the phenix data demonstrated in fig . [ phenix ] justifies our approach to neutron production in upcs . coupling this with the good description of coherent @xmath0 photoproduction at the lhc makes our approach reasonable for the prediction of coherent and incoherent upcs accompanied by neutron emission at the lhc . figure [ cms ] presents the results of our calculation of the rapidity distributions in @xmath0 photoproduction in pb - pb upcs with neutron emission in the 0nxn - channel ( upper row of panels ) and the xnxn - channel ( lower row of panels ) as functions of the rapidity @xmath183 at the lhc energy of @xmath217 gev ; the two panels on the left correspond to the coherent case and the panels on the right are for the incoherent case . our theoretical predictions are made using the leading twist approximation ( lta ) ; the shaded areas represent the dominant lta theoretical uncertainty associated with the uncertainty in the predicted nuclear gluon distributions @xcite . the uncertainty associated with the value of the hard scale @xmath64 is much smaller and , hence , has been neglected . each panel contains two sets of curves : the upper shaded area is a sum of the two terms in eq . ( [ csupc ] ) and the lower shaded area represents the contribution of @xmath0 production by the photon emitted by the nucleus moving with the momentum in the direction of positive @xmath0 rapidity ( the dashed curves labeled `` one side '' ) . we can draw several conclusions from fig . first , one can see from the two left panels that using the data on the 0nxn and xnxn - channels , one can try to extract the high - photon - energy contribution using eq . ( [ eq : disentangling ] ) . in the range of rapidities of @xmath218 , this will allow one to determine nuclear gluon shadowing in pb down to @xmath197 and at the scale of @xmath219 gev@xmath54 from coherent @xmath0 photoproduction in upcs . second , a comparison of the corresponding upper shaded areas in fig . [ cms ] shows that we predict that at central rapidities , ( i ) in the 0nxn - channel , the coherent and incoherent contributions will be practically of the same magnitude and ( ii ) in the xnxn - channel , the coherent contribution exceeds the incoherent one by approximately a factor of two . third , besides the coherent channel , the incoherent cross section can also be used to study nuclear gluon shadowing at small @xmath1 . based on the dominance of the strong interaction mechanism of neutron production in incoherent photoproduction , we predict that there is a good opportunity to separate the low - photon - energy and high - photon - energy contributions in the 0nxn - channel ( see the upper right panel in fig . [ cms ] ) . indeed , since in this case neutrons are emitted by the target nucleus , there should by a correlation between the direction of the produced @xmath0 and the direction of neutrons in upc events . in the kinematics of upcs , the direction of charmonium produced by a high - energy photon and , hence , low-@xmath1 gluons from the target , is opposite to the direction of the target nucleus and , correspondingly , to the direction of neutrons . conversely , in the low - photon - energy production , the direction of charmonium coincides with the direction of the target and neutrons . the standard procedure of the separation of coherent and incoherent events consists in the analysis of momentum transfer distributions . in coherent photoproduction , this distribution is dictated by the nuclear form factor squared and presents several distinct peaks at small @xmath220 mev / c , where @xmath213 is the transverse momentum of produced @xmath0 and @xmath221 . one can see such first three diffractive peaks in fig . [ ptdist ] . the @xmath18 dependence of the incoherent nuclear cross section is the same as in the elementary process , i.e. , @xmath222 $ ] . our predictions for the sum of the coherent and incoherent cross sections of @xmath0 photoproduction in pb - pb upcs accompanied by neutron emission in the 0nxn - channel as a function of @xmath213 in the rapidity range of @xmath218 is shown in the upper panel of fig . [ ptdist ] . as in fig . [ cms ] , our theoretical predictions are based on the lta ; the shaded bands represent the lta uncertainty in the predicted nuclear gluon distribution . in the figure , the peaks at small @xmath213 correspond to the coherent signal . therefore , the @xmath220 mev / c cut will effectively reject incoherent events with good accuracy . as we already discussed above , the incoherent cross section can also probe the small-@xmath1 nuclear gluon distribution . in particular , in the 0nxn - channel , the directions of @xmath0 and the emitted neutrons can be unambiguously correlated with either the low - photon - energy or with the high - photon - energy contributions . at the same time , in the coherent channel , due to the assumption of independence of the processes of coherent photoproduction and electromagnetic dissociation followed by the nucleus neutron decay , it is impossible to assign emitting neutrons to the source or the target , and , hence , the probabilities to detect neutrons in the same and in the opposite direction with respect to the direction of @xmath0 should be equal . these predictions are demonstrated in the middle panel of fig . [ ptdist ] , where the upper band corresponds to the situation when the direction of @xmath0 coincides with that of the neutrons and the lower band is for the opposite directions of @xmath0 and the neutrons . the bottom panel shows the ratio of the two curves from the middle panel . since we identify the events , where the directions of @xmath0 and neutrons are opposite , with high - energy photoproduction , there could be two sources of suppression of the corresponding cross section : ( i ) the falloff of the flux of high - energy photons emitted by the nucleus and ( ii ) stronger nuclear gluon shadowing of the small-@xmath1 gluons at @xmath6 , which we are mainly interested in . we considered @xmath0 photoproduction in ion ion upcs at the lhc and rhic in the coherent and incoherent channels with and without accompanying forward neutron emission and analyzed the role of nuclear gluon shadowing at small @xmath1 , @xmath2 , in these processes . we extended the formalism of leading twist nuclear shadowing characterized by large nuclear gluon shadowing to the incoherent @xmath13 cross section . we found that despite a good agreement between the approaches predicting large nuclear gluon shadowing at @xmath5 and the large nuclear suppression factor extracted from the alice data on coherent @xmath0 photoproduction in pb - pb upcs at the lhc @xcite , in the incoherent channel , the leading twist approximation predicts the amount of nuclear suppression due to gluon shadowing which exceeds that seen in the data by approximately a factor of @xmath3 ( fig . [ cirapalice ] ) . we hypothesize that one source of the discrepancy could be the contribution of incoherent nucleon dissociation , @xmath223 , which could potentially contribute to the alice data @xcite and which is not taken into account in our theoretical analysis . in coherent @xmath0 photoproduction in ion ion upcs , it is problematic ( except for @xmath147 and large @xmath179 ) to separate the contributions of high - energy and low - energy photons to the @xmath4 cross section , which reduces the range of @xmath1 for the studies of small-@xmath1 nuclear gluon shadowing . this problem can be circumvented by considering @xmath0 photoproduction in ion ion upcs accompanied by neutron emission due to electromagnetic excitation of one or both colliding nuclei . using the leading twist approximation for nuclear gluon shadowing , we made predictions for coherent and incoherent nuclear @xmath0 photoproduction in pb - pb upcs accompanied by neutron emission in various channels at the lhc ( fig . [ cms ] ) . in particular , we discuss the strategy allowing one to separate the low - photon - energy and the high - photon - energy contributions to coherent @xmath0 photoproduction performing a joint analysis of the data in the 0nxn and xnxn - channels . this gives an opportunity to shift the study of nuclear gluon shadowing to the lower @xmath1 region of @xmath6 . in addition , in the incoherent case accompanied by neutron emission , we show that the separation between the low - photon - energy and high - photon - energy contributions can be efficiently performed by measuring the correlation between the directions of @xmath0 and the emitted neutrons ( fig . [ ptdist ] ) . in the kinematics where nuclear shadowing is small , we showed ( fig . [ phenix ] ) that theoretical predictions based on the dipole model agree with the phenix data on @xmath0 photoproduction in au - au upcs at @xmath224 gev in the xnxn - channel ( both colliding nuclei emit neutrons detected in the zdcs ) . we would like to thank l. frankfurt for useful discussions . m. strikman s research was supported by doe grant no . de - fg02 - 93er40771 . e. abbas _ et al . _ [ alice collaboration ] , eur . j. c * 73 * , 2617 ( 2013 ) [ arxiv:1305.1467 [ nucl - ex ] ] . b. abelev _ et al . _ [ alice collaboration ] , phys . b * 718 * , 1273 ( 2013 ) [ arxiv:1209.3715 [ nucl - ex ] ] . v. guzey and m. zhalov , jhep * 1310 * , 207 ( 2013 ) [ arxiv:1307.4526 [ hep - ph ] ] . k. j. eskola , h. paukkunen and c. a. salgado , jhep * 0904 * , 065 ( 2009 ) [ arxiv:0902.4154 [ hep - ph ] ] . l. frankfurt , v. guzey and m. strikman , phys . * 512 * , 255 ( 2012 ) [ arxiv:1106.2091 [ hep - ph ] ] . l. frankfurt , m. strikman and m. zhalov , phys . lett . * 102 * , 232001 ( 2009 ) [ arxiv:0811.0368 [ hep - ph ] ] . j. j. sakurai , phys . * 22 * , 981 ( 1969 ) . t. h. bauer , r. d. spital , d. r. yennie and f. m. pipkin , rev . mod . phys . * 50 * , 261 ( 1978 ) [ erratum - ibid . * 51 * , 407 ( 1979 ) ] . n. n. nikolaev and b. g. zakharov , z. phys . c * 49 * , 607 ( 1991 ) . a. h. mueller and b. patel , nucl . b * 425 * , 471 ( 1994 ) [ hep - ph/9403256 ] . b. blaettel , g. baym , l. l. frankfurt and m. strikman , phys . lett . * 70 * , 896 ( 1993 ) . m. g. ryskin , z. phys . * c57 * , 89 ( 1993 ) . v. p. goncalves and m. v. t. machado , phys . c * 84 * , 011902 ( 2011 ) [ arxiv:1106.3036 [ hep - ph ] ] . t. lappi and h. mantysaari , phys . c * 87 * , 032201 ( 2013 ) [ arxiv:1301.4095 [ hep - ph ] ] ; arxiv:1310.6336 [ hep - ph ] . v. n. gribov , sov . jetp * 29 * , 483 ( 1969 ) [ zh . eksp . teor . fiz . * 56 * , 892 ( 1969 ) ] . v. a. abramovsky , v. n. gribov and o. v. kancheli , yad . * 18 * , 595 ( 1973 ) [ sov . j. nucl . phys . * 18 * , 308 ( 1974 ) ] ; also in v.n . gribov , _ gauge theories and quark confinement _ , phasis , moskow , 2002 , p. 67 . j. c. collins , phys . d * 57 * , 3051 ( 1998 ) [ erratum - ibid . d * 61 * , 019902 ( 2000 ) ] . a. aktas _ et al . _ [ h1 collaboration ] , eur . phys . j. c * 48 * , 715 and 749 ( 2006 ) ; + s. chekanov _ et al . _ [ zeus collaboration ] , nucl . b * 831 * , 1 ( 2010 ) . s. afanasiev _ et al . _ [ phenix collaboration ] , phys . b * 679 * , 321 ( 2009 ) [ arxiv:0903.2041 [ nucl - ex ] ] . a. h. rezaeian , m. siddikov , m. van de klundert and r. venugopalan , phys . d * 87 * , no . 3 , 034002 ( 2013 ) [ arxiv:1212.2974 ] . a. j. baltz _ et al . _ , phys . * 458 * , 1 ( 2008 ) [ arxiv:0706.3356 [ nucl - ex ] ] . j. pumplin , d. r. stump , j. huston , h. l. lai , p. m. nadolsky and w. k. tung , jhep * 0207 * , 012 ( 2002 ) [ hep - ph/0201195 ] . l. frankfurt , m. strikman and m. zhalov , phys . b * 670 * , 32 ( 2008 ) [ arxiv:0807.2208 [ hep - ph ] ] . a. takahara for the phenix collaboration , proceedings of xx international workshop on deep inelastic scattering 2012 . university of bonn , 26 - 30 june , 2012 , doi : http://dx.doi.org/10.3204/desy-proc-2012-02/114 .

we apply the low - energy theorems to analyze the recent lattice qcd results for the two - nucleon system at a pion mass of mev obtained by the nplqcd collaboration . we find that the binding energies of the deuteron and dineutron are inconsistent with the low - energy behavior of the corresponding phase shifts within the quoted uncertainties and vice versa . using the binding energies of the deuteron and dineutron as input , we employ the low - energy theorems to predict the phase shifts and extract the scattering length and the effective range in the and channels . our results for these quantities are consistent with those obtained by the nplqcd collaboration from effective field theory analyses but are in conflict with their determination based on the effective - range approximation .

1604.02551

[ sec : intro ] understanding of certain fine - tunings in the parameters of the standard model is an important frontier in modern hadron and nuclear physics . in connection with the anthropic considerations , a question has been raised whether the light quark masses have to take very specific values in order to maintain conditions essential for the development of life , see ref . @xcite for discussion . in particular , the proximity of the hoyle state , the first @xmath3 excited state of @xmath4c , to the triple alpha - particle threshold is known to be crucial for the enhanced resonance formation of the life - important elements @xmath4c and @xmath5o in red giant stars @xcite . the dependence of the excitation energy of the hoyle state on the light - quark masses was analyzed recently within _ ab initio _ nuclear lattice simulations @xcite . it was found that the variation of the light quark masses by a few percent is likely to be not detrimental for the development of life . more conclusive statements would require a better knowledge of the quark mass dependence of the nuclear force or , more precisely , of the nucleon - nucleon ( nn ) @xmath6-wave scattering lengths which is by far the dominant source of theoretical uncertainty in this calculation . the quark mass dependence of the nuclear force also plays an important role for constraining a time variation of the standard model parameters as predicted by various extensions of the standard model at the time of big bang nucleosynthesis by comparing the observed and calculated primordial deuterium and helium abundances @xcite . in recent years , there has been significant progress in lattice - qcd calculations of nuclear systems which constitute the primary source of information about the light - quark or , equivalently , pion mass dependence of nuclear observables . in particular , the fully dynamical calculations at unphysical pion masses as low as @xmath7-@xmath8 mev have been performed , see e.g. refs . @xcite . to connect these results with experimentally observed quantities corresponding to the physical value of the quark masses one can employ chiral effective field theory ( eft ) which is still expected to be applicable at such pion masses notice that not only the binding energies but also the nn scattering observables such as the phase shifts and effective range parameters have been calculated on a lattice , see refs . @xcite for recent analyses of the nn @xmath2 and @xmath1 channels by the nplqcd collaboration at @xmath9 and @xmath10 mev and ref . @xcite for the analysis of the higher partial waves at @xmath11 mev by the callat collaboration . interestingly , the general trend of lattice calculations by different groups suggests a stronger attraction both in the @xmath2 and @xmath1 channels when going away from the physical point towards heavier quark masses @xcite . both the deuteron and dineutron systems in these calculations are found to be bound at unphysically heavy pion masses with the deuteron binding energy being significantly larger than the experimentally observed one . these results are , however , not supported by the hal qcd collaboration which finds no bound states in these channels for pion masses ranging from @xmath12 to @xmath13 mev @xcite . it should be noted that unlike the lattice calculations mentioned above , a different approach is employed by the hal qcd collaboration , which makes use of a two - nucleon potential at the intermediate step . the puzzle is getting even more intriguing given that the chiral eft calculations tend to indicate less attraction at @xmath14 larger than the physical value @xcite , see also ref . @xcite for a related work . these calculations , however , rely on the naturalness assumption and/or make use of resonance saturation estimates for @xmath14-dependent four nucleon contact interactions . in our recent paper @xcite , we have argued that the low - energy theorems ( lets ) in nn scattering can provide important consistency checks of both lattice - qcd results and their chiral extrapolations . specifically , the knowledge of the analytic properties of the scattering amplitude allows one to predict its energy dependence and , under certain circumstances , extract the parameters of the effective range expansion . in ref . @xcite , we have tested this approach by predicting the effective range parameters for @xmath6-wave nn scattering in the spin - triplet and the spin - singlet channels , see also refs . @xcite for earlier studies along this line . further , we generalized the lets to unphysical pion masses and applied the resulting approach to selected lattice - qcd results . in particular , using the linear behavior of the quantity @xmath15 , where @xmath16 refers to the effective range , conjectured in ref . @xcite , and visualized in fig . [ fig:3s1eff_rad_mpi ] , we employed the lets to predict the @xmath14 dependence of the deuteron binding energy and of the other parameters in the effective range expansion of the nn scattering amplitude in the @xmath1 channel . remarkably , the resulting @xmath14-dependence of the deuteron binding energy turned out to be in good agreement with the general trend of different lattice calculations @xcite except for the results of ref . @xcite which do not support the existence of bound states for large pion masses . ( color online ) linear with @xmath14 behavior of @xmath15 in the @xmath17s@xmath18 partial wave conjectured in ref . red solid triangles correspond to the nplqcd results at @xmath11 mev @xcite while the shaded area shows the uncertainty of the suggested linear interpolation . the black squares show the empirical values of the effective range at the physical pion mass @xcite . , scaledwidth=35.0% ] recently , the nplqcd collaboration has reported their new results for the @xmath6-wave nn observables both in the spin - triplet and the spin - singlet channels at @xmath19 mev @xcite . in particular , they have calculated the values of the deuteron and the dineutron binding energies and the phase shifts in the @xmath1 and @xmath2 channels at several values of the center of mass ( cms ) momenta above threshold . in this paper we confront these results with the lets . in particular , we demonstrate that the lattice phase shifts at the two lowest energies in the @xmath1 channel are inconsistent ( within the quoted uncertainties ) with the deuteron binding energy obtained in the same lattice calculation . the situation is less conclusive in the @xmath2 channel due to larger uncertainties of the lets in this channel . however , the results of our analysis in this channel also indicate an inconsistency between the phase shifts and the large value of the dineutron binding energy reported by the nplqcd collaboration . taking the nplqcd results for the deuteron and dineutron binding energies as input , we use the lets to infer the corresponding phase shifts and extract the values of the scattering lengths and effective ranges . we perform a detailed comparison of our results for these quantities with the ones of ref . @xcite and argue that their determination by means of the effective range approximation is not self - consistent . our paper is organized as follows . in section [ sec : letphys ] , we discuss in detail our formalism , explain the meaning of the lets and discuss their generalization to unphysical values of the pion mass . implications of the lets for the recent lattice - qcd results at @xmath20 mev are considered in section [ sec : latticeqcd ] . section [ sec : rangempi ] addresses the dependence of the effective range on the pion mass . finally , our main findings are summarized in section [ sec : summary ] . [ sec : letphys ] the concept of the low - energy theorems for nn scattering and their generalization to unphysical pion masses have been discussed in ref . @xcite , see also refs . @xcite for related earlier studies . in this section we formulate the main idea of the lets from a somewhat different perspective as compared to ref . @xcite , where a quantum mechanical framework of the modified effective range expansion @xcite was employed . we assume that the nn interaction is characterized by two distinct scales @xmath21 and @xmath22 , @xmath23 , so that the potential can be written as @xmath24 with the interaction ranges of the order of @xmath25 and @xmath26 , respectively . the analytic structure of the scattering amplitude near threshold is governed by the long - range interactions . taking into account the discontinuity across the left - hand cut from the long - range potential @xmath27 , the energy dependence of the scattering amplitude can be predicted in a model independent way up to the energies corresponding to the branch point of a more distant left - hand cut associated with the potential @xmath28 . this prediction can be regarded as a low - energy theorem . alternatively , one can view the lets as correlations between the parameters in the effective - range expansion of the inverse scattering amplitude induced by long - range interactions . notice that the inverse scattering amplitude may possess poles in the near - threshold region , whose appearance does not affect the validity range of the lets if the scattering amplitude is kept unexpanded . the longest - range part of the nn force is due to the one - pion exchange potential ( opep ) . thus , the let for nn scattering are expected to be governed by the left - hand cut generated by the opep . the opep is , however , singular at the origin and requires regularization and renormalization . therefore , instead of using the quantum mechanical approach , we formulate the let within the modified weinberg approach of chiral eft @xcite . since the correlations between the effective range expansion parameters are inherently long - range phenomena , the results after renormalization and removing the ultraviolet cutoff should be model and regularization - scheme independent . to be specific , we calculate the scattering amplitude @xmath29 by solving the lippmann - schwinger - type integral equation introduced originally by kadyshevsky @xcite which , for the case of the fully off - shell kinematics , has the form @xmath30 where @xmath31 is the free green function @xmath32 further , @xmath33 ( @xmath34 ) is the incoming ( outgoing ) three - momentum of the nucleon in the cms and @xmath35 with @xmath36 denoting the nucleon mass and @xmath37 being the corresponding ( on - mass - shell ) three - momentum . the @xmath6-wave potential at leading order ( lo ) consists of the opep and two derivative - less contact interactions ( here denoted @xmath38 for each given partial wave ) @xmath39 where @xmath40 ( @xmath41 ) denote the spin ( isospin ) pauli matrices of the nucleon @xmath42 , while @xmath43 and @xmath44 refer to the axial vector coupling of the nucleon and the pion decay constant , respectively . as discussed in ref . @xcite , the lo integral equation in this framework is exactly renormalizable scattering see ref . @xcite . this paper also addresses chiral extrapolations of the @xmath45 binding energy . ] , that is all ultraviolet divergencies appearing from iterations of the lo potential can be removed via an appropriate redefinition of the contact interaction @xmath38 . as a consequence , the cutoff in the integral equation can be put to infinity , so that no finite - cutoff artifacts can affect the let . the integral equation ( [ meqlok0integrated ] ) with the potential ( [ vlo ] ) can , in general , be only solved numerically . however , the correlations between the parameters of the effective range expansion implied by the lets can be demonstrated analytically . for simplicity reasons , we consider the partial wave projected @xmath29-matrix in an uncoupled channel with the zero orbital angular momentum , whereas generalizations to the coupled - channel case and to non - zero angular momenta are straightforward . since the lo contact interaction is separable , it is possible to write the solution for the @xmath29-matrix in the semi - analytic ( operator ) form @xmath46 where @xmath47 denotes the off - shell solution of the projected integral equation ( [ g ] ) with the opep alone ( @xmath48 ) , whereas the functions @xmath49 and @xmath50 , shown graphically in fig . [ fig : blocks ] , involve convolutions of this pionic amplitude with the point - like vertex , namely @xmath51 1 true cm graphical illustration of the functions @xmath49 and @xmath50 in eq . , which correspond to convolutions of the pionic amplitude @xmath52 with the point - like vertices as explained in the text.,title="fig:",scaledwidth=90.0% ] to complete the renormalization program at lo , one has to express the contact interaction @xmath38 in terms of the scattering amplitude at zero momentum , that is the scattering length @xmath53 , which is assumed to be an input quantity . thus , the on - shell nn amplitude reads @xmath54 where we used the shorthand notation @xmath55 . for an uncoupled channel with the zero orbital angular momentum , the scattering amplitude @xmath56 can be expressed in terms of the so - called effective range function @xmath57 via @xmath58 we are now in the position to formulate the lets at lo : using the scattering length as the only input quantity to fix the unknown low - energy constant , we employ eqs . ( [ tmaton])-([fk ] ) to predict the momentum dependence of the scattering amplitude and to calculate the phase shifts @xmath59 . such a prediction is possible because all information about the longest - range ope potential and , in particular , about the discontinuity across the corresponding left - hand cut starting from the momentum @xmath60 , is explicitly taken into account in the calculation . unlike the scattering amplitude , the effective range function does not possess the kinematic unitarity cut and is a real meromorphic function of @xmath61 near the origin @xmath62 @xcite . it can , therefore , be taylor - expanded about the origin leading to the effective range expansion @xmath63 where @xmath16 is the effective range , while @xmath64 are the so - called shape parameters . thus , a single piece of information in the form of the scattering length ( or the energy of a bound / virtual state ) allows one to predict all the coefficients in the effective range expansion @xmath65 where @xmath66 and @xmath67 are polynomials in the inverse scattering length , namely @xmath68 with the coefficients @xmath69 and @xmath70 being calculable from the various quantities appearing in eq . ( [ tmaton ] ) and their derivatives evaluated at @xmath62 . their explicit form can be easily obtained by performing taylor expansion of the inverse amplitude @xmath71 around @xmath72 . given that the left - hand cut from the opep is explicitly included , the convergence radius of the lets is restricted by the next - to - lowest - lying left - hand singularity associated with the two - pion exchange potential ( tpep ) . in the present analysis , we take into account the contributions of the tpep implicitly by including the relevant momentum - dependent short - range interactions at next - to - leading order ( nlo ) , whose strengths have to be adjusted to reproduce the empirical value of the effective range in the corresponding channel . then , the shape parameters can be predicted up - to - and - including the nlo corrections . after renormalization , the predictions for the lets become insensitive to details of the short - range interaction once its strength is adjusted to reproduce the physical observable . as argued in ref . @xcite , it is convenient to employ resonance saturation via a heavy - meson exchange in order to model higher - order contact interaction without destroying explicit renormalizability of the integral equation or having to rely on perturbation theory . specifically , the nlo correction to the potential is taken in the form @xmath73 where the heavy - meson mass @xmath74 is set to be @xmath75 mev and the strength @xmath76 is adjusted to reproduce the empirical value of the effective range in the @xmath2 and the @xmath1 channels . our results are not sensitive to the functional form of the term parameterizing the subleading short - range interaction in eq . ( [ satur ] ) , see ref . @xcite for more details . we now summarize the main findings of ref . @xcite for the physical value of the pion mass : * in the @xmath1 channel , the lets yield very accurate results already at lo . for example , the effective range is predicted with an accuracy better than @xmath77 . the accuracy at lo even appears to be better than one could naively expect from the ratio of scales corresponding to the explicitly included lowest left - hand cut from the opep and the next - to - lowest one from the tpep , which is not considered explicitly at this order . this observation can be understood by noticing that iterations of the opep do actually generate the dominant contributions to the left - hand cuts due to the two- and multiple - pion exchange . * the accuracy of the lets in the @xmath2 channel is much worse than in the spin - triplet case . this has to be expected due to the weakness of the opep in that channel . indeed , the opep contributes less than @xmath78 to the magnitude of the empirical @xmath2 phase shift at its maximum value . * as expected , the predicted values of the shape parameters at nlo show in both channels a clear improvement as compared with the lo results . in particular , the nlo lets appear to be accurate at the level of a few percent for the @xmath1 channel ( except for @xmath79 which is unnaturally small ) , while the accuracy of the predictions in the @xmath2 channel is improved to @xmath80 . a generalization of the lets to the case of unphysical pion masses can be carried out straightforwardly @xcite . the main dynamical effect of changing the pion mass in the opep corresponds to shifts of the branch points of all left - hand cuts . the discontinuity across the left - hand cuts also changes due to the dependence of the ratio @xmath81 , which determines the strength of the opep , on the pion mass . ] finally , the discontinuity across the left - hand cuts is also affected by the @xmath14-dependence of the nucleon mass through its appearance in the integral equation ( [ meqlok0integrated ] ) . to account for the last two effects in a way that minimizes the theoretical uncertainty , we use lattice - qcd results for the @xmath14 dependence of @xmath82 and @xmath36 . in particular , we performed quadratic polynomial regression fits ( as functions of @xmath83 ) of the lattice - qcd data for pion masses up to @xmath84mev as shown in fig . 3 of ref . we refer the reader to this paper for more details on the fits and references to the included lattice - qcd calculations of these quantities . using the above results , we can generalize the lo lets by calculating the effective range function @xmath85 , see eq . ( [ fk ] ) , as a function of the ( inverse ) scattering length at _ arbitrary _ values of the pion mass . using a single input quantity such as the binding energy or the scattering length , we can then predict the phase shifts and extract the effective range @xmath16 and the shape parameters @xmath64 ( provided the effective range function does not have poles near the origin ) . some exemplary results for such predictions are discussed in ref . @xcite for different values of the pion mass . notice that contrary to the chiral extrapolations performed in the framework of chiral eft , no assumptions about the short - range interaction @xmath38 as a function of the pion mass is made when calculating the lets . instead , we perform fully independent calculations at each given value of the pion mass as if we lived in different worlds characterized by a specific value of @xmath14 . thus , for each considered value of the pion mass , the lec @xmath38 has to be adjusted to reproduce the given value of the scattering length used as input . by providing relations between various low - energy observables at unphysical values of the pion mass , the lets may serve as consistency checks of the lattice qcd calculations . as already emphasized above , the extension of the lets to nlo is achieved by including subleading contact interactions parametrized via resonance saturation , see eq . ( [ satur ] ) . retaining the light - quark mass variation in the subleading contact interaction is formally suppressed according to the chiral eft estimates . on the other hand , allowing for a variation of this term with @xmath14 can be used to estimate the theoretical uncertainty of our analysis . following ref . @xcite , this is achieved by adjusting the strength @xmath76 of the short - range interaction to reproduce the effective range at the physical point and by assuming that the @xmath83-dependence of @xmath76 is within the envelope built by the straight lines which go through the physical point and describe a @xmath86 change in the value of @xmath76 for @xmath87 mev , i.e. : @xmath88 with @xmath89 and @xmath90 . such a choice of @xmath91 is motivated by the fact that it would cover the known @xmath14-dependence of @xmath43 , @xmath44 and @xmath36 if the same procedure is applied to these quantities . in the next section , we will also give results corresponding to the more conservative choice of @xmath92 . as explained in the previous section , the lets allow one to perform consistency checks of lattice - qcd results for nn scattering provided more than a single observable is extracted . unfortunately , most of the lattice calculations in the nn sector have so far focused on the determination of the binding energies . one exception is the work by the nplqcd collaboration at the pion mass of @xmath94 mev @xcite , which provides , in addition to the binding energies , also the values of the scattering length , effective range and even the first shape parameter . it is , furthermore , conjectured in that paper that the effective range , expressed in units of the pion mass , may be approximated by a linear function of @xmath14 . while the lets are certainly beyond their range of applicability at such heavy pion masses , this conjecture was tested using the lets in our previous work @xcite , where the resulting @xmath95-dependence of the deuteron binding energy was indeed found to be in good agreement with the general trend of lattice data @xcite . recently , new results for nn scattering in the @xmath1 and @xmath2 channels have been reported by the nplqcd collaboration at @xmath96 mev @xcite . the calculations were performed for @xmath97 flavors of light quarks at three lattice volumes of @xmath98 fm , @xmath99 fm , and @xmath100 fm using the lattice spacing of @xmath101 fm . in analogy to their previous work , the scattering phase shifts for the @xmath1 and @xmath2 partial waves were extracted for several values of the cms nn momenta using the extended lscher approach @xcite as shown by the black filled regions in fig . [ fig:3s1let ] for the case of the @xmath1 channel . channel calculated on the lattice at @xmath102 mev @xcite ( filled black regions ) in comparison with the predictions based on the lets at lo ( orange light - shaded bands ) and nlo ( blue dark - shaded and hatched blue light - shaded bands ) using the nplqcd result for the deuteron binding energy @xmath103 as input . the uncertainty at lo shown by the orange bands is entirely given by the uncertainty of @xmath103 in eq . . the nlo dark - shaded ( hatched light - shaded ) bands correspond to the uncertainty in @xmath103 and the theoretical uncertainty of the lets estimated via the variation of @xmath76 with @xmath89 ( @xmath92 ) combined in quadrature . the grey light- and dark - shaded bands in the right panel depict the fit results of the lattice points of ref . @xcite based on the effective range approximation . the energy of the bound ( virtual ) states corresponds to the intersection points of the effective - range function @xmath104 and the unitarity term @xmath105 , shown by the dotted line in the right panel , in the lower ( upper ) half - plane . the phase shift corresponds to the blatt - biedenharn parametrization of the s - matrix @xcite . [ fig:3s1let ] , title="fig:",scaledwidth=49.0% ] channel calculated on the lattice at @xmath102 mev @xcite ( filled black regions ) in comparison with the predictions based on the lets at lo ( orange light - shaded bands ) and nlo ( blue dark - shaded and hatched blue light - shaded bands ) using the nplqcd result for the deuteron binding energy @xmath103 as input . the uncertainty at lo shown by the orange bands is entirely given by the uncertainty of @xmath103 in eq . . the nlo dark - shaded ( hatched light - shaded ) bands correspond to the uncertainty in @xmath103 and the theoretical uncertainty of the lets estimated via the variation of @xmath76 with @xmath89 ( @xmath92 ) combined in quadrature . the grey light- and dark - shaded bands in the right panel depict the fit results of the lattice points of ref . @xcite based on the effective range approximation . the energy of the bound ( virtual ) states corresponds to the intersection points of the effective - range function @xmath104 and the unitarity term @xmath105 , shown by the dotted line in the right panel , in the lower ( upper ) half - plane . the phase shift corresponds to the blatt - biedenharn parametrization of the s - matrix @xcite . [ fig:3s1let ] , title="fig:",scaledwidth=49.0% ] in addition to the phase shifts , the binding energies of the deuteron and the dineutron were extracted . thus , it is interesting to test whether these results fulfill the lets introduced above . the deuteron binding energy calculated in ref . @xcite at @xmath106 mev at three lattice volumes and extrapolated to the infinite volume is @xmath107 where the errors include statistical and systematic uncertainties as well as the extrapolation uncertainty combined in quadrature . further , the first two coefficients in the effective range expansion , namely the scattering length and the effective range , were determined in ref . @xcite by fitting the effective range approximation of the effective range function , @xmath108 to the two lowest - energy scattering data points and the deuteron binding energy , see the grey bands in the right panel of fig . [ fig:3s1let ] . notice that all three lattice data correspond to nucleon momenta below the branch point @xmath109 of the left - hand cut from the opep . the resulting values for the inverse scattering length and the effective range in units of the pion mass reported in ref . @xcite are @xmath110 where the uncertainties in the first and second brackets are statistical and systematic , respectively . in fig . [ fig:3s1let ] , we confront the lattice - qcd phase shifts of ref . @xcite with the predictions of the lets at lo and nlo . we use the nplqcd result for the deuteron binding energy given in eq . as input to adjust the leading - order contact term @xmath38 . this is sufficient to predict the phase shift at lo . as explained in the previous section , there are no additional parameters at nlo . as shown in the left panel of fig . [ fig:3s1let ] , the change in the phase shifts when going from lo to nlo is reasonably small which confirms a good convergence of the lets in this channel . the expected accuracy of the nlo prediction can be roughly estimated by the width of the blue band generated by the variation of the parameter @xmath76 as described above and appears to be consistent with the shift from lo to nlo . notice that the lo ( orange ) band reflects the uncertainty in the nplqcd prediction of the binding energy and does not include the theoretical uncertainty of the lets . as required by the levinson theorem for the case of a bound deuteron , the phase shifts generated by the lets go through @xmath111 at the origin . comparing the nplqcd results for phase shifts and the effective range function in the @xmath1 channel with those based on the lets as visualized in fig . [ fig:3s1let ] , we end up with the following conclusions : * first , as shown in the right panel of fig . [ fig:3s1let ] , only _ positive _ values of the scattering length appear to be consistent with the nplqcd result for the deuteron binding energy quoted in eq . ( [ bd ] ) as opposed to the negative central value for @xmath112 reported in ref . our results for the inverse scattering length extracted from @xmath103 by means of the lets disagree with the nplqcd ones given in eq . ( [ ere3s1 ] ) as can be inferred from the right panel of fig . [ fig:3s1let ] . * while the lattice phase shifts at higher momenta are in reasonable agreement with the ones predicted by the lets , their low - momentum behavior is incompatible ( within the quoted errors ) with that predicted by the lets as demonstrated in both panels in fig . [ fig:3s1let ] . in particular , the phase shift calculated on the lattice at the lowest considered momentum of @xmath113 mev , @xmath114 degrees , is a factor of three smaller than the corresponding value of @xmath115 degrees extracted from the lets . * an extrapolation of the lattice data to zero momenta in the left panel of fig . [ fig:3s1let ] seems to indicate that the phase shift goes to zero . this would , however , contradict the existence of a bound state in this partial wave as a consequence of the levinson theorem ( or require shifting @xmath116 by @xmath117 degrees in the entire plotted energy range which would be inconsistent with the lets ) . one may raise a question whether the observed inconsistencies between the lattice - qcd results for phase shifts and the lets predictions could originate from underestimating the quark mass dependence of the nlo contact interaction by constraining the function @xmath118 as described in the previous section . to clarify this issue , we have increased the allowed variation of @xmath76 by a factor of two , i.e. we set @xmath119 instead of @xmath89 . this corresponds to the allowed variation of the strength of the short - range term at @xmath87 mev by @xmath120 as compared to its value at the physical point . the resulting predictions for the phase shifts and the effective range function are shown by the hatched blue light - shaded bands in fig . [ fig:3s1let ] . with the resulting uncertainty nearly covering the shift from our lo to nlo results , we expect such an error estimation to be too conservative . still , none of our conclusions appear to be affected by employing this very conservative uncertainty estimation . we are now in the position to employ the lets in order to extract the scattering length and the effective range from the deuteron binding energy calculated by the nplqcd collaboration . such an extraction is possible because the effective range function does not possess poles at low momenta and , therefore , can be taylor expanded around the origin mev where the phase shift crosses zero but these momenta are already beyond the region of the validity of the effective range expansion . ] . in fig . [ fig : lets_3s1 ] , we plot the deuteron binding energy and the effective range as functions of the inverse scattering length in units of the pion mass predicted by the lets at lo ( shown by the lines ) and nlo ( shown by the bands ) . , effective range @xmath16 and the binding energy in the @xmath1 partial wave induced by the one - pion exchange potential . the red solid and dashed magenta lines show the predictions of the lo lets for @xmath121 and @xmath103 . the light - shaded bands between the red solid and dashed magenta lines visualize the predictions of the nlo lets for @xmath121 and @xmath103 , respectively , and reflect the theoretical uncertainty estimated via the variation of @xmath76 with @xmath89 as described in the text . the horizontal dotted lines specify the range of values for @xmath103 consistent with the lattice - qcd results of ref . @xcite for this observable . the solid dark - red circle ( blue rectangle ) shows the lo ( nlo ) let predictions for the effective range . the open black circle gives the result for the inverse scattering length and effective range reported by the nplqcd collaboration @xcite while the grey area around it shows the estimated uncertainty from that paper . all results correspond to the blatt - biedenharn parametrization of the s - matrix @xcite . [ fig : lets_3s1 ] , scaledwidth=70.0% ] specifically , the red band between two solid lines represents the nlo let calculation for the effective range as a function of the inverse scattering length . similarly , the magenta band between two dashed lines shows the deuteron binding energy versus the inverse scattering length at nlo . further , the two horizontal dotted lines separate the region of the binding energies consistent with the nplqcd result of ref . @xcite , eq . , for the binding energy . projecting this area onto the @xmath122-axis , as shown by the vertical lines , one obtains the corresponding values of the scattering length and the effective range from the lets . in particular , we find @xmath123 which correspond to the following values in units of fm @xmath124 here , the errors in the first brackets reflect the uncertainty in the value of the deuteron binding energy in eq . ( [ bd ] ) used as input . for the nlo results , we also give in the second brackets an estimation of the theoretical uncertainty corresponding to the choice of @xmath89 . clearly , the above values are at variance with those extracted by the nplqcd collaboration and given in eq . ( [ ere3s1 ] ) . in particular , our value for the effective range is about a factor of @xmath125 smaller than the one found in ref . interestingly , the nlo let prediction for the effective range is in excellent agreement with the assumed linear in @xmath126 behavior of the quantity @xmath127 conjectured in ref . @xcite , cf . [ fig:3s1eff_rad_mpi ] and the right panel of fig . [ fig : eff_rad_mpi ] . for the sake of completeness , we also give the nlo let results based on a more conservative uncertainty estimation resulting by employing a weaker constraint on the allowed @xmath14-dependence of the subleading contact interaction corresponding to the choice of @xmath128 : @xmath129 or @xmath130 in units of fm . to understand the origin of the disagreement between our results for the scattering length and effective range with those of ref . @xcite , it is instructive to take a closer look at the procedure for their determination employed by the nplqcd collaboration . to this aim , a fit of the lattice phase - shift data at the two lowest energies and the deuteron pole was performed using the effective - range approximation ( [ eq : ere ] ) . note that the considered phase - shifts and the deuteron pole correspond to momenta below the branch point of the @xmath131-channel cut due to the opep . for non - singular potentials of a finite range , the applicability region of the effective range expansion is given by the inverse range of the interaction which determines the position of the first left - hand singularity . consequently , the effective range and shape parameters may be expected to scale with the corresponding powers of the pion mass . for example , for the physical value of the pion mass , one has @xmath132 and @xmath133 . the very large value of the effective range reported by the nplqcd collaboration , @xmath134 , either indicates that the range of the nuclear force is considerably larger than that of the opep or signals the appearance of a pole in the effective - range function in the near - threshold region .- wave channel for neutron - deuteron scattering . ] in both cases , the applicability range of the effective range expansion of @xmath135 would be significantly smaller than one may expect based on the position of the left - hand cut due to the opep . as a consequence , the solution for @xmath136 and @xmath137 reported in ref . @xcite and listed in eq . ( [ ere3s1 ] ) is not self - consistent in the sense that it is obtained by fitting the effective range approximation to the data points outside of its validity region which can be roughly estimated as @xmath138 . specifically , the deuteron binding momentum at @xmath139 mev is of the order of @xmath140 , whereas the phase - shifts data employed in the analysis correspond to @xmath141 and @xmath142 . to get further insights into this issue , consider the two roots of the quadratic equation @xmath143 which determines the pole positions of the scattering amplitude within the effective - range approximation , @xmath144 where we have expanded the square root in powers of @xmath145 and neglected terms of order @xmath146 . this is justified both for the physical value of the pion mass and for the solution given in eq . ( [ ere3s1 ] ) , since in both cases one has @xmath147 . at the physical pion mass , the second root yields the deuteron binding momentum @xmath148 mev while the first root , @xmath149 mev , lies outside of the applicability region of the effective range expansion and is an artifact of the effective range approximation . in particular , it disappears or changes the position upon including higher - order terms in the effective range expansion . on the contrary , for the solution in eq . ( [ ere3s1 ] ) at @xmath150 mev , the deuteron pole corresponds to the first root , @xmath151 mev , where the dominant contribution comes from the effective range . meanwhile , because the scattering length in eq . is negative , the second root corresponds to the momentum , @xmath152 mev , lying on the imaginary axis in the lower half plane . therefore , the results of ref . @xcite imply the existence of a shallow virtual state with the excitation energy less than @xmath153 mev in addition to the deuteron , which is not supported by our analysis based on the lets . finally , it is interesting to compare our results based on the lets with the ones obtained using an alternative approach proposed in ref . @xcite , which will be referred to as bbsvk , where the expansion of the nuclear force around the chiral limit was employed , see , however , ref . @xcite for a criticism . this approach was used in ref . @xcite to calculate the phase shifts and the mixing angle in the @xmath154 channel . a comparison of results from the two approaches is presented in fig . [ fig : comp_3s1 ] . ( color online ) neutron - proton phase shifts in the @xmath1 channel ( left panel ) and the mixing angle @xmath155 ( right panel ) at @xmath93 mev based on the lets at lo and nlo in comparison with the results obtained in ref . @xcite using the eft formulation of ref . @xcite , labelled as bbsvk , at lo ( light - shaded band between pink dotted lines ) and nlo ( light - shaded band between blue dashed lines ) . the results for the @xmath17s@xmath18 phase shift correspond to the blatt - biedenharn parametrization of the s - matrix @xcite while the mixing parameter is shown for the stapp parameterization to allow for the comparison with the results of ref . @xcite . for remaining notation see fig . [ fig:3s1let ] . , title="fig:",scaledwidth=49.0% ] ( color online ) neutron - proton phase shifts in the @xmath1 channel ( left panel ) and the mixing angle @xmath155 ( right panel ) at @xmath93 mev based on the lets at lo and nlo in comparison with the results obtained in ref . @xcite using the eft formulation of ref . @xcite , labelled as bbsvk , at lo ( light - shaded band between pink dotted lines ) and nlo ( light - shaded band between blue dashed lines ) . the results for the @xmath17s@xmath18 phase shift correspond to the blatt - biedenharn parametrization of the s - matrix @xcite while the mixing parameter is shown for the stapp parameterization to allow for the comparison with the results of ref . @xcite . for remaining notation see fig . [ fig:3s1let ] . , title="fig:",scaledwidth=49.0% ] while the @xmath1 phase shift and the mixing angle show a very similar behavior at lo , there are more sizable differences at nlo . notice that apart from the different treatment of pions , the two approaches also differ in the way the nlo short - range interaction is taken into account . in particular , in ref . @xcite , the strength of this subleading short - range term was adjusted to fit the lattice phase shifts . in contrast , in our approach , the strength of the subleading contact interaction @xmath76 is determined by the value of the effective range at the physical point while its allowed @xmath14 dependence at unphysical pion masses is used to estimate the theoretical uncertainty as explained in sec . [ general ] . this procedure ensures that both the lo and nlo let results depend on a single unknown parameter . we further emphasize that the low - energy behavior of the mixing angle found in ref . @xcite and shown in the right panel of fig . [ fig : comp_3s1 ] seems to be at variance with the expected threshold behavior for this quantity , @xmath156 ( for details see , e.g. @xcite ) . irregardless of these differences , the two approaches yield similar numerical results for the @xmath1 phase shift and the mixing angle @xmath155 in the considered range of momenta . the values of the scattering length and effective range extracted in ref . @xcite from the lattice data using the framework of ref . @xcite read @xmath157 where the uncertainties in the first and second brackets correspond to the statistical and systematic uncertainties of the lattice results . as already pointed out , the lo values are in agreement with our lo predictions given in eq . ( [ erenlofm ] ) , while the deviations at nlo and , in particular , the large value of the effective range are presumably caused by an attempt to reproduce the lattice - qcd result for the @xmath1 phase shift at @xmath158 gev within the bbsvk approach . we now turn to the spin - singlet channel . in fig . [ fig:1s0_letseb ] , we confront the phase shifts extracted based on the lets with the lattice - qcd results for the @xmath2 partial wave . here we apply the same procedure as in the @xmath1 channel and use the nplqcd result for the dineutron binding energy , @xcite @xmath159 as input to fix the short - range interaction at lo . the nlo short - range interaction is again taken into account by means of resonance saturation , see eq . ( [ satur ] ) , with the strength @xmath76 being determined by the effective range at the physical point . the allowed @xmath14-dependence of @xmath76 is specified by eq . ( [ beta_range ] ) , and the blue dark - shaded bands in fig . [ fig:1s0_letseb ] correspond to the choice @xmath160 . notice that the shift in the predictions when going from lo to nlo is now much larger than in the spin - triplet channel which is in line with the lower predictive power of the lets in the @xmath2 partial wave . consequently , we believe that a variation of the strength @xmath76 with @xmath89 does not provide a realistic estimation of the theoretical uncertainty at nlo in this channel . to have a more conservative estimation , we will allow for a larger @xmath14-dependence in this channel and set @xmath119 as visualized by the hatched blue light - shaded bands in fig . [ fig:1s0_letseb ] . as shown in fig . [ fig:1s0_letseb ] , we arrive at similar conclusions as in the case of the spin - triplet channel . while our nlo let predictions for @xmath161 mev are in very good agreement with the phase shifts calculated by the nplqcd collaboration , there is a clear discrepancy for the two lowest values of the momentum @xmath162 . in particular , for the lowest momentum of @xmath163 mev , the phase shift from the nlo lets is roughly a factor of two larger than that from the lattice - qcd analysis . similarly to the @xmath1 channel , the predictions of the lets based on the dineutron binding energy are only compatible with positive values of the scattering length , see the right panel of fig . [ fig:1s0_letseb ] . specifically , we obtain @xmath164 which correspond to the following values in units of fm @xmath165 here , the errors at lo and in the first brackets at nlo correspond to the uncertainty in the dineutron binding energy while the ones in the second brackets at nlo reflect the unknown @xmath14-dependence of @xmath76 subject to the constraint @xmath119 . these results are in conflict with the nplqcd determination based on the effective range expansion , namely @xcite : @xmath166 channel calculated on the lattice at @xmath102 mev @xcite in comparison with the predictions based on the lets at lo and nlo using the nplqcd result for the dineutron binding energy @xmath167 as input . for notation see fig . [ fig:3s1let ] . [ fig:1s0_letseb ] , title="fig:",scaledwidth=49.0% ] channel calculated on the lattice at @xmath102 mev @xcite in comparison with the predictions based on the lets at lo and nlo using the nplqcd result for the dineutron binding energy @xmath167 as input . for notation see fig . [ fig:3s1let ] . [ fig:1s0_letseb ] , title="fig:",scaledwidth=49.0% ] again , we believe that the analysis performed by the nplqcd collaboration and based on the effective range approximation is not self - consistent . all arguments given in the previous section apply to the @xmath2 channel too , even though our conclusions in this case are somewhat less stringent due to the lower accuracy of the lets . to further elaborate on this point and to provide an assessment of the robustness of our conclusions , we have re - done the calculations by using the lattice phase shifts instead of the dineutron binding energy as input . specifically , we vary the scattering length , which is now used as input for the lets at nlo , in the range consistent with the lattice - qcd phase shifts at the two lowest energies . the resulting phase shifts , corresponding to the inverse scattering length in the range of @xmath168 are shown in the left panel of fig . [ fig:1s0_letsa ] . here , we set @xmath169 , and the width of the band reflects the uncertainty of the lattice - qcd phase shifts used as input . notice that while the nplqcd value of the inverse scattering length given in eq . ( [ ere1s0nplqcd ] ) is indeed consistent with the range of values in eq . ( [ ainvphases ] ) , the obtained solutions correspond to the bound ( virtual ) state binding energy of @xmath170 mev ( @xmath171 mev ) which is in conflict with the lattice - qcd prediction . the apparent bound state corresponding to the leftmost intersection point of the gray bands with the unitarity term @xmath172 in the right panel of fig . [ fig:1s0_letseb ] is an artifact of the effective range approximation . finally , it is interesting to compare our results for the scattering length and effective range with the values obtained in ref . @xcite within the ksw approach to chiral eft @xcite , namely @xmath173 notice that the effective range vanishes at lo in the ksw approach , and the number of independent parameters fitted to lattice data is equal to one , two and three at lo , nlo and next - to - next - to - leading order ( nnlo ) , respectively . our nlo let results are in excellent agreement with the nlo ksw values and also nearly consistent with the nnlo ksw results . channel calculated on the lattice at @xmath102 mev @xcite in comparison with the predictions based on the lets at nlo ( blue shaded bands ) using the scattering length in eq . ( [ ainvphases ] ) as input . for remaining notation see fig . [ fig:3s1let ] . [ fig:1s0_letsa ] , title="fig:",scaledwidth=49.0% ] channel calculated on the lattice at @xmath102 mev @xcite in comparison with the predictions based on the lets at nlo ( blue shaded bands ) using the scattering length in eq . ( [ ainvphases ] ) as input . for remaining notation see fig . [ fig:3s1let ] . [ fig:1s0_letsa ] , title="fig:",scaledwidth=49.0% ] as already discussed , ref . @xcite conjectured that the effective range calculated on lattice at @xmath11 mev and expressed in units of the pion mass may be extrapolated to the physical point by a linear function of @xmath14 . we are now in the position to test this hypothesis by an explicit calculation based on the lets . using the nn bound state energies calculated on the lattice at @xmath174 mev @xcite , @xmath175 mev @xcite , @xmath139 mev @xcite and @xmath176 mev @xcite , we employ the lets at nlo to predict the values of the effective range in the @xmath177s@xmath178 and @xmath17s@xmath18 partial waves . the results are visualized in fig . [ fig : eff_rad_mpi ] . at @xmath11 mev given in ref . @xcite , @xmath179 , is somewhat different from the one plotted in their fig . 11 and corresponding to the linear extrapolation specified in eq . ( 8) of that work . the lattice - qcd result at @xmath11 mev shown in the left panel of fig . [ fig : eff_rad_mpi ] is based on the linear extrapolation specified in eq . ( 8) of ref . @xcite . ] note that the last point in both panels at @xmath11 mev represents the result of lattice calculations by nplqcd @xcite , while the lets are already beyond their range of validity at such heavy pion masses . as seen from the right panel of fig . [ fig : eff_rad_mpi ] , the nlo lets predictions for the effective range in the @xmath17s@xmath18 partial wave are in very good agreement with the linear in @xmath126 behavior of the quantity @xmath127 . interestingly , the lattice data point at @xmath11 mev is consistent with the nlo let results linearly extrapolated to higher pion masses . the results from the lets for the @xmath177s@xmath178 partial wave , although less conclusive due to larger uncertainties , are also generally consistent with the linear in @xmath14 behavior of @xmath180 . we note at this point that the lattice data at @xmath11 mev were obtained in ref . @xcite by using the effective range approximation . the same procedure was employed by the nplqcd to extract the scattering length and effective range at @xmath139 mev and criticized in this work . it is conceivable that the data at @xmath11 mev might also suffer from underestimated systematic uncertainties . ( color online ) nucleon - nucleon effective range ( @xmath15 ) in the @xmath177s@xmath178 ( left panel ) and @xmath17s@xmath18 ( right panel ) partial waves predicted based on the next - to - leading order lets using the bound state energies calculated on the lattice as input . blue open triangles , orange crosses , green open circles and purple solid circles show @xmath15 for the binding energies at @xmath174 mev @xcite , @xmath175 mev @xcite , @xmath139 mev @xcite and @xmath176 mev @xcite , respectively . the uncertainty of our results for @xmath15 is twofold : the smaller error bars reflect the uncertainty of the lattice results for the binding energies used as input while larger ones correspond to the theoretical uncertainty of the lets estimated by setting @xmath119 and the uncertainty of the lattice results added in quadrature . red solid triangles correspond to the nplqcd results for @xmath15 at @xmath11 mev @xcite . the black squares show the empirical values of the effective range at the physical pion mass @xcite . , scaledwidth=95.0% ] in this paper , we have employed the low - energy theorems for nn scattering , which have been generalized in ref . @xcite to the case of unphysical pion masses , to analyze the recent lattice - qcd results at @xmath9 mev reported by the nplqcd collaboration @xcite . the pertinent results of our work can be summarized as follows . * we have used the lets along with the lattice - qcd results for the deuteron and dineutron binding energies in order to extract the energy behavior of the nn phase shifts in the @xmath1 and @xmath2 partial waves and the mixing angle @xmath181 at @xmath139 mev . our lo and nlo calculations suggest a good ( fair ) convergence of our theoretical approach in the spin - triplet ( spin - singlet ) channel . in both channels , the resulting phase shifts are in good agreement with the lattice - qcd results of ref . @xcite for momenta of @xmath182 mev , but are inconsistent with the lattice - qcd predictions at lower energies . * we have used the lets to extract the values of the scattering length and effective range in the @xmath1 and @xmath2 partial waves from the bound state energies obtained on the lattice . the extracted value of @xmath183 is in excellent agreement with the linear in @xmath14 behavior of this quantity conjectured in ref . @xcite . on the other hand , our results are in strong disagreement with the values obtained by the nplqcd collaboration from fits to the lattice - qcd data based on the effective range approximation . we have argued that the very large values for the effective range found in ref . @xcite make the effective range approximation invalid in the energy region corresponding to the lattice data . * our results for phase shifts , scattering lengths and effective ranges agree reasonably well with those obtained in ref . @xcite by analyzing lattice - qcd data within the various eft approaches . given considerable evidence of a bound dineutron and a stronger bound deuteron at heavy pion masses @xcite , our findings indicate that the lattice - qcd calculations of the nn phase shifts of ref . @xcite using the extended lscher approach may possibly suffer from underestimated systematic errors at the lowest considered energies . in addition , using lattice results for the binding energies of the deuteron and dineutron at various pion masses as input @xcite , we demonstrate that the effective range expressed in units of the pion mass behaves as a linear function of @xmath126 . our work demonstrates that the lets provide a useful tool to analyze lattice qcd results for the nn system by allowing one to extract the scattering phase shifts from the calculated bound state energies and/or test consistency of lattice calculations if several observables are computed . we would like to thank jambul gegelia for sharing his insights into the topic discussed here and for a careful reading of the manuscript . we are also grateful to the organizers of the yipqs long - term and nishinomiya - yukawa memorial international workshop on `` computational advances in nuclear and hadron physics '' at yukawa institute for theoretical physics , kyoto university and to the organizers of the int-16 - 1 program `` nuclear physics from lattice qcd '' at int seattle , where a part of this work was carried out . work supported in part by the erc ( project 259218 nucleareft ) , the bmbf ( verbundprojekt 05p2015 - nustar r&d ) and the dfg ( grant gz : ba 5443/1 - 1 aobj : 616443 ) . meiner , sci . bull . * 60*(1 ) , 43 ( 2015 ) , [ arxiv:1409.2959 [ hep - th ] ] . h. oberhummer , a. csoto and h. schlattl , nucl . a * 689 * , 269 ( 2001 ) , [ nucl - th/0009046 ] . e. epelbaum _ et al . _ , phys . * 110 * , no . 11 , 112502 ( 2013 ) , [ arxiv:1212.4181 [ nucl - th ] ] . e. epelbaum , h. krebs , t. a. lhde , d. lee and u .- g . meiner , eur . j. a * 49 * , 82 ( 2013 ) , [ arxiv:1303.4856 [ nucl - th ] ] . f. bedaque , t. luu and l. platter , phys . c * 83 * , 045803 ( 2011 ) , [ arxiv:1012.3840 [ nucl - th ] ] . j. c. berengut _ et al . _ , phys . d * 87 * , no . 8 , 085018 ( 2013 ) , [ arxiv:1301.1738 [ nucl - th ] ] . s. r. beane _ et al . _ [ nplqcd collaboration ] , phys . d * 85 * , 054511 ( 2012 ) , [ arxiv:1109.2889 [ hep - lat ] ] . t. yamazaki , k. i. ishikawa , y. kuramashi and a. ukawa , phys . d * 92 * , no . 1 , 014501 ( 2015 ) , [ arxiv:1502.04182 [ hep - lat ] ] . k. orginos , a. parreno , m. j. savage , s. r. beane , e. chang and w. detmold , phys . rev . d * 92 * , no . 11 , 114512 ( 2015 ) [ arxiv:1508.07583 [ hep - lat ] ] . s. r. beane , p. f. bedaque , m. j. savage and u. van kolck , nucl . a * 700 * , 377 ( 2002 ) , [ nucl - th/0104030 ] . s. r. beane and m. j. savage , nucl . a * 713 * , 148 ( 2003 ) , [ hep - ph/0206113 ] . e. epelbaum , u .- g . meiner and w. glckle , nucl . a * 714 * , 535 ( 2003 ) , [ nucl - th/0207089 ] . s. r. beane and m. j. savage , nucl . a * 717 * , 91 ( 2003 ) , [ nucl - th/0208021 ] . e. epelbaum , u .- g . meiner and w. glckle , nucl - th/0208040 . j. w. chen , t. k. lee , c .- p . liu and y. s. liu , phys . c * 86 * , 054001 ( 2012 ) , [ arxiv:1012.0453 [ nucl - th ] ] . j. soto and j. tarrus , phys . c * 85 * , 044001 ( 2012 ) , [ arxiv:1112.4426 [ nucl - th ] ] . e. epelbaum and j. gegelia , pos cd * 12 * , 090 ( 2013 ) , [ arxiv:1301.6134 [ nucl - th ] ] . s. r. beane _ et al . _ [ nplqcd collaboration ] , phys . d * 87 * , no . 3 , 034506 ( 2013 ) , [ arxiv:1206.5219 [ hep - lat ] ] . s. r. beane _ et al . _ [ nplqcd collaboration ] , phys . rev . c * 88 * , no . 2 , 024003 ( 2013 ) , [ arxiv:1301.5790 [ hep - lat ] ] . e. berkowitz _ et al . _ , arxiv:1508.00886 [ hep - lat ] . t. yamazaki , k. i. ishikawa , y. kuramashi and a. ukawa , phys . d * 86 * , 074514 ( 2012 ) , [ arxiv:1207.4277 [ hep - lat ] ] . t. inoue _ et al . [ hal qcd collaboration ] , nucl . a * 881 * , 28 ( 2012 ) , [ arxiv:1112.5926 [ hep - lat ] ] . v. v. flambaum and r. b. wiringa , phys . c * 76 * , 054002 ( 2007 ) , [ arxiv:0709.0077 [ nucl - th ] ] . v. baru , e. epelbaum , a. a. filin and j. gegelia , phys . c * 92 * , no . 1 , 014001 ( 2015 ) , [ arxiv:1504.07852 [ nucl - th ] ] . t. d. cohen and j. m. hansen , phys . c * 59 * , 13 ( 1999 ) , [ nucl - th/9808038 ] . t. d. cohen and j. m. hansen , phys . c * 59 * , 3047 ( 1999 ) , [ nucl - th/9901065 ] . j. v. steele and r. j. furnstahl , nucl . a * 645 * , 439 ( 1999 ) , [ nucl - th/9808022 ] . e. epelbaum and j. gegelia , eur . j. a * 41 * , 341 ( 2009 ) , [ arxiv:0906.3822 [ nucl - th ] ] . e. epelbaum and j. gegelia , pos cd * 09 * , 077 ( 2009 ) . e. epelbaum , arxiv:1001.3229 [ nucl - th ] . j. j. de swart , c. p. f. terheggen and v. g. j. stoks , nucl - th/9509032 . h. van haeringen and l. p. kok , phys . * a26 * , 1218 ( 1982 ) . e. epelbaum and j. gegelia , phys . b * 716 * , 338 ( 2012 ) , [ arxiv:1207.2420 [ nucl - th ] ] . e. epelbaum , a. m. gasparyan , j. gegelia and h. krebs , eur . phys . j. a * 51 * , no . 6 , 71 ( 2015 ) , [ arxiv:1501.01191 [ nucl - th ] ] . v. g. kadyshevsky , nucl . b * 6 * , 125 ( 1968 ) . v. baru , e. epelbaum , a. a. filin , j. gegelia and a. v. nefediev , phys . d * 92 * , no . 11 , 114016 ( 2015 ) , [ arxiv:1509.01789 [ hep - ph ] ] . j. m. blatt and j. d. jackson , phys . rev . * 76 * , 18 ( 1949 ) . b. midya _ et al . _ , arxiv:1501.04011 [ quant - ph ] . m. lscher , commun . phys . * 105 * , 153 ( 1986 ) . m. lscher , nucl . b * 354 * , 531 ( 1991 ) . r. a. brice@xmath184o , z. davoudi , t. luu and m. j. savage , phys . d * 88 * , no . 11 , 114507 ( 2013 ) , [ arxiv:1309.3556 [ hep - lat ] ] . j. m. blatt and l. c. biedenharn , rev . * 24 * , 258 ( 1952 ) . d. b. kaplan , m. j. savage and m. b. wise , nucl . b * 534 * , 329 ( 1998 ) , [ nucl - th/9802075 ] . d. b. kaplan , m. j. savage and m. b. wise , phys . b * 424 * , 390 ( 1998 ) [ nucl - th/9801034 ] . m. pavon valderrama and e. ruiz arriola , phys . rev . c * 72 * , 044007 ( 2005 ) .

we analyze the hawking radiation process due to collapsing configurations in the presence of superluminal modifications of the dispersion relation . with such superluminal dispersion relations , the horizon effectively becomes a frequency - dependent concept . in particular , at every moment of the collapse , there is a critical frequency above which no horizon is experienced . we show that , as a consequence , the late - time radiation suffers strong modifications , both quantitative and qualitative , compared to the standard hawking picture . concretely , we show that the radiation spectrum becomes dependent on the measuring time , on the surface gravities associated with different frequencies , and on the critical frequency . even if the critical frequency is well above the planck scale , important modifications still show up .

0807.4147

phenomenological approaches to quantum gravity have recently started to develop in parallel with the more traditional attempts to construct such a theory from first principles . in particular , increasing attention has focused on the consideration that maybe lorentz invariance is not a fundamental law , but an effective low - energy symmetry which is broken at high energies ( see , e.g. , @xcite for a general introduction ) . conceptually speaking , in quantum gravity theories from first principles , it is not really clear whether lorentz invariance is fundamental or effective , and in the latter case , how its breaking scale is related to the planck scale . for example , while many string theory scenarios axiomatically incorporate lorentz invariance , it has been argued that in certain situations , violations of lorentz invariance may occur in a way consistent with world - sheet conformal invariance @xcite , thus leading to acceptable string theory backgrounds . in the context of loop quantum gravity , in @xcite it has been argued that quantum effects should modify the relativistic dispersion relations , although the issue seems far from settled ( see @xcite for some general remarks ) . in scenarios of emergent gravity based on condensed matter analogies @xcite , the situation is clearer : lorentz invariance is a low - energy effective symmetry , and so it is expected to break at some scale , although not necessarily related to ( and therefore possibly much higher than ) the planck scale @xcite . a simple way of modelling a wide range of lorentz violating effects ( and a quite natural one in the case of condensed matter analogies ) consists in modifying the dispersion relations at high energy @xcite . this modification can be subluminal or superluminal , depending on whether high - frequency modes move slower or faster than low - energy ones . we should mention that there also exist ways of modifying the dispersion relations , for example by introducing a minimum length or maximum energy , without violating lorentz invariance @xcite . here , however , we will be interested in modifications that explicitly break lorentz invariance . from extrapolation of current experiments , we know that there exist stringent bounds on the most commonly expected types of lorentz violations at the planck scale ( see , e.g. , @xcite and references therein ) . nevertheless , even violations at much higher energy scales might still significantly affect black hole physics . in this paper we will investigate the effects that a superluminal modification of the dispersion relation would have on the hawking radiation produced by collapsing configurations . hawking s original prediction that black holes radiate thermally @xcite , rested on the implicit assumption that the low - energy laws of physics , and in particular lorentz invariance , are preserved up to arbitrarily large scales , much higher than the planck energy . the question of the robustness of hawking s prediction to modifications of the transplanckian physics has been tackled principally by analyzing effective field theories such that at high energies a modification of the dispersion relation is incorporated ( see however @xcite for a different take on the problem ) . historically , attention has mainly been given to subluminal dispersion relations . important contributions such as @xcite , and more recently @xcite seem to have settled the robustness of hawking radiation with respect to subluminal modifications and truncation of the frequency spectrum , although this conclusion still rests on certain assumptions , usually related to the behaviour of the fields near the horizon . in any case , it is important to remember that this only solves part of the so - called ` transplanckian problem ' with regard to hawking radiation . indeed , subluminal modifications gradually dampen the influence of ultra - high frequencies , and so they do not explore arbitrarily large frequencies . so even assuming that it has been demonstrated that hawking s result can be recovered without appeal to transplanckian frequencies , and that it is robust to subluminal modifications , the question remains whether it is also robust with respect to a ( non - dampened ) modification of the transplanckian physics . superluminal modifications differ conceptually from subluminal ones , in that they gradually magnify the influence of ultra - high energies , and thereby offer an interesting test - case for the transplanckian robustness of hawking radiation . an additional motivation for our study comes from a possible connection with experiment . it is well - known ( see , e.g. , @xcite ) that hawking radiation is not a purely gravitational effect , but a characteristic of quantum field theory in curved spacetimes with a horizon . condensed matter systems such as bose - einstein condensates ( becs ) can , under certain approximations , be described by a relativistic quantum field theory . it has also been shown , again under certain idealizing approximations , that ( acoustic ) black hole horizons in becs are dynamically stable @xcite . therefore , becs are expected to be good candidates for a possible experimental verification of hawking radiation @xcite . but since the physics of becs automatically leads to a superluminal dispersion relation at high energies , the question is again which kind of modifications are to be expected in the laboratory realization of a bec black hole with respect to the standard hawking picture . of the above - mentioned works on the robustness of hawking radiation , a few have also tried to address superluminal modifications @xcite . various problems make this case quite different from the subluminal one . these problems can be related to the fact that the horizon becomes frequency - dependent when modifying the dispersion relations . this is also true in the subluminal case , though qualitatively in a very different way . with superluminal modifications , the horizon lies ever closer to the singularity for increasing frequencies , and asymptotically coincides with it . this causes the ` apparent ' interior of the black hole ( the interior of the zero - frequency horizon ) to be exposed to the outside world . since it seems unreasonable to impose a condition arbitrarily close to the singularity as long as we do not have a solidly confirmed quantum theory of gravity , most of the approaches used for subluminal dispersion are invalid , or at least questionable , for superluminal dispersion . moreover , since it seems reasonable to expect that quantum effects will remove the general relativistic singularity , a critical frequency might appear above which no horizon would be experienced at all . in the analysis that follows we will try to avoid making any further assumptions about the physics near the singularity . for instance we will analyze the characteristics of the radiation at retarded times at which the singularity has not yet formed . our approach will be based on a derivation of hawking radiation through the relation between the asymptotically past ray trajectories and the asymptotically future ones in the case of a collapsing configuration . the only assumption about these asymptotic extremities are the standard ones , namely a minkowski geometry in the asymptotic past , and flatness at spatial infinity also in the asymptotic future . the language we will use is related to the fluid analogy for black holes , which provides the intuitive picture of the spacetime vacuum flowing into the black hole and getting shredded in its center . our main results can be summarized as follows . three crucial elements distinguish the late - time radiation with superluminal dispersion relations from standard relativistic ones . first , at any instant there will be a critical frequency above which no horizon has yet been experienced . this critical frequency will induce a finite limit in the modes contributing to the radiation , which will therefore have a lower intensity than in the standard case , even if the critical frequency is well above the planck scale . second , due to the effective frequency - dependence of the horizon , the surface gravity will also become frequency - dependent and the radiation will depend on the physics inside the black hole . unless special conditions are imposed on the profile to ensure that the surface gravity is nonetheless the same for all frequencies below the critical one , the radiation spectrum also undergoes a strong qualitative modification . depending on the relation between the critical frequency and the planck scale , the radiation from high frequencies is no longer negligible compared to the low - frequency thermal part , but can even become dominant . this effect becomes more important with increasing critical frequency . finally , a third effect is that the radiation will extinguish as time advances . the remainder of this article has the following structure . in section [ s : model ] , we will describe and motivate the classical geometry of our model and the concrete kinds of profiles that we are interested in . in section [ s : sdr ] , we will briefly review how hawking s standard result can be obtained in the formalism of our choice for the case of standard relativistic dispersion relations . this calculation will be adapted in section [ s : mdr ] to superluminal dispersion relations , and we will analytically obtain a formula for the late - time radiation for this case . then , in section [ s : graphics ] , we will present graphics obtained from this analytic formula by numerical integration . these graphics will illustrate the results mentioned in the previous paragraph , which we will discuss in more depth and compare with other results in the recent literature in section [ s : conclusion ] . we will study radiation effects in simple 1 + 1 dimensional collapsing configurations . it is well known that in the hawking process , most of the radiation comes out in the @xmath0-wave sector . hence , a spherically symmetric treatment , effectively 1 + 1 dimensional , suffices to capture the most relevant aspects of the process . we will work in painlev - gullstrand coordinates , where @xmath1\d t^2 - 2v(t , x)\d t \d x + \d x^2~ , \label{acoustic - metric}\end{aligned}\]]and make the further simplification of taking a constant @xmath2 ( we will always write @xmath2 explicitly to differentiate it from the frequency / wave - number dependent @xmath3 that will show up in the presence of a superluminal dispersion relation ) . in this manner all the information about the configuration is encoded in the single function @xmath4 . the painlev - gullstrand coordinates have the advantage , compared to the schwarzschild form , of being regular at the horizon . moreover , they suggest a natural interpretation in terms of the language of acoustic models . in such a model , @xmath2 represents the speed of sound and @xmath5 is the velocity of the fluid ( which corresponds to the velocity of an observer free - falling into the black hole ) . throughout the paper we will consider the fluid to be left - moving , @xmath6 , so that the outgoing particles of light / sound move towards the right . the most relevant aspects of the analysis presented hereafter depend only on the qualitative features of the fluid profile . however , to justify some specific calculations later on it is helpful to use the following concrete profiles , see figs . [ fig : v - constant - kappa ] and [ fig : v - increasing - kappa ] . consider a velocity profile @xmath7 such that , and further decreasing monotonically as . in the fluid image , the fluid nearly stands still at large distances and accelerates inwards , with a sonic point or horizon at @xmath8 . the further decrease in @xmath9 can either be linear until a constant limiting value is almost achieved , as in fig . [ fig : v - constant - kappa ] , or of the form @xmath10again up to a constant limiting value , as in fig . [ fig : v - increasing - kappa ] . let us recall that this second velocity profile corresponds to the schwarzschild line element with @xmath11 the central mass , as can be seen @xcite by reparametrizing the time coordinate and using @xmath12 as radial coordinate . velocity profile of a black hole with a linear slope of the velocity @xmath7 , and hence a constant surface gravity @xmath13 , from the classical horizon at @xmath8 down to some predefined limiting value . the auxiliary function @xmath14 separates each instantaneous velocity profile @xmath4 into a dynamical : @xmath15 and a stationary region ii : @xmath16 . ] velocity profile characteristic of a schwarzschild black hole . the slope of @xmath7 increases from the classical horizon at @xmath8 leftwards up to a constant limiting value , leading to a surface gravity @xmath13 which increases with the frequency . ] up to here , we have described a stationary profile . now , to incorporate the dynamics of the collapse , let us introduce a monotonically decreasing function @xmath14 , , and define @xmath4 as @xmath17 imagining the collapse of a homogeneous star , the function @xmath14 represents the distance from the star surface to its gravitational ( schwarzschild ) radius . the dynamical configuration that we obtain consists of a series of snapshots . in each snapshot , @xmath18 increases ( i.e. , the fluid accelerates ) from @xmath19 for @xmath20 , up to a point @xmath21 , and then remains constant as @xmath22 further decreases towards @xmath23 . for consecutive snapshots , the point @xmath21 moves leftwards , so @xmath4 covers an ever larger part of @xmath7 . from the point of view of an outgoing ( right - moving ) particle of light / sound , the configuration is nicely split up into , first , a dynamical or @xmath24-dependent region , and then a stationary @xmath22-dependent region , as defined in . the difference between the two types of profiles , figs . [ fig : v - constant - kappa ] and [ fig : v - increasing - kappa ] , can best be explained in terms of the surface gravity . when modifying the dispersion relation , the horizon becomes a frequency - dependent concept : each frequency experiences a different horizon , as we will see explicitly in section [ s : mdr ] . in particular , for superluminal modifications , the horizon forms later ( i.e. , at higher values of @xmath25 , or more negative values of @xmath22 ) for increasing frequencies . the surface gravity will then also become frequency - dependent : @xmath26 where @xmath27 is the frequency - dependent location of the horizon . this surface gravity will be seen in section [ s : mdr ] and [ s : graphics ] to play a crucial role . this explains our choice of the two types of profiles : in our first profile , fig . [ fig : v - constant - kappa ] , we have considered a linear velocity profile @xmath28 from the classical or zero - frequency horizon ( @xmath29 ) down to some given limiting value , @xmath30 , from where it stays constant . this maximum velocity defines a critical frequency @xmath31 as we will discuss in more detail in section [ s : mdr ] . then the surface gravity @xmath13 will be frequency - independent up to this critical frequency @xmath31 after which it will rapidly vanish . in the second profile , see fig . [ fig : v - increasing - kappa ] , we have taken a velocity profile typical for a schwarzschild black hole , and therefore @xmath13 increases with @xmath32 , again up to a predefined limiting value corresponding to the horizon for a critical frequency @xmath31 . in the context of the fluid analogy , it seems obvious that some mechanism will avoid the formation of a singularity . but in any case , we will also take into account the limit for an infinite critical frequency , which corresponds to a velocity profile with @xmath33 . in this section we will briefly review a way of deriving hawking s formula for the radiation of a black hole with standard ( relativistic ) dispersion relations of the form . for the sake of simplicity , we will only consider a massless scalar field . we will summarize the main steps : from the klein - gordon inner product and the definition of the bogoliubov @xmath34 coefficients , over the relation between past and future null coordinates , to the black - body radiation in the wave - packet formulation . our aim is to present the key points of the procedure in such a way that they can easily be adapted to superluminally modified dispersion relations the subject of the next sections . the dalembertian or wave equation for a massless scalar field in 1 + 1 dimensions for the metric ( [ acoustic - metric ] ) with constant @xmath2 can be written as @xmath35 this conformally invariant theory is equivalent to the dimensionally - reduced 3 + 1 spherically - symmetric theory if one neglects the backscattering due to the angular - momentum potential barrier ( responsible for the so - called grey - body factors ) . in the space of solutions of this wave equation , we can define the klein - gordon pseudo - scalar product @xmath36which is independent of the choice of the spatial slice @xmath37 . for a slice , and in particular for this becomes @xmath38~.\]]we can define future null coordinates @xmath39 and @xmath40 , such that , when , @xmath41 writing the inner product in terms of these null coordinates , gives , for the limit @xmath42 , @xmath43_{w=+\infty } \right . \nonumber \\ & & \hspace{-10 mm } \left . + \int_{-\infty}^{+\infty } \d w ~ \left[\varphi_1 \partial_w\varphi_2^ * -\varphi_2^ * \partial_w\varphi_1\right]_{u=+\infty } \right\}~. \label{kg - product}\end{aligned}\]]note that , in the derivation of this formula , we have only made use of coordinate transformations , without any appeal to their null character or to geometrical tools such as conformal diagrams or the deformation of cauchy surfaces . this is important because in the case of modified dispersion relations , such geometrical concepts become problematic and actually can only be maintained in the context of a rainbow geometry if at all . similarly we could have used past null coordinates @xmath44 , which obey , when @xmath45 , @xmath46to calculate the inner product in the asymptotic past . the right - moving positive - energy solutions associated with the asymptotic past and with the asymptotic future , normalized in the dirac - delta sense , can be expressed as @xmath47 respectively , where we use primes to indicate asymptotic past values . the relevant bogoliubov @xmath34 coefficients encoding the production of radiation are defined as . the mode mixing relevant for the hawking process occurs in the right - moving sector . therefore , we only need to calculate the first term in the scalar product ( [ kg - product ] ) . then , plugging into the definition of @xmath34 and integrating by parts we obtain the simple expression @xmath48so that all the information about the produced radiation is contained in the relation @xmath49 . for a standard relativistic dispersion relation , it is well known that the relation between @xmath51 and @xmath52 for a configuration that forms a horizon ( in our case at @xmath8 ) can be expressed at late times as ( note that we use a subscript @xmath53 for all quantities associated with the horizon ) , where @xmath54 , @xmath55 and the surface gravity @xmath56 are constants . we can define a threshold time @xmath57 at which an asymptotic observer will start to detect thermal radiation from the black hole . this retarded time corresponds to the moment at which the function @xmath50 enters the exponential regime . we can then rewrite the previous expression , valid for @xmath58 , as @xmath59 plugging this relation into and integrating in @xmath52 gives @xmath60 \exp[-i\frac{c\omega}{\kappa}\ln(\omega'a_0)]\nonumber\\ & & ~~\times \exp(-i\omega u_i)\exp(-{\pi c\omega \over 2 \kappa } ) \gamma(ic\omega/\kappa ) ~.\end{aligned}\ ] ] in order to obtain physically sensible results , it is a good precautionary measure to replace the monochromatic rays described until now by wave packets ( see e.g. @xcite or the discussion in @xcite ) . positive - energy wave packets can be defined as @xmath61 where @xmath62 with @xmath63 an overall reference and @xmath64 an integer phase parameter . the central frequencies of the wave packets are , with @xmath65 their width . then , from the expression and assuming that the wave packets are sufficiently narrow ( @xmath66 ) , we obtain @xmath67 where we have defined @xmath68 finally , integration in @xmath32 gives the number of particles with frequency @xmath69 detected at time @xmath70 by an asymptotic observer : @xmath71which reproduces hawking s formula ( in the absence of backscattering ) and corresponds to a planckian spectrum with temperature @xmath72 . in this section we will indicate how the late - time radiation originating from the formation of a black hole from collapse can be calculated in the case of superluminal dispersion relations . we introduce superluminally modified dispersion relations by adding a quartic term to the wave equation : @xmath73 where @xmath74 ( the ` planck scale ' ) is the scale at which non - relativistic deviations in the associated dispersion relation @xmath75become significant . we use this relation for concreteness and because it is the one that shows up in bose einstein condensates , where @xmath76 , with @xmath77 the healing length of the condensate . however , qualitatively , our results will not depend on the specific form of the deviations from the relativistic dispersion relation but on their superluminal character . this dispersion relation leads to a modification in both the phase velocity @xmath78 and the group velocity @xmath79 . behaviour of the effective phase @xmath80 and group @xmath81 speeds of light / sound with respect to the wave number @xmath82 . due to the non - relativistic dispersion relation , the effective velocities become ` superluminal ' for @xmath83 ( where @xmath74 is the planck scale ) . ] for a right - moving wave , we have @xmath84 where we have introduced the effective @xmath82-dependent phase and group speeds of light / sound @xmath85 respectively ( see fig . [ fig : c_k ] ) . both @xmath80 and @xmath81 become larger than @xmath2 ( i.e. , become ` superluminal ' ) as @xmath82 increases above @xmath74 . at first sight , it seems obvious that the ray equation should be defined in terms of the group velocity @xmath79 . nevertheless , the question of whether the velocity relevant for hawking radiation is the group or the phase velocity seems to be tricky @xcite . for example , below suggests that the phase velocity might be relevant . we limit ourselves to remark that @xmath81 and @xmath80 show the same qualitative behaviour . hence our results are independent of this question and we will simply write @xmath3 ( or @xmath86 when wishing to emphasize the frequency - dependence ) . then , there will be a horizon , which becomes frequency - dependent , when @xmath87 , irrespectively of whether @xmath81 or @xmath80 is used for @xmath3 . moreover , since @xmath3 becomes arbitrarily high for increasing wave number @xmath82 , given a certain @xmath30 at a particular instant of time , there will be a critical @xmath31 such that waves with an initial frequency @xmath88 do not experience a horizon at all . the only exception to this rule occurs when the velocity profile ends in a singularity @xmath89 , which implies @xmath90 . our aim is to calculate the black hole radiation with superluminally modified dispersion relations . we will now repeat the main steps of section [ s : sdr ] , and point out where and how these modifications must be taken into account . the essential point with regard to the pseudo - scalar product is the following . its explicit form for @xmath91constant is not changed by the presence of the @xmath92 term in the wave equation ( see also the discussion in @xcite ) . indeed , it is still a well - defined inner product , and in particular it is still a conserved quantity , since @xmath93 which can be seen , by repeated integration by parts , to vanish under the usual assumption that the fields die off asymptotically . note that the modification of the dispersion relation singles out a preferred time frame : the ` laboratory ' time @xmath24 . changing to another time @xmath94 will in general lead to a mixing between @xmath24 and @xmath22 , and hence the simple relations given here would no longer be valid . using the preferred time @xmath24 , and making exactly the same change of coordinates as in the case of standard dispersion relations , we can again transform the inner product , evaluated at @xmath42 , into the expression in terms of @xmath52 and @xmath95 . as in the standard case , only the first term is relevant for the hawking process : @xmath96_{w=+\infty}~.\end{aligned}\ ] ] note that we are now using @xmath52 and @xmath95 merely as a perfectly good set of ( auxiliary ) coordinates , in order to cast the inner product into a useful form . however , we can not associate them any null character such as they had in the case of standard dispersion relations . let us define some sets of spacetime functions that will prove to be useful in what follows . given a fluid profile of the type described in section [ s : model ] , we can integrate the ray equation @xmath97 starting from the past left infinity towards the right . the ray initially has a frequency @xmath32 and an associated initial wave number @xmath98 , from which we can deduce the value of @xmath99 . in the left region , where the velocity profile is dynamic but position - independent ( @xmath100 , with @xmath14 the auxiliary function introduced in section [ s : model ] ) , @xmath101 can be considered as fixed while the frequency changes ( this is what happens in a mode solution of equation in this region ) . then , we can define the function @xmath102 where @xmath103 is the instantaneous frequency of the particle at each time @xmath24 , defined through the dispersion relation and such that . when the ray reaches the kink it passes into a stationary region in which the velocity profile only depends on the position ( @xmath104 ) . at the kink , the ray still has the initial wave number @xmath98 and a frequency @xmath105 . in its propagation towards the right this frequency now remains fixed while its wave number becomes a decreasing function @xmath106 of the position , such that @xmath107 , with @xmath108 the moment at which the kink is crossed . the final frequency of the ray will then simply be @xmath105 and its final wave number @xmath109 . in this region , then , the function @xmath110 can be expressed as @xmath111 note here that @xmath105 and @xmath112 both depend on the initial @xmath32 . the same can be done by integrating the ray equation starting from the future . in this way we can define @xmath113 functions . ( the same procedure can be used to define @xmath114 and @xmath115 ) . it is worth noting that @xmath116 and @xmath113 are not null coordinates in the usual geometric sense , since they are frequency - dependent , but they nevertheless share many properties with null coordinates . the calculation of the inner product ( [ scalar_u ] ) involves the limit @xmath117 ( equivalently , @xmath118 ) . so , the general form @xmath119 can be simplified to @xmath120 . we can therefore change variables in the inner product from @xmath52 to @xmath121 . the combination of the derivative and the integral means that the form of is preserved in the new integration variable @xmath121 . then we can write the bogoliubov @xmath34 coefficients relevant for the hawking process as @xmath122_{w_{\omega}=+\infty}~. \label{kg - modified - product}\ ] ] now , assuming that the profiles vary slowly ( in scales much larger than the planck distance ) , the right - moving positive - energy modes associated with past and future infinity can be approximated by the following simple expressions : @xmath123 @xmath124 so that the bogoliubov coefficients read @xmath125 in analogy with the standard case , all information about the radiation is seen to be encoded in the relation @xmath126 . in this expression @xmath32 is the initial frequency of a ray at the past left infinity and @xmath127 its final frequency when reaching the future right infinity . the previous approximation for the bogoliubov coefficients amounts to considering profiles that vary slowly both with @xmath22 and @xmath24 . this is equivalent to considering large black holes . in general the quartic term in the wave equation , or equivalently the quartic modification of the dispersion relation , introduce a new source of backscattering , on top of the usual angular - momentum potential barrier which we have already neglected . in our approximation , this additional backscattering ( beyond the standard grey - body factors ) has been neglected . for large black holes this contribution will in any case be very small as has been observed in numerical simulations @xcite . in addition we are also neglecting any reflection caused by the kink . however , let us remark that given an approximative scheme for calculating @xmath126 for general profiles @xmath4 , the same @xmath126 obtained from a profile with a kink would be obtainable from one ( or several ) specific @xmath4 , this time perfectly smooth and thus causing no further backscattering . our results , which rely only on the specific form of the relation @xmath126 , are therefore valid beyond the specific configurations with a kink presented in this paper . our next task is to calculate the uniparametric family of functions @xmath128 and the relation between @xmath128 and @xmath121 for different configurations . as explained in section [ ss : rainbow ] , the relation @xmath126 is obtained by integrating the ray equation using the profiles discussed in section [ s : model ] . these can be described by means of stationary profiles @xmath7 and an auxiliary function @xmath14 , see the definition of @xmath4 in . we will use a straightforward extension of the procedure established in @xcite ( see also @xcite for a summary ) for a relativistic dispersion relation . care must be taken , however , with the quantities that depend on the frequency @xmath32 . in particular : * we will denote by @xmath27 and @xmath129 the position and the time at which the horizon associated with a particular initial frequency @xmath32 is formed . * the surface gravity @xmath13 , defined in , allows to write , for all @xmath130 and for @xmath22 close to @xmath27 : @xmath131 up to higher - order terms in @xmath132 . * the linearization of @xmath14 near @xmath129 also requires the introduction of an @xmath32-dependent parameter @xmath133 : @xmath134 again up to higher - order terms . let us consider laboratory times @xmath135 such that @xmath14 has already crossed the classical horizon at : . as we explained earlier we can define a critical frequency @xmath136 as the minimum initial frequency such that @xmath137 for all @xmath138 , or , in other words , the minimum frequency which at that particular time has not experienced any horizon yet . in the dynamical part of the profile , where @xmath139 , for rays crossing the kink just before the formation of the horizon , integration of the ray equation leads to @xmath140 for small values of @xmath141 ( see @xcite for details of the calculation in the case of a relativistic dispersion relation ) , where @xmath142 is the ( frequency - dependent ) time at which the kink separating the dynamical and the stationary regions is crossed , is the speed of light in the asymptotic past , and @xmath143 is the ray constituting the horizon associated with the frequency @xmath32 . in the stationary part of the profile , where @xmath144 , we obtain , for , again using a method and notation based on @xcite , @xmath145~,\ ] ] where @xmath146 is the speed of light in the asymptotic future , and @xmath147 a bulk integration constant . so @xmath148 both regions are connected by combining and . making use of , this gives @xmath149 this relation is valid for frequencies for which a horizon is experienced , i.e. for @xmath130 and times @xmath150 , where @xmath151 is the threshold time defined in section [ ss : relation_u(u ) ] ( which , unsurprisingly , has become frequency - dependent ) . again , as in the case of standard dispersion relations section [ ss : relation_u(u)]we can write @xmath152 where @xmath153 is essentially @xmath151 ( with possible higher - order corrections ) . assuming that the collapse takes place rapidly , it is a good approximation to replace @xmath153 by @xmath154 . actually , as we will see shortly , this is a conservative estimate , in the sense that it slightly underestimates the superluminal correction to the radiation spectrum . the relation @xmath155 we are considering interpolates between a linear behaviour at early times and an exponential behaviour at late times . it has the same form as the relation @xmath50 for standard dispersion relations , and so we can continue following the steps of the standard case . in particular , the equivalent of is obtained by integrating out @xmath113 . note that we use the subscript @xmath105 to emphasize that the @xmath113 are not the null coordinates of the standard case , but this should not be interpreted as an explicit function of @xmath105 and so does not complicate the integration steps in @xmath113 and @xmath105 . however , when integrating @xmath156 in @xmath32 to obtain @xmath157 , see section [ ss : wave - standard ] , we must carefully consider the frequency - dependence of the relevant terms , i.e. , of @xmath158 , @xmath159 and @xmath13 . the term carrying @xmath160 is moduloed away in , and we have replaced @xmath159 by @xmath161 , so the only relevant frequency - dependent factor that we are left with is the surface gravity @xmath13 . moreover , because of the critical frequency @xmath31 in the horizon formation process , a finite upper boundary will also be induced in the integral . indeed , frequencies @xmath88 do not contribute to the radiation at all , since they do not experience a horizon . this is a delicate but crucial point . it was already observed long ago by jacobson @xcite that trying to solve the transplanckian problem naively by imposing a cut - off frequency would seemingly extinguish hawking radiation on a relatively short time scale . in our case , however , this cut - off is not imposed _ ad hoc _ , but appears explicitly because of the superluminal character of the system at high frequencies . moreover , the critical frequency , and hence the upper boundary induced in the integral , depend directly on the physics inside the horizon . indeed , given a certain velocity profile , and in particular its behaviour near the center of the black hole , the critical frequency can be calculated by setting @xmath162 in eq . and extracting the corresponding critical frequency from the dispersion relation . we will see this effect graphically in section [ s : graphics ] . in analogy with eqs . and , we now obtain the number of particles detected for each frequency @xmath69 as @xmath163 \over \left[{\kappa_0 \over \kappa_{\omega'}}(z -z_{l,\omega'})\right]^2}\;{1 \over \exp({2\pi c\omega_j \over \kappa_{\omega'}})-1}\nonumber,\end{aligned}\ ] ] where now , with ( which corresponds to the standard @xmath164 of ) , and . changing the integration variable from @xmath32 to @xmath165 , we finally obtain the central expression in our analysis : @xmath166 \over \left[{\kappa_0 \over \kappa_{\omega'}}(z -z_{l,\omega'})\right]^2}\ ; { 1 \over \exp({2\pi c\omega_j \over \kappa_{\omega'}})-1}.\ ] ] note that , as we indicated earlier , the use of @xmath154 in the definition of @xmath167 is a conservative stance . indeed , strictly speaking , we should write @xmath168 . for a fixed @xmath169 , a smaller value of @xmath167 , and hence a larger value of @xmath153 , means that a larger part of the central peak of the integrand will be integrated over . since we are replacing @xmath153 by the upper bound @xmath154 , we are overestimating the resulting radiation ( i.e. , underestimating the modification with respect to the standard hawking radiation ) . the expression brings out the two crucial factors mentioned earlier , and a third , corollary one . * first , it shows the dependence of the total radiation on the critical frequency @xmath31 ( through the integration boundary @xmath169 induced by it ) , as discussed just before obtaining formula . * second , it shows the importance of the frequency - dependent @xmath13 ( to be compared with the fixed @xmath164 of the standard case ) . given a concrete profile @xmath4 , the frequency - dependence of @xmath13 can be derived explicitly , as we will discuss next . * finally , as a corollary of the first point , it shows that , as @xmath70 ( and hence @xmath167 ) increases , a smaller part of the central peak of the integrand will be integrated over , so the radiation will die off as @xmath70 advances . a careful analysis has shown that , so far , most formulas for the standard case could be adapted to superluminal dispersion relations by replacing the relevant magnitudes with their frequency - dependent counterparts . for example , @xmath170 was obtained by replacing @xmath51 and @xmath52 by @xmath116 and @xmath121 , and in particular @xmath54 , @xmath57 and @xmath164 by @xmath160 , @xmath159 and @xmath13 , respectively . given a concrete profile , we can explicitly deduce the relation between @xmath13 and @xmath32 , with the planck scale @xmath74 as a parameter , as follows . the horizon for a particular initial frequency @xmath32 is formed when @xmath171 i.e. , when @xmath172 , where we have used the phase velocity for concreteness . but again , qualitatively , our results would be similar if taking the group velocity @xmath79 instead of @xmath78 . taking into account that @xmath173 , the dispersion relation in the asymptotic past can be written as latexmath:[\[\begin{aligned } \label{omega - v_h } \omega'^2= given a concrete profile , @xmath27 can then be obtained and @xmath13 calculated explicitly . for the profiles of the first type , see fig . [ fig : v - constant - kappa ] , the result is trivial . indeed , @xmath28 increases linearly between the horizon corresponding to @xmath175 and the one for @xmath31 , so we obtain a constant @xmath176 for all @xmath130 . for a schwarzschild profile as in fig . [ fig : v - increasing - kappa ] , on the other hand , @xmath177 so we obtain @xmath178 note that @xmath13 remains nearly constant until frequencies of the order of magnitude of the planck scale are reached , and then starts to increase rapidly , see fig . [ fig : kappa - omega ] . as we will see graphically in the next section , this can have an important qualitative influence on the radiation spectrum . surface gravity @xmath13 with respect to the ( logarithm of the ) frequency for a schwarzschild - type black hole as in fig . [ fig : v - increasing - kappa ] , with the planck scale @xmath179 and the critical frequency @xmath180 ( where @xmath181 represents an arbitrary time unit . ) ] we now have all the tools necessary to compute and plot for different parameters of the profiles described in section [ s : model ] . we have calculated eq . numerically using the gauss - chebyshev quadrature method . the results are plotted in figs . [ fig : cut - off][fig : u_l ] and perfectly illustrate the three important factors that we deduced theoretically in section [ ss : wave - mod ] . note that in all the figures we have plotted @xmath182 against @xmath105 to make visual comparison with the usual thermal spectra in 3 + 1 dimensions easier . [ fig : cut - off ] shows the influence of the critical frequency @xmath31 for a profile with constant @xmath13 , as in fig . [ fig : v - constant - kappa ] , and @xmath183 ( i.e. , immediately after the horizons have formed for all @xmath130 ) . the constant @xmath13 guarantees that the form of the thermal spectrum is preserved . however , the intensity of the radiation decreases with decreasing critical frequency . actually , only for extremely high critical frequencies is the original hawking spectrum recovered . for a critical frequency still well above the planck scale , the decrease can be significant . for example , for @xmath184 , the peak intensity decreases by nearly 30% , while for @xmath185 , the decrease is approximately 40% . at the other end , note that extremely low critical frequencies still leave a significant amount of radiation . for example , for @xmath186 , still 20% of the original peak radiation is obtained . so hawking radiation receives significant contributions from an extremely wide range of frequencies . influence of the critical frequency on the radiation spectrum for a black hole with velocity profile such that the surface gravity is constant , as in fig . [ fig : v - constant - kappa ] , and different values of the critical frequency . we have chosen @xmath187 , @xmath188 , @xmath189 , @xmath190 and the planck scale @xmath179 ( where @xmath181 represents an arbitrary time unit ) , which amounts to considering a solar - mass black hole . from top to bottom we have plotted @xmath191 . note that the standard hawking spectrum coincides perfectly with the upper curve , which effectively corresponds to the absence of a critical frequency . ] [ fig : increasing - kappa ] shows the radiation spectrum for a schwarzschild - like profile and hence increasing @xmath13 , as in fig . [ fig : v - increasing - kappa ] , in particular for critical frequencies @xmath31 close to the planck scale . on top of the general decrease of the standard thermal part of the spectrum ( approximately 40% , as in the previous case ) due to the finite integration boundary induced by the critical frequency , the fact that the surface gravity is now frequency - dependent leads to an important qualitative change of the spectrum . the high - frequency tail of the spectrum is totally transformed . actually , if the critical frequency is sufficiently higher than the planck scale , the dominant source of radiation lies in the high - frequency region . note that this effect is truly a consequence of the modification of the physics for frequencies above the planck scale . this can be appreciated by noticing that , at @xmath192 , the whole tail - modifying effect has disappeared and the usual thermal form of the radiation spectrum is recovered ( although still with the quantitative decrease described above ) . influence of a frequency - dependent surface gravity @xmath13 on the radiation spectrum for a black hole with schwarzschild - type velocity profile ( surface gravity increases with frequency ) , as in fig . [ fig : v - increasing - kappa ] , and different values of the critical frequency around the planck scale @xmath74 : ( from top to bottom ) . the standard hawking spectrum is depicted in dashed dotted line for comparison . numerical values of @xmath193 , @xmath74 etc as in fig . [ fig : cut - off ] . ] finally , fig . [ fig : u_l ] shows the influence of the measuring time @xmath70 ( measured with respect to @xmath161 ) for a profile of the second type ( increasing @xmath13 ) . it is clearly seen that the radiation dies off with time , and actually dies off rather fast . for an actual solar - mass black hole and an @xmath31 of the order of the planck scale , the radiation would last only a few tens of milliseconds . note that this effect is a corollary of the existence of a critical frequency @xmath31 , since in its absence , the integration boundaries in would be infinite , and so the integral would be insensitive to a change @xmath194 . the combined effect of fig . [ fig : increasing - kappa ] and fig . [ fig : u_l ] leads to the following qualitative picture for the further collapse towards a singularity once the initial or classical horizon has formed . as the interior gradually uncovers a larger portion of the schwarzschild geometry , two competing processes will take place . on the one hand , the spectrum acquires ever larger contributions associated with higher and higher temperatures . on the other hand , the overall magnitude of the spectrum is damped with time . the question of which process dominates would depend on the fine details of the dynamics of the collapse , and might be further complicated by backreaction effects , which we have not considered in our analysis . influence of the measuring time @xmath70 on the radiation spectrum , for a schwarzschild - type velocity profile , see fig . [ fig : v - increasing - kappa ] and @xmath195 , compare with fig . [ fig : increasing - kappa ] . different values of @xmath70 ( from top to bottom ) : @xmath196 with @xmath181 denoting an arbitrary time unit . the standard hawking spectrum is depicted in dashed dotted line for comparison . numerical values of @xmath193 , @xmath74 etc as in fig . [ fig : cut - off ] . ] we have discussed the hawking radiation for a collapsing configuration with superluminal dispersion relations . modifications of the dispersion relation cause the horizon , and various associated quantities such as the surface gravity @xmath164 , to become frequency - dependent . in particular , a critical frequency @xmath31 naturally appears such that frequencies higher than @xmath31 do not experience a horizon at all . more generally , it also means that the standard geometric concepts traditionally used to study black holes must be handled with care . nevertheless , through a detailed analysis , we have seen that the equations related to the late - time radiation can be adapted quite straightforwardly from standard ( relativistic ) to superluminal dispersion relations . we analytically derived an approximate equation for the particle production at late times with superluminal modifications of the dispersion relations . this equation clearly showed that important modifications in the late - time radiation should be expected , first , due to the existence of the critical frequency @xmath31 and the finite upper boundary it induces in the integration , and second , due to the frequency - dependence of @xmath13 . we integrated numerically , and plotted the resulting spectrum , thereby confirming these expectations . we have seen that the standard hawking spectrum is recovered only in a very particular case : in the limit when the critical frequency goes to infinity ( i.e. , when the profile for the velocity @xmath197 goes to a singularity ) and moreover the surface gravity @xmath13 is constant ( linear velocity profiles of the type of fig . [ fig : v - constant - kappa ] ) . for lower critical frequencies , as long as @xmath164 remains constant , the thermal form of the hawking spectrum is maintained , but the intensity decreases rapidly with decreasing @xmath31 . a non - negligible radiation persists , however , even for extremely low critical frequencies . this clearly establishes the first point , namely the importance of the critical frequency @xmath31 , which illustrates the statement made in the introduction that with superluminal dispersion relations , the interior of the black hole is probed , which can significantly affect the hawking process . for a schwarzschild - like profile , the velocity profile generally leads to a frequency - dependent surface gravity @xmath13 , increasing with the frequency @xmath32 . this means that the standard hawking result can not be recovered for a schwarzschild black hole with superluminal dispersion relations , even when the velocity profile has a singularity . actually , when the velocity profile has a limiting value , the same quantitative decrease of the hawking part of the spectrum as before shows up . moreover , if the critical frequency @xmath31 is above the planck scale , a drastic qualitative change of the radiation spectrum takes place , and for sufficiently high @xmath31 ( roughly a few times the planck scale ) the high - frequency part of the spectrum even becomes dominant . this shows the importance of the surface gravity , which again illustrates the role played by the interior of the black hole . finally , we have also seen that , as a corollary of the existence of a critical frequency and of the finite upper boundary induced by it in the thermal response function , the radiation spectrum dies off as time advances . a few observations might be useful to connect our work with existing results on hawking radiation and its sensitivity to modified dispersion relations . the most general observation is that the ` robustness ' of hawking radiation , which is often considered to be a well - established result , actually depends crucially on a series of assumptions . these assumptions might be reasonable in the case of subluminal dispersion relations . but for the case of superluminal dispersion relations , as we have shown explicitly , the assumptions needed to reproduce the standard hawking result depend on the physics inside the ( zero - frequency ) horizon , and moreover in a way which is not compatible with the schwarzschild geometry . in particular , in @xcite , it was shown for a stationary scenario that hawking radiation is robust with respect to superluminal modifications of the dispersion relations , provided that positive free fall frequency modes were in their ground state just before crossing the horizon . however , it was also admitted that it is not clear whether this is the physically correct quantum state condition . in @xcite , three explicit assumptions were given for the previous condition to hold : freely falling frame , ground state and adiabatic evolution . in the same article , it is similarly pointed out that these assumptions might fail for a superluminal dispersion relation , since ( as we also mentioned in the introduction ) for high - frequency modes this implies that one makes an assumption about the physics at the singularity . rather than imposing any conditions on the state near the horizon , and hence ultimately at the singularity , we have fixed the _ initial _ geometry to be minkowskian and the quantum field to be in the natural associated vacuum state , and evolved this into a black hole configuration . since our results are seemingly in contradiction with the ones obtained in @xcite and @xcite , it is worth examining in which sense the conditions stated by those authors are violated in our approach . if one considers a collapse scenario , for example of a bec , in a laboratory setting , then it is quite natural to assume that the `` freely falling frame '' condition will be violated . indeed , the superluminal modification is in this case associated with the existence of a privileged external reference frame : the black hole rest frame or lab frame ( as we noted in sec . [ ss : generalization_inner_product ] ) , and not the freely falling frame , as assumed in @xcite . note that this violation of the free - fall frame condition automatically implies a violation of the ground state condition in the sense in which this condition is formulated in @xcite , namely as the ground state `` with respect to the freely falling frame '' . indeed , our results suggest the following picture . the low - energy modes experience the classical geometry , are therefore dragged along in the free - falling frame and so at the horizon they occupy the vacuum state associated with this free - falling frame , namely the unruh vacuum . hence they contribute to the black hole radiation in the traditional hawking way . however , the ultra - high - frequency modes ( above the critical frequency ) do not see the horizon . hence they do not couple to the classical geometry of the collapse , but rather remain connected to the external or laboratory frame , and therefore pass through the black hole ( nearly ) undisturbed . the ground state of these high - frequency modes is then not the vacuum associated with the freely falling frame , but the boulware vacuum associated with the original minkowski geometry , or in other words , with the stationary reference frame of the lab . so these frequencies above the critical frequency do not contribute to the thermal output spectrum . the overall radiation is a convolution of the contributions from all the different initial frequencies , where the surface gravity for each frequency can be interpreted as leading to an effective weighting factor . depending on the internal physics of the black hole , this leads to either a spectrum of the traditional hawking form but with a reduced intensity ( in the case of a constant surface gravity ) , or a modified spectrum where the high - frequency contributions dominate ( as for a schwarzschild profile , provided that the critical frequency lies sufficiently above the planck scale ) . in any case , what our analysis shows with respect to the situation treated in @xcite and @xcite is the following . the robustness of hawking radiation discussed by those authors is the robustness with respect to a superluminal modification under the precise conditions occurring in the standard case , in which all modes occupy the ground state associated with the free - falling frame just before crossing the horizon . this assumption implies that all modes , regardless of their frequency , occupy the unruh vacuum . our analysis shows precisely that the superluminality can lead to a spontaneous breaking of these conditions . the natural evolution of a collapsing configuration is such that the frequencies above the critical frequency @xmath31 are not in the unruh vacuum associated with the freely falling frame , but remain in the boulware vacuum associated with the initial asymptotic minkowski condition . it should then come as no surprise that this can indeed have an important impact on the resulting radiation spectrum . we should mention that in a recent paper by schtzhold and unruh @xcite based on an alternative approximation method , a new potential problem with superluminal dispersion relations is highlighted . this `` ultra - violet catastrophe '' is associated with a higher - than - linear growth of the group velocity at large wave numbers , and should therefore not occur for the dispersion relation used in our work . however , we have found that a schwarzschild interior geometry could give place to the same sort of phenomenology . a final observation concerns the recent article @xcite , in which the authors numerically simulate the formation of an acoustic horizon in a bec , and analyze the creation of correlated pairs of phonons through the so - called truncated wigner method . our findings for the case of a constant surface gravity seem to be in qualitative disagreement with the discussion presented in @xcite , since the authors of this paper assert having observed a stationary hawking flux , while our analysis concludes that the radiation should quickly fade off . the source of this apparent discrepancy might reside in the fact that the correlation function which they study is normalized , and therefore probes the form of the spectrum ( which we also find stationary with time ) , but does not provide details about the net amount of particle production . we kindly thank stefano liberati and matt visser for enlightening comments and discussions , and jos luis jaramillo for very useful suggestions on the numerical aspects of this work . g.j . would also like to express his gratitude to fernando barbero , daniel gmez vergel and lorenzo sindoni for useful discussions , and sissa ( trieste ) for hospitality . financial support was provided by the spanish mec through the projects fis2005 - 05736-c03 - 01 , fis2005 - 05736-c03 - 02 and fis2006 - 26387-e . 99 t. jacobson , s. liberati and d. mattingly , `` lorentz violation at high energy : concepts , phenomena and astrophysical constraints , '' annals phys . * 321 * , 150 ( 2006 ) [ arxiv : astro - ph/0505267 ] . n. e. mavromatos , `` lorentz invariance violation from string theory , '' pos qg - ph:027 ( 2007 ) [ arxiv:0708.2250 [ hep - th ] ] . r. gambini and j. pullin , `` nonstandard optics from quantum spacetime , '' phys . d * 59 * , 124021 ( 1999 ) [ arxiv : gr - qc/9809038 ] . a. ashtekar , `` loop quantum gravity : four recent advances and a dozen frequently asked questions , '' arxiv:0705.2222 [ gr - qc ] . c. barcel , s. liberati and m. visser , `` analogue gravity , '' living rev . * 8 * , 12 ( 2005 ) [ arxiv : gr - qc/0505065 ] . g. e. volovik , `` the universe in a helium droplet , '' clarendon press , oxford ( 2003 ) . f. r. klinkhamer and g. e. volovik , `` merging gauge coupling constants without grand unification , '' pisma zh . fiz . * 81 * , 683 ( 2005 ) [ jetp lett . * 81 * , 551 ( 2005 ) ] [ arxiv : hep - ph/0505033 ] . d. mattingly , `` modern tests of lorentz invariance , '' living rev . * 8 * , 5 ( 2005 ) [ arxiv : gr - qc/0502097 ] . j. magueijo and l. smolin , `` lorentz invariance with an invariant energy scale , '' phys . lett . * 88 * , 190403 ( 2002 ) [ arxiv : hep - th/0112090 ] . s. w. hawking , `` black hole explosions , '' nature * 248 * , 30 ( 1974 ) . s. w. hawking , `` particle creation by black holes , '' commun . phys . * 43 * 199 ( 1975 ) . i. agullo , j. navarro - salas , g. j. olmo and l. parker , `` short - distance contribution to the spectrum of hawking radiation , '' phys . d * 76 * , 044018 ( 2007 ) [ arxiv : hep - th/0611355 ] . w. g. unruh , `` experimental black hole evaporation , '' phys . rev . * 46 * , 1351 ( 1981 ) . t. jacobson , `` black hole evaporation and ultrashort distances , '' phys . d * 44 * , 1731 ( 1991 ) . t. jacobson , `` black hole radiation in the presence of a short distance cutoff , '' phys . d * 48 * , 728 ( 1993 ) [ arxiv : hep - th/9303103 ] . w. g. unruh , `` sonic analog of black holes and the effects of high frequencies on black hole evaporation , '' phys . d * 51 * , 2827 ( 1995 ) . r. brout , s. massar , r. parentani and p. spindel , `` hawking radiation without transplanckian frequencies , '' phys . d * 52 * , 4559 ( 1995 ) . [ arxiv : hep - th/9506121 ] . s. corley and t. jacobson , `` hawking spectrum and high frequency dispersion , '' phys . rev . d * 54 * , 1568 ( 1996 ) . [ arxiv : hep - th/9601073 ] . s. corley , `` particle creation via high frequency dispersion , '' phys . d * 55 * , 6155 ( 1997 ) . s. corley , `` computing the spectrum of black hole radiation in the presence of high frequency dispersion : an analytical approach , '' phys . d * 57 * , 6280 ( 1998 ) . [ arxiv : hep - th/9710075 ] . w. g. unruh and r. schtzhold , `` on the universality of the hawking effect , '' phys . d * 71 * , 024028 ( 2005 ) . [ arxiv : gr - qc/0408009 ] . r. schtzhold and w. g. unruh , `` on the origin of the particles in black hole evaporation , '' phys . d * 78 * , 041504 ( 2008 ) [ arxiv:0804.1686 [ gr - qc ] ] . m. visser , `` essential and inessential features of hawking radiation , '' int . j. mod . d * 12 * , 649 ( 2003 ) . [ arxiv : hep - th/0106111 ] . c. barcel , a. cano , l. j. garay and g. jannes , `` stability analysis of sonic horizons in bose - einstein condensates , '' phys . rev . d * 74 * , 024008 ( 2006 ) [ arxiv : gr - qc/0603089 ] . s. corley and t. jacobson , `` black hole lasers , '' phys . d * 59 * , 124011 ( 1999 ) [ arxiv : hep - th/9806203 ] . l. j. garay , j. r. anglin , j. i. cirac and p. zoller , `` black holes in bose - einstein condensates , '' phys . lett . * 85 * , 4643 ( 2000 ) [ arxiv : gr - qc/0002015 ] . l. j. garay , j. r. anglin , j. i. cirac and p. zoller , `` sonic black holes in dilute bose - einstein condensates , '' phys . a * 63 * , 023611 ( 2001 ) [ arxiv : gr - qc/0005131 ] . c. barcel , s. liberati and m. visser , `` analog gravity from bose - einstein condensates , '' class . * 18 * , 1137 ( 2001 ) [ arxiv : gr - qc/0011026 ] . c. barcel , s. liberati and m. visser , `` towards the observation of hawking radiation in bose - einstein condensates , '' int . j. mod . a * 18 * , 3735 ( 2003 ) [ arxiv : gr - qc/0110036 ] . r. schtzhold , `` on the detectability of quantum radiation in bose - einstein condensates , '' phys . * 97 * , 190405 ( 2006 ) [ arxiv : quant - ph/0602180 ] . s. wster and c. m. savage , `` limits to the analogue hawking temperature in a bose - einstein condensate , '' phys . a * 76 * , 013608 ( 2007 ) [ arxiv : cond - mat/0702045 ] . r. balbinot , a. fabbri and s. fagnocchi , `` non - local density correlations as signal of hawking radiation in bec acoustic black holes , '' phys . rev . a * 78 * , 021603 ( 2008 ) [ arxiv:0711.4520 [ cond-mat.other ] ] . i. carusotto , s. fagnocchi , a. recati , r. balbinot and a. fabbri , `` numerical observation of hawking radiation from acoustic black holes in atomic becs , '' new j. phys . * 10 * , 103001 ( 2008 ) [ arxiv:0803.0507 [ cond-mat.other ] ] . s. wster , `` phonon background versus analogue hawking radiation in bose - einstein condensates , '' phys . a * 78 * , 021601(r ) ( 2008 ) [ arxiv:0805.1358 [ cond-mat.other ] ] . m. visser , `` acoustic black holes : horizons , ergospheres , and hawking radiation , '' class . grav . * 15 * , 1767 ( 1998 ) [ arxiv : gr - qc/9712010 ] . a. fabbri and j. navarro - salas , `` modeling black hole evaporation , '' imperial college press , london ( 2005 ) . m. visser and s. weinfurtner , `` analogue spacetimes : toy models for `` quantum gravity '' , '' pos qg - ph:042 ( 2007 ) [ arxiv:0712.0427 [ gr - qc ] ] . m. visser , `` emergent rainbow spacetimes : two pedagogical examples , '' arxiv:0712.0810 [ gr - qc ] . c. barcel , s. liberati , s. sonego and m. visser , `` quasi - particle creation by analogue black holes , '' class . * 23 * , 5341 ( 2006 ) . [ arxiv : gr - qc/0604058 ] . c. barcel , s. liberati , s. sonego and m. visser , `` hawking - like radiation does not require a trapped region , '' phys . lett . * 97 * , 171301 ( 2006 ) . [ arxiv : gr - qc/0607008 ] .

observational evidence exists for the formation of gas giant planets on wide orbits around young stars by disk gravitational instability , but the roles of disk instability and core accretion for forming gas giants on shorter period orbits are less clear . the controversy extends to population synthesis models of exoplanet demographics and to hydrodynamical models of the fragmentation process . the latter refers largely to the handling of radiative transfer in three dimensional ( 3d ) hydrodynamical models , which controls heating and cooling processes in gravitationally unstable disks , and hence dense clump formation . a suite of models using the cooling approximation is presented here . the initial disks have masses of 0.091 and extend from 4 to 20 au around a 1 protostar . the initial minimum toomre values range from 1.3 to 2.7 , while ranges from 1 to 100 . we show that the choice of is equal in importance to the value assumed : high disks can be stable for small , when the initial disk temperature is taken as a lower bound , while low disks can fragment for high . these results imply that the evolution of disks toward low must be taken into account in assessing disk fragmentation possibilities , at least in the inner disk , i.e. , inside about 20 au . the models suggest that if low disks can form , there should be an as yet largely undetected population of gas giants orbiting g dwarfs between about 6 au and 16 au .

1701.04892

two end - member mechanisms have been proposed for gas giant planet formation , core accretion ( e.g. , mizuno 1980 ) and gas disk gravitational instability ( e.g. , cameron 1978 ) . the former depends on the reliable mechanism of the collisional accumulation of solid bodies , widely accepted as the formation mechanism for the terrestrial planets ( e.g. , wetherill 1990 , 1996 ) . if a growing solid core should exceed a critical core mass ( mizuno 1980 ) , its atmosphere will become unstable and the core will rapidly gain mass by accretion of gas and solids from the surrounding disk ( e.g. , pollack et al . 1996 ) , resulting in a gas giant planet . disk instability , on the other hand , relies on the disk being massive and cold enough to become gravitationally unstable , forming spiral arms that grow and collide to form dense , self - gravitating clumps of gas and dust that might collapse and contract to form giant gaseous protoplanets ( e.g. , boss 1997 ) . hybrid mechanisms may also be considered for forming terrestrial and gas giant planets ( e.g. , boss 1998 ; nayakshin 2010 ) . lissauer & stevenson ( 2007 ) presented a review of the core accretion mechanism , while durisen et al . ( 2007 ) reviewed the disk instability mechanism . more recently , helled et al . ( 2014 ) reviewed both mechanisms , listing the strengths and weaknesses of each end - member , without reaching a definite conclusion about either mechanism . rather , helled et al . ( 2014 ) called for considerably more research on specific theoretical problems for both mechanisms . one problem highlighted by helled et al . ( 2014 ) is the question about whether or not disk instability can lead successfully to giant protoplanet formation inside about 20 au , where high optical depths in the disk mean that radiative cooling is crucial for clump formation and survival . that challenge is the focus of this paper . we begin with a brief summary of the current status of observational constraints on giant planet formation models , both from studies of protoplanetary disks and from exoplanet detection surveys , before turning to a reconsideration of the disk radiative cooling problem that is the primary motivation for this new suite of disk instability models . observational constraints on planet formation models range from estimates of the initial conditions for planet formation , namely studies of protoplanetary disks , which provide the feedstock for planetary systems , to exoplanet demographics , the final result of the planet formation process . disk masses , temperatures , and lifetimes are key discriminators for comparing the disk instability and core accretion mechanisms . the most essential constraint is having sufficient disk mass to form a planetary system , by either core accretion or disk instability . the minimum mass solar nebula was estimated by weidenschilling ( 1977 ) to be in the range of 0.01 to 0.07 @xmath1 , but most citations refer only to the _ minimum _ of this range ( 0.01 @xmath1 ) as being the minimum mass solar nebula . this estimate also assumes a 100% efficient planet formation process , and so in reality these estimates of 0.01 to 0.07 @xmath1 represent _ lower bounds _ on the mass of the protoplanetary disk needed to produce our solar system ( weidenschilling 1977 ) . from this point of view , the solar nebula might well have had a mass of @xmath3 or more . considerably more information on the range of protoplanetary disk masses is obtained by millimeter - wave astronomy . millimeter wave observations of 11 low- and intermediate - mass pre - main - sequence stars implied that their disks formed with masses in the range from 0.05 to 0.4 @xmath1 ( isella et al . submillimeter wave observations of t tauri stars in the @xmath4 1 myr old ophiuchus star - forming region imply disk masses in the range from 0.004 to 0.143 @xmath1 ( andrews et al . similar observations of young stars in the even younger ( @xmath4 0.5 myr old ) cluster ngc 2024 found disk masses ranging from @xmath4 0.003 to 0.2 @xmath1 ( mann et al . 2015 ) , with @xmath4 10% having disk masses greater than 0.1 @xmath1 , a considerably higher fraction than in the older taurus star - forming region . millimeter wave observations of class ii disks in the taurus star - forming region imply that the inferred dust disk masses scale roughly proportionally to the masses of their host stars ( andrews et al . however , estimates of total disk masses have long suffered from the uncertain conversion of the observed mass of the growing dust grains ( e.g. , banzatti et al . 2011 ) that produce the submillimeter continuum emission to the inferred gas disk mass , potentially leading to substantial underestimates of the total disk masses ( andrews & williams 2007 ) . high optical depth midplanes , even at millimeter wavelengths , can lead to weak dust continuum emission and hence underestimates of the true disk mass ( forgan et al . disk mass estimates based on observations of molecular species also depend on the assumed grain size distribution , leading to underestimated disk masses by as much as a factor of 10 ( miotello et al . 2014 ; dunham et al . however , observations of the isotopologues of co gas may offer a direct estimate of the mass of gas in protoplanetary disks ( miotello et al . 2016 ) . disks with masses of @xmath4 25% that of their host star will be gravitationally unstable and produce spiral arms similar to those in two protoplanetary disks , mwc 758 and sao 206462 ( dong et al . the young stellar object ( yso ) elias 2 - 27 , with a mass of @xmath5 , has a protoplanetary disk with a mass of @xmath6 and trailing , symmetric spiral arms that extend to the disk midplane ( prez et al . episodic fu orionis outbursts in young solar - type stars are best explained by disks that are at least partially gravitationally unstable ( e.g. , zhu et al . 2007 , 2009 , 2010 ; vorobyov & basu 2010a , b ; liu et al . 2016 ) . taken together , all of these observations suggest that a significant fraction of protoplanetary disks are still gravitationally unstable during their earliest phases of evolution , when massive protostellar disks transition into less - massive protoplanetary disks . disk lifetimes are another key constraint , especially for the core accretion mechanism , where gas disk lifetimes of @xmath4 1 myr or longer are typically required to form gas giant planets ( e.g. , pollack et al . 1996 ; hubickyj et al . 2005 ) . the conventional wisdom , based on ir excesses , is that typical disk lifetimes are in the range of 1 to 10 myr , with an average age of @xmath4 3 myr ( e.g. , figure 11 in hernandez et al . 2008 ) . these statistics refer to young stars of different evolution classes in a variety of stellar groups . when restricting the sample to include only class iii sources , cieza et al . ( 2007 ) found that roughly @xmath4 50% of the youngest of 230 weak - line t tauri stars showed no evidence for ir excesses , suggesting that half of the disks dissipate on time scales of @xmath4 1 myr or less . these observations refer solely to ir excesses , wavelengths where the disks are optically thick and the dust disk mass is difficult to estimate . however , protoplanetary disks are optically thin at submillimeter wavelengths , making submillimeter dust continuum observations capable of producing more reliable dust disk masses , and hence gas disk masses . a submillimeter survey of nearly 300 ysos in the @xmath7 orionis cluster detected only 9 disks , finding a mean disk mass of 0.5 jupiter masses ( @xmath8 ) for these ysos with ages of @xmath4 3 myr ( williams et al . 2013 ) , and implying that gas giant planet formation needs to finish well within 3 myr . a similar survey of the slightly younger ( 2 - 3 myr old ) cluster ic 348 found disk masses in the range of 1.5 to 16 @xmath8 ( cieza et al . 2015 ) for 13 out of the 370 cluster members , implying that disks with at least the mass ( @xmath9 ) of the lower bound on the minimum mass solar nebula ( weidenschilling 1977 ) are very rare , occurring less than about 1% of the time at the age of the ic 348 cluster . in contrast , near - ir excesses occur for roughly half of the stars in ic 348 and @xmath7 orionis ( e.g. , hernandez et al . 2008 ) , showing that the conventional wisdom about gas disk lifetimes has been severely skewed by a reliance on ir excesses . the submillimeter surveys imply a typical gas disk lifetime closer to 1 myr than to 3 myr , at least for disks capable of forming gas giant planets . disk midplane temperatures , along with disk surface densities , determine the toomre ( 1964 ) @xmath10 stability value , and hence the propensity for a disk to be gravitationally unstable . the solar nebula is the protoplanetary disk for which one would think we have the best information about its properties . cometary compositions are often used to place constraints on solar nebula temperatures at their time of formation . molecular abundance ratios determined for the comet 67p imply that if its ice grains agglomerated from clathrates ( mousis et al . 2016 ) , then this jupiter - family comet must have formed in a region of the solar nebula with temperatures in the range of about 32 k to 70 k ( lectez et al . 2015 ) . jupiter - family comets are believed to have been formed beyond neptune , in the kuiper belt , whereas oort cloud comets are thought to have formed inward of neptune , where jupiter and the other giant planets could scatter them outward to much longer period orbits . comet c/1999 s4 is a long period comet from the oort cloud , containing ammonia ices with ortho - to - para ratios ( opr ) that imply formation at a temperature of about 28 k ( kawakita et al . 2001 ) . measurements of the opr for cometary water have often been used to infer even lower nebular temperatures , in the range of 10 k to 20 k ( e.g. , hogerheijde et al . however , hama et al . ( 2016 ) have disputed the commonly assumed relationship between water opr and temperature , casting some doubt on use of the water opr as a cometary thermometer for the solar nebula . other disk temperature estimates come from observations of protoplanetary disks around young solar - type stars . observations of the dm tau outer disk , on scales of 50 to 60 au , imply midplane temperatures of 13 to 20 k ( dartois et al . 2003 ) , with surface temperatures of @xmath4 30 k. theoretical disk models that reproduce the observed spectral energy distributions for t tauri - star disks typically predict disk midplane temperatures of @xmath4 200 k at 1 au , @xmath4 40 k at 10 au , and @xmath4 15 k at 100 au ( lachaume et al . 2003 ; dalessio et al . these midplane temperatures are low enough that provided the disk is massive enough , the disk is likely to be gravitationally unstable . the spectacular images of the hl tau protoplanetary disk ( alma partnership et al . 2015 ) show several gap - like structures centered on the @xmath11 central protostar . hl tau is less than 1 - 2 myr old and has a disk mass of @xmath12 . models suggest that the three main gaps at 15 , 35 , and 70 au could be caused by embedded protoplanets with masses of 0.2 , 0.27 , and 0.55 @xmath8 ( dipierro et al . 2015 ) . a direct imaging search for embedded planets in the 70 au gap placed only upper limits of @xmath13 on the unseen objects ( testi et al . 2015 ) . testi et al . ( 2015 ) noted that boss ( 2011 ) showed that the hl tau disk could form a @xmath14 planet at 70 au by disk instability , a planet massive enough to cause the gap , but not massive enough to have been directly imaged . akiyama et al . ( 2016 ) have shown that all of the observed gaps in the hl tau disk are consistent with gas giant formation by disk instability , coupled with inward migration . embedded planets can also generate spiral arms in marginally gravitationally unstable disks ( pohl et al . carrasco - gonzalez et al . ( 2016 ) have used the vla to study the hl tau disk , finding that the inner disk has fragmented and formed dense clumps , suggesting gravitational instability as the cause ( see also mayer et al . the hl tau system thus may represent the foremost poster child for gas giant planet formation by disk gravitational instability . the hl tau disk is threaded with a magnetic field that is coincident with its major axis ( stephens et al . 2014 ) , implying that the field does not control the dynamics of the disk . in a related vein , recent work on the allende meteorite has shown that its magnetization was derived from its parent body , not the solar nebula ( fu et al . 2014 ) , implying that the dynamics of the midplane of the solar nebula were not dominated by magnetic forces . hence non - magnetic disk instability models appear to be relevant to both the hl tau disk and the solar nebula , and presumably to other protoplanetary disks as well . the ultimate constraints on theoretical models of planetary formation are derived from the results of the ongoing world - wide effort to determine the population statistics of exoplanets with widely varying masses , mean densities , atmospheric compositions , and orbital properties . the spectacular discoveries of thousands of exoplanets by the kepler mission ( borucki et al . 2010 , 2011a , b ) have revolutionized the field , and made possible direct comparisons with the predictions of exoplanet population synthesis ( epops ) models based on the core accretion mechanism . perhaps most dramatically , the kepler detections imply that the number of planets per star appears to be a roughly monotonically increasing function of decreasing exoplanet radius ( fressin et al . 2013 ) , at least down to radii of @xmath15 ( earth radii ) , contrary to the predictions of epops models . early epops models ( e.g. , ida & lin 2005 , 2008 ) were based on the population uncovered primarily by radial velocity ( rv ) and ground - based transit photometry surveys and were able to select model parameters that fit reasonably well the exoplanets known at the time . these models ( e.g. , figure 3 in ida & lin 2008 ) predicted a dearth of planets with masses in the super - earth mass range and short period orbits ( i.e. , inside @xmath4 1 au ) , the so - called _ exoplanet desert _ ( ida & lin 2004 ) , as well as a surplus of hot jupiters and hot super - earths orbiting at @xmath4 0.03 au . on the contrary , kepler found an _ exoplanet oasis _ where the epops predicted a desert : batalha ( 2014 ) showed that kepler had found the highest frequency of exoplanets in the region where the desert had been predicted , and no evidence for a pile - up of hot exoplanets with orbital periods of a few days . the same problem besets the more recent epops models of alibert et al . ( 2011 ) and mordasini et al . epops models with a new prescription for orbital migration have filled in part of the desert , but only at the cost of over - producing massive gas giants and shifting the desert upward to higher masses ( see figure 8 in dittkrist et al . 2014 ; see also coleman & nelson 2014 ) . evidently epops models based solely on core accretion are not yet able to find a combination of parameters that enables them to reproduce the kepler findings . epops models that include gas giant planet formation by disk instability ( e.g. , forgan & rice 2013 ) might offer a means to escape some of the inward orbital migration and surplus gas giant planet problems that beset pure core accretion epops models by shortening gas disk lifetimes , and by stopping cores from migrating inward and growing to become gas giants . the kepler mission s 4-yr survey of the field in cygnus - lyra prohibited the detection of exoplanets with orbital periods greater than about 1 yr . other techniques shed light on the demographics at greater distances . e.g. , an rv survey of 202 solar - type stars that had been followed for 17 yrs found a frequency of @xmath4 6% for giant planets orbiting from 3 to 7 au ( wittenmyer et al . 2016 ) . the combination of adaptive optics ( ao ) imaging and a long - term rv survey found a total occurrence rate of @xmath16% for exoplanet masses in the range of 1 - 20 @xmath8 at distances of 5 to 20 au ( bryan et al . 2016 ) , with a suggestion that most have orbital distance less than 10 au . this fraction is higher than that found earlier by cumming et al . ( 2008 ) , who found that 8 yrs of rv data implied that up to 20% of stars have gas giants within 20 au . evidently longer survey periods find increasing frequencies of long - period exoplanets , as expected . hence even the current estimates should be considered lower bounds . microlensing surveys also sample a different portion of exoplanet discovery space compared to transit photometry surveys like kepler , namely orbital distances comparable to the einstein radius , which means orbital distances of a few au and orbital periods of a few years for solar - type stars . cassan et al . ( 2012 ) found their microlensing data to reveal that @xmath4 17% of stars host planets with masses in the range of 0.3 to 10 @xmath8 , while lower mass planets are even more common . with more recent data , shvartzvald et al . ( 2016 ) found that @xmath4 55% of these low mass ( k - m dwarf ) stars hosted planets , with neptune - mass planets being about 10 times as common as jupiter - mass planets . sumi et al . ( 2011 ) interpreted short - time - scale microlensing events ( less than 2 days ) as evidence for a population of unbound or distant jupiter - mass objects that are about twice as common as main - sequence stars . sumi et al . ( 2011 ) argued that these objects were most likely to be unbound gas giants that were ejected from protoplanetary disks , but veras & raymond ( 2012 ) found that planet - planet scattering could not be expected to form that many objects . their analysis was restricted to gas giants formed by core accretion , i.e. , their planets were initially located at a few au from their stars . ( 2016 ) similarly found that core accretion could not produce the number of ejected planets implied by the sumi et al . ( 2011 ) observations . forgan et al . ( 2015 ) and vorobyov ( 2016 ) , on the other hand , propose that disk fragmentation in the outer disk regions will lead to the frequent ejection of brown dwarfs and gas giants , perhaps explaining the sumi et al . ( 2011 ) results . even if these objects are bound , their orbits must lie beyond 10 au from their stars , a region where disk instability is likely to be the main formation mechanism , rather than core accretion ( see below ) . contrary to the assertions by sumi et al . ( 2011 ) , direct imaging surveys searching for such distant but bound gas giants do not exclude the microlensing population from being bound objects ( quanz et al . either way , gravitational instability seems to be implicated in explaining the sumi et al . ( 2011 ) gas giant population . the discovery of the hr8799 bcde exoplanet system ( marois et al . 2008 , 2010 ) is as remarkable in its own way as the millimeter - wave image of the hl tau disk ( alma partnership et al . this system consists of _ four _ gas giant planets with masses of about 5 to 7 @xmath8 , orbiting at distances of about 14 , 24 , 38 , and 68 au from an a5 star of 1.5 @xmath1 with an age of @xmath4 30 myr . 51 eridani is a younger ( @xmath4 20 myr ) f0iv star with a mass of @xmath17 and with a @xmath18 planet orbiting at a distance of @xmath19 au ( macintosh et al . hd 100546 is an even younger ( @xmath4 10 myr ) herbig ae / be star with a mass of @xmath20 , which has a gas giant planet orbiting at a distance of @xmath21 au , and a possible second planet at @xmath4 14 au ( quanz et al . 2015 ) . finally , lkca 15 is an extremely young ( @xmath4 2 myr ) solar - type star with two likely protoplanets orbiting @xmath4 15 and 19 au away , with planet masses no greater than @xmath22 being required for orbital stability ( sallum et al . 2015 ) . the m0 close binary system roxs 42b is orbited at @xmath4 150 au by a third body with a mass in the range of 6 to 15 @xmath8 ( currie et al . this third object appears to blur the line between exoplanets ( with masses less than @xmath23 ) , such as the hr 8799 bcde planetary system ( marois et al . 2008 ; 2010 ) , and brown dwarfs . it also therefore blurs the line between the formation mechanisms of wide and close binary stars , brown dwarfs , and possibly of gas giant exoplanets , namely protostellar and protoplanetary disk fragmentation ( currie et al . as one example , konopacky et al . ( 2016 ) found a brown dwarf companion with a mass of @xmath24 orbiting 20 au from hr 2562 , a 1.3 @xmath1 star with a debris disk . such brown dwarfs are widely believed to form by disk fragmentation , based on both observational ( e.g. , ma & ge 2014 ) and theoretical ( e.g. , li et al . 2015 ) evidence . given the observational evidence for a growing number of extrasolar giant planets orbiting at distances well beyond the orbits of jupiter and saturn , the question arises as to their formation mechanism . the formation of the solar system s outer giant planets , uranus and neptune , is considered problematical for the core accretion ( ca ) mechanism . levison et al . ( 2015 ) claimed to have solved the problem of forming all four of the solar system s giant planets by relying on the pebble accretion mechanism ( e.g. , bitsch et al . 2015 ) to grow giant planet cores rapidly , before the solar nebula gas disappeared . their ca models led to the formation of 1 to 4 giant planets between 5 and 15 au , as desired . however , their initial conditions consisted of a protoplanetary disk that is assumed to be passive , with a surface density profile ( figure 1 ) that is nearly identical to that of the gravitationally unstable disk models studied by boss ( 2013 , 2015 ) and by this paper . such a moderately massive disk will not be passive , however , and the resulting spiral arms will play havoc with the orbits of the solid bodies that are trying to grow by pebble accretion , as shown by boss ( 2013 , 2015 ) . as a result , the formation of even the solar system s outer giant planets should be considered to remain problematical . bromley & kenyon ( 2016 ) studied ca models in disks as massive as 0.5 @xmath1 in order to try to explain the possible existence of a massive , ninth planet orbiting beyond 100 au , yet another challenge for solar system formation models . chambers ( 2006 ) studied the ca process in a disk with a radius of 50 au , and found that no giant planets more massive than @xmath25 were formed beyond @xmath4 8 au . kenyon & bromley ( 2015 ) found that in a suitably massive disk , ca could lead to the formation of an icy super - earth at distances of 125 to 250 au , but that this would require 1 to 3 gyr , roughly a factor of @xmath26 times too slow to allow such an icy core to accrete a gaseous envelope and form a gas giant . chambers ( 2016 ) studied ca in the pebble accretion scenario for disks with radii of 50 to 100 au , finding that gas giant planet formation was limited to orbital distances of @xmath4 15 au or less . a similar result was found in the ca models of coleman & nelson ( 2016 ) . at best , ca would seem to require the formation of massive icy cores in the inner disk , which are then scattered to the outer disk by close encounters with one or more inner gas giants . once in the outer disk , the icy cores could accrete outer disk gas , which could circularize their initially highly eccentric orbits ( kikuchi et al . however , objects formed by this mechanism are predicted to be extremely rare , occurring in only @xmath4 0.1% of the cases studied by population synthesis models ( kikuchi et al . such a scattering process is also likely to be hazardous to the health of inner terrestrial planets ( kaib & chambers 2016 ) . on the other hand , boss ( 2011 ) found that disk instability ( di ) could form numerous gas giants with initial masses in the range of 1 to 5 @xmath8 , initial distances of 30 to 70 au , and initial eccentricities of @xmath4 0 to 0.35 , around solar - type protostars with masses from 0.1 to 2 @xmath1 . the initial disk mass was assumed to scale with the protostellar mass , with the result that the more massive stars had more giant exoplanets , as expected . forming the hr 8799 exoplanet system thus seems quite possible thorough the di mechanism ( boss 2011 ) . in fact , forming gas giants at large distances by disk instability has now become a part of the conventional wisdom of planet formation theory ( e.g. , boley 2009 ; boley & durisen 2010 ; boley et al . 2010 ; nero & bjorkman 2009 ; meru & bate 2010 ; kratter & murray - clay 2011 ; rogers & wadsley 2012 ; vorobyov et al . 2013 ; madhusudhan et al . 2014 ; rice et al . 2015 ; young & clarke 2016 ) . meru ( 2015 ) found that fragments formed by di in the outer disk could even trigger the formation of more fragments in the inner disk . galvagni et al . ( 2012 ) found that clumps formed by di in the outer disk at @xmath4 100 au could contract fast enough ( @xmath27 yr ) to begin rapid dynamic collapse without being tidally disrupted . nayakshin ( 2015a , b ) showed that pebble accretion by a gaseous protoplanet could accelerate its collapse by changing its opacity , again helping avoid tidal disruption . boss ( 2013 ) found that giant planets orbiting in a marginally gravitationally unstable disk did not undergo monotonic orbital migration , but rather underwent random inward and outward excursions , driven by the gravitational actions of the spiral arms , for several thousand years or more . stamatellos ( 2015 ) found that a giant planet formed in the outer disk would soon open a disk gap , preventing its migration to the inner disk ( cf . , vorobyov 2013 ) , and that it would not accrete sufficient gas to become a brown dwarf , rather than a gas giant planet ( cf . , kratter et al . 2010 ) . given this strong observational evidence that a mechanism similar to disk instability must be able to form at least some exoplanets , it behooves us to consider the current status of disk instability theoretical modeling efforts . we first note the present status of the more computationally challenging models that involve 3d radiative transfer , and then introduce the computationally simpler @xmath0 cooling approach . boss ( 1997 ) began the detailed study of di with 3d models that used simplified thermodynamics for the gaseous disk , namely isothermal or adiabatic gas pressure laws . boss ( 2001 ) then calculated the first di models with 3d radiative transfer ( i.e. , radiation hydrodynamics - rhd ) , finding that disk fragmentation was still possible . boley et al . ( 2006 ) found in their 3d rhd models that the clumps that formed did not last long . durisen et al . ( 2007 ) summarized much of the detailed work on 3d radiative transfer models of the di mechanism , both from the side of models where clump formation occurred ( e.g. , boss 2001 , 2002 , 2007 ; mayer et al . 2007 ) and from models where clump formation was less robust ( e.g. , boley et al . 2006 , 2007 ) . while many numerical factors come to play in these calculations , e.g. , grid resolutions for finite difference ( fd ) codes and smoothing lengths for smoothed - particle hydrodynamics ( sph ) codes ( e.g. , mayer et al . 2007 ) , flux limiters for radiative transfer in the diffusion approximation , and accuracy of the gravitational potential solver , the key issue was determined to be whether a protoplanetary disk could remain sufficiently cold for spiral arms to collide and form self - gravitating clumps that could contract toward planetary densities without re - expanding , being disrupted by tidal forces from the central protostar , or meeting some other equally unkind fate , i.e. , the radiative transfer solver . boley et al . ( 2006 ) showed that their cylindrical coordinate rhd code could reproduce an analytical radiative transfer solution , while boss ( 2009 ) demonstrated the agreement of his spherical coordinate rhd code on two other analytical radiative transfer solutions . helled et al . ( 2014 ) updated the situation in their review paper without reaching a definitive conclusion regarding the validity of the di hypothesis in rhd models ( e.g. , the work by boss 2008 , 2009 , 2010 , 2011 , 2012 ; boley & durisen 2008 , 2010 ; boley 2009 ; cai et al . 2010 ; meru & bate 2010 ) . more recently , steiman - cameron et al . ( 2013 ) studied the effect of spatial resolution on cooling times in rhd models , finding convergence for optically thick , inner regions , but not for optically thin , outer regions . tsukamoto et al . ( 2015 ) used 3d rhd models to follow the formation of disks , starting from collapsing molecular cloud cores , finding that radiative heating from the interstellar medium could have a significant effect on the fragmentation process . gammie ( 2001 ) suggested that the outcome of gravitational instability depended on the beta parameter @xmath28 , where @xmath29 is the disk cooling time and @xmath30 is the local angular velocity . gammie ( 2001 ) found that for @xmath31 , the disk would become gravitoturbulent with a toomre @xmath10 value @xmath32 , whereas for @xmath33 , the disk would fragment . lodato & rice ( 2004 ) and mejia et al . ( 2005 ) also discussed the importance of cooling times for disk fragmentation . the use of the critical value of @xmath34 to predict the fragmentation of protoplanetary disks was first called into question by the 3d sph hydrodynamical models of meru & bate ( 2011a , b ) , who found that when sufficiently high spatial resolution was employed , even disks previously thought to be stable underwent fragmentation into clumps . meru & bate ( 2011b ) even suggested that in the absence of clear indications of numerical convergence , a critical value might not exist . meru & bate ( 2012 ) found evidence for convergence with both 3d sph and 2d fd ( fargo ) codes on a critical value of @xmath35 , and perhaps as large as @xmath4 30 . gammie ( 2001 ) had studied two dimensional ( 2d ) , shearing - sheet disk models with a maximum numerical resolution of @xmath36 . baehr & klahr ( 2015 ) also studied 2d shearing sheet models , but with a refined cooling law , and found that with their highest resolution models ( @xmath37 ) , the 2d disks fragmented even for @xmath38 , but not when the resolution was @xmath39 with @xmath38 . paardekooper ( 2012 ) studied 2d shearing sheet models with the fargo fd code , finding fragmentation to occur for @xmath0 values as high as 20 , provided the integration was followed long enough in time . rice et al . ( 2014 ) , however , argued that numerical problems involving the implementation of the disk cooling in sph codes and the handling of artificial viscosity in both sph and the fargo fd codes may have led to higher critical @xmath0 values . they claimed that their sph simulations converged on a critical value of @xmath0 between 6 and 8 . however , zhu et al . ( 2015 ) have shown that sph models will not converge to the continuum limit if only the total number of particles ( @xmath40 ) is increased to @xmath41 ; formal numerical convergence is only possible when @xmath42 , the smoothing length @xmath43 , and the number of neighbor particles within the smoothing volume @xmath44 . the latter constraint is seldom applied , lending suspicion to the claims of numerical convergence of many previous sph studies . young & clarke ( 2015 ) used both 2d sph and 2d finite difference models to suggest two possible pathways to fragmentation : dynamic collapse when @xmath33 , or quasi - static contraction when @xmath45 , for models with effectively @xmath39 resolutions . evidently even for simplified @xmath0 cooling prescriptions , enough numerical questions and differences remain that the critical @xmath0 value for fragmentation , if there is one , can not be constrained better than to lying in the range of @xmath4 10 to @xmath4 30 . takahashi et al . ( 2016 ) pointed out the crucial importance of the toomre @xmath10 value : fragmentation occurs when the toomre @xmath10 value drops to @xmath4 0.6 inside the spiral arms , regardless of the value of @xmath0 , and that low @xmath0 values do not necessarily result in fragmentation . gammie ( 2001 ) studied 2d disks with @xmath46 initially . the paardekooper ( 2012 ) and baehr & klahr ( 2015 ) 2d models also started with @xmath46 throughout the disk . young & clarke ( 2015 ) started with disks with @xmath47 . meru & bate ( 2012 ) and rice et al . ( 2014 ) started with disks with a minimum of @xmath48 at the outer disk edge . rafikov ( 2015 ) has argued that gravitoturbulent disks without rapid cooling should settle into a quasi - stationary , fluctuating state where @xmath10 remains close to a constant value @xmath49 . the concept of a quasi - steady balance of disk heating and cooling dates back to pringle ( 1981 ) , and has been investigated in many models by the indiana university group ( e.g. , steiman - cameron et al . 2013 ) , yielding @xmath50 to 2 . this balance has been suggested to be an alternative to fragmentation ( e.g. , durisen et al . 2007 ) . backus & quinn ( 2016 ) found that the critical initial value of @xmath10 for fragmentation depended on how stable the initial disk equilibrium model was : less stable initial disk equilibria fragmented for higher @xmath10 values . evidently initial @xmath10 values in the range of 1 to 2 are considered reasonable starting values for di models with @xmath0 cooling . in order to allow a direct comparison with previous work , the new models were calculated with the same basic code that has been used in all of the author s previous studies of disk instability . the primary change was to replace the radiative transfer subroutine that calculates the energy changes due to radiative flux with the @xmath0 cooling formula . the numerical code solves the three dimensional equations of hydrodynamics and the poisson equation for the gravitational potential and is second - order - accurate in both space and time . a complete description of the entire code is given by boss & myhill ( 1992 ) , while the updated energy equation of state is described by boss ( 2007 ) . also , the central protostar is effectively forced to wobble in order to preserve the location of the center of mass of the entire system ( boss 1998 , 2012 ) , which is accomplished by altering the apparent location of the point mass source of the star s gravitational potential in order to balance the center of mass of the disk . central massive sink cell particles behave in exactly this same way to preserve the center of mass of the system when used in , for example , the flash hydrodynamics code . as usual , explicit artificial viscosity ( av ) is not used in the models . boss ( 2006 ) found that small amounts of tensor av had little effect on fragmentation , while large amounts could suppress fragmentation , as was also found by pickett et al . av is designed to stabilize and capture the microphysics of heating in strong shocks , and the effect of its inclusion in @xmath0 cooling models remains to be investigated . the numerical code solves the specific internal energy @xmath51 equation ( boss & myhill 1992 ) : @xmath52 where @xmath53 is the gas density , @xmath54 is time , @xmath55 is the velocity , @xmath56 is the gas pressure , and @xmath57 is the time rate of change of energy per unit volume , which is normally taken to be that due to the transfer of energy by radiation in the diffusion approximation . here we define @xmath57 in terms of the @xmath0 cooling formula ( gammie 2001 ) , as follows . with @xmath28 , @xmath29 is defined as the ratio of the specific internal energy to the time rate of change of the specific internal energy . we thus define @xmath57 to be : @xmath58 where @xmath30 is the local angular velocity of the gas in prograde rotation ( @xmath59 ) . evidently @xmath57 is always negative with this formulation , i.e. , only cooling is permitted . it is important to note that in spite of the @xmath0 cooling formula , the disk temperature at a given radial distance from the central protostar is not allowed to fall below its initial value . this means that initially high @xmath10 disks can not become more gravitationally unstable solely due to cooling to lower temperatures than the initial state , regardless of the cooling parameter @xmath0 , and can only become gravitationally unstable by transporting disk mass such that the local disk surface density increases , thereby lowering @xmath10 locally . this assumption is critical to understanding the results to be presented for initially high @xmath10 disk models . this minimum temperature constraint was also imposed in all of the previous disk instability models in this series , and so was retained here in order to allow a direct comparison with those earlier models . a future paper will address the outcome of models similar to these , but where the disk temperature is allowed to cool below the initial value . an alternative justification for this temperature constraint is to consider the case of disks being heated by protostellar irradiation , e.g. , cai et al . ( 2008 ) and takahashi et al . ( 2016 ) , in which case disks might become hot enough to suppress fragmentation altogether . the equations are solved on a spherical coordinate grid with @xmath60 or 200 radial grid points ( as well as the central grid cell , which contains the central protostar ) , @xmath61 theta grid points , distributed from @xmath62 , and @xmath63 or 1024 azimuthal grid points . @xmath64 and @xmath65 are increased to their higher values once the disk forms dense spiral arms and the jeans and toomre length criteria begin to be violated ( see below ) . the radial grid is uniformly spaced with either @xmath66 au or 0.08 au , for @xmath67 or 200 , respectively . the radial grid extends from 4 to 20 au . the mass of disk gas flowing inside 4 au is added to the central protostar , whereas that reaching the outermost shell at 20 au loses its outward radial momentum but remains on the active hydrodynamical grid . the @xmath68 grid points are compressed into the midplane to ensure adequate vertical resolution ( @xmath69 at the midplane ) . the @xmath70 grid is uniformly spaced in @xmath71 . the number of terms in the spherical harmonic expansion for the gravitational potential of the disk is @xmath72 for all phases of evolution . the numerical resolution is increased during the evolutions in order to avoid violating the jeans length ( e.g. , boss et al . 2000 ) and toomre length criteria ( nelson 2006 ) . both criteria are monitored throughout the evolutions ( e.g. , boss 2011 ) to ensure that any clumps that might form are not numerical artifacts . the jeans length criterion consists of requiring that all of the grid spacings in the spherical coordinate grid remain smaller than 1/4 of the jeans length @xmath73 , where @xmath74 is the local sound speed , and @xmath75 the gravitational constant . similarly , the toomre length criterion consists of requiring that all of the grid spacings remain smaller than 1/4 of the toomre length @xmath76 , where @xmath77 is the mass surface density . when one of these two criteria is violated , the calculation is halted , and the grid spatial resolution is doubled in either the radial or azimuthal direction , as necessary , by dividing each cell into half in the relevant direction . mass and momentum are conserved by this process . once well - defined clumps form , the jeans and toomre length criteria will eventually again be violated at the maximum densities of the clumps , even with the @xmath78 and @xmath79 grids . at this point , the cell with the maximum clump density is drained of 90% of its mass and momentum , which is then used to form a virtual protoplanet ( vp ) initially at the cell center ( e.g. , boss 2005 , 2013 ) . the vps thereafter orbit around the disk , subject to the gravitational forces of the disk gas and the central protostar , as well as those of any other vps . the disk gas is similarly subject to the gravity of the vps . vps that orbit to the inner boundary at 4 au or to the outer boundary at 20 au are simply removed from the remainder of the evolutions . the vps are allowed to gain mass @xmath80 at the rate ( boss 2005 , 2013 ) given by the bondi - hoyle - lyttleton ( bhl ) formula ( e.g. , ruffert & arnett 1994 ) : @xmath81 where @xmath82 is a dimensionless coefficient , @xmath83 is the vp mass , and @xmath84 is the speed of the vp respect to the disk gas . the vps also accrete orbital angular momentum from the disk gas , by accreting an amount of momentum from the local hydrodynamical cell proportional to the mass being accreted from that cell in such a way as to guarantee the conservation of the total orbital angular momentum . this insertion of vps in regions where the jeans and toomre length criteria can no longer be met is identical to the manner in which _ sink particles _ are inserted into adaptive mesh refinement codes , such as flash , when no further sub grids are allowed to be formed ( e.g. , federrath et al . 2010 ; klassen et al . 2014 ) . these models used @xmath85 in order to maximize the gas accretion by the vps and to minimize subsequent jeans or toomre length violations . if the jeans or toomre criterion is nevertheless subsequently violated in the immediate vicinity ( 10 cells ) of an existing vp , the violation is ignored , as the vp is expected to accrete or disrupt this high density gas . violations more than 10 cells away result in the creation of a new vp . along with the maximum spatial resolution required , the number of vps formed thus gives a basic estimate of the extent to which a given set of initial conditions leads to a gravitationally unstable disk capable of fragmenting into clumps . the models all begin from the same initial disk model as that used by boss ( 2001 , 2005 , 2007 , 2008 , 2012 , 2013 ) , differing only in the minimum initial toomre @xmath10 value and the @xmath0 parameter . the initial density is that of an adiabatic , self - gravitating , thick disk in near - keplerian rotation about a stellar mass @xmath86 ( boss 1993 ) : @xmath87,\ ] ] where @xmath88 and @xmath89 are cylindrical coordinates , @xmath90 is the surface density , @xmath91 ( cgs units ) and @xmath92 . the initial midplane density ensures near - keplerian rotation throughout the disk : @xmath93 where @xmath94 g @xmath95 and @xmath96 au , the inner edge of the numerical grid . the resulting initial disk mass is 0.091 @xmath1 , while the initial protostellar mass is @xmath97 . note that this initial approximate equilibrium disk state does not depend explicitly on the outer disk temperature , and was chosen to be a state dominated by rotational support rather than by gas pressure , at least in the radial direction ( boss 1993 ) . as a result , this same initial density distribution is used for all the models , regardless of the variations in the outer disk temperatures . table 1 lists the initial conditions chosen for the models . the initial outer disk temperatures @xmath98 are varied in order to alter the initial minimum values of the toomre ( 1964 ) @xmath10 gravitational stability parameter , decreasing monotonically outward from highly stable @xmath99 values at 4 au to marginally stable @xmath100 values ranging from 1.3 to 2.7 beyond 10 au . values of @xmath0 of 1 , 3 , 10 , 20 , 30 , 40 , 50 , and 100 were used in the models . these choices for the initial disk conditions , @xmath100 , and @xmath0 are consistent with the observational constraints and with the papers on @xmath0 cooling models that were cited previously . figure 1 shows the initial surface density profile for the models , compared to the power law assumed in the core accretion models by levison et al . we shall see that the passive disk assumed in these core accretion models is likely to be gravitationally unstable for realistic disk midplane temperatures . figure 2 displays the initial toomre ( 1964 ) @xmath10 profiles for three representative sets of models , those with outer disk temperatures of 40 k , 90 k , and 180 k , leading to initial minimum @xmath10 values beyond 10 au of 1.3 , 2.0 , and 2.7 , respectively . midplane temperatures at the inner disk edge at 4 au are 600 k for all models , while the outer disk temperatures are varied as listed in table 1 . the initial midplane temperature profiles are based on the 2d rhd models by boss ( 1996 ) , as used by boss ( 2001 , 2005 , 2007 , 2008 , 2012 ) . figure 3 displays the initial midplane density distribution for all of the models , along with the initial midplane temperature distribution for the models with initial minimum toomre @xmath101 , i.e. , the models with outer disk temperatures of 40 k. while the initial temperature distributions are perfectly axisymmetric about the rotation axis ( center of each plot ) , the initial density distributions are non - axisymmetric at a low level : random cell - to - cell noise at the level of 1% has been added to the density distribution , along with @xmath102 and 4 modes of amplitude 0.01 , as in previous models ( e.g. , boss 1998 ) . table 1 lists the basic results for all of the models , namely the final times reached , the final spatial resolution employed , and the maximum number of virtual planets ( @xmath103 ) that formed and existed simultaneously , usually close to the final number of vps . the final times reached ranged from 244 yrs to 978 yrs . for comparison , the initial orbital period of the disk at the inner edge ( 4 au ) was 8.0 yr , whereas the initial orbital period at the outer edge ( 20 au ) was 91 yrs . the key point is that all the models were evolved far enough in time that there was no evidence for any further significant growth of non - axisymmetry , i.e. , no need to further refine the spatial grid . some models still required the creation of additional vps , but this was balanced by the loss of vps that hit either the inner or outer grid boundaries , indicating that a semblance of a steady state configuration had been reached . because the boss & myhill ( 1992 ) code employed has not been parallelized , each model was computed continuously on a single , dedicated cluster core . most models were run for as long as 2.3 yrs on the dtm flash cluster , while others were run for as long as 7 months on the considerably faster carnegie memex cluster at stanford university . figure 4 summarizes the results from table 1 as follows . models where the initial spatial resolution of @xmath60 and @xmath63 sufficed for the entire evolution without violating the jeans or toomre constraints experienced little growth of non - axisymmetry ( red dots ) , while those that needed to be refined to @xmath78 with @xmath63 experienced moderate growth ( green dots ) , and those that ended with @xmath78 and @xmath79 experienced significant growth of spiral arms ( blue dots ) . the remaining models continued to violate the jeans or toomre criteria , even with @xmath78 and @xmath79 , necessitating the creations of vps at the location of the violations . the black symbols in figure 4 represent the maximum number of vps created for those models , ranging from a black circle for a single vp , to a bar for two vps , to a three - pointed star for three vps , etc . figure 4 demonstrates that the assumed initial conditions are just as important for determining the success of a possible disk gravitational instability in forming self - gravitating clumps as is the choice of the @xmath0 cooling parameter . all of the models with @xmath101 fragmented and formed from 4 to 10 vps , regardless of the cooling rates investigated , from @xmath0 = 1 to @xmath0 = 100 . this somewhat surprising result means that if a disk is initially extremely gravitationally unstable , self - gravitating clumps can form rapidly without being stifled by compressional heating during their assembly . however , the time at which the first vp formed depended strongly on the value of @xmath0 , with the first vp in models 1.3 - 1 and 1.3 - 3 forming after 104 yrs and 118 yrs , respectively , compared to forming only after 190 yrs and 212 yrs , respectively , for models 1.3 - 50 and 1.3 - 100 . clearly even for an initially highly unstable disk , the cooling rate affects the time evolution of the fragmentation process . at the same time , the models starting with @xmath104 show that such an initially gravitationally stable disk can not become non - axisymmetric enough to undergo fragmentation , even for @xmath105 . as noted previously , this result is largely a result of the constraint that the disk temperatures can not drop below their initial values , in spite of vigorous cooling , and can only become more gravitationally unstable by increasing the disk surface density in a limited region of the disk , e.g. , by forming a ring . for initial toomre @xmath10 values in between these two extremes , the general result is a smooth progression toward more of a tendency toward fragmentation at a fixed value of @xmath0 ( e.g. , for @xmath0 = 3 and 10 as @xmath2 is lowered ) . however , the transition from stable to unstable with regard to forming vps is not necessarily completely monotonically dependent on @xmath2 ( i.e. , for @xmath0 = 1 and 100 , where as @xmath2 is decreased , successive models can fragment and then resist fragmentation , before eventually fragmenting again ) . this result demonstrates the stochastic nature of gravitationally unstable disks , where the spiral arms that form repeatedly and interact with each other may or may not happen to combine in a constructive wave sense and form a clump dense enough to require the creation of a vp . given this ambiguity , the critical value for fragmentation appears to be @xmath106 for @xmath107 , @xmath108 for @xmath109 , and @xmath110 for @xmath111 , roughly speaking . figure 5 presents the final midplane density distributions for the four models at the extremes of the parameter space investigated , namely models 1.3 - 1 , 1.3 - 100 , 2.7 - 1 , and 2.7 - 100 . the two models with @xmath2 = 1.3 both clearly formed numerous spiral arms and clumps , regardless of whether @xmath0 = 1 or 100 , though in the latter case the spiral arms are broader and nowhere near as sharply defined . neither of the two models with @xmath2 = 2.7 formed any significant non - axisymmetric features , and the model with @xmath0 = 100 looks the most similar to the initial disk model ( figure 3a ) , in spite of having evolved for 414 yrs . figure 6 displays the corresponding midplane temperature distributions for these same four end - member models . the slower cooling rate in model 1.3 - 100 ( figure 6b ) compared to that of 1.3 - 1 results in well - defined , moderately warm spiral arms throughout the disk , whereas the rapid cooling assumed in model 1.3 - 1 ( figure 6a ) allows the disk to cool back down to its initial temperature throughout most of the disk . for the high @xmath2 models , figure 6c shows that model 2.7 - 1 maintains a disk temperature essentially unchanged from its initial conditions , as a result of the rapid cooling , whereas in figure 6d , model 2.7 - 100 shows several distinct rings of somewhat hotter gas than the initial temperature distribution . figure 7 shows the location of the nine vps still active in model 1.3 - 1 at the end of the evolution ( 224 yrs ) , the same time as the midplane density contours shown in figure 5a . it is evident that the nine vps are not aligned with the numerous spiral arm features seen in figure 5a , as is to be expected , given that the vps begin their existence with an initial mass of the order of a fraction of a jupiter mass , and as such are not subject to the tendency for gas drag to force small ( e.g. , cm - size ) particles to remain in the vicinity of spiral arms ( e.g. , boss 2013 , 2015 ) . the time evolution of the ensemble of vps formed in the eight models with @xmath2 = 1.3 is presented in figure 8 , sampled roughly every 40,000 time steps . starting with masses only a small fraction of a jupiter mass , the vps accrete mass at the blh rate and soon reach masses as high as 5 @xmath8 . the vps tend to be created in the sweet spot between 6 au and 10 au , where the disk is cool enough and massive enough to support the growth of strong spiral arms . once formed there , the vps can migrate through disk interactions ( e.g. , boss 2013 ) both inwards and outwards , with significant numbers hitting either the inner or outer disk boundaries and thereafter being removed from the calculations . finally , figure 9 compares the vps formed in the @xmath2 = 1.3 models with the presently known distribution in mass and semi - major axis of exoplanets between 4 au and 20 au . while not intended to be a rigorous exoplanet population synthesis on a par with those presented in section 2.5 , figure 9 hints that the disk instability mechanism may be able to crudely match the exoplanet demographics for gas giants with semi - major axes of 4 au to 6 au , and if so , this approach would predict an as yet mostly undetected significant population of jupiter - mass gas giants orbiting at distances of about 6 au to 16 au . refining the implications of these disk instability models for explaining the known exoplanet demographics , and for predicting what more might be discovered by future exoplanet surveys ( e.g. , by microlensing and direct imaging searches with the nasa wfirst mission ) is a promising subject for further work . these models were intended in part to provide a comparison between the disk stability models previously presented in this series ( e.g. , boss 2001 , 2007 , 2008 , 2012 ) , which employed diffusion approximation radiative transfer , with new models using the @xmath0 cooling approach . the previous models all explored variations on numerical parameters such as spatial grid resolution , equations of state , temperature assumptions , and diffusion approximation flux - limiters , with the result that the disks , with initial @xmath112 , nevertheless fragmented into clumps . given the results found here for the @xmath112 initial disks , which all formed numerous vps , even with @xmath0 as high as 100 , the basic results of the previous diffusion approximation models starting from these initial conditions do not appear to be as contentious as they have at times seemed ( e.g. , durisen et al . 2007 ; helled et al . 2014 ) . section 2.3 noted that several indicators of protoplanetary disk temperatures at distances of @xmath4 10 to 20 au imply midplane temperatures of @xmath4 40 to 70 k , i.e. , values of @xmath2 from about 1.3 to 1.7 for the present disk models . figure 4 shows that such disks are unstable to fragmentation on short time scales , regardless of the assumed value of @xmath0 . figure 1 shows that these disk models have an initial surface density profile quite similar to that assumed in the levison et al . ( 2015 ) models of solar system formation by pebble accretion . evidently the background passive gas disk adopted for the levison et al . ( 2015 ) efforts is unlikely to be passive , calling into question their results regarding the formation of our giant planets . boss ( 2004 ) found evidence in 3d diffusion approximation models of disk instability for convective upwellings that were able to cool the outer disk at a rate equivalent to @xmath113 , leading to fragmentation , as expected based on the results shown in figure 4 [ note , however , that boley & durisen ( 2006 ) and lyra et al . ( 2016 ) have interpreted such upwellings as hydraulic jumps , rather than as the result of convection ) . ] as previously noted , meru & bate ( 2011b ) suggested that a critical value of @xmath0 for fragmentation might not exist , a suspicion that is supported by the weak dependence of the results shown in figure 4 on @xmath0 for a fixed value of @xmath2 , i.e. , the initial conditions assumed are more important in general than the choice of @xmath0 . takahashi et al . ( 2016 ) found a similar result to that seen in figure 4 , namely that fragmentation can occur whenever @xmath10 becomes sufficiently small , regardless of @xmath0 , and that low @xmath0 alone does not guarantee fragmentation . in particular , takahashi et al . ( 2016 ) found that fragmentation would occur when @xmath10 dropped to 0.6 in a spiral arm . figure 10 displays the radial profile of @xmath10 in model 1.3 - 1 at the time ( 104 yrs ) when a vp needed to be inserted into the calculation : it can be seen that the minimum azimuthally averaged value of @xmath10 at that time was 0.61 , seemingly in remarkable agreement with the results of takahashi et al . however , for the seven other models with @xmath101 , the equivalent minimum @xmath10 value ranges from 0.69 to 1.0 . given that these are azimuthally averaged @xmath10 values , however , the basic agreement with takahashi et al . ( 2016 ) holds . in fact , the first vp in model 1.3 - 1 was inserted at 7.2 au , the radius of the innermost local minimum in the azimuthally averaged @xmath10 seen in figure 10 . the @xmath10 value of the actual midplane grid cell where this first vp was inserted was @xmath114 , considerably lower still . the present set of models has shown that the disk instability mechanism for gas giant planet formation depends even more strongly on the initial conditions assumed in the models than on the assumed disk cooling rate @xmath0 , based on a suite of 3d models with identical handling of the spatial resolution of the grid . hence the evolution of protoplanetary disks into gravitationally unstable configurations is as important a factor to consider as the detailed heating and cooling processes in the disks . the models imply that if a significant fraction of protoplanetary disks can form that are similar to those assumed here , then there should be an equally significant population of gas giants remaining to be discovered with separations of order 15 au from solar - type stars . the nasa wfirst mission may very well test this prediction , based on a combination of exoplanet searches by gravitational microlensing and by coronagraphic direct imaging . in order to help determine what is the proper value of @xmath0 to use in disk instability models , the author is currently running a suite of eight flux - limited diffusion approximation ( flda ) models similar to the present suite , with initial minimum @xmath10 values ranging from 1.3 to 2.7 , and with the same approach of adding spatial resolution and vps when demanded by the jeans and toomre criteria , with the goal of comparing these results for varied beta cooling models with those of flda models , for the same spatial resolution and same initial conditions ( i.e. , initial toomre @xmath10 ) . previous flda disk instability models by boss ( 2008 , 2012 ) used different spatial resolutions than the present models and so can not be used for a valid comparison . however , the suite of models currently running will allow the proper value of @xmath0 that should be used in disk instability models to be determined , as opposed to determining the critical value of @xmath0 for fragmentation for a given initial disk model , which was the goal of the present work . in order to better understand the approach to gravitational instability , a second set of eight models with the @xmath0 cooling approximation is also currently underway . these models all start from the same @xmath115 model as in this paper , and with same range of eight @xmath0 values as here , but with the minimum disk temperature values all relaxed to 40 k ( i.e. , relaxed to the outer disk minimum temperature for the highly unstable models with initial @xmath112 ) . these models will thus study the effect of varied cooling rates on the approach to a gravitationally unstable phase of disk evolution . i thank the referee , richard durisen , for a number of perceptive and constructive comments on the manuscript , and sandy keiser for computer systems support . the calculations were performed primarily on the flash cluster at dtm and also on the carnegie memex cluster at stanford university . 1.3 - 1 & 40 & 1.3 & 1 & 244 . & 200 & 1024 & 10 + 1.3 - 3 & 40 & 1.3 & 3 & 291 . & 200 & 1024 & 5 + 1.3 - 10 & 40 & 1.3 & 10 & 275 . & 200 & 1024 & 6 + 1.3 - 20 & 40 & 1.3 & 20 & 378 . & 200 & 1024 & 5 + 1.3 - 30 & 40 & 1.3 & 30 & 302 . & 200 & 1024 & 5 + 1.3 - 40 & 40 & 1.3 & 40 & 285 . & 200 & 1024 & 4 + 1.3 - 50 & 40 & 1.3 & 50 & 306 . & 200 & 1024 & 9 + 1.3 - 100 & 40 & 1.3 & 100 & 342 . & 200 & 1024 & 5 + 1.6 - 1 & 60 & 1.6 & 1 & 312 . & 200 & 1024 & 3 + 1.6 - 3 & 60 & 1.6 & 3 & 270 . & 200 & 1024 & 6 + 1.6 - 10 & 60 & 1.6 & 10 & 263 . & 200 & 1024 & 2 + 1.6 - 20 & 60 & 1.6 & 20 & 319 . & 200 & 1024 & 2 + 1.6 - 30 & 60 & 1.6 & 30 & 263 . & 200 & 1024 & 1 + 1.6 - 40 & 60 & 1.6 & 40 & 293 . & 200 & 1024 & 2 + 1.6 - 50 & 60 & 1.6 & 50 & 295 . & 200 & 1024 & 3 + 1.6 - 100 & 60 & 1.6 & 100 & 799 . & 200 & 512 & 0 + 1.7 - 1 & 70 & 1.7 & 1 & 358 . & 200 & 1024 & 1 + 1.7 - 3 & 70 & 1.7 & 3 & 327 . & 200 & 1024 & 3 + 1.7 - 10 & 70 & 1.7 & 10 & 292 . & 200 & 1024 & 4 + 1.7 - 100 & 70 & 1.7 & 100 & 393 . & 200 & 1024 & 3 + 1.9 - 1 & 80 & 1.9 & 1 & 394 . & 200 & 1024 & 5 + 1.9 - 3 & 80 & 1.9 & 3 & 286 . & 200 & 1024 & 3 + 1.9 - 10 & 80 & 1.9 & 10 & 413 . & 200 & 1024 & 1 + 1.9 - 100 & 80 & 1.9 & 100 & 470 . & 200 & 1024 & 1 + 2.0 - 1 & 90 & 2.0 & 1 & 235 . & 200 & 1024 & 3 + 2.0 - 3 & 90 & 2.0 & 3 & 383 . & 200 & 1024 & 3 + 2.0 - 10 & 90 & 2.0 & 10 & 351 . & 200 & 1024 & 1 + 2.0 - 100 & 90 & 2.0 & 100 & 315 . & 200 & 1024 & 0 + 2.1 - 1 & 100 & 2.1 & 1 & 354 . & 200 & 512 & 0 + 2.1 - 3 & 100 & 2.1 & 3 & 486 . & 200 & 1024 & 0 + 2.1 - 10 & 100 & 2.1 & 10 & 896 . & 200 & 1024 & 0 + 2.1 - 100 & 100 & 2.1 & 100 & 490 . & 200 & 512 & 0 + 2.2 - 1 & 120 & 2.2 & 1 & 415 . & 200 & 1024 & 1 + 2.2 - 3 & 120 & 2.2 & 3 & 697 . & 200 & 512 & 0 + 2.2 - 10 & 120 & 2.2 & 10 & 772 . & 100 & 512 & 0 + 2.2 - 100 & 120 & 2.2 & 100 & 978 . & 100 & 512 & 0 + 2.7 - 1 & 180 & 2.7 & 1 & 450 . & 100 & 512 & 0 + 2.7 - 3 & 180 & 2.7 & 3 & 454 . & 100 & 512 & 0 + 2.7 - 10 & 180 & 2.7 & 10 & 414 . & 100 & 512 & 0 + 2.7 - 100 & 180 & 2.7 & 100 & 414 . & 100 & 512 & 0 +

we show that there exist polynomial endomorphisms of , possessing a wandering fatou component . these mappings are polynomial skew - products , and can be chosen to extend holomorphically of . we also find real examples with wandering domains in . the proof relies on parabolic implosion techniques , and is based on an original idea of m. lyubich .

1411.1188

let @xmath3 be a polynomial mapping . in this article we consider @xmath4 as a dynamical system , that is , we study the behavior of the sequence of iterates @xmath5 . a case of particular interest is when @xmath4 extends as a holomorphic endomorphism on @xmath6 . as a matter of expositional simplicity , let us assume for the moment that this property holds . the _ fatou set _ @xmath7 is classically defined as the largest open subset of @xmath6 in which the sequence of iterates is locally equicontinuous ( or _ normal _ , according to the usual terminology ) . its complement , the _ julia set _ , is where chaotic dynamics takes place . fatou component _ is a connected component of @xmath7 . when the dimension @xmath8 equals 1 , the non wandering domain theorem due to sullivan @xcite asserts that every fatou component is eventually periodic . this result is fundamental for at least two reasons . * first , it opens the way to a complete description of the dynamics in the fatou set : the orbit of any point in the fatou set eventually lands in a ( super-)attracting basin , a parabolic basin or a rotation domain . * secondly , it introduced quasi - conformal mappings as a new tool in this research area , leading to many subsequent developments . there are many variants and generalizations of sullivan s theorem in several areas of one - dimensional dynamics . for instance it was shown by eremenko and lyubich @xcite and goldberg and keen @xcite that entire mappings with finitely many singular values have no wandering domains . on the other hand , baker @xcite , prior to sullivan s result , gave the first example of an entire mapping with a wandering domain . simple explicit entire mappings with fatou components wandering to infinity were given in @xcite and @xcite , while more elaborate examples with varied dynamical behaviors were presented in @xcite . more recently , bishop @xcite constructed an example with a bounded singular set . in all cases , the orbit of the wandering domain is unbounded . in the real setting , the question of ( non-)existence of wandering intervals has a long and rich history . it started with denjoy s theory of linearization of circle diffeomorphisms @xcite : a @xmath9-diffeomorphism of the circle with irrational rotation number has no wandering intervals ( hence it is linearizable ) , whereas this result breaks down for @xmath10 diffeomorphisms . more recent results include homeomorphisms with various degrees of regularity and flatness of critical points . for interval maps , the non - existence of wandering intervals for unimodal maps with negative schwarzian was established by guckenheimer @xcite , and later extended to several classes of interval and circle maps in @xcite . in particular , the result of martens , de melo and van strien implies the non - existence of wandering intervals for polynomials on the real line . one difficulty is to define a notion of fatou set in this context . let us just note here that for a real polynomial , the fatou set as defined in @xcite contains the intersection of the complex fatou set with the real line but could _ a priori _ be larger . the problem was also studied in the non - archimedian setting , in particular in the work of benedetto @xcite and trucco @xcite . @xmath11 the question of the existence of wandering fatou components makes sense in higher dimension , and was put forward by many authors since the 1990 s ( see e.g. @xcite ) . higher dimensional transcendental mappings with wandering domains can be constructed from one - dimensional examples by simply taking products . an example of a transcendental biholomorphic map in @xmath0 with a wandering fatou component oscillating to infinity was constructed by fornss and sibony in @xcite . for higher dimensional polynomials and rational mappings , it is widely acknowledged that quasi - conformal methods break down , so a direct approach to generalize sullivan s theorem fails . besides this observation , little was known about the problem so far . recently , m. lyubich suggested that polynomial skew products with wandering domains might be constructed by using parabolic implosion techniques . the idea was to combine slow convergence to an invariant fiber and parabolic transition in the fiber direction , to produce orbits shadowing those of a lavaurs map ( see below for a more precise description ) . in this paper , we bring this idea to completion , thereby providing the first examples of higher dimensional polynomial mappings with wandering domains . there exists an endomorphism @xmath12 , induced by a polynomial skew - product mapping @xmath13 , possessing a wandering fatou component . let us point out that the orbits in these wandering domains are bounded . the approach is in fact essentially local . a more detailed statement is the following ( see below for the definition of lavaurs maps ) . let @xmath14 and @xmath15 be polynomials of the form @xmath16 if the lavaurs map @xmath17 has an attracting fixed point , then the skew - product map @xmath13 defined by @xmath18 admits a wandering fatou component . notice that if @xmath19 and @xmath20 have the same degree , @xmath4 extends to a holomorphic self - map on @xmath1 . observe also that examples in arbitrary dimension @xmath21 may be obtained from this result by simply considering products mappings of the form @xmath22 , where @xmath23 admits a fixed fatou component . before explaining what the lavaurs map is , let us give some explicit examples satisfying the assumption of the main theorem . [ ex : complex ] let @xmath14 be the cubic polynomial @xmath24 , and @xmath20 be as in . if @xmath25 is sufficiently small and @xmath26 , then the polynomial skew - product @xmath4 defined in admits a wandering fatou component . numerical experiments suggest that the value @xmath27 works ( see figure [ fig:0.95 ] on page ) . in view of the results of @xcite cited above , it is also of interest to look for real polynomial mappings with wandering fatou domains intersecting @xmath28 . our method also provides such examples . [ ex : real ] let @xmath14 be the degree @xmath29 polynomial defined by @xmath30 there exist parameters @xmath31 such that if @xmath20 is as in , the polynomial skew - product @xmath4 defined in admits a wandering fatou component intersecting @xmath28 . numerical experiments suggest that the parameter @xmath32 satisfies this property . we illustrate this phenomenon graphically in figure [ realwandering ] for the mapping @xmath4 defined by @xmath33 the set of points @xmath34 with bounded orbit is contained in the rectangle @xmath35 . the topmost image in figure [ realwandering ] displays this set of points . the bottom right image is a zoom near the point @xmath36 . a wandering component is visible ( in green ) . the bottom left image displays the window @xmath37 and @xmath38 ; the orbit of a point @xmath39 contained in the wandering domain is indicated . the coordinate @xmath40 is close to @xmath41 and we plotted the first @xmath42 and the next @xmath43 iterates . the points are indicated in black except @xmath39 , @xmath44 and @xmath45 which are colored in red . these peculiar values are explained by the proof of the main theorem ( see proposition [ key ] below ) . @xmath11 using skew - products to construct new examples is natural as it allows to build on one - dimensional dynamics . this idea was already used several times in holomorphic dynamics ( see e.g. @xcite ) and beyond . fatou components of polynomial skew - products were studied in several earlier works . lilov showed in his thesis @xcite that skew - products do not have wandering components near a super - attracting invariant fiber . in @xcite it was shown that the argument used in @xcite breaks down near an attracting fiber . the construction in @xcite uses a repelling fixed point in the invariant fiber with multiplier equal to the inverse of the multiplier of the attracting fiber . this resonance between multipliers induces a dynamical behavior that can not occur in one - dimensional dynamics . in the same vein , an invariant fiber at the center of a siegel disk was used in @xcite to construct a non - recurrent fatou component with limit set isomorphic to a punctured disk . in that construction the invariant fiber also contains a siegel disk , but with the opposite rotation number . we note that the construction presented in this paper uses an invariant parabolic fiber which again contains a parabolic point . @xmath11 to explain the notion of lavaurs map and the strategy of the proof , we need to recall some facts on parabolic dynamics ( see appendix [ app ] for more details ) . let @xmath19 be a polynomial of the form @xmath46 and denote by @xmath47 the parabolic basin of @xmath48 . it is known that there is an _ attracting fatou coordinate _ @xmath49 which conjugates @xmath19 to the translation @xmath50 by @xmath51 : @xmath52 this fatou coordinate may be normalized by used in this normalization as well as in the next one is the branch defined in @xmath53 which vanishes at @xmath51 ] @xmath55 likewise , there is a _ repelling fatou parameterization _ @xmath56 satisfying @xmath57 which may be normalized by @xmath58 the ( phase @xmath48 ) lavaurs map @xmath59 is defined by @xmath60 mappings of this kind appear when considering high iterates of small perturbations of @xmath19 : this phenomenon is known as _ parabolic implosion _ , and will play a key role in this paper . the reader is referred to the text of douady @xcite for a gentle introduction to this topic , and to @xcite for a more detailed presentation by shishikura . ( semi-)parabolic implosion was recently studied in the context of dissipative polynomial automorphisms of @xmath0 by bedford , smillie and ueda @xcite ( see also @xcite ) . let us already point out that since our results do not fit into the classical framework , our treatment of parabolic implosion will be essentially self - contained . as it turns out , our computations bear some similarity with those of @xcite . we can now explain the strategy of the proof of the main theorem . let @xmath61 be the parabolic basin of @xmath48 under iteration of @xmath20 . if @xmath62 , then @xmath63 converges to 0 like @xmath64 . we want to choose @xmath65 so that the first coordinate of @xmath66 returns close to the attracting fixed point of @xmath59 infinitely many times . the proof is designed so that the return times are the integers @xmath67 for @xmath68 . so , we have to analyze the orbit segment between @xmath67 and @xmath69 , which is of length @xmath70 . for large @xmath71 , the first coordinate of @xmath4 along this orbit segment is approximately @xmath72 the lavaurs theorem from parabolic implosion asserts that if @xmath73 , then for large @xmath71 , the @xmath74 iterate of @xmath75 is approximately equal to @xmath76 on @xmath77 . our setting is slightly different since @xmath78 keeps decreasing along the orbit . indeed on the first coordinate we are taking the composition of @xmath70 transformations @xmath79 the main step of the proof of the main theorem consists in a detailed analysis of this perturbed situation . we show that the decay of @xmath80 is counterbalanced by taking _ exactly _ one additional iterate of @xmath4 . the precise statement is the following . [ key ] as @xmath81 , the sequence of maps @xmath82 converges locally uniformly in @xmath83 to the map @xmath84 see figure [ fig : key ] for a graphical illustration of this proposition . from this point , the proof of the main theorem is easily completed : if @xmath85 is an attracting fixed point of @xmath59 and if @xmath65 is chosen so that @xmath86 is close to @xmath87 for some large @xmath88 , then @xmath89 gets closer to @xmath87 and we can repeat the process to get that the sequence @xmath90 converges to @xmath87 . since this reasoning is valid on an open set of initial conditions , these points belong to some fatou component . simple considerations show that it can not be preperiodic , and the result follows . let us observe that by construction , the @xmath91-limit set of any point in the wandering fatou component consists of a single two - sided orbit of @xmath92 under @xmath93 , plus the origin . we give two different approaches for constructing examples satisfying the assumptions of the main theorem . the next proposition corresponds to example [ ex : complex ] . [ attracting ] consider the cubic polynomial @xmath14 defined by @xmath94 if @xmath25 is sufficiently close to @xmath48 and @xmath95 , then the lavaurs map @xmath17 admits an attracting fixed point . to prove this proposition we construct a fixed point for the lavaurs map by perturbation from the degenerate situation @xmath96 and estimate its multiplier by a residue computation . to construct the real examples of example [ ex : real ] , we use the intermediate value theorem to find a real lavaurs map with a real periodic critical point . [ superattracting ] consider the degree @xmath29 polynomial @xmath14 defined by @xmath30 then there exists a parameter @xmath31 such that the lavaurs map @xmath97 has a fixed critical point in @xmath98 . in particular this fixed point is super - attracting so we are in the situation of the main theorem . @xmath11 the plan of the paper is the following . in section [ sec : thm ] we show how to deduce the main theorem from proposition [ key ] . in section [ sec : approxcoord ] we develop a formalism of _ approximate fatou coordinates _ in our context , which is necessary for proving proposition [ key ] , which is itself established in section [ sec_key ] . the proofs in sections [ sec : approxcoord ] and [ sec_key ] are rather technical , and are divided into a number of steps . a detailed outline will be given there . propositions [ attracting ] and [ superattracting ] are respectively established in sections [ sec_lavaurs ] and [ sec : real ] . appendix [ app ] summarizes the main properties of fatou coordinates and lavaurs maps . this project grew up from discussions between misha lyubich and han peters in relation with the work @xcite . it evolved to the current list of authors at the occasion of a meeting of the anr project lambda in universit paris - est marne - la - valle in april 2014 . we are grateful to misha lyubich for his crucial input which ultimately led to the present work . in this section , we prove the main theorem assuming proposition [ key ] . let @xmath99 be an attracting fixed point of the lavaurs map @xmath59 . let @xmath100 be a disk centered at @xmath85 , chosen that @xmath101 is compactly contained in @xmath100 . it follows that @xmath102 converges to @xmath85 as @xmath103 . let also @xmath104 be an arbitrary disk . denote by @xmath105 the first coordinate projection , that is @xmath106 according to proposition [ key ] , there exists @xmath107 such that for every @xmath68 , @xmath108 let @xmath109 be a connected component of the open set @xmath110 . [ limit ] the sequence @xmath111 converges locally uniformly to @xmath87 on @xmath109 . an easy induction shows that for every integer @xmath68 , @xmath112 indeed this holds by assumption for @xmath113 . now if the inclusion is true for some @xmath68 , then @xmath114 from which follows . from this we get that the sequence @xmath111 is uniformly bounded , hence normal , on @xmath109 . also , any cluster value of this sequence of maps is constant and of the form @xmath115 for some @xmath116 . in addition , @xmath115 is a limit value ( associated to a subsequence @xmath117 ) if and only if @xmath118 is a limit value ( associated to the subsequence @xmath119 ) . we infer that the set of cluster limits is totally invariant under @xmath120 , therefore it must be reduced to the attracting fixed point @xmath85 of @xmath59 , and we are done . the domain @xmath109 is contained in the fatou set of @xmath4 . it is well - known in our context that the sequence @xmath121 is locally bounded on @xmath109 if and only if there exists a subsequence @xmath122 such that @xmath123 has the same property . indeed since @xmath124 is compact , there exists @xmath125 such that if @xmath126 , then for every @xmath127 , @xmath128 escapes locally uniformly to infinity under iteration . the result follows . let @xmath129 be the component of the fatou set @xmath7 containing @xmath109 . according to lemma [ limit ] , for any integer @xmath130 , the sequence @xmath131 converges locally uniformly to @xmath132 on @xmath109 , hence on @xmath129 . therefore , the sequence @xmath111 converges locally uniformly to @xmath133 on @xmath134 . as a consequence , if @xmath135 are nonnegative integers such that @xmath136 , then @xmath137 , from which we deduce that @xmath138 . indeed , @xmath85 belongs to the parabolic basin @xmath77 , and so , it is not ( pre)periodic under iteration of @xmath19 . this shows that @xmath129 is not ( pre)periodic under iteration of @xmath4 : it is a wandering fatou component for @xmath4 . in this section we study the phenomenon of _ persistence of fatou coordinates _ in our non - autonomous setting . as in the main theorem , we consider polynomial mappings @xmath139 of the form @xmath140 we put @xmath141 we want to show that there exist changes of variables @xmath142 and @xmath143 which are in a sense approximations to the fatou coordinates of @xmath19 in appropriate domains , and such that @xmath144 is close to a translation . these change of variables are normalized so that @xmath144 is roughly defined in a vertical strip of width 1 and the translation vector is @xmath145 . they will be given by explicit formulas : in this respect our approach is similar to that of @xcite . precise error estimates are required in order to ultimately prove proposition [ key ] in the next section . the following notation will be used throughout this section and the next one ( see also figure [ notation ] ) . first , choose @xmath125 large enough so that @xmath146 is univalent on @xmath147 , @xmath148 denote by @xmath149 the right half - plane @xmath150 and by @xmath151 the left half - plane @xmath152 . define the _ attracting petal _ @xmath153 by @xmath154 then , * the restriction of the attracting fatou coordinate @xmath155 to the attracting petal @xmath153 is univalent , and * the restriction of the repelling fatou parameterization @xmath156 to the left half - plane @xmath151 is univalent . we use the notation @xmath157 only for the inverse branch @xmath158 of @xmath156 on the repelling petal @xmath159 . recall that @xmath160 @xmath161 next , for @xmath25 we set @xmath162 and fix @xmath25 small enough that @xmath163 for the remainder of the article , we assume that @xmath164 , whence @xmath165 as @xmath166 . the notation @xmath167 stands for the branch of the square root on @xmath168 that has positive real part . finally , we fix a real number @xmath169 the relevance of this range of values for @xmath170 will be made clearer during the proof . for @xmath164 , we let @xmath171 define @xmath172 to be the rectangle @xmath173 and let @xmath174 and @xmath175 be the disks @xmath176 in this section , the notation @xmath177 means a quantity that is defined for @xmath164 close enough to zero and is bounded by @xmath178 for a constant @xmath179 which does not depend on @xmath180 . as usual , @xmath181 means a quantity such that @xmath182 converges to zero as @xmath183 . our purpose in this paragraph is to state precisely the properties of the approximate fatou coordinates , in an axiomatic fashion . the actual definitions as well as the proofs will be detailed afterwards . the claim is that there exists a family of domains @xmath184 and univalent maps @xmath185 parameterized by @xmath164 , satisfying the following three properties . [ compatt ] as @xmath186 in @xmath187 , @xmath188 this means that @xmath189 approximates the fatou coordinate on the attracting side . a similar result holds on the repelling side . [ comprep ] as @xmath186 in @xmath187 , we have that @xmath190 and @xmath191 finally , the last property asserts that @xmath192 is almost a fatou coordinate . [ approx ] as @xmath186 in @xmath187 , we have that @xmath193 and @xmath194 to improve the readability of the proof , which involves several changes of coordinates , we adopt the following typographical convention : * block letters ( like @xmath195 , @xmath196 , ) are used for objects which are thought of as living in the initial coordinates ; * script like letters ( like @xmath197 , @xmath172 , ) are used for objects living in approximate fatou coordinates ; * the coordinate @xmath198 is used for the actual fatou coordinate . this gives rise to expressions like @xmath199 or @xmath200 . the skew - product @xmath4 fixes the origin and leaves the line @xmath201 invariant . we may wonder whether there are other _ parabolic invariant curves _ near the origin , in the sense of calle @xcite and hakim @xcite . does there exist holomorphic maps @xmath202 such that @xmath203 as @xmath186 and such that @xmath204 for @xmath164 ? we shall content ourselves with the following weaker result . [ lem : zetapm ] let @xmath205 be defined by @xmath206 then , @xmath207 an elementary computation shows that @xmath208 on the other hand , @xmath209 so @xmath210 thus the result follows from our choice of @xmath211 and @xmath212 since @xmath213 let @xmath214 be the universal cover defined by @xmath215 this universal cover restricts to a univalent map on the vertical strip @xmath216 with inverse given by @xmath217 in this formula , @xmath218 stands for the branch of the logarithm defined on @xmath219 and such that @xmath220 . for @xmath164 , let @xmath221 be the map defined by @xmath222 where in this formula the branch of @xmath223 is defined on @xmath224 and vanishes at @xmath51 . now introduce @xmath225 and its image under @xmath226 : @xmath227 [ lem : chiw ] if @xmath164 is close enough to @xmath48 , then @xmath228 is univalent . in addition , @xmath229 is close to the identity in the following sense : @xmath230 observe that @xmath231 as a consequence , @xmath232 if @xmath233 , then @xmath234 } \bigl|\chi_w'({\boldsymbol{\mathpzc{z}}})-1\bigr| \cdot |{\boldsymbol{\mathpzc{z}}}_2-{\boldsymbol{\mathpzc{z}}}_1| .\ ] ] when @xmath180 is sufficiently close to @xmath48 , the supremum is smaller than @xmath51 and we necessarily have @xmath235 . the second assertion of the lemma follows directly from the definition of @xmath229 and the fact that on @xmath236 , @xmath237 for some positive constant @xmath238 . from now on , we assume that @xmath180 is sufficiently close to @xmath48 so that @xmath228 is univalent . the approximate fatou coordinates @xmath142 are the maps @xmath239 we need to prove that these approximate fatou coordinates satisfy properties [ compatt ] , [ comprep ] and [ approx ] . in this paragraph , we prove that the approximate fatou coordinate @xmath240 satisfies property [ compatt ] , namely when @xmath241 in @xmath187 , @xmath242 and @xmath243 recall that @xmath244 , @xmath245 , and @xmath246 . + * step 1 . * let us first prove that @xmath242 . note that @xmath247 as @xmath186 , hence @xmath248 for @xmath180 close to @xmath48 . if @xmath249 , then @xmath250 in addition , @xmath251 in particular , @xmath252 since @xmath253 , we have that @xmath254 , and it follows that @xmath255 so , the real part of @xmath197 is comparable to @xmath256 and since @xmath257 , we get that @xmath258 , whence @xmath259 for @xmath164 close enough to @xmath48 . * we now establish . note that @xmath260 observe first that when @xmath180 tends to @xmath48 , the domain @xmath261 also tends to @xmath48 . so , if @xmath249 , then @xmath262 on the other hand , @xmath263 thus , together with the estimate we infer that @xmath264 which completes the proof . in this paragraph , we deal with property [ comprep ] , that is , we wish to prove that as @xmath186 in @xmath187 , @xmath265 and @xmath191 the proof is rather similar to that of property [ compatt ] . + * step 1 . * let us first prove that for @xmath266 close enough to @xmath48 , the disk @xmath267 is contained in @xmath268 . note that with @xmath269 as before , we have @xmath270 since @xmath253 , we have that @xmath271 . furthermore , @xmath272 for @xmath164 , hence @xmath273 in addition , by lemma [ lem : chiw ] , @xmath274 for @xmath275 , so @xmath276 * step 2 . * given @xmath277 , we set @xmath278 note that @xmath279 has modulus equal to @xmath256 . also , put @xmath280 by lemma [ lem : chiw ] we have that @xmath281 hence @xmath282 remembering that @xmath283 , from we deduce that @xmath284 so when @xmath164 is close enough to @xmath48 , we find that @xmath285 and @xmath286 moreover , @xmath287 finally , as in ( [ phiw ] ) , we compute @xmath288 this completes the proof of property [ comprep ] . in this paragraph , we prove that the approximate fatou coordinate @xmath142 satisfies property [ approx ] , that is : as @xmath186 in @xmath187 , the inclusions @xmath289 and @xmath290 hold ( recall that the rectangle @xmath172 was defined in ) , and @xmath194 let @xmath291 to handle the fact that @xmath292 is not exactly equal to @xmath293 ( see lemma [ lem : zetapm ] ) , rather than dealing directly with @xmath294 , we introduce an intermediate change of coordinates @xmath295 let @xmath296 be the horizontal strip @xmath297 we will see that there are lifts @xmath298 , @xmath299 and a map @xmath300 such that the following diagram commutes : @xmath301^{\boldsymbol{\mathpzc{f } } } & & & & \c \\ { \boldsymbol{\mathpzc{q}}}_w \rrto^{{\boldsymbol{\mathpzc{f}}}^0}\uto_{\chi^0}\dto^{\psi^0 } & & { \boldsymbol{\mathpzc{f}}}^0({\boldsymbol{\mathpzc{q}}}_w ) \dto^{\psi^1}\rrto^{{\boldsymbol{\mathpzc{f}}}^1 } & & { \boldsymbol{\mathpzc{s}}}_{g(w ) } \dto_{\psi^2}\uto^{\chi^2}\\ \psi^0({\boldsymbol{\mathpzc{q}}}_w ) \rrto_{f_w } \ar@/^2.5pc/[uu]^{\varphi_w } & & v_{g(w ) } \rrto_{{\rm i d } } & & v_{g(w ) } \ar@/_2.5pc/[uu]_{\varphi_{g(w)}}. \enddiagram\ ] ] * step 1 . * we prove that @xmath302 and that @xmath303 define @xmath304 . * step 2 . * we define @xmath305 on @xmath306 and prove that for @xmath307 , @xmath308 in particular , for @xmath180 close enough to @xmath48 , @xmath309 , hence @xmath310 * step 3 . * we define @xmath311 on @xmath296 and prove that for @xmath312 , @xmath313 in particular , for @xmath180 close enough to @xmath48 , @xmath314 , from which we deduce that @xmath315 * step 4 . * we use @xmath316 and @xmath317 to eliminate the third term on the right hand side of . specifically , we define @xmath318 on @xmath172 and prove that for @xmath319 , @xmath320 thus property [ approx ] is established . we prove that @xmath321 and that @xmath322 let @xmath323 be the rectangle @xmath324 with , as before , @xmath325 . we see that @xmath326 and the distance between the boundaries is @xmath327 . on the other hand , by lemma [ lem : chiw ] , for @xmath328 , @xmath329 . from this it follows that @xmath330 surrounds @xmath172 , whence @xmath331 as desired . to prove the estimate on @xmath332 , define @xmath333 . since @xmath334 , we see that @xmath335 . the above sequence of inclusions shows that @xmath336 . thus from @xmath337 we infer that @xmath338 this completes the proof of step 1 . we define @xmath305 on @xmath306 and prove that for @xmath307 , @xmath339 * step 2.1 . * we first . it will be convenient to set @xmath340 so that expansions with respect to @xmath167 become expansions with respect to @xmath78 . set @xmath341 choose @xmath342 small enough so that the only preimage of @xmath48 by @xmath19 within @xmath343 is @xmath48 . choose @xmath344 so that for @xmath345 , the only preimage of @xmath346 under @xmath347 within @xmath348 is @xmath349 . the function @xmath350 extends holomorphically to @xmath351 and does not vanish there . in addition , it identically takes the value @xmath51 for @xmath352 . we set @xmath353 where the branch of logarithm is chosen so that @xmath354 . consider the map @xmath305 defined on @xmath306 by @xmath355 then , for @xmath258 , set @xmath356 as @xmath197 ranges in @xmath306 , @xmath195 avoids the points @xmath349 and remains in a small disk around @xmath48 , thus @xmath357 avoids the points @xmath358 . so we can define @xmath359 where the branch is chosen so that @xmath360 we therefore have @xmath361 in other words , the following diagram commutes : @xmath362 * step 2.2 . * we now prove that for @xmath363 , the following estimate is true : @xmath364 indeed , observe that @xmath365 whence @xmath366 recall that @xmath367 , so @xmath368 since @xmath369 is holomorphic on @xmath370 and @xmath371 , it admits an expansion of the form @xmath372 on @xmath370 . to find @xmath373 and @xmath374 we write @xmath375 and taking the difference between these two expressions we conclude that @xmath376 * step 2.3 . * we finally establish ( [ eq : f0 ] ) . if @xmath307 and @xmath377 it then follows from step 1 of the proof ( see ) that @xmath378 . since @xmath253 we see that @xmath379 .. ] thus , for @xmath307 , @xmath380 since @xmath381 , we arrive at the desired estimate , and the proof of step 2 is complete . we define @xmath311 on the horizontal strip @xmath296 and prove that @xmath313 the idea is simply that since the distance @xmath382 is much smaller than @xmath383 , @xmath384 is very close to the identity . * step 3.1 . * we first define @xmath311 . let @xmath385 and @xmath386 be the mbius transformations defined by ( recall that @xmath387 ) @xmath388 the mbius transformation @xmath389 sends @xmath390 to 0 , @xmath391 to @xmath392 and fixes 1 . set @xmath393 note that @xmath394 therefore @xmath395 thus , the image of the horizontal strip @xmath396 under the exponential map @xmath397 avoids @xmath398 and @xmath399 and @xmath400 lifts to a map @xmath401 such that the following diagram commutes : @xmath402_{\psi^1 } \dto^\exp & \c \ar@/^1.5pc/[dd]^{\psi^2 } \dto_\exp \\ \p^1(\c ) \rto^{\mu } & \p^1(\c ) \\ \p^1(\c ) \rto_{{\rm i d } } \uto_{\mu_1 } & \p^1(\c)\uto^{\mu_2}. \enddiagram\ ] ] since @xmath403 , the choice of lift is completely determined by requiring @xmath404 . * step 3.2 . * we estimate @xmath405 . since @xmath406 , @xmath407 and @xmath403 , we infer that @xmath408 as a consequence , @xmath409 where @xmath223 is the principal branch of the logarithm on @xmath410 ( the arguments of the four logarithms are close to 1 ) . since @xmath411 and @xmath412 we conclude that @xmath413 as desired . * step 3.3 . * let us show that @xmath414 . first , we saw in the course of step 1 that @xmath336 . now , when @xmath180 is small , @xmath415 hence @xmath416 and the distance between the boundaries is comparable to @xmath417 . since @xmath418 on @xmath419 and @xmath420 , we see that @xmath414 , as claimed . from this we deduce that @xmath421 which finishes the proof of step 3 . we define @xmath318 on @xmath172 and prove that @xmath320 let @xmath422 so that the following diagram commutes : @xmath423^{{\boldsymbol{\mathpzc{f}}}}&&&&\c\\ { \boldsymbol{\mathpzc{q}}}_w \rrto_{{\boldsymbol{\mathpzc{f}}}^0 } \uto^{\chi^0}&&{\boldsymbol{\mathpzc{f}}}^0({\boldsymbol{\mathpzc{q}}}_w ) \rrto_{{\boldsymbol{\mathpzc{f}}}^1 } & & { \boldsymbol{\mathpzc{s}}}_{g(w)}.\uto_{\chi^2 } \enddiagram\ ] ] for @xmath307 , define @xmath424 where the second equality follows from the fact that @xmath425 on @xmath419 . now we write @xmath426 using @xmath427 and arguing as in lemma [ lem : chiw ] , we see that the logarithm in is @xmath428 . thus we infer that @xmath429 as a result : @xmath430 from the estimate @xmath431 , we deduce @xmath432 so finally , @xmath433 this completes the proof of step 4 , and accordingly of property [ approx ] of the approximate fatou coordinates . this section is devoted to the proof of proposition [ key ] . let @xmath434 be a compact subset of @xmath77 and @xmath435 be a compact subset of @xmath61 . we need to prove that the sequence of maps @xmath82 converges uniformly on @xmath436 to the map @xmath437 for @xmath438 and for @xmath439 , set @xmath440 this sequence converges uniformly to @xmath48 on @xmath435 so the difficulty consists in proving that the first coordinate converges uniformly to @xmath76 . to do this , we will have to estimate various quantities which depend on an integer @xmath441 $ ] ( corresponding to an iterate @xmath442 $ ] ) . we adopt the convention that the notation @xmath443 or @xmath444 stands for an estimate that is uniform on @xmath436 , and depends only on @xmath71 , meaning that it is uniform with respect to @xmath441 $ ] . for @xmath445 , we set @xmath446 by convention , an empty composition is the identity , whence @xmath447 . then , @xmath448 so we must prove that @xmath449 let us recall that @xmath125 was chosen so large that @xmath146 satisfies @xmath450 the repelling petal @xmath451 is the image of @xmath151 under the univalent map @xmath156 , and the notation @xmath157 is reserved for the inverse branch @xmath452 . set @xmath453 the proof will be divided into four propositions that we state independently , corresponding to three moments of the transition between @xmath67 and @xmath69 . the proofs will be given in [ sec : eps ] to [ sec : leaving ] . we first show that for the first @xmath454 iterates , the orbit stays close to an orbit of @xmath19 ( the bound @xmath455 is important here ) . [ entering ] assume @xmath456 and let @xmath457 be defined by @xmath458 . stands for _ incoming _ , and in proposition [ leaving ] below , @xmath459 stands for _ outgoing_. this convention was used in @xcite . ] then , @xmath460 , whence @xmath461 for @xmath71 large enough . moreover , @xmath462 the next two propositions concern the iterates between @xmath463 and @xmath464 . [ eps ] as @xmath465 , @xmath466 [ passing ] let @xmath467 be a sequence such that @xmath460 , whence @xmath461 for @xmath71 large enough . set @xmath468 then , @xmath469 , whence @xmath470 for @xmath71 large enough . moreover , as @xmath471 , @xmath472 the last one is similar to proposition [ entering ] . [ leaving ] let @xmath473 be a sequence contained in @xmath451 such that @xmath474 as @xmath81 . then , @xmath475 we start with proposition [ entering ] : if @xmath456 , then @xmath476 satisfies @xmath460 and as @xmath81 , @xmath477 according to proposition [ passing ] , @xmath478 satisfies @xmath469 and as @xmath81 , @xmath479 finally , since @xmath480 , proposition [ leaving ] implies that as @xmath81 , @xmath481 this in turn finishes the proof of proposition [ key ] because @xmath482 propositions [ entering ] , [ passing ] and [ leaving ] are valid if instead of using the sequence @xmath483 , we use the sequence @xmath484 defined by @xmath485 in that case , the only modification is for proposition [ eps ] which has to be replaced by : @xmath486 following the proof of proposition [ key ] , we get @xmath487 we thus see that in our non - autonomous context , where the dynamics slowly decelerates as the orbit transits between the eggbeaters , it takes exactly one more iteration to make the transition than in the classical case . in this paragraph , we prove proposition [ eps ] , which concerns the dynamics of @xmath20 only . we need to show that @xmath488 with @xmath489 denoting the attracting fatou coordinate of @xmath20 , for all @xmath490 , we have @xmath491$.}\ ] ] as a consequence , for @xmath492 $ ] it holds @xmath493 and @xmath494 it follows that @xmath495 and we are done . in this paragraph , we prove proposition [ entering ] , that is : if @xmath456 and @xmath496 then , as @xmath81 , @xmath497 choose @xmath498 sufficiently large so that @xmath499 for every fixed @xmath490 , the sequence of polynomials @xmath500 converges locally uniformly to @xmath19 . it follows that for every @xmath501 $ ] , the sequence @xmath502 converges uniformly to @xmath503 on @xmath434 . in particular , if @xmath71 is large enough , then @xmath504 , \text { and } \f_{n^2+\kappa_0 , n^2}(c_f)\subset \pa.\ ] ] in addition , since @xmath505 is close to @xmath506 , then for large @xmath71 we also have @xmath507 we now prove that if @xmath71 is large enough and @xmath508 $ ] , then @xmath509 . for this purpose , we work in the coordinate @xmath510 . for @xmath439 , consider the rational map @xmath511 defined by @xmath512 ^ 2}{4+\pi^2 w_m \cdot f(z)}.\ ] ] this has to be understood as a perturbation of @xmath513 . notice however that the remainder term @xmath514 is _ not _ negligible with respect to @xmath515 as @xmath516 , so we have to control precisely for which values of @xmath198 the remainder is indeed small . since @xmath517 as @xmath518 and since @xmath519 for @xmath520 $ ] , we get @xmath521 in particular , according to the maximum principle and the choice of @xmath522 - see ( [ eq : choicer ] ) - for @xmath71 large enough , if @xmath520 $ ] then @xmath523 an easy induction on @xmath8 shows that for every @xmath508 $ ] and every @xmath456 , @xmath524 indeed , the induction hypothesis clearly holds for @xmath525 and if it holds for some @xmath526 $ ] , then @xmath527 if @xmath198 belongs to the latter disk , then @xmath528 and using ( [ kappa0 ] ) , @xmath529 this shows that @xmath530 for all @xmath508 $ ] and all @xmath456 . we finally prove that for @xmath520 $ ] @xmath531 this is clear for @xmath501 $ ] since for each fixed @xmath8 , the sequence @xmath532 converges uniformly to @xmath503 on @xmath434 . so it is enough to prove the estimate for @xmath508 $ ] . we have that @xmath533 as @xmath534 also , we saw in that for @xmath508 $ ] , @xmath535 it follows that for @xmath526 $ ] and @xmath456 , @xmath536,\ ] ] whence @xmath537 as a consequence , for @xmath508 $ ] , @xmath538 where the second equality follows from the estimate @xmath539 and the fact that @xmath540 since @xmath541 . taking @xmath542 , we conclude that @xmath543 and so , @xmath544 as required . the proof of proposition [ entering ] is completed . in this paragraph , we prove proposition [ passing ] , that is : if @xmath467 is a sequence such that @xmath460 and if @xmath545 then , as @xmath81 , @xmath546 the proof relies on the formalism of approximate fatou coordinates introduced in section [ sec : approxcoord ] and notation thereof ( in particular properties [ compatt ] , [ comprep ] and [ approx ] ) . figure [ quasitrans ] illustrates the proof . let @xmath547 which belongs to @xmath548 for @xmath71 large enough . * step 1 . * if @xmath460 , then for @xmath71 large enough , @xmath549 . set @xmath550 and note that @xmath551 according to property [ compatt ] , for @xmath71 large enough , @xmath552 and @xmath553 * step 2 . * we will now prove by induction on @xmath554 that for all @xmath555 $ ] , @xmath556 and @xmath557 indeed , for @xmath558 , we have that @xmath559 and according to step 1 , @xmath560 so the induction hypothesis holds in this case . now , assume the induction hypothesis holds for some @xmath561 $ ] . according to step 1 , @xmath562 in addition , @xmath563 it follows that @xmath564 and therefore @xmath565 since @xmath566 , we see that for large @xmath71 , @xmath567 . according to property [ approx ] , @xmath568 and @xmath569 * step 3 . * we now specialize to the case @xmath570 and set @xmath571 according to step 2 of the proof , @xmath572 and @xmath573 in particular , by using and proposition [ eps ] we get @xmath574 set @xmath575 since @xmath576 , from we deduce that @xmath577 . since in addition @xmath578 it follows that for @xmath71 large enough , @xmath579 . thus we compute @xmath580 where we deduce the first line by property 2 , the second line follows by , the third line holds thanks to property 1 , and the final line holds since @xmath581 . this completes the proof . in this paragraph , we prove proposition [ leaving ] , that is : if @xmath473 is a sequence contained in the repelling petal @xmath451 and if @xmath582 then , as @xmath81 , @xmath583 set @xmath584 and for @xmath585 $ ] , set @xmath586 let @xmath587 be an integer such that for all @xmath588 , @xmath589 we prove by induction on @xmath8 that for @xmath71 large enough , if @xmath590 $ ] , then @xmath591 first , the induction hypothesis clearly holds for @xmath592 . so , let us assume it is true for some @xmath593 $ ] . as in proposition [ entering ] , since @xmath541 and @xmath594 , @xmath595 it follows that for large @xmath71 , @xmath596 since @xmath597 and @xmath598 , taking @xmath71 larger if necessary , @xmath599 belongs to the repelling petal @xmath451 . next , since @xmath600 as in proposition [ entering ] we see that @xmath601 and the proof of the induction is completed . by the previous step we have that @xmath602 applying @xmath156 on both sides yields @xmath603 since the sequence of polynomials @xmath604 converges locally uniformly to @xmath605 , we deduce that @xmath606 thereby concluding the proof of proposition [ leaving ] . this section is devoted to the proof of proposition [ attracting ] . given @xmath607 , let @xmath608 be the cubic polynomial defined by @xmath609 we must show that if @xmath25 is sufficiently close to @xmath48 and @xmath95 , then the lavaurs map @xmath610 admits an attracting fixed point . notice that since @xmath611 commutes with @xmath612 , it therefore has infinitely many of them . set @xmath613 this is an open set containing an upper half - plane and a lower half - plane . note that @xmath614 semi - conjugates @xmath615 to @xmath611 . since @xmath616 is univalent in a left half - plane , it is enough to show that @xmath615 has an attracting fixed point with real part arbitrarily close to @xmath617 . since @xmath615 commutes with the translation by @xmath51 , it is therefore enough to show that @xmath618 has an attracting fixed point . the open set @xmath619 is invariant by @xmath50 and the map @xmath615 commutes with @xmath50 . the set @xmath620 is an open subset of @xmath0 and the map @xmath621 is analytic . the universal cover @xmath622 semi - conjugates @xmath615 to a map @xmath623 the open set @xmath624 is a neighborhood of @xmath48 and @xmath392 in @xmath625 . the map @xmath626 has removable singularities at @xmath48 and @xmath392 , see the proof of lemma [ multhorn ] below , thus it extends as a map @xmath627 , where @xmath628 is the riemann sphere and @xmath629 . we set @xmath630 [ multhorn ] the points @xmath48 and @xmath392 in @xmath631 are fixed points of @xmath632 . both fixed points have multiplier @xmath633 . as @xmath634 , we have that @xmath635 where @xmath223 is the principal branch of logarithm . note that @xmath636 as @xmath634 . thus , @xmath637 as a consequence , as @xmath638 , we have that @xmath639 thus we conclude that @xmath48 is a fixed point of @xmath626 with multiplier @xmath640 . a similar argument shows that @xmath392 is also a fixed point of @xmath626 with multiplier @xmath640 . in particular , we see that for @xmath96 , the map @xmath641 has multiple fixed points at @xmath48 and @xmath392 . the multiplicity of @xmath48 and @xmath392 as fixed points of @xmath641 is @xmath642 . the mapping @xmath643 is a finite type analytic map in the sense of epstein ( see appendix [ sec : finitetype ] ) . therefore each attracting petal at @xmath48 or at @xmath392 must attract the infinite orbit of a singular value of @xmath641 . indeed if not , the component @xmath644 of the immediate basin containing this petal would avoid the singular values of @xmath641 . the restriction @xmath645 would then be a covering and the corresponding attracting fatou coordinate would extend to a covering map @xmath646 . this would force @xmath644 to be isomorphic to @xmath647 , which is not possible since @xmath644 is contained in @xmath625 . according to proposition [ prop : sing ] , the finite type map @xmath641 admits exactly two critical values ( the images of the critical values of @xmath648 under the map @xmath649 ) and two singular values which are respectively fixed at @xmath48 and @xmath392 . it follows that the number of attracting petals at @xmath48 plus the number of attracting petals at @xmath392 is at most @xmath642 . so this number must be equal to @xmath642 and the result follows . as we perturb @xmath650 away from @xmath51 , the multiple fixed point at @xmath48 splits into a pair of fixed points of @xmath626 : one at @xmath48 with multiplier @xmath640 and another one denoted by @xmath651 , with multiplier @xmath652 . we use a classical residue computation to estimate this multiplier . let @xmath653 be a small loop around @xmath48 . the cauchy residue formula yields @xmath654 from this it follows that @xmath655 now observe that @xmath656 and similarly , @xmath657 as a consequence when @xmath25 is sufficiently close to @xmath48 and @xmath658 , we deduce that @xmath659 , so @xmath651 is an attracting fixed point . let finally @xmath660 be a preimage of @xmath651 under @xmath661 , that is @xmath662 . we claim that for @xmath650 sufficiently close to @xmath51 , @xmath660 is a fixed point of @xmath615 . indeed observe first that @xmath663 is an integer which does not depend on the choice of preimage @xmath660 . therefore it is sufficient to prove that @xmath664 this may be seen as follows . the function @xmath665 is periodic of period @xmath51 , hence of the form @xmath666 for some function @xmath667 . the function @xmath668 is analytic , in particular continuous . so , @xmath669 the last equality follows from the proof of lemma [ multhorn ] . this shows that for @xmath650 sufficiently close to @xmath51 , @xmath660 is a fixed point of @xmath615 with multiplier @xmath652 , and the proof of proposition [ attracting ] is complete . in this section , we prove proposition [ superattracting ] , which shows the existence of real polynomial maps in two complex variables with wandering fatou components intersecting @xmath28 . let us consider the polynomial @xmath670 . we seek a parameter @xmath31 such that the lavaurs map @xmath97 has a fixed critical point in @xmath98 . set @xmath671\and f_c(z):= z+z^2-\frac{1 + 2c}{4c^3 } z^4.\ ] ] as @xmath238 increases from @xmath672 to @xmath673 , the corresponding parameter @xmath674 decreases from @xmath675 to @xmath676 and as @xmath238 increases from @xmath673 to @xmath677 , the parameter @xmath674 increases from @xmath676 to @xmath48 . the point @xmath238 is a critical point of the polynomial @xmath678 . as a consequence , @xmath679 so , @xmath238 is a critical point of @xmath680 , whence a critical point of @xmath681 . * claim 1 : * when @xmath682 $ ] , the critical point @xmath238 belongs to the parabolic basin @xmath683 . * claim 2 : * the function @xmath684\to \r$ ] defined by @xmath685 is continuous . * claim 3 : * @xmath686 . * claim 4 : * there is a sequence @xmath687 converging to @xmath672 with @xmath688 . these four claims are enough to get the desired conclusion . indeed , the function @xmath689 takes a positive value at @xmath690 and takes negative values arbitrarily close to @xmath672 . since it is continuous , it follows from the intermediate value theorem that it must vanish somewhere in @xmath691 . figure [ graphl ] shows the graph of the function @xmath692 which intersects the diagonal . as @xmath238 tends to @xmath693 , @xmath694 accumulates the whole interval @xmath695 $ ] with @xmath696 . a numerical experiment suggests that the function @xmath697 vanishes for a value of @xmath238 close to @xmath698 . accordingly , for @xmath699 the lavaurs map @xmath17 has a real attracting fixed point . it is enough to show that @xmath700 for @xmath701 . indeed , if so , then the sequence @xmath702 stays in @xmath703 and it is non - decreasing , so it must converge to the unique fixed point in @xmath704 $ ] , namely to the parabolic fixed point @xmath48 . to see that @xmath705 on @xmath703 , note that if @xmath706 $ ] , then @xmath707 $ ] and if @xmath701 , then @xmath708 thus , @xmath709 to see that @xmath710 on @xmath703 , it is enough to see that @xmath711 on @xmath703 . as above , for @xmath706 $ ] and @xmath701 , we have @xmath712 thus , @xmath20 is increasing on @xmath703 and since @xmath713 the proof of claim 1 is completed . for @xmath714 we may consider the attracting fatou coordinate @xmath680 and the repelling fatou parameterization @xmath715 of @xmath678 ( normalized according to our usual convention , see appendix [ app ] ) . the formulas and defining @xmath680 and @xmath716 as limits show that @xmath680 and @xmath716 take real values on the real axis . define @xmath717 propositions [ prop : phicontinuous ] and [ prop : psicontinuous ] imply that @xmath718 are continuous , as well as their composition @xmath719 now for @xmath720 $ ] , the point @xmath721 belongs to @xmath722 so we conclude that the function @xmath723 is continuous on @xmath724 $ ] . assume @xmath690 so that @xmath725 is the quadratic polynomial @xmath726 . the repelling fatou parametrization sends points on @xmath727 which are sufficiently close to @xmath617 to points on @xmath728 which are close to @xmath48 . since @xmath729 and since @xmath730 for all @xmath439 , we see that @xmath731 . as a consequence , @xmath732 . let us first study the behavior of @xmath733 when @xmath238 is close to @xmath672 . putting @xmath734 we compute @xmath735 let @xmath736 . then the asymptotic expansion of @xmath680 ( see [ subs : appatt ] ) at 0 yields @xmath737 thus the sequence of maps @xmath738 defined by : @xmath739 converges uniformly to the identity on compact intervals of @xmath727 . now let us consider the map @xmath740 for @xmath741 . figure [ graph ] shows the graph of @xmath19 . the fixed points of @xmath19 are @xmath48 , @xmath742 and @xmath743 . the critical points of @xmath19 are @xmath744 , @xmath745 and @xmath746 . we see that @xmath747 thus , @xmath19 sends the interval @xmath748 into itself and the orbit of any point in this interval escapes to @xmath617 . in particular , the orbit of @xmath749 escapes to @xmath617 . in addition , @xmath750 on @xmath751 $ ] , and @xmath19 is increasing on @xmath751 $ ] . so , we can define a sequence @xmath752 recursively by @xmath753 and for @xmath439 , @xmath754 this sequence is decreasing and converges to a fixed point of @xmath19 , thus to @xmath48 . choose @xmath755 large enough so that @xmath756 belongs to the repelling petal @xmath451 . hence @xmath757 . choose @xmath758 small enough so that for all @xmath759 , the point @xmath760 belongs to the repelling petal of @xmath678 and @xmath761 where @xmath762 is the leftmost fixed point of @xmath678 in @xmath727 . in particular , for all @xmath763 , we have that @xmath764 . for @xmath759 , set @xmath765 and @xmath766 note that @xmath767 since @xmath760 lies in the repelling petal of @xmath678 and @xmath768 varies continuously with @xmath238 . therefore we also have that @xmath769 . together with the intermediate value theorem , this implies that for large enough @xmath71 , the equation @xmath770 admits at least one solution @xmath771 , where @xmath772 denotes the fractional part of @xmath773 . now set @xmath774 we have that @xmath775 thus , @xmath776 finally , since @xmath777 for @xmath71 large enough and since @xmath778 for all @xmath763 and all @xmath759 , we deduce that @xmath779 for @xmath71 large enough . this completes the proof of claim 4 . throughout this section @xmath14 is a polynomial of the form @xmath780 in the coordinate @xmath781 , the expression of @xmath19 becomes @xmath782 choose @xmath125 sufficiently large so that @xmath513 is univalent on @xmath783 , and the estimates hold . denote by @xmath149 the right half - plane @xmath150 and by @xmath151 the corresponding left half - plane . as @xmath785 , @xmath786 the map @xmath513 is univalent on the right half - plane @xmath149 and if @xmath787 , then @xmath788 it follows that the sequence of univalent maps @xmath789 is normal and converges locally uniformly to a univalent map @xmath790 satisfying @xmath791 in addition , @xmath792 transferring this to the initial coordinate , we see that the sequence of mappings @xmath793 defined by @xmath794 converges locally uniformly to an _ attracting fatou coordinate _ @xmath49 which semi - conjugates @xmath795 to @xmath796 , that is @xmath797 , and satisfies @xmath798 the restriction of @xmath155 to the _ attracting petal _ @xmath799 coincides with @xmath800 , hence it is univalent . figure [ extendedattfatou ] illustrates the behavior of the extended fatou coordinate for the cubic polynomial @xmath802 which has two critical points @xmath803 . the basin of attraction @xmath77 is colored according to the following scheme : as @xmath807 , @xmath808 it follows that if @xmath125 is sufficiently large and @xmath809 , then @xmath810 in that case , the sequence of univalent maps @xmath811 converges locally uniformly to a map @xmath812 satisfying @xmath813 in addition , @xmath814 transferring this to the initial coordinate , we see that the sequence of maps @xmath815 converges locally uniformly to an _ repelling fatou parameterization _ @xmath816 which semi - conjugates @xmath796 to @xmath14 , that is @xmath817 , and satisfies @xmath818 the restriction of @xmath156 to the left half - plane @xmath151 coincides with @xmath819 , whence is univalent . the image @xmath820 is called a _ repelling petal_. for @xmath821 , the lavaurs map with phase @xmath822 is the map @xmath823 in this article , we are only concerned by the lavaurs map @xmath824 with phase @xmath48 . the relevance of lavaurs maps is justified by the following result due to pierre lavaurs @xcite . [ th : lavaurs ] let @xmath14 be a polynomial such that @xmath825 as @xmath826 . for @xmath827 , set @xmath828 . let @xmath829 be a sequence of complex numbers and @xmath830 be a sequence of integers , such that @xmath831 then , the sequence of polynomials @xmath832 converges locally uniformly on @xmath77 to @xmath833 . it is also relevant to consider the map @xmath834 the repelling parameterization @xmath156 semi - conjugates @xmath835 to @xmath17 . figure [ ef ] illustrates the behavior of the map @xmath836 for @xmath837 . note that @xmath836 commutes with @xmath50 : @xmath838 so , the universal cover @xmath839 semi - conjugates @xmath835 to a map @xmath840 the map @xmath841 has removable singularities at @xmath48 and @xmath392 , thus it extends as a map @xmath842 , where @xmath628 is the riemann sphere and @xmath843 . the map @xmath842 is called the _ horn map _ associated to @xmath19 . as observed by adam epstein in his phd thesis @xcite , this horn map is a finite type analytic map ( see definition [ def : finite ] below ) . let @xmath844 be an analytic map of complex 1-manifolds , possibly disconnected . an open set @xmath845 is _ evenly covered _ by @xmath846 if @xmath847 is a homeomorphism for each component @xmath100 of @xmath848 ; we say that @xmath849 is a _ regular value _ for @xmath846 if some neighborhood @xmath109 of @xmath760 is evenly covered , and a _ singular value _ for @xmath846 otherwise . note that the set @xmath850 of singular values is closed . recall that @xmath127 is a _ critical point _ if the derivative of @xmath846 at @xmath180 vanishes , and then @xmath851 is a _ critical value_. we say that @xmath852 is an _ asymptotic value _ if @xmath846 approaches @xmath760 along some path tending to infinity relative to @xmath853 . it follows from elementary covering space theory that the critical values together with the asymptotic values form a dense subset of @xmath850 . in particular , every isolated point of @xmath850 is a critical or asymptotic value . when @xmath844 is a finite type analytic map with @xmath855 , we say that @xmath846 is a finite type analytic map on @xmath854 . the reason why finite type analytic maps are relevant when studying lavaurs maps is the following . let @xmath856 be a rational map , let @xmath49 be an attracting fatou coordinate defined on the parabolic basin of some fixed point of @xmath19 with multiplier @xmath51 and let @xmath857 be a repelling fatou parameterization associated to some fixed point of @xmath19 with multiplier @xmath51 . define @xmath858 finally set @xmath859 and let @xmath842 be defined by @xmath860 the following result is stated as ( * ? ? ? 7.3 ) . * @xmath48 and @xmath392 , which are fixed asymptotic values of @xmath841 , and * the images by @xmath861 of the critical orbits of @xmath19 contained in @xmath77 , which are critical values of @xmath841 . @xmath862 m. lyubich , _ non - existence of wandering intervals and structure of topological attractors of one dimensional dynamical systems 2 . the smooth case _ ergodic theory dynam . systems * 9 * ( 1989 ) , no . 4 , 751758 . , _ non - existence of wandering intervals and structure of topological attractors of one dimensional dynamical systems 1 . the case of negative schwarzian derivative . _ ergodic theory dynam . systems * 9 * ( 1989 ) , no . 4 , 737749 .

in the strong - coupling bec region where a feshbach resonance gives rise to tightly - bound dimer molecules , we show that a spin - polarized fermi superfluid reduces to a simple bose - fermi mixture of bose - condensed dimers and the leftover unpaired fermions . using a many - body functional integral formalism , the gaussian fluctuations give rise to an induced dimer - dimer interaction mediated by the unpaired fermions , with the dimer - fermion vertex being given by the ( mean - field ) born approximation . treating the pairing fluctuations to quartic order , we show how the action for a spin - polarized fermi superfluid reduces to one for a bose - fermi mixture . this bose - fermi action includes an expression for the effective dimer - unpaired fermion interaction in a spin - polarized fermi superfluid beyond the born approximation , in the superfluid phase at finite temperatures . in the low - density limit , we show how this dimer - fermion interaction gives the-wave scattering length ( is the-wave fermion scattering length ) , a result first derived by skorniakov and ter - martirosian in 1957 for three interacting fermions . 0 * 0 *

cond-mat0703433

since superfluidity was first realized in ultracold trapped two - component atomic fermi gases @xcite , there has been increasing interest in the properties of spin - polarized fermi superfluids @xcite . as in unpolarized fermi gas superfluids , by making use of a feshbach scattering resonance between fermions , the effective scattering length @xmath3 that characterizes low - energy scattering between fermions in different hyperfine states ( i.e. , @xmath4 ) can be continuously tuned from negative to positive values . we use @xmath5 to denote the two species of atomic fermions prepared in different hyperfine states . when @xmath6 , the system is in the bcs region and pairing can occur about the fermi surface between fermions in different hyperfine states . passing through unitarity ( where @xmath7 ) , the scattering length becomes positive and one enters the bec region , characterized by the disappearance of the fermi surface ( the fermion chemical potential becomes negative ) and the appearance of strongly bound molecular pairs . in the spin - polarized fermi superfluid , there is an excess of one of the two - species of fermions : @xmath8 . in the bec region , all of the minority species of fermions are paired up , leaving a gas of the remaining spin @xmath9 fermions . thus , in the bec region , a spin - polarized fermi gas should behave like a bose - fermi mixture of the dimer molecules and unpaired excess fermions . in this paper , we prove that the effective action for a spin - polarized two - component fermi superfluid in the bec region reduces to an effective action for a bose - fermi mixture . the idea that a spin - polarized fermi superfluid reduces to a bose - fermi mixture in the bec region has been explored in refs . we extend those studies by considering quartic pairing fluctuations of the effective bose action for a spin - polarized fermi superfluid . at this order , we show that a spin - polarized fermi superfluid can still be described in terms of an equivalent bose - fermi mixture with a renormalized dimer - fermion interaction that is valid beyond the born approximation value @xmath10 @xcite , where @xmath11 is the reduced mass of the molecular dimer - atomic fermion system . our discussion of the effective action of a spin - polarized fermi gas is based on the functional integration treatment @xcite of the bcs - bec crossover problem developed in refs . @xcite . extending the approach of ref . @xcite , we expand this action in powers of pairing fluctuations about the bcs - type mean - field saddle - point up to quartic order in fluctuations . at the level of quadratic ( gaussian ) fluctuations , we derive the bogoliubov theory of excitations of a dimer molecular condensate in the spin - polarized gas in sec . [ bogspectrum ] . we show how the bogoliubov excitation energy involves a dimer - dimer interaction mediated by the unpaired @xmath9 fermions through a lindhard response function . this result is compared with the bogoliubov excitation spectrum in a bose - fermi mixture in sec . [ gaussianbf ] . in sec . [ quartic ] we discuss the quartic fluctuations of an effective bose action for a bose - fermi mixture as well as for a spin - polarized fermi gas in the bec limit . comparing the two , we prove their equivalence and derive an expression for the momentum- and energy - dependent dimer - fermion interaction in a spin - polarized . in order to make contact with the extensive literature on two - body dimer - fermion scattering in free space @xcite , in sec . [ tbsa ] we show how our theory reproduces the well - known result @xmath1 due to skorniakov and ter - martirosian @xcite for the dimer - fermion scattering length . we first review the functional integral formalism for an ultracold two - component spin - polarized fermi gas , as well as discuss some of the essential physics . at the low temperatures of interest , @xmath0-wave scattering between fermions in different hyperfine states dominates all other partial - wave amplitudes . the @xmath0-wave interaction can be calculated using the pseudopotential u _ ( ) = -u ( ) . the partition function for the spin - polarized fermi gas can be written as a functional integral over fermionic grassmann fields @xmath12 @xcite : = [ ^*,]e^-s[^*,],[z0]where the euclidean ( imaginary time ) action is given by s = ^_0 d , and the hamiltonian @xmath13 is h&=&d_^*_(x)(--_)_(x ) + & & -ud^*_(x)^*_(x)_(x)_(x).[h]here , @xmath14 is a four - vector denoting the spatial coordinate @xmath15 and the imaginary time variable @xmath16 . @xmath17 is the inverse temperature . we set @xmath18 and the volume to be unity . we restrict ourselves to a uniform gas in this paper . the chemical potentials @xmath19 in eq . ( [ h ] ) are determined from the standard thermodynamic relation @xmath20 , where @xmath21 . a spin polarized fermi gas ( with @xmath22 ) is distinguished from an unpolarized gas by having different chemical potentials for each species : @xmath23 . introducing the bose ( superfluid ) pairing field @xmath24 through the hubbard - stratonovich transformation and integrating out the fermi grassmann fields , the partition function in eq . ( [ z0 ] ) becomes : = [ , ^ * ] e^-s _ [ , ^*],[bpartition ] where s _ [ , ^ * ] _ 0^dd - . [ seffq ] the inverse of the @xmath25 nambu - gorkov bcs green s function is ^-1(x , x ) & = & ( cc - _ + + _ & ( x ) + ^*(x ) & - _ - - _ ) + & & ( x - x ) . [ g ] this approach has been used in ref . @xcite for a spin - polarized fermi gas . expanding the action given by eq . ( [ seffq ] ) up to quadratic order in fluctuations @xmath26 about the saddle - point value @xmath27 for the bose - pairing field , and fourier - transforming the resulting expression , the bcs mean - field @xmath28 and gaussian fluctuation @xmath29 contributions are given by @xcite s^(0 ) & = & - _ k[s0k]and s^(2 ) & = & _ q + & & + _ k , q[_0(k)(-q ) _ 0(k+q)(q ) ] + & & _ q^ . [ s2 kb ] here , _ 0(k ) & & + & & ( cc ik_n + _ , & -_0 + -_0 & ik_n - _ , ) [ g0q ] is the bcs @xmath25 matrix green s function for the spin - polarized fermi superfluid , and the fluctuation self - energy is defined as ( q ) = ( cc 0 & _ q + ^*_-q & 0 ) , [ bsig]and @xmath30 . the poles of @xmath31 are given by e^ _ = ^-_,[epm ] where @xmath32 . the green s function in eq . ( [ g0q ] ) describes the bcs excitation spectrum @xcite in a spin - polarized fermi superfluid . it reduces to the standard result when @xmath33 , in which case @xmath34 and @xmath35 . in the last line of eq . ( [ s2 kb ] ) , we have defined the spinor @xmath36 . also , @xmath37 and @xmath38 are four - vectors denoting the wave vectors @xmath39 and @xmath40 as well as the fermi and bose matsubara frequencies , @xmath41 and @xmath42 , respectively . throughout this paper , @xmath43 shall refer to a fermi frequency and @xmath44 a bose frequency . the matrix elements of the inverse @xmath45 matrix pair fluctuation propagator @xmath46 defined in eq . ( [ s2 kb ] ) are given by @xcite & & + & & + _ k g_0,11(k+q)g_0,22(k ) [ m11q]and m_12(q)=m_21(q)=_k g_0,12(k+q)g_0,12(k).[m12q ] here , @xmath47 denotes the @xmath48-th element of the green s function defined by eq . ( [ g0q ] ) . the number of fermions in state @xmath49 is given by @xmath50 ) , where the thermodynamic potential is @xcite - _ k+_q(q ) . + in the bec limit [ @xmath51 , the chemical potential for the minority species is large and negative , roughly corresponding to the binding energy of a dimer molecule @xcite . using this fact , one can show in the bec limit and above the superfluid transition temperature @xmath52 that ( see also the related discussion in ref . @xcite ) n _ = _ f ( _ , ) + _ n_b(_b,),[nsig]where @xmath53 with the boson chemical potential @xmath54 @xcite and molecular mass @xmath55 . here , @xmath56 and @xmath57 are the fermi and bose thermal distribution functions , respectively . the fermi distribution function describes the unpaired free fermions while the bose distribution function describes the fermions in bound states . deep in the bec region , there will be no unpaired @xmath58 fermions , and thus @xmath59 . this means that the number of excess unpaired @xmath9 fermions is given by [ using eq . ( [ nsig ] ) ] n_f n_-n _ & = & _ + & & _ f ( _ , ) . [ deltan2]this shows that the chemical potential @xmath60 for the majority species determines the number of unpaired excess fermions . one can think of @xmath60 as being roughly equal to the fermi energy of a gas of unpaired fermions of density @xmath61 , i.e. , @xmath62 . deep in the bec region , the dimer - unpaired fermion interactions are weak and eq . ( [ deltan2 ] ) will also be valid in the superfluid phase . in the strong - coupling bec limit , @xmath27 is much smaller than the dimer binding energy @xcite and consequently , @xmath63 . in this limit , the two quasiparticle branches in eq . ( [ epm ] ) can be approximated by e^+ _ - _ + g()[ep]and e^- _ & & -_+ g ( ) + & & + |e_b| + g(),[em ] where g().[term]in sec . [ quartic ] , we will show that @xmath64 is the low - density limit of the dimer - fermion interaction for a fermion of momentum @xmath39 interacting with a dimer boson , multiplied by the condensate density of dimers . thus , @xmath64 is the expected self - energy correction for the fermions due to interactions with the bose condensate of dimer molecules . hence , @xmath65 in eq . ( [ ep ] ) is precisely the spectrum one would expect for an unpaired fermion interacting with a bose - condensate of dimer molecules . the second fermionic branch , given by eq . ( [ em ] ) , involves the break - up of a dimer pair and this excitation is frozen out in the bec limit , where the dimer binding energy @xmath66 is very large . in this section , we will derive the bogoliubov excitation energy in a spin - polarized fermi gas in the bec region using the results of sec . [ spfg ] . summing over the fermion matsubara frequencies in eqs . ( [ m11q ] ) and ( [ m12q ] ) , the matrix elements of the inverse matrix propagator for pair fluctuations are given by ( see refs . @xcite ) & & + & & + _ [ m11b]and & & + & & _ , [ m12b]where f^ _ f(e^ _ ) are the fermi distribution functions and e _ = e^+ _ -^-_.here , @xmath67 and @xmath68 are the bogoliubov quasiparticle amplitudes for a spin - polarized fermi superfluid . we note that @xmath69 is not the physical quasiparticle dispersion ( given by @xmath70 ) , except when the polarization vanishes ( @xmath71 ) . the processes described by the first two lines ( on the right - hand side ) in eqs . ( [ m11b ] ) and ( [ m12b ] ) correspond to pair creation / destruction of two excitations @xmath72 , while the last two terms describe creation / destruction of particle - hole excitations @xmath73 . following the usual approach for unpolarized fermi superfluids @xcite , in the bec limit , we expand the first and second lines of eqs . ( [ m11b ] ) and ( [ m12b ] ) in powers of @xmath74 . for an unpolarized gas ( @xmath75 ) , @xmath76 and @xmath77 since @xmath78 in the bec region ( recall that @xmath79 in the bec region while @xmath80 is much larger when @xmath81 ) . thus , the third and fourth lines in eqs . ( [ m11b ] ) and ( [ m12b ] ) , describing particle - hole excitations , vanish . however , for the polarized superfluid , using eq . ( [ em ] ) we see that in the bec limit we still have that @xmath82 since the chemical potential for the minority species is large and negative . on the other hand , @xmath83 is _ positive _ , where @xmath84 is the fermi energy of the excess fermions , with density @xmath85 . thus , f^+ _ & & f(e^+ _ ) + & & f ( _ , + g ( ) ) + & & f(^_,),[xip]does _ not _ vanish in the bec limit . this means that for a polarized fermi superfluid in the bec limit , the particle - hole terms given by the third lines in both eqs . ( [ m11b ] ) and ( [ m12b ] ) do not vanish . we will show that these terms ( which we denote as @xmath86 and @xmath87 , respectively ) can be written in terms of a fermionic particle - hole density response function in the bec limit [ see eq . ( [ r11r12 ] ) ] . following the previous discussion , in the bec limit , the matrix elements of the inverse pair fluctuation propagator in eqs . ( [ m11b ] ) and ( [ m12b ] ) are approximated by m_11 a + b^2 - c(iq_m ) + r_11(q ) [ m11f]and m_12 a + d^2 + r_12(q),[m12f]where the expansion coefficients @xmath88 , and @xmath89 ) describe the small-@xmath90 expansion of the first two lines in eqs . ( [ m11b ] ) and ( [ m12b ] ) . to leading order in @xmath91 , these coefficients reduce to ( see refs . @xcite and @xcite for similar calculations ) a = _ ( 1-f(_,)),[a0]b = _ ( 1-f(_,)),and c = _ ( 1-f(_,)),[c0]where @xmath92 . @xmath93 is vanishingly small in the bec limit @xcite . the mean - field bcs gap equation follows from the requirement that the saddle - point action @xmath28 be stationary with respect to fluctuations @xmath26 about @xmath27 : @xmath94 . this gives the usual expression = _ kg_0,12(k).[gap0 ] in the bec limit , the off - diagonal bcs green s functions in eq . ( [ g0q ] ) are given by g_0,12(k ) = g_0,21(k)-_0_0,(k)_0,(k),[g12]where we have introduced the normal single - particle green s functions g_0,(k ) = , g_0,(k ) = . [ gup]using eq . ( [ g12 ] ) , eq . ( [ gap0 ] ) reduces to & = & -_k _ 0,(k)_0,(k ) + & = & _ .[gap ] introducing the standard regularized two - body potential @xcite = - + _ , [ ls]we can replace @xmath95 in eq . ( [ gap ] ) in terms of @xmath2 . in order to solve eq . ( [ gap ] ) analytically , we use the approximation @xmath96 , valid at the mean - field level for low densities of the unpaired @xmath9 fermions since @xmath97 , where @xmath98 is the fermi wave vector for the unpaired @xmath9 fermions . we find & & + & & .[gap1]since @xmath99 for @xmath100 and @xmath101 , we have _ _ f( _ , ) n_f , + [ approx]and thus eq . ( [ gap1 ] ) reduces to |_| - _ |e_b|(1 - 8(n_fa^3_f ) - 2).[gap2]recall that @xmath61 is the density of the unpaired excess @xmath9 fermions . the corrections on the right - hand side of eq . ( [ gap2 ] ) , due to a finite density of unpaired fermions as well as a finite condensate density , are both higher - order corrections ( @xmath102 $ ] and @xmath103 $ ] , respectively @xcite ) . in the bec limit , we ignore these corrections . we note that @xmath104 . thus , the contributions we drop from eq . ( [ gap2 ] ) are much smaller than @xmath60 in the bec limit and the gap equation reduces to |_| - _ = ( 1+[n_fa^3_f]).[gap3]in eqs . ( [ a0])-([c0 ] ) , the terms containing the fermi thermal factor @xmath105 are also @xmath106 $ ] , as can be shown by using the approximation in eq . ( [ approx ] ) . using eq . ( [ gap3 ] ) , the expansion coefficients reduce to a = ( 1+[n_fa^3_f]),[a]b = ( 1+[n_fa^3_f]),[b]and c = ( 1+[n_fa^3_f]).[c ] we now turn to the evaluation of the @xmath107 and @xmath108 terms in eqs . ( [ m11f ] ) and ( [ m12f ] ) . using the expansion e _ & & ^+ _ + ^2_0/2^+ _ + & & ^+ _ + g()in the third lines of eqs . ( [ m11b ] ) and ( [ m12b ] ) , we obtain r_11(q)r_12(q ) _ .[r11r12]as we shall show later , this term is of the order ^2_0a^3_f m^3(n_fa^3_f)^1/3.[magnitude]thus , comparing with the finite - density corrections to the coefficients given in eqs . ( [ a])-([c ] ) , we see that this is the leading - order finite - density correction . in summary , for small energy and momentum @xmath90 , the matrix elements of the inverse pair fluctuation propagator @xmath46 defined in eqs . ( [ m11b ] ) and ( [ m12b ] ) are given by m_11 & = & a + b^2 - c(iq_m ) + & & + _ + & = & + & = & [ m11c]and m_12&= & a + _ + & = & , [ m12c]where we have defined the effective interaction ^(0)_bb(,iq_m ) u^(0)_bb + _ . + [ ubbtilde]the direct dimer - dimer interaction is u^(0)_bb = , [ ubb0]where the molecular scattering length is given by the born approximation @xmath109 @xcite . the condensate density @xmath110 of dimer molecules which appears in eq . ( [ ubbtilde ] ) is given in the bec limit by @xcite n_c(t ) = . [ nc ] substituting the results in eqs . ( [ m11c ] ) and ( [ m12c ] ) into eq . ( [ s2 kb ] ) and defining the bose fluctuation field c_q _ q,[cq]the action describing gaussian fluctuations past the bcs mean - field is given by s^(2)&= & -^_q ^-1(q)_q,[s2sp]where @xmath111 and @xmath112 is the @xmath25 inverse matrix propagator for the bogoliubov excitations of a dimer condensate in a spin - polarized superfluid : & & + & & ( cc -iq_m + + n_c^(0)_bb(q ) & n_c^(0)_bb(q ) + n_c^(0)_bb(q ) & iq_m + + n_c^(0)_bb(q ) ) . + [ dtilde]equation ( [ dtilde ] ) is identical to the propagator for bogoliubov excitations in an unpolarized fermi superfluid in the bec limit @xcite , apart from the renormalized interaction @xmath113 defined in eq . ( [ ubbtilde ] ) . we now discuss the bogoliubov excitations of the dimer condensate which are described by the propagator defined in eq . ( [ dtilde ] ) . if we replace @xmath64 by its long - wavelength value @xmath114 , @xmath115 in eq . ( [ ubbtilde ] ) becomes ^(0)_bb(q ) u^(0)_bb + ( ) ^2(q , t),[utilde]where ( , iq_m;t ) _ , [ chifull]is the lindhard response function @xcite for the excess @xmath9 fermions . this function ( see , for example , ref . @xcite ) describes the induced fermionic density fluctuation in response to a density fluctuation in the bosons with energy @xmath116 and momentum @xmath40 . the fermion density fluctuation acts back on the bosons to give rise to a fermion - mediated boson - boson interaction . at low temperatures ( @xmath117 ) , one finds eq . ( [ chifull ] ) gives ( 0,0)-n(_f ) -,[chi]where @xmath118 is the fermionic density of states and @xmath119 is the fermi energy of the unpaired @xmath9 fermions . in eq . ( [ chi ] ) we have used the density of states @xmath120 of an ideal single - component fermi gas of density @xmath61 . using @xmath121 in eq . ( [ chi ] ) leads to the result given in eq . ( [ magnitude ] ) [ giving the magnitude of @xmath122 . we thus see that this correction in eq . ( [ utilde ] ) due to the effects of the unpaired excess fermions is much larger than the corrections we dropped in eqs . ( [ a])-([c ] ) . these higher - order terms due to a finite density of unpaired fermions would renormalize the parameters ( @xmath123 ) appearing in the bogoliubov propagator . however , we do not discuss this extension here . the long - wavelength bosonic spectrum of a spin - polarized fermi superfluid in the bec limit [ found from the solution of @xmath124 given by eq . ( [ dtilde ] ) ] , are the usual bogoliubov excitations , _ = , where the renormalized sound velocity @xmath125 is defined by c^2 = .[sound]since the induced interaction @xmath126 is attractive , the sound velocity becomes smaller as the density @xmath127 of the excess unpaired fermions is increased . above a critical value of the density corresponding to @xmath128 , the sound velocity becomes imaginary , signalling an instability towards phase - segregation of the bound dimers and unpaired fermions . using eqs . ( [ utilde ] ) , ( [ chi ] ) , and ( [ sound ] ) , one can show that the sound velocity is real as long as n_f^1/3 .[conditionps]thus , eq . ( [ conditionps ] ) gives the stability condition for the uniform superfluid phase of a spin - polarized fermi gas in the bec region [ see also eq . ( [ conditionps2 ] ) for further discussion ] . summarizing the results of this section , we have given an explicit expression for the spectrum of bogoliubov excitations of the dimer condensate in the bec limit of a spin - polarized fermi superfluid . since the renormalized phonon sound speed becomes imaginary above a critical density of the unpaired @xmath9 fermions , we showed how the bogoliubov spectrum also gives the condition for the system to become unstable towards phase segregation . in order to better understand the results given here , in sec . [ gaussianbf ] we discuss the gaussian fluctuations of a bose - fermi mixture and show that they are equivalent to the results of this section for a spin - polarized fermi superfluid . in order to understand our expression for the effective dimer - dimer interaction in eq . ( [ ubbtilde ] ) that arises in the bogoliubov excitation spectrum for a spin - polarized fermi superfluid in the bec limit , in this section we consider the gaussian fluctuations of a bose - fermi mixture . we show that at the gaussian level , a spin - polarized fermi superfluid in the bec limit can be modelled as a bose - fermi mixture and identify the effective dimer - unpaired fermion interaction ( within the born approximation ) . we introduce the following action for a bose - fermi mixture : s[a , c ] & = & _ k(-ik_n + -_f)a^*_k , a_k , + & & + _ q(-iq_m + - _ b)c^*_qc_q + & & + _ k , k,p u_bf(k , k,p ) c^*_p - kc_p - ka^*_k , a_k , + & & + _ q , q,ku_bb(q , q,k)c^*_+qc^*_-qc_-qc_+q . + [ seff]here , @xmath129 and @xmath130 is a fermi matsubara frequency representing the centre - of - mass degree of freedom of a boson - fermion pair . as with before , @xmath131 are also fermi frequencies , while @xmath132 and @xmath44 are bose frequencies . equation ( [ seff ] ) describes a system of bosons ( @xmath125 ) of mass @xmath133 and fermions ( @xmath134 ) of mass @xmath135 interacting via the potential @xmath136 , and bosons interacting with each other through @xmath137 . @xmath125 is a complex number bose field , while @xmath134 is a fermi grassmann field . @xmath138 and @xmath139 refer to the chemical potentials of the boson and fermions , respectively . in contrast to the spin - polarized gas in sec . [ spfg ] , only one species of fermions with chemical potential @xmath139 appears in eq . ( [ seff ] ) . this chemical potential is the same as @xmath60 in the spin - polarized gas , however , being determined by eq . ( [ deltan2 ] ) . integrating out the fermi grassmann fields from the partition function @xmath140e^{-s[a , c]}$ ] , we obtain an effective bose action which can be compared directly with the effective action for the bosonic fluctuations of a spin - polarized fermi superfluid . the model bose - fermi action given in eq . ( [ seff ] ) is bilinear in the fermi grassmann fields @xmath141 , and the gaussian integral can be evaluated in the usual way to give s_[c ] = s_b - , [ seffb]where s_b & & _ q(-iq_m + _ b,)c^*_qc_q + & & + _ q , q,ku_bb(q , q,k)c^*_+qc^*_-qc_-qc_+q + [ sb ] and a(k , k ) & & ( ik_n - _ f,)_k , k + & & - _ pu_bf(k , k,p ) c^*_p - kc_p - k,with @xmath142 and @xmath143 . treating fluctuations of the bose condensate as small , we apply the bogoliubov shift @xmath144 to @xmath145 and expand about its saddle point value @xmath146 : a(k , k ) = a_0(k ) + ^(1)(k , k ) + ^(2)(k , k),where a_0(k ) & = & ik_n - _ f , - n_cu_bf(k , k , k ) + & & ik_n - _f , , [ a_02 ] & & + & & , and ^(2)(k , k ) = - _ p u_bf(k , k,p)c^*_p - kc_p - k.[lambda2]using this expansion in eq . ( [ seffb ] ) , the saddle - point action is given by s^(0 ) = -_b n_c + - . [ s0]the gaussian fluctuations are described by s^(2 ) = s^(2)_b - [ a^-1_0^(2 ) ] + [ ( a^-1_0^(1))^2 ] , [ s2]where @xmath147 gives the gaussian fluctuation terms from @xmath148 in eq . ( [ sb ] ) . we now show that eq . ( [ s2 ] ) is equivalent to the analogous terms in eq . ( [ s2sp ] ) for a spin - polarized fermi superfluid . performing the frequency sums in the two terms in eq . ( [ s2 ] ) and replacing terms like @xmath149 by its free - space ( @xmath150 ) , long - wavelength , static ( @xmath90=0 ) value , u_bf()u_bf(k , k , k;ik_n=^2/2 m ) , we find & & + & & -_q c^*_q c_q_u_bf()f(_f,)[tra0l2b ] and & & + & & _ q(c^*_qc_q + c_qc_-q + c^*_qc^*_-q+c_-qc^*_-q ) + & & _ n_cu^2_bf ( ) , [ tra0l1l1b]where @xmath151 is defined in eq . ( [ a_02 ] ) : _f , = _ f , + n_cu_bf(k , k , k).[xif]using the born approximation value @xmath152 for the boson - boson interaction [ to facilitate comparison with eq . ( [ s2sp ] ) ] , the gaussian fluctuation terms @xmath147 of the bose action given in eq . ( [ sb ] ) reduce to s^(2)_b & = & _ q(-iq_m + _ b,)c^*_qc_q + & & + _ qu^(0)_bb(c_qc_-q + 4c^*_qc_q + c^*_-qc^*_q).[s2b ] the requirement that the saddle - point action given in eq . ( [ s0 ] ) be stationary ( @xmath153 ) gives _ b = u^(0)_bbn_c + _ u_bf()f(_f,).[mub ] using eqs . ( [ tra0l2b])-([mub ] ) , it is straightforward to show that the gaussian action in eq . ( [ s2 ] ) for the bose fluctuations of a bose - fermi mixture is the same as the gaussian action of a spin - polarized superfluid given by eq . ( [ s2sp ] ) . for the bose - fermi mixture , the effective boson - boson interaction is then given by the expression in eq . ( [ ubbtilde ] ) with u_bf ( ) = .[ubfgauss ] we note that this expression for @xmath136 has only been derived within a gaussian approximation . in sec . [ quartic ] , we show how the inclusion of quartic fluctuations leads to an improved result . using eqs . ( [ term ] ) and ( [ nc ] ) as well as @xmath154 , eq . ( [ ubfgauss ] ) reduces to u^(0)_bf .[ubf0]this corresponds to the born - approximation value of the dimer - fermion interaction @xcite , with reduced mass @xmath11 . substituting eq . ( [ ubf0 ] ) into eq . ( [ conditionps ] ) , we find that the stability condition for the spin - polarized fermi superfluid becomes n_f^1/3 .[conditionps2]this requirement is identical to the stability condition for atomic bose - fermi mixtures ( see , for instance , eq . ( 9 ) in ref . @xcite ) . by considering the gaussian fluctuations in the action for a spin - polarized fermi superfluid , we have shown how the bogoliubov excitations of the dimer condensate are affected by the remaining unpaired @xmath9 fermions . at the gaussian level , however , the renormalized dimer - dimer interaction only involves the born - approximation value for the dimer - fermion interaction @xmath136 as given in eq . ( [ ubf0 ] ) . to derive a better expression for the dimer - fermion interaction , we need to include quartic fluctuation terms . this is done in sec . [ quartic ] . in this section , we consider the gaussian fluctuations of a spin - polarized gas in the normal phase . the advantage of doing this is that we do not need to work with the small @xmath90 expansion for the two - fermion vertex function , as we did in the superfluid phase discussed in sec . [ bogspectrum ] . we use the results obtained in this section in our derivation of the effective dimer - fermion interaction given in sec . [ quartic ] . even though we carry out our analysis in the normal state , we restrict ourselves to temperatures well below the dimer binding energy . thus , we are still dealing with a bose - fermi mixture . the dimers , however , are not bose - condensed . we emphasize , though , that in the extreme bec limit , @xmath27 is much smaller than the binding energy of the diatomic molecules even at @xmath155 @xcite . consequently , in this region , the effective interaction is the same whether the molecules ( cooper pairs ) are bose - condensed ( @xmath156 ) or not . in the normal state , the gaussian action given in eq . ( [ s2 kb ] ) reduces to s^(2 ) & = & _ q^*_q_q^-1_ff(q),[s2kcn]where the inverse of the two - fermion vertex function @xmath157 is given by @xcite ^-1_ff(q ) & & + _ kg_0,(k)g_0,(k+q ) + & = & + _ .[gamma]in the discussion to follow , we only consider the low - density limit of the unpaired @xmath9 fermions and thus we omit the @xmath158 term in eq . ( [ gamma ] ) . recall that the fermi distribution factor @xmath159 for the @xmath58 fermions vanishes at low temperatures since @xmath160 is large and negative . as before , we use the regularized two - body potential given by eq . ( [ ls ] ) to replace @xmath95 in terms of @xmath2 . we thus obtain & & + & & = \{_- 1}. + [ gamma2b]the @xmath39-sum in eq . ( [ gamma2b ] ) can be carried out analytically ( assuming @xmath161 ) and we find & & + & & . [ gamma3 ] after some rearranging , the vertex function in eq . ( [ gamma3 ] ) can be written as _ ff(,iq_m)= , [ gamma2]where @xmath53 with the boson chemical potential @xmath54 @xcite . for zero polarization , ( [ gamma2 ] ) reduces to the standard particle - particle vertex function @xmath162 given in refs . @xcite . in the bec region , the two - fermion vertex function @xmath157 in eq . ( [ gamma2 ] ) has a pole arising from bosonic dimer molecules . substituting eq . ( [ gamma2 ] ) into eq . ( [ s2kcn ] ) , the gaussian action becomes s^(2 ) & = & _ q^*_q_q^-2_q(-iq_m + _ b , ) , [ s2kdn ] where the strength of the bound state pole in eq . ( [ gamma2 ] ) is ^2_q & & ^2(,iq_m ) + & & = .[lambda ] defining the bose fields c_q ^-1_q_q , c^*_q ^-1_q^*_q,[c]the gaussian action in eq . ( [ s2kdn ] ) assumes the more familiar form s^(2 ) & = & _ , q_m ( -iq_m + _ b,)c^*_q c_q + & & -_q c^*_q d_0 ^ -1(q)c_q,[s2ken ] where we have defined the free - space molecular bose propagator d_0(q ) . [ dn ] note that at energies well - below the molecular binding energy ( i.e. , @xmath163 ) , the bose fluctuation fields defined in eq . ( [ c ] ) reduce to @xmath164 , which is the same field we used in our discussion of the superfluid state , given by eq . ( [ cq ] ) . our definition of @xmath165 here represents a very natural separation of the two energy scales in the problem . the propagator @xmath166 in eq . ( [ dn ] ) describes the low - energy dynamics of the dimer molecules through their effective single - particle energy , @xmath167 . the momentum- and energy - dependence that enters through @xmath168 , in contrast , describes the high - energy internal degrees of freedom of the molecules characterized by the binding energy @xmath169 . we now consider the quartic fluctuations to the effective bose actions for a spin - polarized superfluid as well as a bose - fermi mixture . we use this comparison to identify an expression for the effective dimer - unpaired fermion interaction @xmath136 in spin - polarized superfluids that is valid beyond the born approximation . to make the physics as clear as possible , we work in the normal state . we start by discussing the quartic fluctuations of the effective bose - fermi theory . the quartic fluctuation terms in eq . ( [ seffb ] ) are s^(4 ) & = & s^(4)_b + [ ( a^-1_0^(2))^2 ] + & & - [ ( a^-1_0^(1))^2a^-1_0^(2)]+ [ ( a^-1_0^(1))^4 ] . [ s4]in the normal state @xmath170 vanishes and eq . ( [ s4 ] ) reduces to s^(4 ) = _ q , q,kv^(4)_bf(q , q,k)c^*_+qc^*_-qc_-qc_+q . + [ s4bf]the effective boson - boson interaction @xmath171 in this expression is v^(4)_bf(q , q,k)=u_bb(q , q,k ) + ^(4)_bf(q , q,k ) , [ v4bf ] where , using the shorthand @xmath172 , & & + & & _ k u_bf(k+q^+ + , k - q^++,k + + q^- ) + & & u_bf(k - q^++ , k+q^++,k+-q^- ) + & & a^-1_0(k + q^++ ) a^-1_0(k - q^++).[v1bf ] in this expression , the @xmath173 green s functions are given by their @xmath174 value in eq . ( [ a_02 ] ) : a^-1_0(k ) & = & .[a_03 ] taken together , eqs . ( [ v4bf ] ) and ( [ v1bf ] ) describe the leading - order terms in the effective boson - boson interaction in a bose - fermi mixture . the first term in eq . ( [ v4bf ] ) is the bare boson - boson interaction @xmath137 while the second term , @xmath175 , given by eq . ( [ v1bf ] ) , describes an effective boson - boson interaction mediated by the fermions . we will show that the quartic fluctuation action in eq . ( [ s4bf ] ) has the same form as the quartic fluctuation action for a spin - polarized superfluid . the quartic fluctuation terms in the action for a spin - polarized gas given by eq . ( [ seffq ] ) are & & + & & _ q , q,kv^(4)(q , q,k)c^*_+qc^*_-qc_-qc_+q . [ s4c]here , we have used the bose field @xmath165 defined in eq . ( [ c ] ) . as before , @xmath176 is a four - vector denoting the momentum @xmath177 and bose matsubara frequency @xmath178 , corresponding to the centre - of - mass degree of freedom of two bosons . the interaction @xmath179 in eq . ( [ s4c ] ) is found to be & & + & & _ k/2+q_k/2-q_k/2-q_k/2+q _ k + & & g_0,(k + q^- - ) g_0,(k -q^- - ) + & & g_0,(k + q^+ + ) g_0,(k -q^+ + ) . + [ v1f ] in the case of an unpolarized fermi gas , this expression reduces to the interaction discussed by ohashi @xcite when we approximate @xmath168 by @xmath180 . since a spin - polarized fermi superfluid in the bec region is equivalent to a bose - fermi mixture , when @xmath23 ( i.e. , @xmath22 ) , the effective interaction given in eq . ( [ v1f ] ) includes an effective dimer - dimer interaction mediated by the gas of unpaired @xmath9 fermions . in addition , eq . ( [ v1f ] ) includes a direct dimer - dimer interaction which , unlike the mediated interaction , does not vanish when @xmath181 . however , unlike the expression given in eq . ( [ v4bf ] ) for the effective boson - boson interaction in a bose - fermi mixture , in eq . ( [ v1f ] ) these two different contributions are not explicitly separated . we will show that the mediated interaction in eq . ( [ v1f ] ) is given by the poles of the @xmath182 for the unpaired @xmath9 fermions . the direct contribution is given by the poles of the green s functions @xmath183 for the paired fermions . we shall use the notation @xmath184 to denote the contribution to the fermi matsubara frequency sum from the residue of the poles of the @xmath182 and @xmath183 green s functions in eq . ( [ v1f ] ) . using this , the direct and mediated interactions can be separated as follows : v^(4)(q , q,k)=v^(4)_(q , q,k ) + v^(4)_(q , q,k ) , + [ vseparatef]where v^(4)_(q , q,k ) & & _ _ g_0,\{},[vdir]v^(4)_(q , q,k ) & & _ _ g_0,\{},[vmed]and @xmath185 denotes the integrand in eq . ( [ v1f ] ) . we now show that @xmath186 can be identified as a direct " dimer - dimer interaction , whereas @xmath187 is an interaction between dimer molecules that is mediated by unpaired fermions . recalling that the chemical potential @xmath60 for the majority species of fermions is just the chemical potential for a gas of the excess unpaired fermions , we see that @xmath188 , where @xmath189 is given by eq . ( [ a_03 ] ) and @xmath190 is given by eq . ( [ gup ] ) . using this identification to compare eqs . ( [ vmed ] ) and ( [ v1bf ] ) leads to the result u_bf(k , k,p ) & = & _ p - kg_0,(k+k-p)_p - k,[ubf2]where @xmath168 is defined in eq . ( [ lambda ] ) and @xmath191 and @xmath192 all involve fermi frequencies . this is a key result of the present paper . note that in making the identification in eq . ( [ ubf2 ] ) , one must assume that the poles of the frequency - dependent boson - fermion interaction @xmath136 in eq . ( [ v1bf ] ) can not contribute to the frequency sum in that expression . this would of course be the case for a boson - fermion interaction between elementary " bosons and fermions , in which case @xmath136 would be frequency - independent . in the present context , however , where @xmath136 must be frequency - dependent owing to the fact that the bosons are dimer molecules , we must impose this restriction on the frequency sum in the term describing interactions between bosons and fermions in the bose - fermi model introduced in eq . ( [ seff ] ) . following the identification of eq . ( [ vmed ] ) with eq . ( [ v1bf ] ) , one sees that the direct " dimer - dimer interaction given by eq . ( [ vdir ] ) is equivalent to @xmath137 in eq . ( [ v4bf ] ) , u_bb(q , q,k ) = v^(4)_(q , q,k).[ubb3 ] we now consider what happens to the direct and mediated interactions identified in eqs . ( [ vdir ] ) and ( [ vmed ] ) as the polarization vanishes , that is , in the limit where @xmath193 . at @xmath155 in the bec region , all fermions are paired up to form dimers so there are no unpaired fermions and the mediated interaction @xmath187 in eq . ( [ vmed ] ) vanishes as expected . on the other hand , at finite temperatures @xmath187 does not vanish . even in the absence of polarization , there is a dimer - dimer interaction mediated by unpaired fermions since some of the dimer molecules will be thermally dissociated . as long as @xmath194 , however , the number of dissociated molecules is small in an unpolarized gas , even close to the transition temperature @xmath52 @xcite . thus , in the bec region of strongly - bound dimer molecules , the mediated interaction in eq . ( [ vmed ] ) is only significant in a spin - polarized gas . in contrast , the direct interaction given in eq . ( [ vdir ] ) remains finite even for zero polarization . at @xmath155 , one can show that ( using @xmath195)v^(4)_(0,0,0 ) & = & ( ) ^2 _ + & = & .[vdirborn]this reproduces the born - approximation value for the direct dimer - dimer interaction given in eq . ( [ ubb0 ] ) . this confirms that eq . ( [ vdir ] ) is the direct interaction between dimer molecules , i.e. , the part of the interaction that is not mediated by the unpaired excess @xmath9 fermions . to summarize , we have compared the quartic fluctuation terms in the action for a spin - polarized fermi gas with those for the action given in eq . ( [ seff ] ) which describes a bose - fermi mixture . we have thus proven that a spin - polarized fermi gas in the bec limit reduces to a bose - fermi mixture described by eq . ( [ seff ] ) . an important caveat is that the poles of @xmath136 [ given in eq . ( [ ubf2 ] ) ] are not allowed to contribute to the frequency sum in the dimer - fermion interaction term in eq . ( [ seff ] ) . in closing this section , we discuss the expression for @xmath64 in eq . ( [ term ] ) . with the expression for @xmath136 given in eq . ( [ ubf2 ] ) , we can rewrite @xmath64 as ( in the limit where @xmath196 ) g ( ) & & ( ) g_0,(;ik_n=^2/2 m ) + & = & n_cu_bf(k , k , k;_0=0;ik_n=^2/2m).[gk]note that @xmath64 arises in the effective theory of a spin - polarized fermi superfluid after performing an expansion in powers of @xmath27 , treating @xmath197 [ see eq . ( [ ubbtilde ] ) ] . thus , it is not surprising that @xmath64 is given by the free - space value of the dimer - fermion interaction @xmath136 , with @xmath198 , ( i.e. , @xmath199 and @xmath200 ) and with the corresponding energy @xmath201 for a free - space fermion . as a simple application of our expression for the dimer - fermion scattering interaction in a spin - polarized fermi gas given by eq . ( [ ubf2 ] ) , we discuss the scattering properties of a dimer - fermion pair in free - space . this allows us to make contact with the extensive literature discussing this problem @xcite and the well - known exact result @xmath1 due to skorniakov and ter - martirosian @xcite for the dimer - fermion scattering length . for free - space scattering of a single dimer - fermion pair , we can set @xmath200 , and @xmath202 in eq . ( [ ubf2 ] ) to give & & + & & ^1/2 + & & ^1/2 , + [ ubf3 ] where , @xmath203 and @xmath204 . in order to calculate the @xmath0-wave dimer - fermion scattering amplitude associated with the interaction given in eq . ( [ ubf3 ] ) , we need to calculate a renormalized low - energy interaction @xmath162 which is obtained from the bare " interaction @xmath136 given by eq . ( [ ubf3 ] ) by integrating out short - wavelength ( i.e. , smaller than @xmath2 ) degrees of freedom . following the usual prescription , @xmath162 is given by the bethe - salpeter equation ( within the ladder approximation @xcite ) for dimer - fermion scattering ( see fig . [ bsdiagram ] ) : & & + & -&_,p_nu_bf(p , k,p ) d_0(p - p)g_0,(p ) + & & ( p - p , p;p - k , k).[bs ] @xmath205 is the free - space molecular propagator for the dimers , d_0(q ) = , and @xmath206 is the propagator for unpaired @xmath9 fermions in free - space . the sum in eq . ( [ bs ] ) is over the fermi matsubara frequency @xmath207 . consistent with our discussion in sec . [ quartic ] , the poles of the green s function @xmath183 involved with @xmath136 do not contribute to the frequency sum over @xmath207 [ see eq . ( [ bs2 ] ) ] . the @xmath0-wave boson - fermion scattering length @xmath208 is given in the usual way by ( see also refs . @xcite ) a_bf ( 0,0;0,0),[abfdef]where @xmath209 is defined below eq . ( [ ubf0 ] ) . to calculate @xmath208 , we need only consider @xmath210 , i.e. , we set the outgoing fermion energy and momentum @xmath211 equal to zero as well as the centre - of - mass energy and momentum @xmath192 ( see fig . [ bsdiagram ] ) . in addition , for scattering in free space , we can use the on - shell energy value @xmath212 . using these , eq . ( [ bs ] ) reduces to & & + & + & _ , p_n + & & , [ bs2 ] where we have defined @xmath213 and _ & & ( , ik_n=_f , ) + & = & ^1/2 . + since we are dealing with the limit of vanishing density of unpaired fermions , we work at @xmath155 . analytically continuing the imaginary fermi frequency @xmath214 to the real frequency @xmath215 , and carrying out the integration over @xmath215 , eq . ( [ bs2 ] ) reduces to & & + & - & _ + & & ( -,).[bstilde5 ] equation ( [ bstilde5 ] ) is an integral equation for the momentum - dependent dimer - fermion scattering vertex @xmath162 . following the definition used in ref . @xcite , we define the scattering amplitude as a_bf( ) ( -,).[abfkn]the @xmath0-wave dimer - fermion scattering length @xmath208 defined in eq . ( [ abfdef ] ) corresponds to @xmath216 . one can show that eq . ( [ bstilde5 ] ) leads to the following integral equation for @xmath217 : & & + & -&4_. [ abfvacuo ] this is precisely the skorniakov - ter - martirosian integral equation for the scattering amplitude @xcite . solving this integral equation for @xmath217 , one finds @xmath218 . in the bec limit of a spin - polarized fermi superfluid , fermions in different hyperfine states pair up to form dimer molecules . when the binding energy of these molecules is larger than any other energy scale in the problem ( i.e. , @xmath219 ) the spin - polarized fermi superfluid reduces to a bose - fermi mixture , where the fermions are the unpaired excess fermions with chemical potential @xmath60 . using a functional integral approach and comparing the fluctuation terms in the effective bose actions of a bose - fermi mixture and spin - polarized fermi superfluid , we have shown precisely how this equivalence emerges in the bec region . at the gaussian level , the fluctuations of the spin - polarized superfluid are described by the usual bec bogoliubov propagator in eq . ( [ s2sp ] ) . the new dimer - dimer interaction in eq . ( [ ubbtilde ] ) includes a contribution that is mediated by the unpaired fermions and is described by a lindhard response function . going beyond the gaussian level by including quartic fluctuations about the mean - field , we derived an explicit expression in eq . ( [ ubf2 ] ) for the interaction between the dimers and unpaired fermions in the spin - polarized fermi gas . taking the free - space value of this interaction , we showed how our results reproduce the well - known value for the @xmath0-wave scattering length @xmath1 , first derived in 1957 @xcite . more recently this same result has been derived for three interacting fermions using a diagrammatic approach by brodsky and co - workers @xcite as well as by levinsen and gurarie @xcite . thanks dr . erhai zhao for helpful discussions . e.t . and a.g . were supported by nserc of canada . was financially supported by grant - in - aid for scientific research from the ministry of education , culture , sports , science and technology of japan and crest(jst ) . 99 c. a. regal , m. greiner , and d. s. jin , phys . lett . * 92 * , 040403 ( 2004 ) . m. bartenstein , a. altmeyer , s. riedl , s. jochim , c. chin , j. h. denschlag , and r. grimm , phys . lett * 92 * , 120401 ( 2004 ) . m. w. zwierlein , c. a. stan , c. h. schunck , s. m. f. raupach , a. j. kerman , and w. ketterle , phys . * 92 * , 120403 ( 2004 ) . m. w. zwierlein , a. schirotzek , c. h. schunck , and w. ketterle , science * 311 * , 492 ( 2006 ) . g. b. partridge , w. li , r. i. kamar , y. liao , and r. g. hulet , science * 311 * , 503 ( 2006 ) . for a review and further references , see d. e. sheehy and l. radzihovsky , e - print cond - mat/0607803 ; ann . ( to be published ) . p. pieri and g. c. strinati , phys . lett . * 96 * , 150404 ( 2006 ) . p. pieri and g. c. strinati , in _ proceedings of the international school of physics enrico fermi " - course clxiv ultra - cold fermi gases " , varenna , june 2006 _ ( to be published ) ; see also , e - print cond - mat/0610675 . m. iskin and c. a. r. s de melo , phys lett . * 97 * , 100404 ( 2006 ) . v. n. popov , _ functional integrals and collective excitations _ ( cambridge university press , cambridge , 1987 ) . c. a. r. s de melo , m. randeria , and j. r. engelbrecht , phys . lett . * 71 * , 3202 ( 1993 ) . j. r. engelbrecht , m. randeria , and c. a. r. s de melo , phys . b * 55 * , 15153 ( 1997 ) . y. ohashi , j. phys . japan * 74 * , 2659 ( 2005 ) . g. v. skorniakov and k. a. ter - martirosian , sov . jetp * 4 * , 648 ( 1957 ) . e. taylor , a. griffin , n. fukushima , and y. ohashi , phys . a * 74 * , 063626 ( 2006 ) . xia - ji liu and hui hu , europhys . lett . * 75 * , 364 ( 2006 ) . in the bec limit and for small polarizations , @xmath220 @xcite , where @xmath221 is the total number of fermions . thus , @xmath222 $ ] . n. fukushima , y. ohashi , e. taylor , and a. griffin , phys . rev . a * 75 * , 033609 ( 2007 ) . gerald d. mahan , _ many - particle physics _ ( plenum press , new york , 1990 ) , chap . c. j. pethick and h. smith , _ bose - einstein condensation in dilute gases _ ( cambridge university press , cambridge , 2002 ) , chap . 14.3.2 . h. p. bchler and g. blatter , phys . a * 69 * , 063603 ( 2004 ) . l. viverit , c. j. pethick , and h. smith , phys . a * 61 * , 053605 ( 2000 ) . r. haussmann , z. phys . b : condens . matter * 91 * , 291 ( 1993 ) . pieri and g. c. strinati , phys . b * 61 * , 15370 ( 2000 ) . i. v. brodsky , a. v. klaptsov , m. yu . kagan , r. combescot , and x. leyronas , jetp lett . * 82 * , 273 ( 2005 ) . i. v. brodsky , m. yu . kagan , a. v. klaptsov , r. combescot , and x. leyronas , phys . a * 73 * , 032724 ( 2006 ) . j. levinsen and v. gurarie , phys . a * 73 * , 053607 ( 2006 ) . p. f. bedaque and u. van kolck , phys . b * 428 * , 221 ( 1998 ) . a. l. fetter and j. d. walecka , _ quantum theory of many - particle systems _ ( mcgraw hill , new york , 1971 ) .

we give several equivalent combinatorial descriptions of the space of states for the box - ball systems , and connect certain partition functions for these models with the-weight multiplicities of the tensor product of the fundamental representations of the lie algebra . as an application , we give an elementary proof of the special case of the haglund haiman loehr formula . also , we propose a new class of combinatorial statistics that naturally generalize the so - called energy statistics . + mathematics subject classification ( 2000 ) 05e10 , 20c35 . + key words and phrases : crystals , paths , energy and tau functions , box ball systems , kostka macdonald polynomials . + + + + + + + + +

0811.1085

the purpose of the present paper is two - fold . first of all , we would like to give an introduction to the beautiful combinatorics related with box - ball systems , and secondly , to relate the latter with the `` classical combinatorics '' revolving around transportation matrices , tabloids , the lascoux schtzenberger statistics charge , macdonald polynomials , @xcite,@xcite , haglund haiman loehr s formula @xcite , and so on . as a result of our investigations , we will prove that two statistics naturally appearing in the context of box - ball systems , namely _ energy function _ and _ tau - function _ , have nice combinatorial properties . more precisely , the statistics energy @xmath4 is an example of a generalized machonian statistics @xcite , section 2 , whereas the statistics @xmath5 related with kostka macdonald polynomials , see section 5.2 of the present paper . box - ball systems ( bbs for short ) were invented by takahashi satsuma @xcite as a wide class of discrete integrable soliton systems . in the simplest case , bbs are described by simple combinatorial procedures using box and balls . despite its simple outlook , it is known that the bbs have various remarkably deep properties ; * time evolution of the bbs coincides with isomorphism of the crystal bases @xcite . thus the bbs possesses quantum integrability . * bbs are ultradiscrete ( or tropical ) limit of the usual soliton systems @xcite . thus the bbs possesses classical integrability at the same time . * inverse scattering formalism of the bbs coincides with the rigged configuration bijection originating in completeness problem of the bethe states @xcite . let us say a few words about the main results of our paper . * in the case of statistics _ , our main result can be formulated as a computation of the corresponding partition function for the bbs in terms of the values of the kostka macdonald polynomials at @xmath6 * in the case of the statistics _ energy _ , our result can be formulated as an interpretation of the corresponding partition function for the bbs as the @xmath0-weight multiplicity in the tensor product of the fundamental representations of the lie algebra @xmath7 we _ expect _ that the same statement is valid for the bbs corresponding to the tensor product of rectangular representations . + we are reminded that a _ @xmath0-analogue of the multiplicity _ of a highest weight @xmath8 in the tensor product @xmath9 of the highest weight @xmath10 @xmath11 irreducible representations @xmath12 of the lie algebra @xmath13 is defined as @xmath14 = \sum_{\eta}~k_{\eta , r}~k_{\eta , \lambda}(q),\ ] ] where @xmath15 stands for the parabolic kostka number corresponding to the sequence of rectangles @xmath16 see e.g. @xcite , @xcite . * we give several equivalent descriptions of _ paths _ which appear in the description of partition functions for bbs : in terms of transportation matrices , tabloids , plane partitions . we expect that such interpretations may be helpful for better understanding connections of the bbs and other integrable models such as melting crystals @xcite , @xmath0-difference toda lattices @xcite , ... . our result about connections of the energy partition functions for bbs and @xmath0-weight multiplicities suggests a deep hidden connections between partition functions for the bbs and characters of the demazure modules , solutions to the @xmath0-difference toda equations , cf.@xcite . as an interesting open problem we want to give raise a question about an interpretation of the sums @xmath17 where @xmath18 denotes the kostka macdonald polynomials @xcite , as _ refined partition functions _ for the bbs corresponding to the tensor product of rectangular representations @xmath19 . see conjecture [ conj : tau ] . in other words , one can ask : what is a meaning of the second statistics ( see @xcite ) in the kashiwara theory @xcite of crystal bases ( of type a ) ? organization of the present paper is as follows . in section 2 , we review necessary facts from the kirillov reshetikhin crystals . especially we explain an explicit algorithm to compute the combinatorial @xmath20 and the energy function . in section 3 , we introduce several combinatorial descriptions of paths . then we define several statistics on paths such as haglund s statistics , energy statistics @xmath21 and tau statistics @xmath22 . in section 4 , we collect necessary facts from the bbs which will be used in the next section . in section 5 , we present our main result ( theorem [ th : main ] ) as well as several relating conjectures . we conjecture that @xmath22 gives independent statistics depending on one parameter @xmath23 although they all give rise to the unique generating function up to constant shift of power . in section 6 , we show that the energy statistics @xmath21 belong to the class of statistics @xmath22 ( theorem [ th : e = tau ] ) . therefore @xmath22 gives a natural extension of the energy statistics @xmath21 . let @xmath25 be a @xmath26 kirillov reshetikhin module , where we shall consider the case @xmath27 . the module @xmath25 is indexed by a dynkin node @xmath28 and @xmath29 . as a @xmath30-module , @xmath25 is isomorphic to the irreducible module corresponding to the partition @xmath31 . for arbitrary @xmath23 and @xmath32 , the module @xmath25 is known to have crystal bases @xcite , which we denote by @xmath33 . as the set , @xmath33 is consisting of all column strict semi - standard young tableaux of depth @xmath23 and width @xmath32 over the alphabet @xmath34 . for the algebra @xmath35 , let @xmath36 be the weight lattice , @xmath37 be the fundamental roots , @xmath38 be the simple roots , and @xmath39 be the simple coroots . as a type @xmath35 crystal , @xmath40 is equipped with the kashiwara operators @xmath41 and @xmath42 ( @xmath43 ) satisfying @xmath44 here @xmath45 is the natural pairing and we set @xmath46 and @xmath47 . actions of the kashiwara operators @xmath48 , @xmath49 for @xmath43 coincide with the one described in @xcite . since we do not use explicit forms of these operators , we omit the details . see @xcite for complements of this section . note that in our case @xmath35 , we have @xmath50 and @xmath51 where @xmath52 is the @xmath53-th canonical unit vector of @xmath54 . we also remark that @xmath55 is the weight of @xmath56 , i.e. , @xmath57 counts the number of letters @xmath53 contained in tableau @xmath56 . for two crystals @xmath58 and @xmath59 , one can define the tensor product @xmath60 . the actions of the kashiwara operators on tensor product have simple form . namely , the operators @xmath61 act on @xmath62 by @xmath63 and @xmath64 . we assume that @xmath65 and @xmath66 as @xmath67 . then it is known that there is a unique crystal isomorphism @xmath68 . we call this map ( classical ) combinatorial @xmath20 and usually write the map @xmath20 simply by @xmath69 . let us consider the affinization of the crystal @xmath58 . as the set , it is @xmath70\ , |\ , b\in b,\ , d\in\mathbb{z}\}.\ ] ] there is also explicit algorithm for actions of the affine kashiwara operators @xmath71 , @xmath72 in terms of the promotion operator @xcite . for the tensor product @xmath73\otimes b'[d']\in \mathrm{aff}(b)\otimes\mathrm{aff}(b')$ ] , we can lift the ( classical ) combinatorial @xmath20 to affine case as follows : @xmath74\otimes b'[d']\stackrel{r}{\simeq } \tilde{b}'[d'-h(b\otimes b')]\otimes \tilde{b}[d+h(b\otimes b')],\ ] ] where @xmath75 is the isomorphism of ( classical ) combinatorial @xmath20 . the function @xmath76 is called the energy function . we will give explicit forms of the combinatorial @xmath20 and energy function in the next section . we give explicit description of the combinatorial @xmath20-matrix ( combinatorial @xmath20 for short ) and energy function on @xmath77 . to begin with we define few terminologies about young tableaux . denote rows of a young tableaux @xmath78 by @xmath79 from the top to bottom . then row word @xmath80 is defined by concatenating rows as @xmath81 . let @xmath82 and @xmath83 be two partitions . we define concatenation of @xmath84 and @xmath85 by the partition @xmath86 . [ prop : shimozono ] @xmath87 is mapped to @xmath88 under the combinatorial @xmath20 , i.e. , @xmath89 if and only if @xmath90 moreover , the energy function @xmath76 is given by the number of nodes of @xmath91 outside the concatenation of partitions @xmath31 and @xmath92 . @xmath93 we define another normalization of the energy function @xmath94 such that for @xmath95 , @xmath96 for special cases of @xmath97 , the function @xmath98 is called unwinding number and @xmath94 is called winding number in @xcite . explicit values for the case @xmath99 are given by @xmath100 where @xmath101 and @xmath102 . in order to describe the algorithm for finding @xmath103 and @xmath104 from the data @xmath91 , we introduce a terminology . let @xmath78 be a tableau , and @xmath105 be a subset of @xmath78 such that @xmath105 is also a tableau . consider the set theoretic subtraction @xmath106 . if the number of nodes contained in @xmath107 is @xmath23 and if the number of nodes of @xmath107 contained in each row is always 0 or 1 , then @xmath107 is called vertical @xmath23-strip . given a tableau @xmath108 , let @xmath105 be the upper left part of @xmath78 whose shape is @xmath31 . we assign numbers from 1 to @xmath109 for each node contained in @xmath106 by the following procedure . let @xmath110 be the vertical @xmath111-strip of @xmath107 as upper as possible . for each node in @xmath110 , we assign numbers 1 through @xmath111 from the bottom to top . next we consider @xmath112 , and find the vertical @xmath111 strip @xmath113 by the same way . continue this procedure until all nodes of @xmath107 are assigned numbers up to @xmath109 . then we apply inverse bumping procedure according to the labeling of nodes in @xmath107 . denote by @xmath114 the integer which is ejected when we apply inverse bumping procedure starting from the node with label 1 . denote by @xmath115 the tableau such that @xmath116 . next we apply inverse bumping procedure starting from the node of @xmath115 labeled by 2 , and obtain the integer @xmath117 and tableau @xmath118 . we do this procedure until we obtain @xmath119 and @xmath120 . finally , we have @xmath121 consider the following tensor product : @xmath122 from @xmath56 , we have @xmath123 , hence we have @xmath124 here subscripts of each node indicate the order of inverse bumping procedure . for example , we start from the node @xmath125 and obtain @xmath126 next we start from the node @xmath127 of @xmath115 . continuing in this way , we obtain @xmath128 and @xmath129 . since @xmath130 , we obtain @xmath131 note that the energy function is derived from the concatenation of shapes of @xmath56 and @xmath132 , i.e. , @xmath133 . let @xmath134 be a positive integer , @xmath135 and @xmath136 be two compositions of the same size . denote by @xmath137 the set of matrices @xmath138 such that @xmath139 remind that a tabloid of shape @xmath140 and weight @xmath141 is a filling of the shape @xmath140 by the numbers @xmath142 in such a way that the number of @xmath53 s appearing in the filling is equal to @xmath143 . it is clear that the number of tabloids of shape @xmath140 and weight @xmath141 is equal to the multinomial coefficient @xmath144 a row ( column ) weakly strict tabloid of shape @xmath140 and weight @xmath141 is a filling of the shape @xmath140 by numbers @xmath142 such that the numbers along each row ( column ) are weakly increasing and @xmath143 is equal to the number of @xmath53 s appearing in the filling . take @xmath145 , @xmath146 , then @xmath147 is a row weakly strict tabloid of shape @xmath140 and weight @xmath141 . @xmath93 we denote by @xmath148 the set of all row weakly strict tabloids of shape @xmath140 and weight @xmath141 . it is well - known that there exists a bijection between the sets @xmath137 and @xmath148 . namely , given a matrix @xmath149 , we fill the row @xmath150 of the shape @xmath140 by the numbers @xmath151 . for example , let @xmath152 then the corresponding row weakly strict tabloid is @xmath153 to each tabloid @xmath154 , one can associate the reading word , namely the word obtained by reading the filling of tabloid @xmath154 from the right to the left starting from the top row . for example , for the tabloid @xmath154 displayed above , @xmath155 . if weight @xmath156 of a tabloid @xmath154 is a partition , we define the charge @xmath157 of tabloid @xmath154 to be the charge @xmath158 of the reading word . see page 242 of @xcite for the definition of the lascoux schtzenberger charge @xcite . take standard tabloid @xmath159 then @xmath160 and therefore @xmath161 . @xmath93 let @xmath8 be a partition . _ a plane partition of shape @xmath8 _ is a tabloid @xmath162 of shape @xmath8 such that the numbers in each row and each column are weakly decreasing . for example , @xmath163 is a plane partitions of shape @xmath164 a plane partition @xmath162 has a three - dimensional diagram , consisting of the points @xmath165 with integer coordinates such that @xmath166 and @xmath167 where @xmath168 is the number that is located in the box @xmath169 by definition , the size @xmath170 of a plane partition @xmath162 of shape @xmath8 is @xmath171 let @xmath140 and @xmath141 be two compositions of the same size . denote by @xmath172 the set of plane partitions @xmath162 such that @xmath173 finally , let us remind two classical results 1 . ( p. macmahon , see e.g. @xcite , page 81 ) let @xmath174 be three positive integers , and @xmath58 be the box with side - lengths @xmath175 then @xmath176 2 . ( robinson schensted knuth , see e.g. @xcite , chapter 3 ) there are bijections @xmath177 let @xmath140 be a partition of size @xmath134 . a path @xmath179 of type @xmath178 and weight @xmath180 is a sequence of positive integers @xmath181 such that @xmath182 . we denote by @xmath183 the set of all paths of type @xmath178 and weight @xmath140 . a path @xmath179 is called a highest weight path if the sequence @xmath181 satisfies the yamanouchi condition . we denote by @xmath184 the set of all @xmath178 type highest paths with weight @xmath140 . it is well known that the total number of @xmath178 type paths of weight @xmath185 is equal to the multinomial coefficient @xmath186 and there are bijections @xmath187 let us describe the general prescription to get the corresponding tabloid from a given path . let the path @xmath188 , we recursively add letters to the tabloid according to @xmath189 , @xmath190 , @xmath191 , @xmath192 as follows . starting from the empty tabloid , assume that we have done up to @xmath193 and have gotten a tabloid @xmath194 . then we add the letter @xmath53 to the right of the @xmath195-th row of @xmath194 and get @xmath196 . for example , the path @xmath197 can be related to the following transportation matrix and row strict tabloid @xmath198 more generally , we define path to be an arbitrary element of tensor product of crystals @xmath199 . recall that for type @xmath24 case , @xmath33 is , as the set , consisting of semi - standard tableaux over alphabet @xmath200 , and tensor product of crystals @xmath62 is , as a set , cartesian product of two sets @xmath58 and @xmath59 . crystal graph structure on the set @xmath62 is is given according to @xcite . weight @xmath201 of a path @xmath202 is given by @xmath203 for example , @xmath204 is a path of rectangular shape @xmath205 , and its weight is @xmath206 . note that the number of standard ( i.e. , weight of @xmath207 ) rectangular shape @xmath208 paths is equal to the generalized multinomial coefficient @xmath209}{0pt}{}{n}{r_1,\cdots , r_l } = \frac{n!}{\prod_a h_{r_a}(1)},\ ] ] where @xmath210 , and for any diagram @xmath8 , @xmath211 denotes the hook polynomial ( see definition on page 45 of @xcite ) corresponding to diagram @xmath8 . [ [ comments . ] ] comments . + + + + + + + + + summarizing , one has the following ( equivalent ) combinatorial descriptions of the set of ( crystal ) paths of type @xmath212 and weight @xmath213 as the set of @xmath214 transportation matrices @xmath215 @xmath216 row weakly increasing tabloids @xmath217 @xmath218 plane partitions @xmath219 for the given path @xmath220 , the corresponding element in @xmath221 is determined as follows . staring from the empty tabloid , we recursively add letters to the tabloid according to @xmath222 , @xmath223 , @xmath191 , @xmath224 as follows . assume we have done up to @xmath225 and have gotten the tabloid @xmath194 . denote the number of @xmath226 contained in @xmath227 by @xmath228 . then , for all @xmath226 , we add letters @xmath53 for @xmath228 times to the right of the @xmath226-th row of @xmath194 and get @xmath196 . consider the path @xmath229 of type @xmath230 and weight @xmath231 . the corresponding tabloid and transportation matrix are @xmath232 to find the plane partition which corresponds to the tabloid @xmath154 ( or matrix @xmath233 ) , one can apply the robinson schensted knuth algorithm @xcite to the multi - permutation @xmath234 which corresponds to the tabloid @xmath154 . one has @xmath235 finally , the plane partition we are looking for , can be obtained from the pair of semi - standard young tableaux displayed above by gluing the gelfand tsetlin patterns that correspond to the young tableaux in question : ( 25,7.5)(0,-2.5 ) ( 0,2)@xmath236 ( 16,2 ) @xmath237 ( 17,0.4)(0.2,-0.2)10 ( 17,1.6)(0.2,-0.2)16 ( 17,2.8)(0.2,-0.2)22 ( 17,4.0)(0.2,-0.2)28 ( 18.2,4.0)(0.2,-0.2)22 ( 19.4,4.0)(0.2,-0.2)16 ( 20.6,4.0)(0.2,-0.2)10 ( 19,-2.2)2 ( 20.2,-2.2)6 ( 21.4,-2.2)8 ( 22.7,-2.2)11 ( 22.7,-1.0)11 ( 22.7,0.2)5 ( 22.7,1.4)3 the result is a plane partition from the set @xmath238 . @xmath93 for the reader s convenience , let us recall the way to get the corresponding gelfand tsetlin pattern from a given semi - standard tableaux . by looking contents of a semi - standard tableau @xmath154 , one can define a sequence of partitions @xmath239 such that each skew diagram @xmath240 @xmath241 is a horizontal strip , see e.g. , chapter i of @xcite . starting from the sequence of partitions ( [ def : gt ] ) , one can define the corresponding gelfand tsetlin pattern @xmath242 by the following rule @xmath243 it is known that thus obtained @xmath84 indeed satisfies the defining properties of the gelfand tsetlin patterns . one of the basic properties of the bbs is that the second young tableau ( of weight @xmath141 ) obtained by means of the robinson schensted knuth algorithm , is conserved under the dynamics of the bbs @xcite ( see also @xcite for the other connections between the simplest bbs and the rsk algorithm ) . nowadays , conserved quantities and linearization parameters ( or angle variables ) of the bbs are completely determined in the most general settings @xcite , and , surprisingly enough , they are elegantly described by the so - called ( unrestricted ) rigged configurations @xcite . the latter result is a consequence of a deep theorem stated in lemma 8.5 of @xcite . @xmath93 for a path @xmath244 , let us define elements @xmath245 for @xmath246 by the following isomorphisms of the combinatorial @xmath20 ; @xmath247 where we have written @xmath248 assuming that @xmath249 . for a given path @xmath250 , define statistics @xmath251 by @xmath252 define the statistics @xmath253 by @xmath254 for example , consider a path @xmath255 . in this case , we have @xmath256 , since the combinatorial @xmath20 act on @xmath257 as identity . therefore , we have @xmath258 define another statistics _ tau _ as follows . [ def : generaltau ] for the path @xmath259 , define @xmath22 by @xmath260 where @xmath261 is the highest element of @xmath33 . @xmath93 we use abbreviation @xmath5 for the statistics @xmath262 on @xmath178 type paths @xmath263 , i.e. , @xmath264 where @xmath189 denotes the first letter of the path @xmath265 . this @xmath5 is a special case of the tau functions for the box - ball systems @xcite which originate as ultradiscrete limit of the tau functions for the kp hierarchy @xcite . [ def : taumu ] for composition @xmath266 , write @xmath267}=\sum_{j=1}^i \mu_j$ ] with convention @xmath268}=0 $ ] . then we define a generalization of @xmath5 by @xmath269}),\ ] ] where @xmath270}= a_{\mu_{[i-1]}+1}\otimes a_{\mu_{[i-1]}+2}\otimes\cdots\otimes a_{\mu_{[i ] } } \in ( b^{1,1})^{\otimes\mu_i}.\ ] ] @xmath93 note that we have @xmath271}\otimes a_{[2]}\otimes\cdots\otimes a_{[n]}$ ] , i.e. , the path @xmath265 is partitioned according to @xmath156 . it is convenient to identify @xmath272 as statistics on a tabloid of shape @xmath156 whose reading word coincides with the partitioned path according to @xmath156 . for example , to path @xmath273 and the composition @xmath274 one associates the tabloid @xmath275 let @xmath179 be a path of type @xmath276 denote by @xmath277 the corresponding row strict tabloid and by @xmath278 its reading word . define the charge of a path @xmath179 to be the charge of tabloid @xmath279 i.e. the charge of the reading word @xmath280 for example , take @xmath281 then @xmath282 and therefore , @xmath283 if @xmath156 is a composition , define @xmath284}),$ ] where @xmath285}= p_{\mu_{[i-1]}+1}~p_{\mu_{[i-1]}+2}~\cdots~p_{\mu_{[i]}},\ ] ] cf . definition 3.3 . one has @xmath286 } } \biggr),\ ] ] where @xmath287}$ ] denotes the first letter of the path @xmath288}.$ ] @xmath93 proof follows from two simple observations that @xmath289 where @xmath290 denotes the length of path @xmath291 [ [ comments.-1 ] ] comments . + + + + + + + + + it follows from @xcite that on the set of semi - standard young tableaux , i.e. , on the set of highest weight paths , the statistics tau coincides with statistics _ cocharge_. therefore , one can consider the statistics tau as a natural extension of the statistics cocharge from the set of semi - standard tableaux to the set of tabloids , or on the set of transportation matrices . @xmath93 [ [ tableaux - language - description ] ] tableaux language description + + + + + + + + + + + + + + + + + + + + + + + + + + + + + for a given path @xmath255 , associate tabloid @xmath292 of shape @xmath156 whose reading word coincides with @xmath265 . this correspondence is the same as those used in definition [ def : taumu ] . denote the cell at the @xmath53-th row , @xmath293-th column ( we denote the coordinate by @xmath294 ) of the tabloid @xmath292 by @xmath295 . attacking region of the cell at @xmath294 is all cells @xmath296 with @xmath297 or @xmath298 with @xmath299 . in the following diagram , gray zonal regions are the attacking regions of the cell @xmath294 . ( 10,9)(-2,1.5 ) ( 0,2)(0,1)8 ( 0,2)(1,0)5 ( 0,10)(1,0)25 ( 5,2)(5,2)4(1,0)5 ( 10,2)(5,2)4(0,1)2 ( 7,6)(0,1)2(1,0)1 ( 7,6)(1,0)2(0,1)1 ( 0,6 ) ' '' '' ( 8,5 ) ' '' '' ( 9.2,7.8)(-4,-3)1.7 ( 9.5,8)@xmath294 follow @xcite , define @xmath300 by @xmath301 then we define @xmath302 , then the cell @xmath294 is called by _ descent_. then define @xmath303 note that @xmath304 is the arm length of the cell @xmath294 . [ [ path - language - description ] ] path language description + + + + + + + + + + + + + + + + + + + + + + + + + consider two paths @xmath305 . we denote by @xmath306 . then we define @xmath307 for more general cases @xmath308 and @xmath309 satisfying @xmath310 , we define @xmath311 then the above definition of @xmath312 is equivalent to @xmath313 for example , consider the following tabloid ( @xmath314 ) ; @xmath315 and @xmath316 . then @xmath317 consider two paths @xmath308 and @xmath309 satisfying @xmath318 . denote @xmath319 . then define @xmath320 for the tableau @xmath154 of shape @xmath156 corresponding to the path @xmath265 , we define @xmath321}\otimes a_{[i+1]}).\ ] ] for a path @xmath265 , statistics @xmath322 is defined by @xmath323 and @xmath324 is defined by @xmath325 @xmath93 if we associate to a given path @xmath326 with the shape @xmath156 tabloid @xmath154 , we sometimes write @xmath327 and @xmath328 . [ ex:222_42 ] for highest weight paths of weight @xmath329 and shape @xmath330 , the following is a list of the corresponding tabloids associated with data in the form @xmath331 @xmath332\hline \yvcentermath1\young(1321,32)\,\,\,\,(1,4 ) & \yvcentermath1\young(3211,32)\,\,\,\,(0,6 ) & \begin{picture}(0,1.7)\end{picture } \\[4mm]\hline \end{array}\ ] ] let us observe that @xmath333 which is different from @xmath334 another interesting choice is @xmath329 and @xmath335 . the following is a list of all such paths with corresponding statistics : @xmath336\hline \yvcentermath1\young(121,323)\,\,\,\,(2,3 ) & \yvcentermath1\young(321,321)\,\,\,\,(0,6 ) & \begin{picture}(0,1.7)\end{picture } \\[4mm]\hline \end{array}\ ] ] thus we have @xmath337 which is again different from @xmath338 @xmath93 in this section , we summarize basic facts about the box - ball system which will be used in the next section . for our purpose , it is convenient to express the isomorphism of the combinatorial @xmath20 @xmath339 by the following vertex diagram : ( 4,4 ) ( 0,2.0)(1,0)3.2 ( 1.6,1.0)(0,1)2 ( -0.6,1.8)@xmath265 ( 1.4,0)@xmath132 ( 1.4,3.2)@xmath56 ( 3.4,1.8)@xmath340 ( 4.1,1.7 ) . successive applications of the combinatorial @xmath20 is depicted by concatenating these vertices . following @xcite , we define time evolution of the box - ball system @xmath341 . let @xmath342 be the highest element and @xmath343 . here the highest element @xmath344 is the tableau whose @xmath53-th row is occupied by integers @xmath53 . for example , @xmath345 . define @xmath346 and @xmath347 by the following diagram . @xmath348 @xmath346 are usually called _ carrier _ and we set @xmath349 . then we define operator @xmath350 by @xmath351 recently @xcite , operators @xmath341 have used to derive crystal theoretical meaning of the rigged configuration bijection . it is known ( @xcite theorem 2.7 ) that there exists some @xmath352 such that @xmath353 if the corresponding path is @xmath354 , we have the following combinatorial description of the box - ball system @xcite . we regard @xmath355 as an empty box of capacity 1 , and @xmath356 as a ball of label ( or internal degree of freedom ) @xmath53 contained in the box . then we have : [ prop : bbs ] for a path @xmath354 of type @xmath24 , @xmath357 is given by the following procedure . 1 . move every ball only once . 2 . move the leftmost ball with label @xmath358 to the nearest right empty box . 3 . move the leftmost ball with label @xmath358 among the rest to its nearest right empty box . 4 . repeat this procedure until all of the balls with label @xmath358 are moved . 5 . do the same procedure 24 for the balls with label @xmath134 . repeat this procedure successively until all of the balls with label @xmath359 are moved . @xmath93 there are extensions @xcite of this box and ball algorithm corresponding to generalizations of the box - ball systems with respect to each affine lie algebra @xcite . using this box and ball interpretation , our statistics @xmath360 admits the following interpretation . [ th : tau = rho ] for a path @xmath354 of type @xmath24 , @xmath360 coincides with number of all balls @xmath361 contained in paths @xmath56 , @xmath357 , @xmath191 , @xmath362 . @xmath93 consider the path @xmath363 where @xmath364 , @xmath365 . we compute @xmath366 in two ways . \(i ) first we compute by eq.([eq : taumu ] ) . @xmath367 thus we obtain @xmath368 . \(ii ) next we use theorem [ th : tau = rho ] . according to proposition [ prop : bbs ] , the time evolutions of the paths @xmath265 and @xmath56 are as follows : here the left and right tables correspond to @xmath265 and @xmath56 , respectively . rows of left ( resp . right ) table represent @xmath265 , @xmath369 , @xmath191 , @xmath370 ( resp . , those for @xmath56 ) from top to bottom . note that we omit all frames of tableaux of @xmath178 and symbols for tensor product . counting letters 2 , 3 and 4 in each table , we have @xmath371 , @xmath372 and again we get @xmath373 . meanings of the above two dynamics corresponding to paths @xmath265 and @xmath56 are summarized as follows : 1 . dynamics of the path @xmath265 . in the first row , there are two solitons ( length two soliton @xmath374 and length one soliton 2 ) , and in the second row , there are also two solitons ( length one soliton 4 and length two soliton 32 ) . this is scattering of two solitons . after the scattering , soliton 4 propagates at velocity one and soliton 32 propagates at velocity two without scattering . dynamics of the path @xmath56 . this shows free propagation of one soliton of length three 432 at velocity three . let @xmath375 be the ( integral form ) modified macdonald polynomials where @xmath84 stands for infinitely many variables @xmath376 . here @xmath375 is obtained by simple plethystic substitution ( see , e.g. , section 2 of @xcite ) from the original definition of the macdonald polynomials @xcite . schur function expansion of @xmath375 is given by @xmath377 where @xmath378 stands for the following transformation of the kostka macdonald polynomials : @xmath379 here we have used notation @xmath380 . then the celebrated haglund haiman loehr ( hhl ) formula is as follows . let @xmath381 be the filling of the young diagram @xmath156 by positive integers @xmath382 , and define @xmath383 . then the macdonald polynomial @xmath375 have the following explicit formula : @xmath384 @xmath93 from the hhl formula , we can show the following formula . for any partition @xmath156 and composition @xmath140 of the same size , one has @xmath385 where @xmath386 runs over all partitions of size @xmath387 . * proof . * indeed , @xcite , if two fillings @xmath388 and @xmath389 belong to the same @xmath390 orbit , then @xmath391 , @xmath392 . @xmath393 the ( modified ) macdonald polynomial @xmath375 have the following expansion in terms of the monomial symmetric functions @xmath394 : @xmath395 where @xmath8 runs over all partitions of size @xmath387 . @xmath93 to find combinatorial interpretation of the kostka macdonald polynomials @xmath396 remains significant open problem . among many important partial results about this problem , we would like to mention the following theorem also due to haglund loehr : if @xmath397 , we have @xmath398 @xmath93 it is interesting to compare this formula with the formula obtained by s. fishel @xcite , see also @xcite , @xcite . concerning validity of the formula eq.([eq : hhl_kosmac ] ) , we state the following conjecture . [ conj : hhl_kostka ] explicit formula for the kostka macdonald polynomials @xmath399 is valid if and only if at least one of the following two conditions is satisfied . 1 . @xmath400 and @xmath401 . 2 . @xmath8 is a hook shape . @xmath93 as an example , the following is the list of the tabloids associated with the highest weight paths of weight @xmath402 and shape @xmath403 . here we also include the values of the statistics in the form @xmath331 @xmath404\hline \yvcentermath1\young(1211,23)\,\,\,\,(2,2 ) & \yvcentermath1\young(2211,31)\,\,\,\,(1,4 ) & \yvcentermath1\young(2211,13)\,\,\,\,(1,5 ) \begin{picture}(0,1.7)\end{picture } \\[4mm]\hline \yvcentermath1\young(3211,21)\,\,\,\,(0,6 ) & \yvcentermath1\young(3211,12)\,\,\,\,(0,5 ) & \yvcentermath1\young(1121,32)\,\,\,\,(2,2 ) \begin{picture}(0,1.7)\end{picture } \\[4mm]\hline \yvcentermath1\young(1121,23)\,\,\,\,(2,1 ) & \yvcentermath1\young(2121,31)\,\,\,\,(1,3 ) & \yvcentermath1\young(2121,13)\,\,\,\,(1,4 ) \begin{picture}(0,1.7)\end{picture } \\[4mm]\hline \yvcentermath1\young(3121,21)\,\,\,\,(0,5 ) & \yvcentermath1\young(3121,12)\,\,\,\,(1,4 ) & \yvcentermath1\young(1321,21)\,\,\,\,(1,3 ) \begin{picture}(0,1.7)\end{picture } \\[4mm]\hline \yvcentermath1\young(1321,12)\,\,\,\,(0,4 ) & \begin{picture}(0,1.7)\end{picture } & \begin{picture}(0,1.7)\end{picture } \\[4mm]\hline \end{array}\ ] ] then the generating function is @xmath405 which is different from @xmath406 even if we consider the special value @xmath407 these two polynomials are distinct . yet other examples which show that the formula ( [ eq : hhl_kosmac ] ) does not hold if the condition ( i ) and ( ii ) of conjecture [ conj : hhl_kostka ] break down , is given in example [ ex:222_42 ] . let us remark that the choice @xmath329 and @xmath335 will give an example of both specializations @xmath408 and @xmath2 give distinct polynomials . our main result is an elementary proof for special case @xmath2 of the formula eq.([eq : hhl ] ) in the following form . [ th : main ] let @xmath140 be a composition and @xmath156 be a partition of the same size . then , @xmath409 @xmath93 [ con : main ] let @xmath140 be a composition and @xmath156 be a partition of the same size . then , @xmath410 @xmath93 here @xmath411 is equal to the number of letters other than 1 contained in each path @xmath412 . let us remark that in view of general definition of @xmath22 , our @xmath272 is related with @xmath262 , whereas @xmath322 is related with @xmath413 where @xmath23 is bigger than the length of weight @xmath140 ( see section 6 ) . as for intermediate @xmath22 , see conjecture [ conj : tau ] for some further information . let us consider case @xmath414 and @xmath403 . the following is a list of paths @xmath179 and the corresponding value of tau function @xmath415 . for example , the top left corner means @xmath416 and @xmath417 . @xmath418 summing up , lhs of eq.([eq : main ] ) is @xmath419 in the rhs of eq.([eq : main ] ) , nontrivial contributions come from the following 4 terms : @xmath420 by setting @xmath2 , we get eq.([eq : main ] ) . @xmath93 * proof of theorem [ th : main ] * let @xmath156 be a composition and @xmath154 be a tabloid of size @xmath421 . denote by @xmath196 the part of @xmath154 which is filled by numbers from the interval @xmath422 + 1},\mu_{[i]}]$ ] . then @xmath423 @xmath93 @xmath424 where the sum in the right hand side runs over standard tabloids @xmath154 of shape @xmath140 . * proof . * recall the following three formulas from @xcite , chapter vi . _ formula 1 . _ @xmath425 and thus @xmath426 as a corollary of the formulas above , @xmath427 _ formula 2 . _ @xmath428 _ formula 3 . _ @xmath429 therefore , @xmath430 and after the plethystic change of variables @xmath431 , we obtain @xmath432 @xmath393 @xmath433 where the second sum runs over all standard tabloids @xmath154 of the size @xmath23 , and @xmath434 denotes either the charge of @xmath154 , or the value of tau function on the path corresponding to tabloid . in the case @xmath435 the charge of tabloid @xmath154 , this result is due to lascoux schtzenberger ; in the case of @xmath436 this statement is a corollary of theorem 7.4 and corollary 6.13 from @xcite , where identification of tau function and cocharge is given . @xmath437 where the second sum runs over the set of standard tabloids of shape @xmath8 , and @xmath438 . on the other hand , @xmath439 and therefore @xmath440 where the sum in the right hand side runs over the set of standard @xmath8-tabloids finally , @xmath441 where the the third sum runs over the set of all standard @xmath140-tabloids . it remains to observe that according to lemma 3.4 , @xmath442 } ) \biggr)=\tau_{\mu}-\sum_{i}~\mu_{i}~(1-\delta_{1,p_{1}^{[i ] } } ) = \mathrm{maj}_{\mu } ( p).\ ] ] @xmath393 in order to clarify nature of tau statistics , we consider possible generalizations of the results in section [ sec : sumoftau ] . the main objective of this section is to give an interpretation of the energy statistics partition function for the bbs as the value of a certain parabolic kostka polynomial , see e.g. , @xcite . this observation allows to write a fermionic formula for the parabolic kostka polynomials in question , see e.g. @xcite , as well as appears to be useful in the study of the bbs , see e.g. @xcite . let @xmath179 be a path of type @xmath443 and weight @xmath8 , define regularization @xmath444 of the path @xmath179 to be @xmath445 where @xmath446 , and we have omitted all symbols @xmath447 . @xmath93 let @xmath448 be semi - standard young tableau ( i.e. , highest weight ) corresponding to the regularized path @xmath449 . assume that all @xmath450 , then @xmath451 where @xmath452 and @xmath453 @xmath93 let @xmath454 , then @xmath455 we have @xmath456 , @xmath457 , and @xmath458 . on the other hand , let @xmath459 , then @xmath460 and @xmath461 , @xmath462 , so that @xmath463 , as expected . time evolution of @xmath464 is @xmath93 @xmath465 where @xmath466 is a constant and partitions @xmath467 and @xmath468 are defined as follows ; @xmath469 , and @xmath470 , @xmath471 ; @xmath472 , @xmath473 , @xmath474 . * proof . * indeed , we have @xmath475 where summation in the third term runs over all littlewood richardson tableaux of shape @xmath467 and weight @xmath476 . @xmath393 let @xmath477 be a sequence of rectangles , then @xmath478 where @xmath467 denotes partition @xmath479 , @xmath480 . @xmath93 _ it s _ well - known that parabolic kostka polynomials @xmath481 , where @xmath482 is a dominant ( i.e. , @xmath483 ) sequence of rectangular shape partitions , satisfy the so - called duality theorem @xmath484 where @xmath485 denotes a dominant rearrangement of the sequence of rectangular shape partitions @xmath486 and @xmath487 [ [ questions-1 . ] ] questions 1 . + + + + + + + + + + + + can one define a set @xmath488 of paths @xmath162 of rectangular type @xmath489 and rectangular weight @xmath490 , and energy function @xmath491 such that @xmath492 @xmath93 [ [ questions-2 . ] ] questions 2 . + + + + + + + + + + + + are there some physical interpretations " of the duality theorems @xmath493 and @xmath494 for polynomials @xmath481 and @xmath495 ? @xmath93 let us consider generating functions of the generalized tau statistics : @xmath496 where @xmath497 is the highest element and @xmath498 . let @xmath386 be a partition and @xmath20 be a sequence of rectangles , denote by @xmath499 the corresponding parabolic kostka polynomial . for any partition @xmath8 , let @xmath500 be the corresponding kostka number . then we have : [ conj : tau ] let @xmath501 be a sequence of rectangles . denote by @xmath502 the set of all paths ( including non - highest elements ) corresponding to the tensor product of crystals @xmath503 . then , @xmath504 where @xmath505 , @xmath506 . @xmath93 in other words , the statistics @xmath22 defines the essentially unique class of polynomials although its definition depends on choice of @xmath261 . as we see in the following example , this is not obvious from definition of @xmath22 . let us consider the case @xmath507 and @xmath508 . then we have total of 759 paths , and by direct computations , we have the following summation over all 759 paths . @xmath509 where @xmath510 and , in the last expression , @xmath511 . however , if we look at specific paths , for example , @xmath512 @xmath513 then we have @xmath514 in particular , dependences of @xmath515 on @xmath23 are different for each @xmath56 . @xmath93 we use the same notations of conjecture [ conj : tau ] . then , as we will see in corollary [ cor : e = maj ] below , we have @xmath516 for all @xmath517 , where @xmath518 is length of weight @xmath8 . @xmath93 let us consider generating function with respect to @xmath21 statistics . this is a more traditional problem compared with it for @xmath22 . indeed , special cases of this problem as well as its restriction to the set of highest weight paths were considered by several authors , see , e.g. , @xcite and references therein . let @xmath501 is a sequence of rectangles . denote by @xmath502 the set of all paths corresponding to tensor product of crystals @xmath503 . then , @xmath519 @xmath93 * comments . * an algebra geometric definition of the parabolic kostka polynomials ( also known as generalized kostka polynomials ) has been introduced in @xcite ( see also @xcite ) as a natural generalization of the well known formula , @xcite , p.244 , ( 1 ) , for the kostka foulkes polynomials in terms of a @xmath0-analogue of the kostant partition function . based on the study of combinatorial properties of the algebraic bethe ansatz , a fermionic formula for the parabolic kostka polynomials has been discovered by the first author in the middle of 80 s of the last century and has been proved in the full generality in @xcite . a `` path realization '' of the kostka foulkes polynomials has been obtained in @xcite , and finally , the formula @xmath520 has been proved in @xcite . @xmath93 let us consider the case @xmath521 and @xmath522 , i.e. , @xmath523 . then we have the following nine paths : @xmath524\hline \yvcentermath1\young(11,22)\otimes\young(22,34)\otimes\young(11,22,33)&10 & \yvcentermath1\young(12,23)\otimes\young(11,22)\otimes\young(12,23,34 ) \begin{picture}(0,2 ) \end{picture}&11 \\[15pt]\hline \yvcentermath1\young(12,23)\otimes\young(12,23)\otimes\young(11,22,34)&10 & \yvcentermath1\young(12,23)\otimes\young(12,24)\otimes\young(11,22,33 ) \begin{picture}(0,2 ) \end{picture}&11 \\[15pt]\hline \yvcentermath1\young(12,24)\otimes\young(12,23)\otimes\young(11,22,33)&11 & \yvcentermath1\young(22,33)\otimes\young(11,22)\otimes\young(11,22,34 ) \begin{picture}(0,2 ) \end{picture}&12 \\[15pt]\hline \yvcentermath1\young(22,34)\otimes\young(11,22)\otimes\young(11,22,33 ) \begin{picture}(0,2 ) \end{picture}&12 & & \\[15pt]\hline \end{array}\ ] ] therefore , the lhs of eq.([eq : conj ] ) is @xmath525 on the other hand , non - zero contributions for rhs of eq.([eq : conj ] ) comes from @xmath526 summing up , @xmath527 which coincides with the lhs . @xmath93 so far in this paper , we have considered several statistics including generalized tau statistics @xmath22 and more traditional one @xmath21 . let us investigate several further aspects of these two statistics . our main results in this section are ( i ) to show that @xmath21 belong to the class of statistics @xmath22 and ( ii ) to show that @xmath22 stabilize when we increase the value of @xmath23 . the following proposition will be a key property . [ prop : large_r ] let @xmath528 be the integer at the @xmath53-th row , @xmath293-th column of the tableau representation of @xmath529 , and let the highest element of @xmath33 be @xmath530 . then we have @xmath531 if @xmath532 and @xmath533 . * proof . * according to the algorithm presented in proposition [ prop : shimozono ] , we have to compute the insertion @xmath535 . this is worked out in lemma [ lem : large_r ] below , and shape of the resulting tableau coincides with the concatenation of two tableaux @xmath56 and @xmath530 . hence @xmath536 due to proposition [ prop : shimozono ] . @xmath393 * proof . * this insertion procedure can be divided into two steps . first , we show @xmath539 & \underbrace{\hspace{92pt}}_{s ' } \underbrace{\hspace{92pt}}_{s } \nonumber\end{aligned}\ ] ] where @xmath540 and @xmath541 . note that we have @xmath542 since ( i ) by the semi - standard property of @xmath56 , we have @xmath543 and ( ii ) by the assumption @xmath532 , thus @xmath544 . again from the assumption @xmath532 , we have @xmath545 by the semi - standard property @xmath546 . consider the insertion @xmath547 . from eq.([eq : proof1 ] ) , we have @xmath548 . thus the first row of @xmath547 is @xmath549 , and the remaining rows are identical to the corresponding rows of @xmath56 . if @xmath550 , this finishes the proof ( i.e. , by @xmath532 we have @xmath551 ) , therefore let us consider the case @xmath552 . assume that we have , for some @xmath553 and @xmath554 , @xmath555 where @xmath556 . we insert @xmath557 into this @xmath558 . from eq.([eq : proof1 ] ) and the assumption @xmath559 , we have @xmath560 , thus @xmath561 . therefore inserted @xmath557 bumps @xmath226 at @xmath562-th column of @xmath558 . as the next step , we have to insert @xmath226 to the second row of @xmath558 . by the similar reasoning , it bumps @xmath563 at @xmath562-th column . in this way , insertion of @xmath557 causes downward shift of @xmath562-th column of @xmath558 and addition of @xmath557 to the first row of @xmath562-th column . similarly , we see that the second insertion of @xmath557 causes shift and addition of @xmath557 to the @xmath564-th column . continuing in this way , we see that successive insertions @xmath565 gives the tableau @xmath558 with replacement @xmath226 by @xmath566 . by induction , we can show eq.([eq : proof0 ] ) . next , we consider the insertion of @xmath567 into @xmath103 : @xmath568 we denote the @xmath53-th row of @xmath103 ( resp . @xmath569 ) by @xmath570 ( resp . @xmath571 ) . although we independently repeat insertions of @xmath572 for @xmath32 times , we can argue more systematically as follows . let us consider the first row of @xmath103 : @xmath573 as we saw in the last paragraph , we have @xmath574 from eq.([eq : proof1 ] ) . suppose there are @xmath575 letters @xmath576 in the first row of @xmath56 , i.e. , @xmath577 . thus there are @xmath578 letters @xmath576 in the first row of @xmath103 . after inserting @xmath579 , precisely @xmath32 letters @xmath576 are bumped and go down to the second row , and the first row becomes @xmath580 in particular , @xmath575 letters @xmath576 on the right of the first row of @xmath56 are precisely reproduced in the right part of @xmath581 . now we have to consider the insertion of @xmath32 letters @xmath576 which are bumped from @xmath582 into the second row of @xmath103 : @xmath583 again , from eq.([eq : proof1 ] ) , we have @xmath584 , and suppose that there are @xmath585 letters @xmath586 in the second row of @xmath56 , i.e. , @xmath587 . after inserting @xmath588 , precisely @xmath32 letters @xmath589 are bumped and go down to the third row , and the second row becomes @xmath590 again , @xmath585 letters @xmath586 on the right of the second row of @xmath56 are precisely reproduced in the right part of @xmath591 . as we see in the above discussions , this procedure can be continued recursively . let the number of @xmath592 contained in the @xmath593-th row of @xmath56 be @xmath594 . then @xmath595 is @xmath596 and the right @xmath594 letters @xmath597 are copy of those originally contained in the right of @xmath593-th row of @xmath56 . in this way , @xmath569 contains copy of the letters in @xmath56 . as we have investigated insertion of @xmath579 , let us consider the insertion of @xmath598 into @xmath569 . consider the row @xmath581 . suppose that there are @xmath599 letters @xmath572 within the first row of @xmath56 , i.e. , @xmath600 . thus there are total of @xmath601 letters @xmath602 in the row @xmath581 . therefore , after insertion of @xmath603 , the row @xmath581 becomes @xmath604 in particular , @xmath599 letters @xmath572 and @xmath575 letters @xmath576 are copy of the corresponding letters of the first row of @xmath56 . as the result , @xmath32 letters @xmath572 are bumped from @xmath581 therefore , for the row @xmath591 , we have to insert @xmath579 , and again obtain copy of the letters @xmath576 and @xmath586 contained in the second row of @xmath56 . we can continue this procedure until the bottom row of @xmath569 . therefore each row of the resulting tableau @xmath605 contains copy of at most two species of letters in the original @xmath56 . we can recursively continue insertions of @xmath606 , @xmath607 , @xmath191 , @xmath608 , and each insertion generates copy of part of letters of @xmath56 . finally we get the result eq.([eq : concatenation ] ) . @xmath393 * proof . * recall the definition of @xmath22 @xmath614 where maj is defined by eq.([def : maj ] ) and @xmath530 is the highest element of @xmath33 . within the definition of @xmath22 , let us first consider the energy functions involving @xmath530 and next consider the remaining ones . as for the terms involving @xmath530 , we have @xmath615 by proposition [ prop : large_r ] . on the other hand , for the remaining contributions , recall that @xmath343 and @xmath616 . then from the definition of normalizations , we have @xmath617 combining both contributions , we obtain the sought relation . @xmath393 [ cor : e = maj ] let @xmath609 , and denote by @xmath610 the largest integer contained in tableau representation of @xmath179 , and define @xmath618 . then we have @xmath619 for all @xmath611 and @xmath620 . @xmath93 therefore conjecture [ conj : tau ] means that we can define at most @xmath610 independent statistics @xmath621 , where @xmath610 is the largest integer contained in the path @xmath179 , and these statistics have essentially unique generating function . let us remark physical interpretation of @xmath22 . for the paths ( including non - highest elements ) of shape @xmath622 and statistics @xmath623 ( @xmath618 ) , there is a straightforward generalization of theorem [ th : tau = rho ] , see section 4.1 of @xcite . under the same assumptions , @xmath623 is identified with cocharge of the unrestricted rigged configurations . it will be an interesting problem to find a physical interpretation of the more general @xmath621 . m. haiman , macdonald polynomials and geometry , _ new perspectives in geometric combinatorics _ ( billera , bjrner , greene , simion and stanley , eds . ) , msri publications , * 38 * cambridge univ . press ( 1999 ) 207254 . a. n. kirillov , combinatorics of young tableaux and rigged configurations ( russian ) proceedings of the st . petersburg math . soc . , * 7 * ( 1999 ) , 23 - 115 ; + translation in proceedings of the st . petersburg math . volume vii , amer . ser.2 , * 203 * , 17 - 98 , ams , providence , ri , 2001

this paper develops and analyzes an interior penalty discontinuous galerkin ( ipdg ) method using piecewise linear polynomials for the indefinite time harmonic maxwell equations with the impedance boundary condition in the three dimensional space . the main novelties of the proposed ipdg method include the following : first , the method penalizes not only the jumps of the tangential component of the electric field across the element faces but also the jumps of the tangential component of its vorticity field ; second , the penalty parameters are taken as complex numbers of negative imaginary parts . for the differential problem , we prove that the sesquilinear form associated with the maxwell problem satisfies a generalized weak stability ( i.e. , inf - sup condition ) for star - shaped domains . such a generalized weak stability readily infers wave - number explicit a priori estimates for the solution of the maxwell problem , which plays an important role in the error analysis for the ipdg method . for the proposed ipdg method , we show that the discrete sesquilinear form satisfies a coercivity for all positive mesh size and wave number and for general domains including non - star - shaped ones . in turn , the coercivity easily yields the well - posedness and stability estimates ( i.e. , a priori estimates ) for the discrete problem without imposing any mesh constraint . based on these discrete stability estimates , by adapting a nonstandard error estimate technique of , we derive both the energy - norm and the-norm error estimates for the ipdg method in all mesh parameter regimes including pre - asymptotic regime ( i.e. , ) . numerical experiments are also presented to gauge the theoretical results and to numerically examine the pollution effect ( with respect to ) in the error bounds . time harmonic maxwell equations , impedance boundary condition , interior penalty discontinuous galerkin methods , absolute stability , error estimates 65n12 , 65n15 , 65n30 , 78a40

1210.5837

this paper develops and analyzes interior penalty discontinuous galerkin ( ipdg ) methods for the following time harmonic maxwell problem : @xmath4 where @xmath5 is a bounded domain with lipschitz continuous boundary @xmath6 and of diameter @xmath7 . @xmath8 denotes the unit outward normal to @xmath9 , @xmath10 , the imaginary unit , and @xmath11 , the _ tangential component _ of the electric field @xmath12 . @xmath1 , called _ wave number _ , is a positive constant and @xmath13 is known as the impedance constant . is the standard impedance boundary condition . assume that @xmath14 , hence , @xmath15 . problem is a prototypical problem in electromagnetic scattering ( cf . @xcite and the references therein ) and has been used extensively as a model ( and benchmark ) problem to develop various numerical discretization methods including finite element methods @xcite and discontinuous galerkin methods @xcite , and to develop fast solvers ( cf . @xcite and the references therein ) . the above maxwell problem with large wave number @xmath1 is numerically difficult to solve mainly because of the following two reasons . first , the large wave number @xmath1 implies the small wave length @xmath16 , that is , the wave is a short wave and very oscillatory . it is well known that , in every coordinate direction , one must put some minimal number of grid points in each wave length in order to resolve the wave . using such a fine mesh evidently results in a huge algebraic problem to solve regardless what discretization method is used . practically , the rule of thumb " is to use @xmath17 grid points per wave length , which means that the mesh size @xmath0 must satisfy the constraint @xmath18 . to the best of our knowledge , no numerical method in the literature has been proved to be uniquely solvable and to have an error bound under the mesh constraint @xmath18 for the above maxwell problem . moreover , numerical experiments have shown that under the mesh condition @xmath18 the errors of all existing numerical methods grow as the wave number @xmath1 increases . this means that the error is not completely controlled by the product @xmath19 and it provides strong evidences of the existence of so - called pollution " in the error bounds . it is known now @xcite that the existence of pollution is related to the loss of stability of numerical methods with large wave numbers for the scalar wave equation , which is also expected to be the case for the vector wave equations . second , for large wave number @xmath1 , the maxwell operator is strongly indefinite . such a strong indefiniteness certainly passes onto any discretization of the maxwell problem . in other words , the stiffness matrix of the discrete problem is not only very large but also strongly indefinite . solving such a large , strongly indefinite , and ill - conditioned algebraic problem is proved to be very challenging and all the well - known iterative methods were proved numerically to be either ineffective or divergent for indefinite wave problems in the case of large wave number ( cf . @xcite and the references therein ) . this paper is an attempt to address the first difficulty mentioned above for the maxwell equations . in particular , our goal is to design and analyze discretization methods which have superior stability properties and give optimal rates of convergence for the maxwell problem . motivated by our previous experiences with the helmholtz equation @xcite , we again try to accomplish the goal by developing some interior penalty discontinuous galerkin method for problem . the focus of the paper is to establish the rigorous stability and error analysis for the proposed ipdg method , in particular , in the preasymptotic regime ( i.e. , when @xmath20 ) . for the ease of presentation and to better present ideas , we confine ourselves to only consider the linear element in this paper and will discuss its high order extensions in a forthcoming paper . the remainder of this paper is organized as follows . section [ sec-2 ] is devoted to the study of the coercivity of the maxwell operator and the wave - number explicit estimates for the solution of . we show that the sesquilinear form associated with the maxwell problem satisfies a generalized weak coercivity ( i.e. , inf - sup condition ) . this coercivity in turn readily infers the wave - number explicit solution estimates which were proved in @xcite . we note that the proofs of both results given in this paper are of independent interest and refer the reader to @xcite for further discussions in the direction . section [ sec-3 ] presents the construction of our ipdg method and some simple properties of the proposed discrete sesquilinear form . section [ sec-4 ] studies the coercivity of the discrete sesquilinear form and derives stability estimates for the ipdg solutions . it is proved that the discrete sesquilinear form satisfies a coercivity for all mesh size @xmath21 and all wave number @xmath22 and for general domains including non - star - shaped ones , which is stronger than the generalized weak coercivity satisfied by its continuous counterpart . all these are possible because of the special design of the discrete sesquilinear form and the special property @xmath23 ( element - wise ) for all piecewise linear functions @xmath24 . this coercivity in turn readily infers the well - posedness and stability estimates for the discrete problem without imposing any mesh constraint . section [ sec-5 ] devotes to the error analysis for the proposed ipdg method . by using the discrete stability estimates and adapting a nonstandard error estimate technique of @xcite , we derive both the energy - norm and the @xmath2-norm error estimates for the ipdg method in all mesh parameter regimes including pre - asymptotic regime ( i.e. , @xmath3 ) . finally , we present some numerical experiment results in section [ sec-6 ] to gauge the theoretical results and to numerically examine the pollution effect ( with respect to @xmath1 ) in the error bounds . the standard space , norm and inner product notation are adopted in this paper . their definitions can be found in @xcite . in particular , @xmath25 and @xmath26 for @xmath27 and @xmath28 denote the @xmath2-inner product on _ complex - valued _ @xmath29 and @xmath30 spaces , respectively . for a given function space @xmath31 , let @xmath32 . in particular , @xmath33 and @xmath34 . we also define @xmath35 throughout this paper , the bold face letters are used to denote three - dimensional vectors or vector - valued functions , and @xmath36 is used to denote a generic positive constant which is independent of @xmath0 and @xmath1 . we also use the shorthand notation @xmath37 and @xmath38 for the inequality @xmath39 and @xmath40 . @xmath41 is a shorthand notation for the statement @xmath37 and @xmath42 . we now recall the definition of star - shaped domains . [ def1 ] @xmath43 is said to be a _ star - shaped _ domain with respect to @xmath44 if there exists a nonnegative constant @xmath45 such that @xmath46 @xmath43 is said to be _ strictly star - shaped _ if @xmath45 is positive . where @xmath47 denotes the unit outward normal to @xmath48 . throughout this paper , we assume that @xmath49 is a strictly star - shaped domain . introduce the following sesquilinear form on @xmath50 @xmath51 then the weak formulation for the maxwell system is defined as seeking @xmath52 such that @xmath53 using the fredholm alternative principle it can be shown that problem has a unique solution ( cf . @xcite ) . note that choosing @xmath54 with @xmath55 shows that @xmath56 or @xmath57 next , we prove that the sesquilinear form @xmath58 satisfies a generalized weak coercivity which is expressed in terms of a generalized _ inf - sup _ condition . [ inf - sup ] let @xmath59 be a bounded star - shaped domain with the positive constant @xmath60 and the diameter @xmath61 . then for any @xmath62 there holds the following generalized inf - sup condition for the sesquilinear form @xmath58 : @xmath63 where @xmath64}\\ & { \vert{\hskip -1.3pt}\vert { \mathbf{u}}\vert{\hskip -1.3pt}\vert_{{\mathbf{l}}^2({\omega } ) } } : = \bigl ( k^2\|{\mathbf{u}}\|_{{\mathbf{l}}^2({\omega})}^2 + k^2 c_{\omega}\|{\mathbf{u}}\|_{{\mathbf{l}}^2(\gamma)}^2 \bigr)^{\frac12 } , { \displaybreak[0]}\\ & \|{\mathbf{u}}\|_{e } : = \bigl ( k^2 \|{\mathbf{u}}\|_{{\mathbf{l}}^2({\omega})}^2 + k^2 c_{\omega}\|{\mathbf{u}}\|_{{\mathbf{l}}^2(\gamma)}^2+\|\operatorname{\mathbf{curl}}{\mathbf{u}}\|_{{\mathbf{l}}^2({\omega})}^2 + c_{\omega}\|\operatorname{\mathbf{curl}}{\mathbf{u}}\|_{{\mathbf{l}}^2(\gamma)}^2 \bigr)^{\frac12}.\label{eq2.0c}\end{aligned}\ ] ] let @xmath65 . setting @xmath66 in and taking the real and imaginary parts we get @xmath67 alternatively , setting @xmath68 in ( notice that @xmath69 is a valid test function for @xmath70 ) , taking the real part , and using the following integral identity ( cf . @xcite ) @xmath71 and the assumption that @xmath72 , we get @xmath73 from and and using the following integral identity ( cf . @xcite ) @xmath74 we have @xmath75}\\ & = -\langle { \mathbf{w}}\cdot{\boldsymbol{\nu } } , |\operatorname{\mathbf{curl}}{\mathbf{u}}|^2 \rangle_\gamma -k^2 \langle { \mathbf{w}}\cdot { \boldsymbol{\nu } } , |{\mathbf{u}}|^2 \rangle_\gamma \nonumber\\ & \qquad + 2k^2 \operatorname{{re}}\langle { \mathbf{w}}\times { \mathbf{u } } , { \mathbf{u}}\times { \boldsymbol{\nu}}\rangle_\gamma -2\lambda \operatorname{{im}}\langle { \mathbf{u}}_t,{\mathbf{v}}_t \rangle_\gamma \nonumber \\ & \qquad + 2\operatorname{{re}}a({\mathbf{u}},{\mathbf{v}})-\operatorname{{re}}a({\mathbf{u}},{\mathbf{u } } ) \nonumber { \displaybreak[0]}\\ & = -\langle { \mathbf{w}}\cdot{\boldsymbol{\nu } } , |\operatorname{\mathbf{curl}}{\mathbf{u}}|^2 \rangle_\gamma -k^2 \langle { \mathbf{w}}\cdot { \boldsymbol{\nu } } , |{\mathbf{u}}|^2 \rangle_\gamma \nonumber \\ & \qquad -2k^2 \langle { \mathbf{w}}\cdot { \boldsymbol{\nu } } , |{\mathbf{u}}\times { \boldsymbol{\nu}}|^2 \rangle_\gamma + 2k^2 \operatorname{{re}}\langle { \mathbf{w}}_t\times { \mathbf{u } } , { \mathbf{u}}\times { \boldsymbol{\nu}}\rangle_\gamma \nonumber \\ & \qquad -2\lambda\operatorname{{im}}\langle { \mathbf{u}}_t,{\mathbf{v}}_t \rangle_\gamma + 2\operatorname{{re}}a({\mathbf{u}},{\mathbf{v}})-\operatorname{{re}}a({\mathbf{u}},{\mathbf{u } } ) . \nonumber\end{aligned}\ ] ] here we have used the decomposition @xmath76 to obtain the last equality . on noting that @xmath77 , @xmath78 , and that @xmath79 , using the star - shaped domain assumption and schwarz inequality we obtain @xmath80}\\ & \leq -\frac{c_{\omega}}2 \|\operatorname{\mathbf{curl}}{\mathbf{u}}\|_{{\mathbf{l}}^2(\gamma)}^2 -\frac{k^2 c_{\omega}}{2 } \|{\mathbf{u}}\|_{{\mathbf{l}}^2(\gamma)}^2 -2 k^2 c_{\omega}\|{\mathbf{u}}_t\|_{{\mathbf{l}}^2(\gamma)}^2 \nonumber\\ & \quad + \frac{2r^2(k^2+\lambda^2)}{c_{\omega } } \|{\mathbf{u}}_t\|_{{\mathbf{l}}^2(\gamma)}^2 + 2\operatorname{{re}}a({\mathbf{u}},{\mathbf{v}})- \operatorname{{re}}a({\mathbf{u}},{\mathbf{u } } ) . \nonumber\end{aligned}\ ] ] finally , it follows from , and that @xmath81 where @xmath68 and @xmath82 is defined in . it is easy to check that there holds for @xmath68 @xmath83 hence , it follows from that @xmath84 where @xmath85 as defined in . the proof is complete . an immediate consequence of the above generalized _ inf - sup _ condition is the following stability estimate for solutions of problem . [ stability ] in addition to the assumptions of theorem [ inf - sup ] , assume that @xmath86 and @xmath87 . let @xmath88 be a solution of the variational problem . then there holds following stability estimate : @xmath89 for all @xmath90 . where @xmath91 let @xmath92 solve @xmath93 set @xmath94 and @xmath95 , where @xmath12 is a solution to . trivially , we have @xmath96 and @xmath97 in @xmath49 , and @xmath98 on @xmath99 . by we also have @xmath100 . hence , @xmath101 . moreover , since @xmath12 satisfies , it is easy to verify that @xmath102 satisfies @xmath103 testing by @xmath104 and integrating by parts on both sides of the resulting equation yield @xmath105 hence , @xmath106 alternatively , testing by @xmath107 with @xmath108 , using the following rellich identity for the laplacian ( cf . @xcite ) : @xmath109 and integrating by parts we get ( note that @xmath110 ) @xmath111 hence , by and the star - shaped domain assumption we obtain @xmath112 finally , by and schwarz inequality we get @xmath113 it follows from the generalized inf - sup condition , and that @xmath114 which together with and the relation @xmath95 as well as the definition of the energy norm @xmath115 infer that ( again , note that @xmath110 ) @xmath116 hence , holds . the proof is complete . we conclude this section with a few remarks . since problem is linear , the stability estimate immediately implies the uniqueness of the problem in the function class in which the estimate is derived . this provides an alternative method ( to the traditional integral equation method and the unique continuation method ) for establishing uniqueness ( and existence ) for the maxwell problem . \(a ) the generalized inf - sup condition is a stronger result than a stability estimate for the solution of the maxwell problem . the reason to restrict @xmath117 in is that @xmath118 operator has a non - trivial kernel . \(b ) stability estimates similar to were established independently early in @xcite and @xcite . also explicitly shows the dependence on the _ size _ and the _ shape constant _ of the domain . such an estimate plays an important role for designing multilevel schwarz preconditioners for discretizations of and for doing practical simulations because in practice the size of the computational domain @xmath49 is often taken to be proportional to the wave length . in addition , not only the sharp wave number - explicit and domain size - explicit stability estimate is obtained as a corollary of the generalized inf - sup condition , but also the derivation reveals some deep insights about the dependence of the solution on the datum functions and the domain . \(c ) the generalized inf - sup condition provides a guideline for constructing good " numerical schemes for the maxwell equations . we shall call a discretization method a coercivity preserving method " if it satisfies a discrete inf - sup condition which mimics the continuous inf - sup condition . constructing such a coercivity preserving ipdg method is one of primary goals of this paper . \(d ) generalized inf - sup conditions similar to also hold for the scalar helmholtz equation and the elastic helmholtz equations ( cf . @xcite ) . based on the above stability estimates in lower norms , one can also derive stability estimates in higher norms when the solution @xmath12 is sufficient regular . we state an @xmath119-estimate for @xmath120 below without giving a proof ( cf . * remark 4.9 ) ) . [ high_stability ] suppose that @xmath121 and the solution @xmath12 of problem satisfies @xmath122 for @xmath123 . then there holds estimate @xmath124 where @xmath125 to formulate our ipdg methods , we first need to introduce some notation . let @xmath126 be a family of partitions ( into tetrahedrons and/or parallelepipeds ) of the domain @xmath49 parameterized by @xmath21 . for any element " @xmath127 , we define @xmath128 . similarly , for each face @xmath129 of @xmath127 , define @xmath130 . we assume that the elements of @xmath131 satisfy the minimal angle condition . let @xmath132 we define the jump @xmath133 and average @xmath134 of @xmath135 on an interior face @xmath136 as @xmath137 if @xmath138 , set @xmath139 and @xmath140 . for every @xmath141 , let @xmath142 be the unit outward normal to the face @xmath129 of the element @xmath143 if the global label of @xmath143 is bigger and of the element @xmath144 if the other way around . for every @xmath138 , let @xmath145 the unit outward normal to @xmath146 . to formulate our ipdg methods , we recall the following ( local ) integration by parts formula : @xmath147 where @xmath148 . next , multiplying equation by a test function @xmath149 , integrating over @xmath127 , using the integration by parts formula , and summing the resulted equation over all @xmath127 we get @xmath150 to deal with the boundary terms in the big sum , we appeal to the following algebraic identity . for each interior face @xmath151 there holds @xmath152 substituting identity into after dropping the first term on the right - hand side of ( because @xmath153 if @xmath12 is sufficiently regular ) yields @xmath154 utilizing the boundary condition in the third term on the left - hand side and adding a symmetrization " term then lead to the following equation : @xmath155 where @xmath156 . the most important and tricky issue for designing an ipdg method is how to introduce suitable _ interior penalty _ term(s ) on the left - hand side of . obviously , different interior penalty terms will result in different numerical methods . as it was proved in @xcite , using the standard interior penalty terms will lead to ipdg methods which require a restrictive mesh constraint to ensure the stability and accuracy in the case of large wave number @xmath1 . inspired by our previous work @xcite on ipdg methods for the helmholtz equation and guided by our stability analysis ( see section [ sec-4 ] ) , here we introduce some non - standard interior penalty terms into , which we shall describe below , and the ipdg method so constructed will be proved to be absolutely stable ( with respect to wave number @xmath1 and mesh size @xmath0 ) in the next section . to define our ipdg methods , we first introduce the energy " space @xmath157 and the sesquilinear form @xmath158 on @xmath159 as follows . @xmath160 @xmath161}\\ { \mathcal{j}}_0({\mathbf{u}},{\mathbf{v}}):=&\sum_{{\mathcal{f}}\in{\mathcal{e}}_h^i}\frac{{\gamma}_{0,{\mathcal{f}}}}{h_{\mathcal{f}}}\ , \bigl\langle \jump{{\mathbf{u}}_t},\jump{{\mathbf{v}}_t}\bigr\rangle_{\mathcal{f}},\label{e3.6}{\displaybreak[0]}\\ { \mathcal{j}}_1({\mathbf{u}},{\mathbf{v}}):=&\sum_{{\mathcal{f}}\in{\mathcal{e}}_h^i } { \gamma}_{1,{\mathcal{f } } } h_{\mathcal{f}}\bigl\langle\jump{\operatorname{\mathbf{curl}}{\mathbf{u}}\times{\mathbf{n}}_{\mathcal{f}}},\jump{\operatorname{\mathbf{curl}}{\mathbf{v}}\times{\mathbf{n}}_{\mathcal{f } } } \bigr\rangle_{\mathcal{f } } , \label{cj1}\end{aligned}\ ] ] where @xmath162 and @xmath163 are nonnegative numbers to be specified later . \(a ) clearly , @xmath158 is a consistent discretization for @xmath164 since @xmath165 for all @xmath166 and @xmath167 with @xmath168 . \(b ) the terms in @xmath169 are called penalty terms . the penalty parameters @xmath170 and @xmath171 are pure imaginary numbers with _ negative _ imaginary parts . our analysis still applies if they are taken as complex numbers of negative imaginary parts . \(c ) the @xmath172 term penalizes the jumps of the vector field @xmath102 and the @xmath173 term penalizes the jumps of the tangential component of the vector field @xmath174 . which , to the best of our knowledge , has not been used before in the context of ipdg methods for the maxwell equations . they play a vital role for our ipdg methods being absolutely stable , see section [ sec-4 ] . \(d ) @xmath175 correspond to the nonsymmetric , incomplete , and symmetric ipdg methods for the poisson problem . in the remainder of this paper , we shall only consider the symmetric case @xmath176 and set @xmath177 for notation brevity . with the help of the sesquilinear form @xmath178 we now introduce the following weak formulation for : find @xmath179 such that @xmath180 where @xmath181 from , it is clear that , if @xmath182 is the solution of , then holds for all @xmath183 . for any @xmath127 , let @xmath184 denote the set of all complex - valued polynomials whose degrees in all variables ( total degrees ) do not exceed @xmath185 . we define our ipdg approximation space @xmath186 as @xmath187 clearly , @xmath188 . but @xmath189 . we are now ready to define our ipdg methods based on the weak formulation : find @xmath190 such that for all @xmath191 @xmath192 we note that defines a family of ipdg methods for @xmath193 . for the ease of presentation and to better present ideas , in the rest of this paper we only consider the case @xmath194 , the linear element case . in the next two sections , we shall study the stability and error estimates for the above ipdg method with @xmath194 . especially , we are interested in knowing how the stability constants and error constants depend on the wave number @xmath1 ( and mesh size @xmath0 , of course ) and what are the optimal " relationship between mesh size @xmath0 and the wave number @xmath1 . we remark that the ipdg method with @xmath194 uses piecewise linear polynomials even for cartesian meshes . by contrast , for the corresponding linear conforming edge element method on cartesian meshes , the trial functions have to be chosen as piecewise trilinear polynomials . we also note that the linear system resulted from is ill - conditioned and strongly indefinite because the coefficient matrix has many eigenvalues with very large negative real parts . solving such a large linear system is another challenging problem associated with time harmonic maxwell problems , which will be addressed in a future work . for further analysis we introduce the following semi - norms / norms on @xmath157 : @xmath195}\\ { \left\vert{\mathbf{v}}\right\vert}_{dg}^2 : = & \|\operatorname{\mathbf{curl}}{\mathbf{v}}\|_{{\mathbf{l}}^2({\mathcal{t}}_h)}^2 + { \vert{\mathbf{v}}\vert_{{\mathbf{l}}^2({\omega})}}^2 \label{e3.12 } \\ & + \sum_{{\mathcal{f}}\in{\mathcal{e}}_h^i } \bigl ( \frac{{\gamma}_{0,{\mathcal{f}}}}{h_{\mathcal{f } } } { \vert{\left[{\mathbf{v}}_t\right]}\vert_{{\mathbf{l}}^2({\mathcal{f}})}}^2 + { \gamma}_{1,{\mathcal{f } } } h_{\mathcal{f}}{\vert{\left[\operatorname{\mathbf{curl}}{\mathbf{v}}\times { \boldsymbol{\nu}}_{\mathcal{f}}\right]}\vert_{{\mathbf{l}}^2({\mathcal{f}})}}^2 \bigr ) \nonumber{\displaybreak[0]}\\ = & \|\operatorname{\mathbf{curl}}{\mathbf{v}}\|_{{\mathbf{l}}^2({\mathcal{t}}_h)}^2 + { \vert{\mathbf{v}}\vert_{{\mathbf{l}}^2({\omega})}}^2+{\mathcal{j}}_0({\mathbf{v}},{\mathbf{v } } ) + { \mathcal{j}}_1({\mathbf{v}},{\mathbf{v } } ) , { \nonumber}{\displaybreak[0]}\\ { \vert{\hskip -1.3pt}\vert { \mathbf{v}}\vert{\hskip -1.3pt}\vert_{dg}}^2 : = & { \left\vert{\mathbf{v}}\right\vert}_{dg}^2 + \sum_{{\mathcal{f}}\in{\mathcal{e}}_h^i } \frac{h_{\mathcal{f}}}{{\gamma}_{0,{\mathcal{f}}}}{\vert{\left\{\operatorname{\mathbf{curl}}{\mathbf{v}}\times { \boldsymbol{\nu}}_{\mathcal{f}}\right\}}\vert_{{\mathbf{l}}^2({\mathcal{f}})}}^2 . \label{e3.13}\end{aligned}\ ] ] clearly , the sesquilinear form @xmath178 satisfies : for any @xmath167 @xmath196 in this section we shall prove that the discrete sesquilinear form @xmath197 satisfies a discrete coercivity , which is slightly stronger than the generalized inf - sup condition proved in the previous section for the sesquilinear form @xmath58 . such a discrete coercivity is possible for the linear element because @xmath23 ( defined element - wise ) for all @xmath198 . as an immediate corollary of the discrete coercivity , we shall derive a priori estimates for solutions of for all @xmath199 , which then infer the well - posedness of . we state the first main theorem of this section which establishes a coercivity for the discrete sesquilinear form @xmath197 . [ discrete_coercivity ] let @xmath200 , @xmath201 , and @xmath202 . then there exists a constant @xmath203 such that @xmath204 for all @xmath205 . where @xmath206 for any @xmath207 , define @xmath208 . by , , and the following trace inequality @xmath209 for some @xmath210-independent positive constant @xmath36 , we get @xmath211\|_{{\mathbf{l}}^2({\mathcal{f}})}\nonumber\\ & \hskip 0.5 in \leq\|\operatorname{\mathbf{curl}}{\mathbf{u}}_h\|_{{\mathbf{l}}^2({\mathcal{t}}_h)}^2 - k^2\|{\mathbf{u}}_h\|_{{\mathbf{l}}^2({\omega})}^2 \nonumber\\ & \hskip 1.2 in + c \sum_{{\mathcal{f}}\in { \mathcal{e}}_h^i } h_{\mathcal{f}}^{-\frac12 } \|\operatorname{\mathbf{curl}}{\mathbf{u}}_h\|_{{\mathbf{l}}^2({\omega}_{\mathcal{f } } ) } \|[({\mathbf{u}}_h)_t]\|_{{\mathbf{l}}^2({\mathcal{f}})}\nonumber\\ & \hskip 0.5 in \leq \frac32 \|\operatorname{\mathbf{curl}}{\mathbf{u}}_h\|_{{\mathbf{l}}^2({\mathcal{t}}_h)}^2 - k^2\|{\mathbf{u}}_h\|_{{\mathbf{l}}^2({\omega})}^2 + c \sum_{{\mathcal{f}}\in { \mathcal{e}}_h^i } h_{\mathcal{f}}^{-1 } \|[({\mathbf{u}}_h)_t]\|_{{\mathbf{l}}^2({\mathcal{f}})}^2 . \nonumber \end{aligned}\ ] ] since @xmath212 is piecewise linear , then @xmath213 in each @xmath127 . by integrating by parts and using the trace inequality we obtain @xmath214}\\ & \qquad = -\sum_{{\mathcal{f}}\in{\mathcal{e}}_h^b } { \bigl\langle}\operatorname{\mathbf{curl}}{\mathbf{u}}_h\times{\boldsymbol{\nu}}_k,({\mathbf{u}}_h)_t{\bigr\rangle}_{{\mathcal{f } } } { \nonumber}\\ & \qquad\qquad -\sum_{{\mathcal{f}}\in{\mathcal{e}}_h^i}\bigl({\bigl\langle}[\operatorname{\mathbf{curl}}{\mathbf{u}}_h\times{\boldsymbol{\nu}}_{\mathcal{f}}],\{({\mathbf{u}}_h)_t\ } { \bigr\rangle}_{{\mathcal{f}}}+{\bigl\langle}\{\operatorname{\mathbf{curl}}{\mathbf{u}}_h\times{\boldsymbol{\nu}}_{\mathcal{f}}\},[({\mathbf{u}}_h)_t]{\bigr\rangle}_{{\mathcal{f}}}\bigr ) { \nonumber}{\displaybreak[0]}\\ & \qquad \leq c\sum_{{\mathcal{f}}\in{\mathcal{e}}_h^b } h_{\mathcal{f}}^{-\frac12 } \|\operatorname{\mathbf{curl}}{\mathbf{u}}_h\|_{{\mathbf{l}}^2({\omega}_{\mathcal{f } } ) } \|({\mathbf{u}}_h)_t\|_{{\mathbf{l}}^2({\mathcal{f } } ) } { \nonumber}\\ & \qquad\qquad + c\sum_{{\mathcal{f}}\in{\mathcal{e}}_h^i } h_{\mathcal{f}}^{-\frac12 } \|[\operatorname{\mathbf{curl}}{\mathbf{u}}_h\times{\boldsymbol{\nu}}_{\mathcal{f}}]\|_{{\mathbf{l}}^2({\mathcal{f } } ) } \|{\mathbf{u}}_h\|_{{\mathbf{l}}^2({\omega}_{\mathcal{f } } ) } { \nonumber}\\ & \qquad\qquad + c\sum_{{\mathcal{f}}\in{\mathcal{e}}_h^i } h_{\mathcal{f}}^{-\frac12 } \|\operatorname{\mathbf{curl}}{\mathbf{u}}_h\|_{{\mathbf{l}}^2({\omega}_{\mathcal{f } } ) } \|[({\mathbf{u}}_h)_t]\|_{{\mathbf{l}}^2({\mathcal{f } } ) } { \nonumber}{\displaybreak[0]}\\ & \qquad \leq \frac13 \|\operatorname{\mathbf{curl}}{\mathbf{u}}_h\|_{{\mathbf{l}}^2({\mathcal{t}}_h)}^2 + \frac{k^2}{6 } \|{\mathbf{u}}_h\|_{{\mathbf{l}}^2({\omega})}^2 + c \sum_{{\mathcal{f}}\in{\mathcal{e}}_h^b } h_{\mathcal{f}}^{-1 } \|({\mathbf{u}}_h)_t\|_{{\mathbf{l}}^2({\mathcal{f}})}^2 { \nonumber}\\ & \qquad\qquad + c\sum_{{\mathcal{f}}\in{\mathcal{e}}_h^i } h_{\mathcal{f}}^{-1}\bigl ( \|[({\mathbf{u}}_h)_t]\|_{{\mathbf{l}}^2({\mathcal{f}})}^2 + k^{-2 } \|[\operatorname{\mathbf{curl}}{\mathbf{u}}_h\times{\boldsymbol{\nu}}_{\mathcal{f}}]\|_{{\mathbf{l}}^2({\mathcal{f}})}^2 \bigr ) . { \nonumber}\end{aligned}\ ] ] hence , @xmath215\|_{{\mathbf{l}}^2({\mathcal{f}})}^2 + k^{-2 } \|[\operatorname{\mathbf{curl}}{\mathbf{u}}_h\times{\boldsymbol{\nu}}_{\mathcal{f}}]\|_{{\mathbf{l}}^2({\mathcal{f}})}^2 \bigr ) . { \nonumber}\end{aligned}\ ] ] adding and and rearranging the terms yield @xmath216}\vert_{{\mathbf{l}}^2({\mathcal{f}})}}^2 { \nonumber}\\ & \quad+ \frac{c}{\gamma_1 k^2 h_{\min}^2 } \sum_{{\mathcal{f}}\in{\mathcal{e}}_h^i}{\gamma}_{1,{\mathcal{f } } } h_{\mathcal{f}}{\vert{\left[\operatorname{\mathbf{curl}}{\mathbf{u}}_h\times{\boldsymbol{\nu}}_{\mathcal{f}}\right]}\vert_{{\mathbf{l}}^2({\mathcal{f}})}}^2 .{\nonumber}\end{aligned}\ ] ] therefore , by the definitions of @xmath217 and @xmath218 and the identity we get @xmath219}\\ & \qquad \leq -2\operatorname{{re}}a_h({\mathbf{u}}_h , { \mathbf{u}}_h ) + c{\gamma}_h\big({\mathcal{j}}_0({\mathbf{u}}_h,{\mathbf{u}}_h)+{\mathcal{j}}_1({\mathbf{u}}_h,{\mathbf{u}}_h)+\lambda \|({\mathbf{u}}_h)_t\|_{{\mathbf{l}}^2(\gamma)}^2\big ) { \nonumber}{\displaybreak[0]}\\ & \qquad = -2\operatorname{{re}}a_h({\mathbf{u}}_h , { \mathbf{u}}_h ) - c \gamma_h \operatorname{{im}}a_h({\mathbf{u}}_h , { \mathbf{u}}_h ) { \nonumber}{\displaybreak[0]}\\ & \qquad \leq c \gamma_h \bigl(|\operatorname{{re}}a_h({\mathbf{u}}_h,{\mathbf{u}}_h)|+|\operatorname{{im}}a_h({\mathbf{u}}_h , { \mathbf{u}}_h)|\bigr ) { \nonumber}{\displaybreak[0]}\leq c \gamma_h |a_h({\mathbf{u}}_h , { \mathbf{u}}_h)| , { \nonumber}\end{aligned}\ ] ] where @xmath220 is defined by . hence , holds . the proof is completed . \(a ) the discrete sesquilinear form @xmath197 satisfies a stronger coercivity than its continuous counterpart @xmath58 does , see theorem [ inf - sup ] . moreover , the proof of theorem [ discrete_coercivity ] is simpler than that of theorem [ inf - sup ] , all these are possible because of the special form of @xmath197 and the fact that @xmath23 in @xmath127 for all piecewise linear functions @xmath198 . however , a weak coercivity is only expected to hold in the case of high order elements . \(b ) it is also important to point out that theorem [ discrete_coercivity ] holds without assuming that @xmath49 is a star - shaped domain . an immediate consequence of the above discrete coercivity are the following a priori estimates for solutions to the ipdg method . [ discrete_stability ] every solution @xmath221 of the ipdg method satisfies the following stability estimates . @xmath222 by and schwarz inequality we get @xmath223 the desired estimates follow from combining the above inequality with . the proof is completed . the above discrete stability estimates in turn immediately imply the well - posedness of the ipdg method . [ existence ] there exists a unique solution to for any fixed set of parameters @xmath224 . in what follows , we suppose @xmath225 and @xmath226 for brevity . for simplicity , we assume that @xmath121 and that @xmath131 is a quasi - uniform partition of @xmath49 consisting of tetrahedrons . let @xmath227 . let @xmath12 be the solution to problem and @xmath229 be its ipdg @xmath228-elliptic projection defined as follows . @xmath230 the following lemma establishes the continuity and coercivity for the discrete sesquilinear form @xmath178 . [ lem5.1 ] for any @xmath231 , the sesquilinear form @xmath178 satisfies @xmath232 in addition , there exists a positive constant @xmath233 such that , for @xmath234 , @xmath235 clearly , follows from the definitions , , and schwarz inequality . it remains to prove . from we have @xmath236\right\rangle_e}\\ & -\sum_{{\mathcal{f}}\in{\mathcal{e}}_h^i } \frac{h_{\mathcal{f}}}{{\gamma}_{0,e}}{\vert{\left\{\operatorname{\mathbf{curl}}{\mathbf{v}}_h\times { \boldsymbol{\nu}}_{\mathcal{f}}\right\}}\vert_{{\mathbf{l}}^2({\mathcal{f}})}}^2.\end{aligned}\ ] ] it follows from the derivation of that there exists a constant @xmath237 such that @xmath238\right\rangle_e}\\ & \leq \frac{1}{4}{\vert\operatorname{\mathbf{curl}}{\mathbf{v}}_h\vert_{{\mathbf{l}}^2({\mathcal{t}}_h)}}^2 + \frac{c_0}{{\gamma}_0}\sum_{{\mathcal{f}}\in{\mathcal{e}}_h^i}\frac{{\gamma}_{0,{\mathcal{f}}}}{h_{\mathcal{f } } } { \vert{\left[({\mathbf{v}}_h)_t\right]}\vert_{{\mathbf{l}}^2({\mathcal{f}})}}^2.\end{aligned}\ ] ] on the other hand , from , there exists a constant @xmath239 such that , @xmath240 therefore , @xmath241 which gives if @xmath242 large enough . the proof is completed . [ rem5.1 ] the coercivity and continuity of @xmath178 ensure that the above @xmath228-elliptic projection is well defined . the following lemma establishes error estimates for @xmath243 . [ lem5.3 ] suppose problem is @xmath244-regular , then , under the conditions of lemma [ lem5.1 ] , there hold the following estimates : @xmath245 where @xmath246 _ step 1 : _ it follows from @xcite that there exists @xmath247 ( i.e. , the conforming ndlec interpolation of @xmath12 ) such that the following estimates hold : @xmath248 where can be proved by , the commuting property between the curl - conforming interpolation operator and the div - conforming interpolation operator ( * ? ? ? * lemma 8.13 ) , and the trace inequality . let @xmath249 and @xmath250 , then @xmath251 . by we have @xmath252 from lemma [ lem5.1 ] and we get @xmath253 therefore , it follows from that @xmath254 which together with the relation @xmath255 and the triangle inequality immediately infer . to show , we first need the following results that can be proved by following the proof of ( * ? ? ? * proposition 4.5 ) and their proofs are omitted : for any @xmath198 there exists @xmath256 such that @xmath257}\vert_{{\mathbf{l}}^2({\mathcal{f}})}}^2 , \label{e5.11c } \\ \|\operatorname{\mathbf{curl}}({\mathbf{v}}_h-{\mathbf{v}}_h^c)\|_{{\mathbf{l}}^2({\mathcal{t}}_h)}^2 & \leq c\sum_{{\mathcal{f}}\in{\mathcal{e}}_h^i } h_{\mathcal{f}}^{-1 } { \vert{\left[({\mathbf{v}}_h)_t\right]}\vert_{{\mathbf{l}}^2({\mathcal{f}})}}^2 . \label{e5.11d}\end{aligned}\ ] ] let @xmath258 be the conforming approximation of @xmath259 as defined above . then it follows from the definition of the norm @xmath260 ( cf . ) , the above two estimates , and that @xmath261 noting that @xmath262 we have @xmath263 which together with and yields @xmath264 we need to bound the last term on the right - hand side of . notice that @xmath265 , by using a standard duality argument , see appendix , based on the helmholtz decomposition of @xmath266 , we can show that @xmath267 the desired estimate follows from combing and . finally , follows from @xmath268 , , the trace inequality , , and . the proof is complete . the @xmath269-conforming ndlec edge element ( of second type ) projection @xmath270 of @xmath12 is introduced and used in the proof to simplify the analysis at the expense of requiring @xmath131 to be a quasi - uniform and conforming mesh . we note that the proof is still valid if one replaces the @xmath269-conforming ndlec edge element projection by the @xmath269-ipdg projection without assuming @xmath131 is a quasi - uniform or conforming mesh . as expected , the new proof will be more complicated and technical , and is left for the interested reader to explore . the goal of this subsection is to derive error estimates for scheme . instead of using the well - known schatz argument @xcite , which is the ( only ) technique of choice for deriving error estimates for indefinite problems in the literature , we shall obtain our error estimates by exploiting the linearity of the maxwell equations and making strong use of the discrete stability estimates proved in theorem [ discrete_stability ] and the projection error estimates established in lemma [ lem5.3 ] . this new technique , which is adapted from @xcite , allows us to derive error estimates for @xmath271 without imposing any mesh constraint . it is easy to check that there holds the following error equation : @xmath272 let @xmath273 and @xmath274 , then @xmath275 . from we get @xmath276 the above equation implies that @xmath277 is the solution of scheme with the source functions @xmath278 and @xmath279 . hence , an application of theorem [ discrete_stability ] and lemma [ lem5.3 ] immediately infers the following estimate for @xmath280 . [ lem5.4 ] under the conditions of lemma [ lem5.1 ] , there holds @xmath281 where @xmath282 and @xmath283 is defined by . by lemmas [ lem5.3 ] and [ lem5.4 ] and the triangle inequality we then obtain the following main theorem of this section . [ main_theorem ] let @xmath12 and @xmath221 be the solutions to problem and scheme , respectively . assume @xmath182 . then , under the conditions of lemma [ lem5.1 ] , there hold the following error estimates : @xmath284 to bound @xmath285 in terms of the source functions @xmath286 and @xmath287 , we need to bound @xmath288 and @xmath289 by the source functions . to the end , we appeal to the solution estimate to get @xmath290 substituting into and yields the following explicit in all parameter error bounds for @xmath291 . [ cor_main ] suppose @xmath292 , and @xmath293 . under the assumptions of theorem [ main_theorem ] , there exist constants @xmath294 and @xmath295 independent of @xmath296 , and @xmath0 such that @xmath297 [ r5.3 ] ( a ) if @xmath298 and @xmath0 is in the pre - asymptotic range given by @xmath20 , then @xmath299 and the @xmath300-estimate becomes @xmath301 \(b ) for asymptotic error estimates we refer to @xcite . when @xmath302 is small , it is possible to improve the discrete stability estimates as well as the error estimates via the technique of stability - error iterative improvement from @xcite . throughout this section , we consider the following maxwell problem on the unit cube @xmath303 : @xmath304 where @xmath287 is so chosen that the exact solution is @xmath305 notice that we have chosen @xmath306 for simplicity . for any positive integer @xmath307 , let @xmath308 denote the cartesian mesh that consists of @xmath309 congruent cubes of edge length @xmath310 . we adopt the ipdg method using piecewise linear polynomials . we remark that the number of total dofs of the ipdg method on @xmath308 is @xmath311 which is the about twice of that of the corresponding conforming edge element method ( eem ) which uses piecewise trilinear polynomials . given a cartesian mesh @xmath131 , recall that @xmath221 denotes the ipdg solution . let @xmath312 denotes the trilinear conforming edge element approximation of the problem . in this subsection , we use the following penalty parameters in the ipdg method : @xmath313 we plot in figure [ fsta ] the following two ratios @xmath314 versus @xmath1 for @xmath315 with @xmath316 , respectively . it is shown that @xmath317 which is also implied by theorem [ discrete_stability ] and theorem [ stability ] . the @xmath318-norm of the edge element solution oscillates for @xmath1 near @xmath319 but is still bounded by @xmath320 . ( left ) and @xmath321 ( right ) versus @xmath1 for @xmath315 with @xmath316 , respectively.,title="fig : " ] ( left ) and @xmath321 ( right ) versus @xmath1 for @xmath315 with @xmath316 , respectively.,title="fig : " ] in this subsection , we use the same penalty parameters as given in . in the left graph of figure [ ferr1 ] , the relative @xmath318-error of the ipdg solution and the relative @xmath318-error of the edge element interpolant are displayed in one plot . when the mesh size is decreasing , the relative error of the ipdg solution stays around @xmath322 before it is less than @xmath322 , then decays slowly on a range increasing with @xmath1 , and then decays at a rate greater than @xmath323 in the log - log scale but converges as fast as the edge element interpolant ( with slope @xmath323 ) for small @xmath0 . the relative error grows with @xmath1 along line @xmath324 by contrast , as shown in the right of figure [ ferr1 ] , the relative error of the finite element solution first stay around @xmath322 but oscillates for large @xmath1 , then decays at a rate greater than @xmath323 in the log - log scale but converges as fast as the edge element interpolant ( with slope @xmath323 ) for small @xmath0 . the relative error of the edge element solution also grows with @xmath1 along line @xmath325 . -norm for @xmath326 and @xmath327 , respectively . the dashed line gives reference slope of @xmath323 . right graph : corresponding plots for edge element solutions.,title="fig : " ] -norm for @xmath326 and @xmath327 , respectively . the dashed line gives reference slope of @xmath323 . right graph : corresponding plots for edge element solutions.,title="fig : " ] unlike the error of the edge element interpolant , both the error of the ipdg solution and that of the edge element solution are not controlled by the magnitude of @xmath328 as indicated by the two graphs in figure [ ferr2 ] . it is shown that when @xmath0 is determined according to the rule of thumb " , the relative error of the ipdg solution keeps less than @xmath322 which means that the ipdg solution has some accuracy even for large @xmath1 , while the edge element solution is unusable for large @xmath1 . we remark that the accuracy of the ipdg solution can be further improved by tuning the penalty parameter @xmath329 , see subsection [ ssec-3 ] below . -norm computed for @xmath330 with mesh size @xmath0 determined by @xmath331.,title="fig : " ] -norm computed for @xmath330 with mesh size @xmath0 determined by @xmath331.,title="fig : " ] next we verify more precisely the pollution errors . to do so , we recall the definition of the critical mesh size with respect to a given relative tolerance ( cf . * definition 7.1 ) ) . given a relative tolerance @xmath332 and a wave number @xmath1 , the critical mesh size @xmath333 with respect to the relative tolerance @xmath332 is defined by the maximum mesh size such that the relative error of the ipdg solution ( or the edge element solution ) in @xmath318-norm is less than or equal to @xmath332 . it is clear that , if the pollution terms are of order @xmath334 , then @xmath333 should be proportional to @xmath335 for @xmath1 large enough . figure [ ferr3 ] which plots @xmath336 versus @xmath1 for the ipdg solution ( left ) with parameters given in and for the edge element solution ( right ) , respectively . they all decay at a rate of @xmath337 , just like the linear fem for the helmholtz problem ( cf . the results of this subsection indicate that both methods satisfy the following pre - asymptotic error bounds ( cf . remark [ r5.3](a ) ) : @xmath338 versus @xmath1 for the ipdg solution ( left ) with parameters given in and for the edge element solution ( right ) , respectively . the dotted lines give lines of slope @xmath339 in the log - log scale.,title="fig : " ] versus @xmath1 for the ipdg solution ( left ) with parameters given in and for the edge element solution ( right ) , respectively . the dotted lines give lines of slope @xmath339 in the log - log scale.,title="fig : " ] in this subsection , we show that appropriate choice of the penalty parameters can significantly reduce the pollution error of the ipdg method . we use the following parameters : @xmath340 we remark that @xmath341 is simply chosen from the set @xmath342 to minimize the relative error of the ipdg solution in @xmath318-norm with @xmath343 for wave number @xmath344 and mesh size @xmath345 . the optimal penalty parameter can also be obtained by the dispersion analysis ( cf . @xcite ) and will be considered in a future work . the relative error of the ipdg solution with parameters given in and the relative error of the edge element interpolant are displayed in the left graph of figure [ ferr12b ] . the ipdg method with parameters given in is much better than both the ipdg method using parameters given in and the eem ( cf . figure [ ferr1 ] and figure [ ferr2 ] ) . the relative error does not increase much with the change of @xmath1 along line @xmath325 for @xmath346 . but this does not mean that the pollution error has been eliminated . -norm for @xmath326 and @xmath327 , respectively . right graph : the relative error of the ipdg solution with parameters given in in @xmath318-norm computed for @xmath347 with mesh size @xmath0 determined by @xmath331.,title="fig : " ] -norm for @xmath326 and @xmath327 , respectively . right graph : the relative error of the ipdg solution with parameters given in in @xmath318-norm computed for @xmath347 with mesh size @xmath0 determined by @xmath331.,title="fig : " ] for more detailed observation , the relative error of the ipdg solution with parameters given in in @xmath318-norm computed for @xmath330 with mesh size @xmath0 determined by @xmath331 , are plotted in the right graph of figure [ ferr12b ] . it is shown that the pollution error is reduced significantly . figure [ ferr3b ] plots @xmath336 , the critical mesh size with respect to the relative tolerance @xmath348 , versus @xmath1 for the ipdg method with parameters given in . we recall that @xmath336 is the maximum mesh size such that the relative error of the ipdg solution in @xmath318-norm is less than or equal to @xmath348 . the decreasing rate of @xmath336 in the log - log scale is less than @xmath339 , which means that the pollution effect is reduced . versus @xmath1 for the ipdg method with parameters given in . the dotted line gives a line of slope @xmath339 in the log - log scale . ] for more detailed comparison between the continuous interior penalty finite element method ( cip - fem ) and the fem , we consider the problem with wave number @xmath349 . the real parts of @xmath350 with parameters given in ( left , solid ) , @xmath351 ( right , solid ) , and @xmath352 @xmath353 ( dotted ) with mesh sizes @xmath354 and @xmath355 are plotted in figure [ fz ] . here @xmath356 , @xmath357 , and @xmath358 are the @xmath359 components of the ipdg solution , the edge element solution , and the exact solution , respectively . the shape of the ipdg solution is roughly same as that of the exact solution for @xmath354 and matches very well for @xmath360 . while the edge element solution has a wrong shape for @xmath354 and @xmath361 and has a correct shape for @xmath360 but suffers an apparent phase error . with parameters given in ( left , solid ) , @xmath351 ( right , solid ) , and @xmath362 ( dotted ) for @xmath349 and @xmath354 , @xmath355 , respectively.,title="fig : " ] with parameters given in ( left , solid ) , @xmath351 ( right , solid ) , and @xmath362 ( dotted ) for @xmath349 and @xmath354 , @xmath355 , respectively.,title="fig : " ] with parameters given in ( left , solid ) , @xmath351 ( right , solid ) , and @xmath362 ( dotted ) for @xmath349 and @xmath354 , @xmath355 , respectively.,title="fig : " ] with parameters given in ( left , solid ) , @xmath351 ( right , solid ) , and @xmath362 ( dotted ) for @xmath349 and @xmath354 , @xmath355 , respectively.,title="fig : " ] table [ table ] shows the numbers of total dofs needed for @xmath363% relative errors in @xmath318-norm for the edge element interpolant , the ipdg solution with parameters given in , and the edge element solution , respectively . the ipdg method needs less dofs than the eem does for @xmath364 and much less for large wave number @xmath1 . .numbers of total dofs needed for 50% relative errors in @xmath318-norm for the edge element interpolant , the ipdg solution with parameters given in , and the edge element solution respectively . [ cols="<,>,>,>,>,>",options="header " , ] * acknowledgments . * the authors would like to thank dr . huangxin chen of xiamen university of china for his helpful suggestions on the construction and analysis of the @xmath228-elliptic projection in section [ sec-5.1 ] . 99 , _ dispersive properties of high order ndlec / edge element approximation of the time - harmonic maxwell equations _ a , 362 ( 2004 ) , pp . 471491 . i. m. babuka and s. a. sauter . is the pollution effect of the fem avoidable for the helmholtz equation considering high wave numbers ? , 42(3):451484 , 2000 . s. brenner and r. scott . the mathematical theory of finite element methods . springer - verlag , new york , 1994 . p. g. ciarlet . the finite element method for elliptic problems . north - holland , amsterdam , 1978 . b. cockburn , f. li and c .- w . shu . locally divergence - free discontinuous galerkin methods for the maxwell equations . , 194:588 - 610 , 2004 . d. l. colton and r. kress . . springer , new york , 1999 . p. cummings and x. feng . sharp regularity coefficient estimates for complex - valued acoustic and elastic helmholtz equations . , 16:139160 , 2006 . x. feng . wave number - explicit a priori estimates for the time - harmonic maxwell equations , preprint , july 15 , 2010 . x. feng and c. lorton . generalized inf - sup conditions and wave - number and domain - size explicit a priori estimates for the time - harmonic acoustic , elastic and electromagnetic wave equations , in preparation . x. feng and h. wu . discontinuous galerkin methods for the helmholtz equation with large wave number . , 47:28722896 , 2009 . x. feng and h. wu . @xmath365-discontinuous galerkin methods for the helmholtz equation with large wave number . , 80:9972024 , 2011 . gatica and s. meddahi . finite element analysis of a time harmonic maxwell problem with an impedance boundary condition . , 32:534552 , 2011 . r. hiptmair , a. moiola and i. perugia . stability results for the time - harmonic maxwell equations with impedance boundary conditions . 21:22632287 , 2011 . r. hiptmair , a. moiola and i. perugia . error analysis of trefftz - discontinuous galerkin methods for the time - harmonic maxwell equations 82:247268 , 2013 p. houston , i. perugia , a. schneebeli and d. schtzau . interior penalty method for the indefinite time - harmonic maxwell equations . , 100:485518 , 2005 . p. houston , i. perugia and d. schtzau . mixed discontinuous galerkin approximation of the maxwell operator . , 42:434459 , 2004 . p. monk . . oxford university press , new york , 2003 . ndlec . a new family of mixed finite elements in @xmath366 . , 50:5781 , 1986 . n. c. nguyena , j. perairea and b. cockburn . hybridizable discontinuous galerkin methods for the time - harmonic maxwell s equations . , 230:71517175 , 2011 . f. rellich . darstellung der eigenwerte von @xmath367 durch ein randintegral . , 46:635636 , 1940 . a. h. schatz . an observation concerning ritz galerkin methods with indefinite bilinear forms . , 28:959962 , 1974 . a. toselli and o. widlund , domain decomposition methods - algorithms and theory . springer , new york , 2005 . pre - asymptotic error analysis of cip - fem and fem for helmholtz equation with high wave number . part i : linear version . to appear . ( see also arxiv:1106.4079v1 ) . l. zhong , s. shu , g. wittum and j. xu . optimal error estimates for ndlec edge elements for time - harmonic maxwell s equations . , 27:563572 , 2009 . the proof follows the same lines as those given in @xcite and in @xcite . let @xmath368 be the @xmath300-conforming linear finite element space . it follows from that @xmath369 is discrete divergence - free , that is , @xmath370 notice that @xmath265 , we have the following discrete helmholtz decomposition of @xmath266 : @xmath371 where @xmath372 and @xmath373 is also discrete divergence - free . it is easy to check that @xmath374 then from ( * ? ? ? * lemma 7.6 ) and on noting that the domain @xmath49 is convex , there exists @xmath375 such that @xmath376 on @xmath377 and @xmath378 thus , it follows from the identity @xmath379 that @xmath380 to estimate @xmath382 , we appeal to a duality argument to be described next . let @xmath383 be the solution of the following auxiliary problem : @xmath384 noting that @xmath49 is convex , the above problem attains a unique solution @xmath385 and satisfies the following regularity estimate ( cf . @xcite ) @xmath386 define sesquilinear forms @xmath387 let @xmath388 and @xmath389 denote the edge finite element approximation and the ipdg approximation to @xmath383 , respectively , that is , @xmath390 it can be shown that there hold the following estimates ( cf . ): @xmath391 since @xmath392 on noting that @xmath393 , from and , we have @xmath394 on the other hand , @xmath395}\\ & = a({\boldsymbol{\phi}}_h^c,{\mathbf{z}}_h^c ) + \bigl({\mathbf{w}}-({\mathbf{w}}_h+{\nabla}r_h),{\mathbf{z}}_h^c\bigr)_{\omega}=a({\boldsymbol{\phi}}_h^c,{\mathbf{z}}_h^c)+({\mathbf{w}}-{\mathbf{w}}_h,{\mathbf{z}}_h^c)_{\omega}.\end{aligned}\ ] ] from the definitions of @xmath396 and @xmath397 , we get @xmath398 therefore @xmath399 since @xmath400 we have from lemma [ lem5.1 ] and the local trace inequality , @xmath401 moreover , from , , @xmath402 , and the local trace inequality , we get @xmath403 thus , it follows from , , , and the above estimate that @xmath404 then we obtain the following estimates for @xmath382 : @xmath405 which together with and implies that holds . the proof is complete .

we construct quasitoric manifolds of dimension 6 and higher which are not equivariantly homeomorphic to any toric origami manifold . all necessary topological definitions and combinatorial constructions are given and the statement is reformulated in discrete geometrical terms . the problem reduces to existence of planar triangulations with certain coloring and metric properties .

1409.6855

origami manifolds appeared in differential geometry recently as a generalization of symplectic manifolds @xcite . toric origami manifolds are in turn generalizations of symplectic toric manifolds . toric origami manifolds are a special class of @xmath0-dimensional compact manifolds with an effective action of a half - dimensional compact torus @xmath1 . in this paper we consider the following question . how large is this class ? which manifolds with half - dimensional torus actions are toric origami manifolds ? since the notion of a manifold with an effective half - dimensional torus action is too general to deal with , we restrict to quasitoric manifolds . this class of manifolds is large enough to include many interesting examples , and small enough to keep statements feasible . in @xcite masuda and park proved [ thm4dim ] any simply connected compact smooth @xmath2-manifold @xmath3 with an effective smooth action of @xmath4 is equivariantly diffeomorphic to a toric origami manifold . in particular , any @xmath2-dimensional quasitoric manifold is toric origami . the same question about higher dimensions was open . in this paper we give the negative answer . [ thmqtornotorigami ] for any @xmath5 there exist @xmath0-dimensional quasitoric manifolds , which are not equivariantly homeomorphic to any toric origami manifold . we will describe an obstruction for a quasitoric @xmath6-manifold to be toric origami and present a large series of examples , where such an obstruction appears . existence of such examples in higher dimensions follows from @xmath6-dimensional case . in spite of topological nature of the task , the proof is purely discrete geometrical : it relies on metric and coloring properties of planar graphs . thus we tried to separate the discussion of established facts in toric topology which motivated this study , from the proof of the main theorem to keep things comprehensible for the broad audience . the paper is organized as follows . in section [ secttopology ] we briefly review the necessary topological objects , and describe the standard combinatorial and geometrical models which are used to classify them . the objects are : quasitoric manifolds , symplectic toric manifolds , and toric origami manifolds . the corresponding combinatorial models are : characteristic pairs , delzant polytopes , and origami templates respectively . in section [ sectweightedspheres ] we introduce the notion of a weighted simplicial cell sphere , which , in a certain sense , unifies all these combinatorial models . we define a connected sum of weighted spheres along vertices . this operation is dual to the operation of producing an origami template from delzant polytopes . it plays an important role in the proof . section [ sectproof ] contains the combinatorial statement from which follows theorem [ thmqtornotorigami ] , and the proof of this statement . the interaction of our study with the study of the brownian map allows to prove that asymptotically most simple 3-polytopes admit quasitoric manifolds which are not toric origami . we describe this interaction as well as other adjacent questions in the last section [ sectconclusion ] . authors are grateful to the anonymous referee for his comments on the previous version of the paper . the subject of this subsection originally appeared in the seminal work of davis and januszkiewicz @xcite . the modern exposition and technical details can be found in ( * ? ? ? * ch.7 ) . let @xmath1 be a compact @xmath7-dimensional torus . the standard representation of @xmath1 is a representation @xmath8 by coordinate - wise rotations , i.e. @xmath9 for @xmath10 , @xmath11 . the action of @xmath1 on a smooth manifold @xmath12 is called locally standard , if @xmath3 has an atlas of standard charts , each isomorphic to a subset of the standard representation . more precisely , a standard chart on @xmath3 is a triple @xmath13 , where @xmath14 is a @xmath1-invariant open subset , @xmath15 is an automorphism of @xmath1 , and @xmath16 is a @xmath15-equivariant homeomorphism @xmath17 onto a @xmath1-invariant open subset @xmath18 ( i.e. @xmath19 for all @xmath20 , @xmath21 ) . in the following @xmath3 is supposed to be compact . since the orbit space @xmath22 of the standard representation is a nonnegative cone @xmath23 , the orbit space of any locally standard action obtains the structure of a compact smooth manifold with corners . recall that a manifold with corners is a topological space locally modeled by open subsets of @xmath24 with the combinatorial stratification induced from the face structure of @xmath24 . there are many technical details about the formal definition which we left beyond the scope of this work ( the exposition relevant to our study can be found in @xcite or @xcite ) . the orbit space @xmath25 of a locally standard action carries an additional information about stabilizers of the action , called a characteristic function . let @xmath26 denote the set of facets of @xmath27 ( i.e. faces of codimension @xmath28 ) . for each facet @xmath29 of @xmath27 consider a stabilizer subgroup @xmath30 of points in the interior of @xmath29 . this subgroup is @xmath28-dimensional and connected , thus it has the form @xmath31 , for some primitive integral vector @xmath32 , defined uniquely up to a common sign . thus , a primitive integral vector ( up to sign ) @xmath33 is associated with any facet @xmath29 of @xmath27 . this map @xmath34 is called a _ characteristic function _ ( or a characteristic map ) . it satisfies the following so called -condition : @xmath35{10cm}{\center { if facets $ f_1,\ldots , f_s$ intersect , then the vectors $ \lambda(f_1),\ldots,\lambda(f_s)$ span a direct summand of $ { \mathbb{z}}^n$.}}\tag{$*$}\ ] ] here we actually take not a class @xmath36 , but one of its two particular representatives in @xmath37 . obviously , the condition does not depend on the choice of sign , thus is well defined . the same convention appears further in the text without special mention . it is convenient to view the characteristic function @xmath38 on @xmath25 as a generalized coloring of facets . we assign primitive integral vectors to facets instead of simple colors , and condition is the requirement for this `` coloring '' to be proper . the general idea , which simplifies many considerations in toric topology is that the combinatorial structure of the orbit space @xmath27 together with the assigned coloring completely encodes the equivariant homeomorphism type of @xmath3 in many cases . the precise statement also involves the so called euler class of the action , which is an element of @xmath39 , and allows to classify all compact smooth manifolds with locally standard torus actions . the reader may find this general statement in @xcite . anyway , we will work with only a special type of locally standard torus actions , namely quasitoric manifolds . in this special case euler class vanishes , so we will not care about it . [ definqtoric ] a manifold @xmath12 with a locally standard action of @xmath1 is called quasitoric , if the orbit space @xmath40 is homeomorphic to a simple polytope as a manifold with corners . recall that a convex polytope @xmath41 of dimension @xmath7 is called simple if any of its vertices lies in exactly @xmath7 facets . in other words , a simple polytope is a polytope which is at the same time a manifold with corners . considering manifolds with corners , simple polytopes are the simplest geometrical examples one can imagine . this makes the definition of quasitoric manifold very natural . let @xmath41 be a simple polytope and @xmath38 be a characteristic function , i.e. any map @xmath42 satisfying -condition . the pair @xmath43 is called a _ characteristic pair_. according to @xcite , there is a one - to - one correspondence @xmath44 up to equivariant homeomorphism on the left - hand side and combinatorial equivalence on the right - hand side . given a characteristic pair , one can construct the corresponding quasitoric manifold explicitly . [ conmodelqtoric ] let @xmath43 be a characteristic pair , @xmath45 . consider a topological space @xmath46 the equivalence @xmath47 is generated by relations @xmath48 where @xmath49 lies in a facet @xmath50 and @xmath51 . the torus @xmath1 acts on @xmath52 by rotating second coordinate and the orbit space @xmath53 is isomorphic to @xmath41 . condition implies that the action is locally standard . there is a smooth structure on @xmath52 and the action of @xmath1 is smooth ( the construction of smooth structure can be found in @xcite ) . therefore , @xmath52 is a quasitoric manifold . let @xmath54 denote the projection to the orbit space @xmath55 . each facet @xmath50 determines a smooth submanifold @xmath56 of dimension @xmath57 , called characteristic submanifold . on its own , the manifold @xmath58 is again a quasitoric manifold with the orbit space @xmath29 . in the following subsections we recall the definitions and properties of toric origami manifolds and origami templates . more detailed exposition of this theory can be found in @xcite , @xcite or @xcite . a _ folded symplectic form _ on a @xmath0-dimensional smooth manifold @xmath3 is a closed @xmath59-form @xmath60 such that * its top power @xmath61 is transversall to the zero section of @xmath62 . as a consequence @xmath61 vanishes on a smooth submanifold @xmath63 of codimension 1 . * the restriction of @xmath60 to @xmath64 has maximal rank . the hypersurface @xmath64 where @xmath60 is degenerate is called the _ fold_. the pair @xmath65 is called a _ folded symplectic manifold_. if @xmath64 is empty , @xmath60 is a genuine symplectic form and @xmath65 is a genuine symplectic manifold according to classical definition . the reader may get a feeling of this notion by working locally . darboux s theorem says that any symplectic form can be written locally as @xmath66 in appropriate coordinates . the folded forms are exactly the forms written as @xmath67 in appropriate coordinates ( for this analogue of darboux s theorem see @xcite and references therein ) . the fold @xmath64 is thus a hypersurface given locally by @xmath68 . since the restriction of @xmath60 to @xmath64 has maximal rank , it has a one - dimensional kernel at each point of @xmath64 . this determines a line field on @xmath64 called the _ null foliation_. if the null foliation is the vertical bundle of some principal @xmath69-fibration @xmath70 over a compact base @xmath71 , then the folded symplectic form @xmath60 is called an _ origami form _ and the pair @xmath72 is called an _ origami manifold_. the action of a torus @xmath73 ( of any dimension ) on an origami manifold @xmath72 is called hamiltonian if it admits a moment map @xmath74 to the dual lie algebra of the torus , which satisfies the conditions : ( 1 ) @xmath75 is equivariant with respect to the given action of @xmath73 on @xmath3 and the coadjoint action of @xmath73 on the vector space @xmath76 ( by commutativity of torus this action is trivial ) ; ( 2 ) @xmath75 collects hamiltonian functions , that is , @xmath77 , where @xmath78 is the function on @xmath3 , taking the value @xmath79 at a point @xmath80 , @xmath81 is a vector flow on @xmath3 , generated by @xmath82 , and @xmath83 is its dual 1-form . a toric origami manifold @xmath84 , abbreviated as @xmath3 , is a compact connected origami manifold @xmath72 equipped with an effective hamiltonian action of a torus @xmath73 with @xmath85 and with a choice of a corresponding moment map @xmath75 . when the fold @xmath64 is empty , a toric origami manifold is a _ symplectic toric manifold_. in this case the image @xmath86 of the moment map is a delzant polytope in @xmath76 , and the map @xmath87 itself can be identified with the map to the orbit space @xmath88 . a classical theorem of delzant @xcite says that symplectic toric manifolds are classified by the images of their moment maps in @xmath89 . in other words , there is a one - to - one correspondence @xmath90 up to equivariant symplectomorphism on the left - hand side , and affine equivalence on the right - hand side . let us recall the notion of delzant polytope . a simple convex polytope @xmath91 is called delzant , if its normal fan is smooth ( with respect to a given lattice @xmath92 ) . in other words , all normal vectors to facets of @xmath41 have rational coordinates , and , whenever facets @xmath93 meet in a vertex of @xmath41 , the primitive normal vectors @xmath94 form a basis of the lattice @xmath37 . let @xmath95 be the symplectic toric manifold corresponding to delzant polytope @xmath41 . we do not need the construction of delzant correspondence in full generality , but we need to review the topological construction of symplectic toric manifold corresponding to a given delzant polytope . forgetting the symplectic structure , any symplectic toric manifold , as a smooth manifold with @xmath1-action , is a quasitoric manifold . [ conmodelsymplectic ] let @xmath41 be a delzant polytope in @xmath96 . for a facet @xmath50 consider its outward primitive normal vector @xmath97 . consider the corresponding vector modulo sign : @xmath98 . by the definition of delzant polytope , @xmath99 satisfies , thus provides an example of a characteristic function . the manifold @xmath95 is equivariantly diffeomorphic to quasitoric manifold @xmath52 corresponding to the characteristic pair @xmath43 . delzant theorem provides a one - to - one correspondence between symplectic toric manifolds and delzant polytopes . to generalize this correspondence to toric origami manifolds we need a notion of an origami template , which we review next . let @xmath100 denote the set of all ( full - dimensional ) delzant polytopes in @xmath96 ( w.r.t . a given lattice ) and @xmath101 the set of all their facets . [ definorigamitemplate ] an origami template is a triple @xmath102 , where * @xmath103 is a connected finite graph ( loops and multiple edges are allowed ) with the vertex set @xmath104 and edge set @xmath105 ; * @xmath106 is a map , which associates to any vertex of @xmath103 a full - dimensional delzant polytope , @xmath107 ; * @xmath108 is a map , which associates to any edge of @xmath103 a facet of delzant polytope , @xmath109 ; subject to the following conditions : 1 . if @xmath110 is an edge of @xmath103 with endpoints @xmath111 , then @xmath112 is a facet of both polytopes @xmath113 and @xmath114 , and these polytopes coincide near @xmath112 ( this means there exists an open neighborhood @xmath115 of @xmath112 in @xmath96 such that @xmath116 ) . 2 . if @xmath117 are two edges of @xmath103 adjacent to @xmath118 , then @xmath119 and @xmath120 are disjoint facets of @xmath121 . the facets of the form @xmath112 for @xmath110 are called the fold facets of the origami template . for convenience in the following we call the vertices of graph @xmath103 the nodes . one can simply view an origami template as a collection of ( possibly overlapping ) delzant polytopes @xmath122 in the same ambient space , with some gluing data , encoded by a template graph @xmath103 . when @xmath123 , the picture looks like a folded sheet of paper on a flat plane , which is one of the explanations for the term `` origami '' ( see fig . [ fig : templates ] ) . nevertheless , to avoid the confusion , we should mention that most flat origami models in a common sense are not origami templates in the sense of definition [ definorigamitemplate ] . ( 0,-0.05)(1,0)(0,1)cycle ; ( -0.05,0)(1,0)(0,1)cycle ; ( 1,0)(0,1 ) ; ( 1,0)(2,-0.1)(2,5.9)(1,6)(0,5)(0,1)cycle ; ( 4,-0.1)(5,0)(6,1)(6,5)(5,6)(4,5.9)cycle ; ( -0.1,2)(0,1)(1,0)(5,0)(6,1)(5.9,2)cycle ; ( -0.1,4)(5.9,4)(6,5)(5,6)(1,6)(0,5)cycle ; ( 1,0)(0,1 ) ; ( 5,0)(6,1 ) ; ( 6,5)(5,6 ) ; ( 0,5)(1,6 ) ; the following is a generalization of delzant s theorem to toric origami manifolds . [ theo : classiforigami ] assigning the moment data of a toric origami manifold induces a one - to - one correspondence @xmath124 up to equivariant origami symplectomorphism on the left - hand side , and affine equivalence on the right - hand side . let @xmath125 be the toric origami manifold corresponding to origami template @xmath126 . as in the case of symplectic toric manifolds , we do not need the construction of this correspondence in full generality . but we give a topological construction of the toric origami manifold from a given origami template . [ conmodelorigami ] consider an origami template @xmath127 , @xmath128 . for each node @xmath118 the delzant polytope @xmath129 gives rise to a symplectic toric manifold @xmath130 , see construction [ conmodelsymplectic ] . now do the following procedure : 1 . take a disjoint union of all manifolds @xmath130 for @xmath131 ; 2 . for each edge @xmath110 with distinct endpoints @xmath132 and @xmath133 take an equivariant connected sum of @xmath134 and @xmath135 along the characteristic submanifold @xmath136 ( which is embedded in both manifolds by construction [ conmodelqtoric ] ) ; 3 . for each loop @xmath110 based at @xmath118 take a real blow up of normal bundle to the submanifold @xmath136 inside @xmath130 . step 2 makes sense because of pt.1 of definition [ definorigamitemplate ] . indeed , the polytopes @xmath113 and @xmath114 agree near @xmath112 , thus @xmath134 and @xmath135 have equivariantly homeomorphic neighborhoods around @xmath136 , so the connected sum is well defined . 2 of definition [ definorigamitemplate ] ensures that surgeries do not touch each other , so all the connected sums and blow ups can be taken simultaneously . the smooth manifold obtained by the above procedure is the origami manifold @xmath137 up to equivariant diffeomorphism . by definition , the operation of equivariant connected sum consists in cutting small equal @xmath1-invariant tubular neighborhoods of @xmath138 in @xmath139 and @xmath140 , and then gluing the resulting manifolds by identity isomorphism of the boundaries . the image of the moment map under this operation becomes smaller . thus the construction described above is certainly not enough to prove the classificational theorem . in the theorem one should not only take a connected sum but also attach collars of the form @xmath141 ( see details in @xcite ) . nevertheless , both constructions , with collars and without collars , lead to the same result , up to equivariant diffeomorphism . [ exsfour ] let us construct a toric origami manifold @xmath142 , corresponding to the origami template , made of two triangles ( fig . [ fig : templates ] , left ) . the symplectic toric 4-manifold corresponding to a triangle is known to be the complex projective plane @xmath143 . the characteristic submanifold corresponding to the fold facet is a projective line @xmath144 . thus , @xmath142 is a connected sum of two copies of @xmath143 along the line @xmath145 , which lies in both . this has a simple geometrical interpretation . if we consider @xmath144 as a projective line at infinity , and denote the tubular neighborhood of this line by @xmath146 , then @xmath147 is a 4-disk @xmath148 . thus @xmath142 is a result of gluing two copies of @xmath148 by the identity diffeomorphism of the boundary . thus @xmath149 . the action of @xmath4 on @xmath150 is also easily described . consider the space @xmath151 , and let @xmath4 act on @xmath152 by coordinate - wise rotations , and trivially on @xmath153 . the unit sphere @xmath154 is invariant under this action . this gives a required action of @xmath4 on @xmath149 . an origami template @xmath155 is called orientable if the template graph @xmath103 is bipartite , or , equivalently , @xmath59-colorable . it is not hard to prove that the origami template @xmath156 is orientable whenever @xmath157 is an orientable manifold @xcite . an origami template @xmath155 ( and the corresponding manifold @xmath157 ) is called corientable if @xmath103 has no loops ( i.e. edges based at one point ) . any orientable template ( resp toric origami manifold ) is corientable , because a graph with loops is not @xmath59-colorable . if @xmath157 is corientable , then the action of @xmath1 on @xmath157 is locally standard ( * ? ? ? * lemma 5.1 ) . the converse is also true . if the template graph has a loop , then the real normal blow up in step 3 of construction [ conmodelorigami ] implies existence of @xmath158-components in stabilizer subgroups . therefore non - corientable toric origami manifolds are not locally standard . in the following we consider only corientable templates and toric origami manifolds . [ conorbitoforigami ] the orbit space @xmath159 of a ( corientable ) toric origami manifold is a smooth manifold with corners . its homeomorphism type can be described as a topological space obtained by gluing polytopes @xmath121 along fold facets . more precisely , @xmath160 where @xmath161 if there exists an edge @xmath162 with endpoints @xmath163 and @xmath164 , and @xmath165 . facets of @xmath27 are given by non - fold facets of polytopes @xmath121 identified in the same way . to make this precise , let us call non - fold facets @xmath166 and @xmath167 elementary neighboring w.r.t . to the edge @xmath110 ( with endpoints @xmath132 and @xmath133 ) if @xmath168 . the relation of elementary neighborliness generates an equivalence relation @xmath169 on the set of all non - fold facets of all polytopes @xmath121 . define the facet @xmath170 $ ] of the orbit space @xmath27 as a union of facets in one equivalence class : @xmath171 \quad{\stackrel{\text{\tiny def}}{=}}\quad \bigsqcup_{\mathclap{\substack{v\in v , g\in{\mathcal{f}}({\psi_v}(v)),\\ g \text { is not fold } , g\leftrightarrow f}}}\,\ , ( v , g ) { \scalerel*{/}{\sum}}\sim,\qquad [ f ] \in { \mathcal{f}}(q),\ ] ] where @xmath47 is the same as in . let us define a primitive normal vector to the facet @xmath170 $ ] of @xmath27 by @xmath172){\stackrel{\text{\tiny def}}{=}}\nu(f)\in{\mathbb{z}}^n/\pm$ ] . it is well defined since @xmath173 for @xmath174 . note that the relation of elementary neighborliness determines a connected subgraph @xmath175}$ ] of @xmath103 . all facets @xmath176 are delzant and lie in the same hyperplane @xmath177}$ ] . thus we obtain an induced origami template @xmath178}=(\gamma_{[f]},{\psi_v}|_{\gamma_{[f]}}\cap h_{[f]},{\psi_e}|_{\gamma_{[f]}}\cap h_{[f]})\ ] ] of dimension @xmath179 . in particular , if @xmath180 denotes the projection to the orbit space , then the characteristic submanifold @xmath181)$ ] is the toric origami manifold of dimension @xmath57 generated by the origami template @xmath182}$ ] . we had defined the facets of the orbit space @xmath183 . all other faces are defined as connected components of nonempty intersections of facets . on the other hand , faces can be defined similarly to facets by gluing faces of polytopes @xmath121 which are neighborly in the same sense as before . extending the origami analogy , we can think of the orbit space @xmath27 as `` unfolding '' the origami template and then forgetting the angles adjacent to the former fold facets ( remember that we have to identify neighboring faces ! ) . ( 0,0)(2,0)(2,2)(0,2)cycle ; ( 0,2)(2,0 ) ; at ( 2.5,1 ) @xmath184 ; ( 3,0) .. controls ( 4.5,0 ) and ( 5,0.5) .. (5,2) .. controls ( 3.5,2 ) and ( 3,1.5) .. (3,0)cycle ; it is easy to see that the orbit space @xmath159 has the same homotopy type as the graph @xmath103 , thus @xmath27 is either contractible ( when @xmath103 is a tree ) or homotopy equivalent to a wedge of circles . this observation shows that whenever the template graph @xmath103 has cycles , the corresponding toric origami manifold can not be quasitoric ( recall that the orbit space of quasitoric manifold is a polytope , which is contractible ) . as an example , the origami template shown on fig . [ fig : templates ] , at the right corresponds to the origami manifold which is not quasitoric . since we want to find a quasitoric manifold which is not toric origami , we need to consider only the cases when the orbit space is contractible . thus in the following we suppose @xmath103 is a tree . in the previous section we have seen that quasitoric manifolds are encoded by the orbit spaces ( which are simple polytopes ) and characteristic functions ( which are colorings of facets by elements of @xmath185 ) . it will be easier , however , to work with the dual objects , which we call weighted simplicial spheres . to some extent this approach is equivalent to multi - fans , used to study origami manifolds in @xcite , but it is more suitable for our geometrical considerations . recall that a _ simplicial poset _ or _ simplicial cell complex _ @xcite is a finite partially ordered set @xmath186 such that : \(1 ) there is a unique minimal element @xmath187 , \(2 ) for each @xmath188 the interval subset @xmath189{\stackrel{\text{\tiny def}}{=}}\{j\in s\mid j\leqslant i\}$ ] is isomorphic to the poset of faces of @xmath190-dimensional simplex ( i.e. boolean lattice of rank @xmath191 ) for some @xmath192 . in this case the element @xmath193 is said to have rank @xmath191 and dimension @xmath194 . the elements of @xmath186 are called simplices and elements of rank @xmath28 are called vertices . the set of vertices of @xmath186 is denoted @xmath195 . a simplicial poset is called pure , if all maximal simplices have the same dimension . a simplicial poset @xmath186 is called a simplicial complex , if for any subset of vertices @xmath196 , there exists at most one simplex whose vertex set is @xmath197 . it is convenient to visualize simplicial posets using their geometrical realizations . to define the geometrical realization we assign the geometrical simplex @xmath198 of dimension @xmath199 to each @xmath188 and attach them together according to the order relation in @xmath186 . more formally , the geometric realization of @xmath186 is the topological space @xmath200 where @xmath201 if @xmath202 and @xmath203 . see details in @xcite or @xcite . a simplicial poset @xmath186 is called a _ simplicial cell sphere _ if @xmath204 is homeomorphic to a sphere . @xmath186 is called a pl - sphere if it is pl - homeomorphic to the boundary of a simplex . in dimension 2 , which is the most important case for us , these two notions are equivalent . simplicial complex , whose geometric realization is homeomorphic to a sphere , is called a _ simplicial sphere_. thus simplicial sphere is a simplicial cell sphere which is also a simplicial complex . [ conlinkstarsumposets ] we want to define a connected sum of two simplicial cell spheres along their vertices . the topological meaning of this operation is clear : cut the small open neighborhoods of vertices and attach the boundaries if possible . however , an attempt to define the connected sum combinatorially for the most general simplicial posets leads to some technical problems . to keep things manageable , we exclude certain degenerate situations . for every @xmath205 in @xmath186 there is a complementary simplex @xmath206 , since the interval @xmath207 $ ] is identified with the boolean lattice . in other words , @xmath208 is the face of @xmath209 complementary to the face @xmath193 . link _ of a simplex @xmath188 as a partially ordered set @xmath210 with the order relation induced from @xmath186 . define an _ open star _ of a simplex @xmath188 as a subset @xmath211 . there is a natural surjective map of sets @xmath212 sending @xmath209 to @xmath208 . we call a simplex @xmath193 _ admissible _ if @xmath213 is injective . ( 0,2) .. controls ( 0,0.5 ) and ( 0.5,0) .. (2,0) .. controls ( 2,1.5 ) and ( 1.5,2) .. (0,2)cycle ; ( 2,0 ) circle(2pt ) node[anchor = west ] @xmath214 ; ( 0,2 ) circle(2pt ) node[anchor = east ] @xmath215 ; note that in a simplicial complex every simplex is admissible . an example of non - admissible simplex is shown on fig . [ fig : nonadmiss ] . there are two simplices containing the vertex @xmath214 , and the complement of @xmath214 in both of them is the same vertex @xmath215 . thus @xmath214 is a non - admissible vertex . let us define the connected sum of two simplicial posets @xmath216 and @xmath217 along admissible vertices . let @xmath218 and @xmath219 be admissible vertices , and suppose there exists an isomorphism of posets @xmath220 ( thus an isomorphism of open stars , by admissibility ) . consider a poset @xmath221 where @xmath222 is identified with @xmath223 whenever @xmath224 . the order relation on @xmath225 is induced from @xmath216 and @xmath217 in a natural way . it can be easily checked that the connected sum @xmath225 is again a simplicial poset . if @xmath226 are simplicial spheres , then so is @xmath225 . this statement would fail if we do not impose the admissibility condition . a connected sum of two simplicial complexes may not be a simplicial complex ( fig . [ fig : consumtriangles ] ) . this is the main reason why we consider a class of simplicial posets instead of simplicial complexes . ( 0,0)(1.5,1)(0,2)cycle ; ( 2.5,1)(4,0)(4,2)cycle ; ( 5.5,0) .. controls ( 5,0.7 ) and ( 5,1.3) .. (5.5,2) .. controls ( 6,1.3 ) and ( 6,0.7) .. (5.5,0)cycle ; at ( 2,1 ) @xmath227 ; at ( 4.5,1 ) @xmath184 ; ( 0,0 ) circle(2pt ) ; ( 0,2 ) circle(2pt ) ; ( 1.5,1 ) circle(2pt ) ; ( 2.5,1 ) circle(2pt ) ; ( 4,0 ) circle(2pt ) ; ( 4,2 ) circle(2pt ) ; ( 5.5,0 ) circle(2pt ) ; ( 5.5,2 ) circle(2pt ) ; let @xmath186 be a pure simplicial poset of dimension @xmath179 . a map @xmath228 is called a characteristic function if , for every simplex @xmath188 with vertices @xmath229 , the vectors @xmath230 span @xmath37 . the pair @xmath231 is called a weighted simplicial poset . let @xmath232 and @xmath233 be weighted simplicial posets . let @xmath234 be admissible vertices of @xmath235 such that there exists an isomorphism @xmath236 preserving characteristic functions : @xmath237 . then @xmath238 induce the characteristic function @xmath38 on the connected sum @xmath225 . the weighted simplicial poset @xmath239 is called a weighted connected sum of @xmath232 and @xmath233 . let @xmath43 be a characteristic pair ( see section [ secttopology ] ) . let @xmath240 be the dual simplicial sphere to a simple polytope @xmath41 . since there is a natural correspondence @xmath241 we get the characteristic function @xmath242 . this defines a weighted sphere @xmath243 . in particular , any delzant polytope @xmath41 defines a weighted sphere @xmath244 , where @xmath245 is the normal vector to @xmath246 modulo sign ( construction [ conmodelsymplectic ] ) . [ conaugposetoforigami ] let @xmath127 be an origami template and @xmath157 be the corresponding toric origami manifold . suppose that @xmath103 is a tree . the orbit space @xmath183 is homeomorphic to an @xmath7-dimensional disc . the face structure of @xmath27 defines a poset @xmath247 , whose elements are faces of @xmath27 ordered by reversed inclusion ( it is easy to show that such poset is simplicial ) . in particular , normal vectors to facets of @xmath27 ( construction [ conorbitoforigami ] ) determine the characteristic function @xmath249 , @xmath172)=\nu(f)$ ] . thus there is a weighted simplicial poset @xmath250 associated with a toric origami manifold @xmath251 . our next goal is to describe the weighted simplicial cell sphere @xmath250 of a toric origami manifold as a connected sum of elementary pieces , corresponding to delzant polytopes of the origami template . [ conaugposetoforigamiascsum ] if @xmath103 is a tree , then the simplicial poset @xmath247 is the connected sum of simplicial spheres @xmath252 along vertices , corresponding to fold facets : @xmath253 let us introduce some notation to make this precise . let @xmath162 be an edge of @xmath103 , and @xmath164 be its endpoint . let @xmath254 be the vertex of @xmath252 corresponding to the facet @xmath255 . then denotes the connected sum of all simplicial spheres @xmath252 along vertices @xmath254 , @xmath256 for all edges @xmath257 of graph @xmath103 . this simultaneous connected sum is well defined . indeed , if @xmath258 are two edges emanating from @xmath131 , then the vertices @xmath259 and @xmath260 are not adjacent in @xmath252 by pt.2 of definition [ definorigamitemplate ] . therefore , open stars @xmath261 and @xmath262 , which we remove in , do not intersect . also note that all vertices @xmath254 are admissible , since the spheres @xmath252 are simplicial complexes . each sphere @xmath252 comes equipped with a characteristic function @xmath263 , since @xmath121 is delzant . by pt.1 of definition [ definorigamitemplate ] these characteristic functions agree on the links which we identify . therefore we have an isomorphism of weighted spheres @xmath264 suppose that a quasitoric manifold @xmath52 is equivariantly homeomorphic to the origami manifold @xmath265 . as was mentioned earlier , in this situation @xmath103 is a tree . first , the orbit spaces should be isomorphic as manifolds with corners : @xmath266 . second , @xmath267 implies that stabilizers of the torus actions coincide for the corresponding faces of orbit spaces . thus characteristic functions on @xmath41 and @xmath27 taking values in @xmath185 are the same . hence , the weighted simplicial cell spheres @xmath243 and @xmath268 are isomorphic . so far to prove theorem [ thmqtornotorigami ] for @xmath269 it is sufficient to prove the following statement . [ propthmreform ] there exists a 3-dimensional simple polytope @xmath41 and a characteristic function @xmath270 such that the dual weighted sphere @xmath243 can not be represented as a connected sum , along a tree , of weighted spheres dual to delzant polytopes . the proof of this proposition takes most part of this section . we proceed by steps . at first notice that any simplicial 2-sphere is dual to some simple 3-polytope by steinitz s theorem ( see e.g. @xcite ) . so it is sufficient to prove that there exists a weighted 2-dimensional simplicial sphere @xmath271 which can not be represented as a connected sum of weighted spheres dual to delzant polytopes . [ conslicingregions ] we introduce some notation in addition to that of construction [ conaugposetoforigamiascsum ] , see fig . [ pictconsumgraph ] . as before , let @xmath128 be a tree . suppose that a simplicial cell @xmath272-sphere @xmath273 is associated with each node @xmath118 , and for each edge @xmath110 with an endpoint @xmath118 there is an admissible vertex @xmath274 subject to the following conditions : ( 1 ) @xmath275 is isomorphic to @xmath276 for any edge @xmath162 with endpoints @xmath277 ; ( 2 ) vertices @xmath278 are different and not adjacent in @xmath273 for any two edges @xmath279 emanating from @xmath164 . then we can form a connected sum along @xmath103 as in construction [ conaugposetoforigamiascsum ] : @xmath280 . for each @xmath118 consider the simplicial subposet @xmath281 this subposet will be called a region . denote @xmath275 by @xmath282 . by construction , @xmath282 is attached to @xmath283 if @xmath257 . the resulting @xmath284-dimensional simplicial subposet of @xmath271 is denoted by @xmath285 . since @xmath254 is admissible , the subposet @xmath286 is a homological @xmath284-sphere as follows from a standard argument in combinatorial topology ( see , e.g. , ( * ? ? ? * prop.2.2.14 ) ) . we get a collection of @xmath284-dimensional cycles @xmath287 , dividing the @xmath272-sphere @xmath271 into regions @xmath288 . if @xmath257 , then @xmath289 and @xmath290 share a common border @xmath285 . note that cycles @xmath285 are mutually ordered , meaning that each @xmath285 lies at one side of any other cycle . though the cycles may have common points ( as schematically shown on fig . [ pictconsumgraph ] ) and even coincide ( in this case the region between them coincides with both of them ) . on the other hand , any collection of mutually ordered @xmath284-dimensional spherical cycles in @xmath271 determines the representation of @xmath271 as a connected sum of smaller simplicial cell spheres . a representation @xmath280 will be called a _ slicing_. define the _ width _ of a slicing @xmath291 to be the maximal number of vertices in its regions : @xmath292 define the _ fatness _ of a sphere @xmath271 as the minimal width of all its possible slicings : @xmath293 the essential idea in the proof of proposition [ propthmreform ] is the following . [ lembigfatsmalcol ] let @xmath271 be an @xmath272-dimensional simplicial cell sphere and @xmath294 a characteristic function . let @xmath295 denote the number of different values of this characteristic function , @xmath296 . suppose that @xmath297 . then @xmath298 can not be represented as a connected sum , along a tree , of simplicial spheres dual to delzant polytopes . assume the converse . then @xmath299 , where @xmath121 are delzant polytopes . forgetting characteristic functions gives a slicing @xmath291 of @xmath271 . the width of every slicing of @xmath271 is greater than @xmath300 by the definition of fatness . in particular , thus there exists a node @xmath164 of @xmath103 such that @xmath302 . the region @xmath289 is a subcomplex of @xmath252 . the restriction of @xmath38 to the subset @xmath303 coincides with the restriction of @xmath304 to @xmath303 . recall , that @xmath97 is the outward normal vector to the facet @xmath305 , and @xmath306 is its class modulo sign . the outward normal vectors to facets of a convex polytope are mutually distinct , thus @xmath307 and , therefore , @xmath308 . thus @xmath309 , the contradiction , since @xmath295 is the total number of values of @xmath38 . so far we may find counterexamples to origami realizability among polytopes , which are @xmath37-colored with a small number of colors , but whose dual simplicial spheres have large fatness . of course such examples do not appear when @xmath123 this would contradict theorem [ thm4dim ] . a simplicial 1-sphere is a cycle graph @xmath310 . by considering diagonal triangulations of a @xmath191-gon , one can easily check that @xmath310 can be represented as a connected sum of several cycle graphs of the form @xmath311 or @xmath312 , giving the slicing of width @xmath313 . hence fatness of any @xmath28-dimensional simplicial sphere is at most @xmath313 , while any characteristic function takes at least @xmath59 values , so the conditions of lemma [ lembigfatsmalcol ] are not satisfied if @xmath123 . the existence of 2-spheres satisfying conditions of lemma [ lembigfatsmalcol ] is thus our next and primary goal . at first , we prove that any @xmath59-sphere admits a characteristic function with few values . [ lemfourcolor ] any simplicial 2-sphere @xmath271 admits a characteristic function @xmath314 such that @xmath315 . four color theorem states that there exists a proper vertex - coloring : @xmath316 . now replace colors by integral vectors @xmath317 , @xmath318 , @xmath319 , @xmath320 . any three of these four vectors span the lattice . therefore the map @xmath321 thus obtained is a characteristic function . it takes at most @xmath2 values . this is a standard trick in toric topology . classically , it is applied to prove that any simple 3-polytope admits a quasitoric manifold @xcite . though for our purpose we just need @xmath59-spheres @xmath271 with @xmath322 , it seems intuitively clear that in dimension @xmath59 and higher there exist spheres of arbitrarily large fatness . but it is not a priori clear how to describe such spheres explicitly in combinatorial terms . we present one possible approach below , but some steps of our construction do not generalize to dimensions greater than @xmath59 . [ propexistfatpoly ] for any @xmath323 there exists a simplicial 2-sphere @xmath271 such that @xmath324 . the underlying idea is the following . suppose that a @xmath59-sphere @xmath271 is `` thin '' i.e. @xmath325 . then there exists a slicing @xmath291 of @xmath271 into pieces with small numbers of vertices . in particular , the discrete length of any cycle @xmath285 in a slicing @xmath291 should be small . then the sphere @xmath271 is `` tightened '' , like the one shown on fig . [ pictcactus ] . it has the feature that small cycles can bound large areas . to measure this property , we introduce a natural metric on @xmath326 in which all edges have length @xmath28 , and then compare the metric space @xmath326 with a ( @xmath59-dimensional ) round sphere @xmath327 of a constant radius . if there is a bijection @xmath328 with close lipschitz constants , then @xmath326 is not thin . the reason why @xmath327 suits well for this consideration is that small curves on @xmath327 can not bound large areas , which follows from the isoperimetric inequality . quite similar considerations and ideas are used in the theory of planar separators . some results of this theory can be used to prove proposition [ propexistfatpoly ] directly . if @xmath329 is a planar graph with @xmath191 vertices , it is known that there exists a set of @xmath330 vertices which separates @xmath329 into two parts of roughly equal size ( such separating sets are called planar separators ) . it is also known that the asymptotic @xmath330 is the best possible for planar separators @xcite . if all @xmath59-spheres were `` thin '' , then every planar graph would have a separator of size , bounded by some constant , which contradicts the aforementioned asymptotic . anyway the deduction of proposition [ propexistfatpoly ] from the known theory requires some additional work , so we give an independent proof . now that we described the intuitive idea beyond our approach let us get to technical work . [ conmetmeasonk ] let @xmath271 be a @xmath59-dimensional simplicial complex . define a ( piecewise riemannian ) metric @xmath331 and measure @xmath75 on @xmath326 in such a way that each triangle @xmath332 becomes an equilateral euclidian triangle with the standard metric and edge length @xmath28 . thus the area of each triangle is @xmath333 . let @xmath334 denote the length of a piecewise smooth curve @xmath335 in @xmath326 . if @xmath336 is a closed @xmath28-dimensional cycle ( simplicial subcomplex ) , then , obviously , @xmath337 a cycle @xmath338 divides @xmath271 into two subcomplexes @xmath339 and @xmath340 , each homeomorphic to a closed @xmath59-disc ( we suppose @xmath341 ) . let us estimate the number of vertices in @xmath340 in terms of its area ( @xmath339 is similar ) . let @xmath342 denote the number of vertices , edges and triangles in @xmath340 . by the definition of measure , @xmath343 . we have @xmath344 ( euler characteristic of @xmath340 ) and @xmath345 ( by counting pairs @xmath346 , where @xmath162 is an edge and @xmath347 is a triangle ) . therefore , @xmath348 let @xmath349 be a @xmath59-dimensional round sphere of radius @xmath350 , with the standard metric @xmath351 and measure @xmath352 . a piecewise smooth closed curve @xmath353 without self - intersections divides @xmath349 into two regions @xmath354 , @xmath355 . the isoperimetric inequality on a sphere ( see e.g. ( * ? ? ? * ch.4 ) ) has the form @xmath356 where @xmath357 is the length of @xmath335 . since @xmath358 we may assume that @xmath359 ( otherwise consider @xmath355 instead ) , thus @xmath360 notice that this inequality does not depend on the sphere radius . let @xmath271 be a @xmath59-dimensional simplicial sphere and @xmath361 be positive real numbers . suppose there exists a bijective piecewise smooth map @xmath362 such that @xmath363 for each piecewise smooth curve @xmath364 and measurable set @xmath365 . numbers @xmath366 will be called lipschitz constants of the map @xmath16 . [ lemsmallpartofcyclek ] suppose there exists a mapping @xmath16 of 2-sphere @xmath271 into a round sphere @xmath349 with lipschitz constants @xmath366 . let @xmath338 be a cycle of @xmath271 dividing it into two closed regions @xmath339 and @xmath340 . if @xmath338 is contains at most @xmath367 vertices , then either @xmath339 or @xmath340 contains at most @xmath368 vertices . among two regions @xmath369 let @xmath370 be the one with the smaller area . combining , , , , and , we get @xmath371 which was to be proved . suppose @xmath372 . then by definition there exists a slicing @xmath280 , encoded by a tree @xmath103 , such that each region @xmath289 has at most @xmath367 vertices ( see construction [ conslicingregions ] ) . let us show that the degree of each node @xmath164 of @xmath103 is bounded from above . [ lemdegreenodebound ] if @xmath291 is a slicing @xmath280 and @xmath373 , then @xmath374 for any node @xmath164 of @xmath103 . denote @xmath375 by @xmath376 . by construction , the region @xmath289 is obtained from a sphere @xmath273 by removing @xmath376 open stars which correspond to the edges of @xmath103 emanating from @xmath164 . the complex @xmath289 itself can be considered as a plane graph . denote the numbers of its vertices , edges and faces by @xmath377 respectively . by the definition of the width , we have @xmath378 . we also have @xmath379 , and @xmath380 ( each region has at least @xmath313 edges ) . thus , @xmath381 . notice that each removed open star represents a face of graph @xmath289 , therefore , @xmath382 . [ lemmanyvertimplyfat ] let @xmath271 be a @xmath59-dimensional simplicial sphere endowed with the map @xmath16 to the round sphere , satisfying lipschitz bounds and . for a natural number @xmath367 set @xmath383 and @xmath384 . if @xmath385 , then @xmath324 . assume the contrary : @xmath372 . then there is a slicing @xmath280 in which every region @xmath289 has at most @xmath367 vertices . consequently , any cycle @xmath287 has at most @xmath367 vertices . by lemma [ lemsmallpartofcyclek ] , the cycle @xmath285 divides @xmath271 into two parts , one of which has @xmath386 vertices . since @xmath387 , the other part has @xmath388 vertices . assign a direction to each edge @xmath162 of @xmath103 in such a way that @xmath162 points from the larger component of @xmath389 to the smaller , where the `` size '' means the number of vertices . @xmath103 is a tree , therefore there exists a source , i.e. a node @xmath163 from which all adjacent edges emanate . speaking informally , this node represents a `` big sized bubble '' , meaning that the part of a sphere , lying across each border has a small size . let @xmath376 denote the degree of the chosen node @xmath163 . denote by @xmath390 the connected components of the graph @xmath391 . by lemma [ lemdegreenodebound ] we have @xmath392 . by the construction of the directions of edges , @xmath393 for each @xmath394 . thus @xmath395 the contradiction . [ lemexistbigandlips ] for any @xmath323 there exists a @xmath59-dimensional simplicial sphere @xmath271 such that : 1 . there exists a piecewise smooth map @xmath396 satisfying lipschitz bounds and for some constants @xmath397 2 . @xmath385 , where @xmath398 and @xmath399 are defined in lemma [ lemmanyvertimplyfat ] . start with the boundary of a regular tetrahedron with edge length @xmath28 : @xmath400 . the projection from the center of @xmath401 to the circumsphere @xmath402 is obviously lipschitz for some constants @xmath403 . now subdivide each triangle of @xmath404 into @xmath405 smaller regular triangles as shown on fig . [ fig : subdiv ] . = [ line width=1mm , draw = black!35,-triangle 90,postaction = draw , line width=3 mm , shorten > = 2 mm , - ] iin 0,1, ... ,4 in 0 , ... , ( 0.5*i+ , i ) ( 0.5*i+ + 1,i ) ( 0.5*i+ + 0.5,i+ ) cycle ; ( -7,0 ) ( -2,0 ) ( -4.5,5 * ) cycle ; ( -2,2.5*)(0,2.5 * ) ; this results in a simplicial complex @xmath406 . as a space with metric and measure @xmath407 is homothetic to @xmath404 with a linear scaling factor @xmath408 ( recall that the metric on simplicial complexes is introduced in such way that each edge has length @xmath28 ) . thus there exists a map @xmath409 with the same lipschitz constants as @xmath16 . the number of vertices @xmath410 can be made arbitrarily large . lemmas [ lemexistbigandlips ] and [ lemmanyvertimplyfat ] conclude the proof of proposition [ propexistfatpoly ] . actually , in the proof of lemma [ lemexistbigandlips ] we could have started from any simplicial sphere @xmath401 , take any piecewise smooth map @xmath411 , find lipschitz constants @xmath412 ( they exist by the standard calculus arguments ) , and then apply the same subdivision procedure . we used the boundary of a regular simplex , because in this case lipschitz map is constructed easily and allows for an explicit computation . we give a concrete example of a quasitoric manifold which is not toric origami , by performing this computation . the calculations themselves are elementary thus omitted . it is sufficient to construct a simplicial sphere for @xmath413 . for a projection map from the boundary of a regular tetrahedron to the circumscribed sphere we have lipschitz constants @xmath414 , @xmath415 . thus @xmath416 . subdivide each triangle in the boundary of a regular tetrahedron in @xmath405 small triangles where @xmath417 . this gives a simplicial sphere @xmath271 with at least @xmath418 vertices and the same lipschitz constants as @xmath419 . thus @xmath420 . now take the dual simple polytope @xmath41 of @xmath271 , consider any proper coloring of facets in four colors and assign a characteristic function @xmath38 , as described in lemma [ lemfourcolor ] . this gives a characteristic pair @xmath43 , whose corresponding quasitoric manifold is not toric origami . of course , all our estimations are very rough , and , probably , there are better ways to construct fat spheres . for sure , there exist @xmath59-spheres of fatness @xmath421 with less than @xmath418 vertices . note that in dimension @xmath313 and higher there is no simplicial subdivision of a regular simplex into smaller regular simplices . this is one of two places in the proof , where the dimension restriction is crucial . the second place is the four color theorem in lemma [ lemfourcolor ] . proposition [ propthmreform ] proves theorem [ thmqtornotorigami ] for @xmath269 . now we need to make the remaining cases @xmath422 . [ propexisthighdim ] there exist quasitoric manifolds of any dimension @xmath0 , @xmath422 , which are not toric origami . let @xmath52 be any quasitoric manifold , which is not toric origami . take the product of @xmath52 with @xmath423 ( the circle @xmath424 acts on @xmath423 by axial rotations ) . on the level of orbit spaces , this corresponds to multiplying @xmath41 with a closed interval @xmath425 . we claim that quasitoric manifold @xmath426 is not toric origami . if @xmath426 were a toric origami manifold , then all its characteristic submanifolds should be toric origami as well ( see construction [ conorbitoforigami ] ) . but @xmath52 is one of them . this gives a contradiction . thus taking products with @xmath423 produces examples for all @xmath422 . sphere @xmath423 is the simplest example of a quasitoric manifold . in the proof of proposition [ propexisthighdim ] we could have used any other quasitoric manifold instead of @xmath423 . if @xmath52 and @xmath427 are quasitoric manifolds and one of them is not toric origami , then the quasitoric manifold @xmath428 is not toric origami as well . on the other hand , new toric origami manifolds can be produced from a given one in a similar way as we used for constructing non - examples . it is easy to observe that if @xmath3 is a toric origami manifold and @xmath429 is a toric symplectic manifold , then @xmath430 is again a toric origami manifold . we would also like to mention that projective bundles over toric origami manifolds are again toric origami manifolds . more precisely , if @xmath12 is toric origami , and @xmath431 are complex line bundles over @xmath3 , each having an @xmath69 action on fibers , then the projectivization @xmath432 with the induced action of @xmath433 is also a toric origami manifold . we already mentioned a relation of our study to the theory of planar separators in section [ sectproof ] . we also want to mention another connection to the theory of random infinite planar maps . this rapidly developing part of probability theory aims , among other things , to give a firm foundation for some facts in statistical physics and quantum gravity . the basic idea of this study is the following @xcite . fix a number @xmath191 , a parameter of the whole construction . for a given @xmath7 consider all possible ( rooted ) plane @xmath191-angulations with @xmath7 faces . for @xmath434 , these are roughly the same as simplicial spheres . every plane graph has a standard metric , turning it into a metric space . by letting the number of faces tend to infinity , and renormalizing the diameter of graphs in a correct way , one considers the limits of converging sequences of graphs . the limits are taken with respect to the gromov hausdorff metric defined on the set of isometry classes of metric spaces . since there is only a finite number of such graphs with a fixed number @xmath7 of faces , we can take a uniform distribution on this set of graphs . the uniform distributions on the sets of prelimit metric spaces give rise to a limiting distribution , which is viewed as a random compact metric space ( of course , here we omit a lot of technicalities , needed to state everything precisely ) . the resulting random metric space is called a brownian map and considered as a good @xmath59-dimensional analogue of the brownian motion . a wonderful thing is that a brownian map does not actually depend on the parameter @xmath191 , if @xmath191 is either @xmath313 or even @xcite . it is also known that the brownian map is almost surely homeomorphic to a @xmath59-sphere @xcite . this suggests the following for each @xmath323 almost all simplicial @xmath59-spheres @xmath271 have @xmath324 . more precisely , if @xmath435 denotes the set of all simplicial @xmath59-spheres with @xmath436 triangles , and @xmath437 the subset of simplicial spheres having @xmath324 , then @xmath438 the reason is as follows ( cf . * cor.5.3 ) ) . if there were a lot of `` thin '' simplicial spheres , they all would have bottlenecks small cycles , dividing them into macroscopic regions . after taking a limit as @xmath439 and rescaling the metric , these bottlenecks would collapse to points . thus the limiting metric space would be non - homeomorphic to a sphere with non - zero probability . this question is motivated by the following fact . there exist a simple 3-polytope @xmath41 such that any quasitoric manifold @xmath52 over @xmath41 is not equivariantly homeomorphic to a symplectic toric manifold . stating shortly : there exist combinatorial types of simple 3-polytopes which do not admit delzant realizations . it was proved in @xcite that any 3-dimensional delzant polytope has at least one triangular or quadrangular face . consequently , in particular , a dodecahedron does not admit a delzant realization . an origami template is a generalization of a single delzant polytope , thus a realizability of a given combinatorial polytope by an origami template is a more complicated task . problem [ problcombrestr ] can be restated in different terms : are there any combinatorial restrictions on the orbit spaces of toric origami manifolds ? the examples of non - origami quasitoric manifolds in high dimensions were constructed from the @xmath313-dimensional case . on the other hand , lemma [ lembigfatsmalcol ] applies for any dimension . the problem of finding higher - dimensional polytopes whose dual spheres have large fatness may be of independent interest . actually , even if we find such a fat sphere , to make use of the developed technique we should also construct a characteristic function with a small range of values . this constitutes a certain problem , since characteristic function may not even exist , if @xmath440 ( this happens for dual neighborly polytopes , see @xcite ) . nevertheless , there is a big class of simplicial @xmath272-spheres , so called balanced spheres , which admit a proper vertex - coloring in @xmath7 colors . such colorings give rise to characteristic functions , which have exactly @xmath7 values , i.e. minimal possible . such characteristic functions and the corresponding quasitoric manifolds were called linear models in @xcite . passing to a barycentric subdivision makes every simplicial sphere into a balanced sphere . we suppose that passing to a barycentric subdivision does not strongly affect the fatness . if so , given any fat sphere dual to a simple polytope , one can pass to its barycentric subdivision , provide it with a linear model characteristic function , and finally obtain a quasitoric manifold which is not toric origami . another problem , which naturally arises from lemma [ lembigfatsmalcol ] is to find , for a given polytope @xmath41 , a characteristic function @xmath38 with the minimal possible range of values @xmath441 , if at least one characteristic function is known to exist . this minimal number seems to be an analogue of buchstaber invariant ( see the definition in @xcite or @xcite ) , as was noted to us recently by n.erokhovets . it may happen that an interesting theory hides beyond this subject . there exist obstructions to origami realizability , other than those described in section [ sectproof ] . if a weighted simplicial sphere @xmath271 can be represented as a connected sum , along a tree , of simplicial spheres dual to delzant polytopes , this does not mean automatically that @xmath271 corresponds to an origami template . the reason is that a convex polytope contains more information than its normal fan ( or , in our terminology , dual weighted simplicial sphere ) . it can be impossible to assemble an origami template from a collection of delzant polytopes , even if their dual weighted spheres suit together well . any projective toric variety corresponds to a convex polytope . thus any smooth projective toric variety is a symplectic toric manifold , which is a particular case of toric origami . thus , to prove the conjecture , one should consider non - projective examples . translating the problem into combinatorial language , the task is to find a complete smooth fan , which is not a normal fan of any polytope and , moreover , not dual to any origami template . the simplest non - polytopal fan is the fan corresponding to a famous non - projective oda s 3-fold @xcite . so it is natural to start with a more concrete question : in section [ secttopology ] we mentioned that a toric origami manifold @xmath137 is not quasitoric if its template graph has cycles . even if the orbit space of @xmath137 is contractible , the manifold @xmath137 may not be quasitoric . the simplest example of this kind is the sphere @xmath150 ( example [ exsfour ] ) . the orbit space of @xmath150 is a 2-gon , shown on fig . [ fig : orbitspace ] , which is not a convex polytope . excluding situations of these two kinds we may ask the following question .

we recall the main features of the regge approach used to understand soft interactions at lhc and higher energies . unitarity tames the power growth of the elastic proton - proton scattering amplitude with energy , and leads to the migration of the secondary particles produced in high - energy proton - proton collisions to larger transverse momenta . we discuss , in qualitative terms , the role of processes containing large rapidity gaps ( lrg ) , and the probability that the gaps survive population by secondaries produced in additional soft interactions . we explain how the regge diagram corresponding to a lrg event simultaneously describes events with different ( single , double , etc . ) particle density in the same rapidity interval . we show that the role of these , enhanced , multi - pomeron diagrams can be studied by measuring multiplicity fluctuations and long - range rapidity correlations between secondaries produced at the tevatron and the lhc . finally , we make a list of the characteristic features of the multi - pomeron description of soft interactions that may be observed at the high energies accessible at the tevatron and the lhc . ippp/09/24 + dcpt/09/48 + + * soft physics at the lhc * + m.g . ryskin , a.d . martin , v.a . khoze and a.g . shuvaev + institute for particle physics phenomenology , university of durham , durham , dh1 3le + petersburg nuclear physics institute , gatchina , st . petersburg , 188300 , russia

0907.1374

`` soft '' physics is manifest in various places in different guises . the interest in soft interactions at _ very high _ energies is stimulated , theoretically , by the fact that the observed hadronic cross sections are growing with energy and , experimentally , by the advent of the lhc . at the lhc energy we have to allow for unitarity effects , which are necessary to provide the consistency of the strong interaction . in fact , when we face a cross section which grows as a power of the energy , there are two possible ways to restore unitarity . first , we may introduce a new particle ( sufficiently heavy to have not been observed yet ) which cancels the growth of the cross section at energies larger than the mass of the particle ( for example , higgs boson , super - partners , etc . ) . after this , we deal with a _ weak _ interaction , where the main contribution comes from one , or a few , simple diagrams . more complicated feynman graphs just describe corrections to the lowest - order amplitude . another possibility is to consider a _ strong _ interaction . here , from the beginning , we start with a hermitian lagrangian , which should already account for unitarity . when the born amplitude ( or an amplitude obtained via the summation of some group of diagrams ) becomes too large and the interaction becomes strong , then new more complicated diagrams must enter the game . it is the contribution due to these new graphs which tames the growth of cross section such that the final result satisfies unitarity . in the case of qcd , where the bfkl amplitude ( @xmath3 ) grows as a non - integer power ( @xmath4 ) of energy , @xmath5 , we expect to observe the second scenario by introducing a new particle . ] . therefore it is important to observe , and to study , the role of more complicated diagrams ( arising from multiple interactions ) at lhc energies , and to trace how the theory restores the unitarity of the _ strong _ interaction . that is , from a theoretical viewpoint , it is of great interest to observe experimentally how unitarity tames the growth of high - energy hadronic amplitudes , leading to saturation both in the transverse momenta @xmath6 and in the impact parameter @xmath7 distributions of the produced particles . in other words , unitarity replaces the growth of the amplitude by the growth of configuration space occupied by the particles , both in @xmath7 and in @xmath6 . here , we do not present new results but attempt to recall , using modern language , the ideas and understanding of high - energy soft interactions which originated some 40 or more years ago . the discussion may be divided into three main topics : ( i ) elastic scattering and the total cross section , ( ii ) the cross sections of processes with gaps in rapidity ( including high - mass diffractive dissociation ) , and , ( iii ) the saturation of particle densities in transverse momenta @xmath6 . soft high - energy @xmath8 interactions are clearly important at the lhc . moreover , they can complicate our ability to observe new physics . for example , topic ( iii ) is relevant to the search for new physics , since particles with a rather large @xmath6 from the underlying inclusive event will affect the jet searching algorithm , as is already the case for jets at hera and the tevatron . on the other hand , the evaluation of the cross section for an exclusive process , such as @xmath9 , requires knowledge of the gap survival probability @xmath10 , which is the subject of topic ( ii ) , see , for example , @xcite . here , we focus on the qualitative features of high energy soft interactions . these features are quite general . with the advent of the lhc , it is timely to gather them together and to emphasize the underlying physics . the behaviour of scattering amplitudes in the high energy , @xmath5 , small momentum - transfer squared , @xmath11 , domain is well described by regge theory ; that is , by the singularities of amplitudes in the complex angular momentum , @xmath12 , plane , see , for example , @xcite . for instance , the measured @xmath13 amplitude behaves as a(s , t ) s^_(t ) where the @xmath14-trajectory , @xmath15 ( with @xmath16 in @xmath17 ) , passes through the spin-1 @xmath14-meson resonance in the ` crossed ' @xmath16-channel @xmath18 ; that is , @xmath19 . on the other hand , total cross sections are observed to grow slowly with energy and are associated with the exchange of a trajectory with vacuum quantum numbers . the simplest possibility is to assume that at high energy these cross sections , such as the @xmath8 total cross section @xmath20 , are driven by an isolated pole at @xmath21 , which gives an ( @xmath8 ) elastic amplitude a(s , t ) s^_p ( t ) , and , via the optical theorem of fig . [ fig : opt ] , a total cross section _ tot s^_p ( 0)-1 . the pole with the largest intercept , originally with @xmath22 since high energy total cross sections were thought to have constant asymptotic behaviour , was called the _ pomeron_. here we are interested in tevatron and higher energies which are sufficiently large to be able to neglect the contributions of all secondary trajectories ( which all have intercepts @xmath23 ) . total cross section to the imaginary part of the forward elastic @xmath8 amplitude ; ( b ) the application of the optical theorem to obtain the cross section for the dissociation of the ` beam ' proton into a high - mass system , see sections [ sec : sd ] and [ sec : lrg2 ] . the high - energy pomeron - exchange expressions that are shown for the _ bare _ amplitudes have sizeable absorptive ( that is , unitarity or screening ) corrections ; @xmath24 and @xmath25 are the pomeron - proton and triple - pomeron couplings respectively . , height=264 ] a popular parametrization of elastic @xmath8-scattering amplitude by donnachie - landshoff ( dl ) is the regge form @xcite @xmath26 where the signature factor @xmath27 gives the complex phase , @xmath28 is the electromagnetic form factor of the proton , and the _ effective _ soft pomeron trajectory _ p=1++ t 1 + 0.08 + 0.25 t , [ eq : dl ] where @xmath16 is in gev@xmath29 . the intercept @xmath30 just above 1 reproduces the observed slow growth of the total hadron - hadron cross sections at high energies . indeed , the effective pomeron pole amplitude , ( [ eq : dl ] ) , gives a good description of the total and elastic differential cross section data up to tevatron energies . however , this simple parametrization is expected to become increasingly deficient at higher energies . unitarity is an easy way to see this . to discuss unitarity effects it is convenient to work in impact parameter , @xmath7 , space , since at high energy the position of the particle in @xmath7 is practically frozen . thus the value of @xmath7 determines the orbital momentum @xmath31 of the incoming proton , @xmath32 ; that is , fixed @xmath7 corresponds to a particular partial wave @xmath31 . in @xmath7 space the elastic unitarity equation . ] @xmath33 limits the value of @xmath34 . however , this limit corresponds to a pure elastic interaction with phase of elastic amplitude , @xmath35 equal to @xmath36 . normally at high energies the inelastic contribution @xmath37 dominates , leading to a large inelasticity ; that is , to a large imaginary part , @xmath38 . in this so - called ` black disk ' limit is not forbidden . the corresponding model , so - called @xmath39-matrix unitarisation , was discussed in @xcite . however , such a model which asymptotically leads to a pure elastic interaction , without any particle production , does not look probable . ] , we have @xmath40 . recall that , if we were to neglect the small contribution from the real part , then im@xmath41 can be measured directly from experiment as the fourier transform a(b ) = j_0(qb ) , where @xmath42 , @xmath43 is a bessel function and @xmath44 . note that the dl amplitude @xmath45 crosses the black disk limit between the tevatron and lhc energies , see fig . [ fig : dl ] . thus we expect that at the lhc the cross section will be less than that given by the dl parametrization , while the slope of elastic cross section , @xmath46 , should be larger . the partons occurring in the proton wave function will be pushed away from the centre to the periphery . this suppression of the amplitude at low values of @xmath7 automatically increases the elastic slope @xmath46 , since @xmath47 where @xmath48 is the interaction radius . for this reason the value of @xmath49 gev@xmath50 in ( [ eq : dl ] ) should not be considered as the slope of the bare pomeron trajectory . part of this slope is generated by the stronger absorptive corrections at smaller @xmath7 . indeed , the shrinkage of the diffractive cone observed in deep - inelastic scattering at hera ( say , in @xmath14-meson or in @xmath51 diffractive production ) is smaller than that observed in @xmath8-scattering . [ t ] gev relevant to the cern - isr , the tevatron and the lhc respectively.,height=377 ] actually to satisfy elastic unitarity it is sufficient to consider an eikonal model which sums up the exchanges of any number of pomerons between the incoming protons . the model gives a result analogous to ( [ eq : el1 ] ) @xmath52 where , up to a normalization factor , the opacity @xmath53 plays the role of the scattering phase @xmath54 . in the eikonal model , ( [ eq : eik ] ) , it is the opacity @xmath55 which is now described ( or parametrised ) by pomeron exchange , rather than the final amplitude @xmath56 . the situation is sketched symbolically in fig . [ fig : epip](a ) . however this is not the whole story . the eikonal model does not include events with large rapidity gaps ( lrg ) , arising from the dissociation of one or both of the protons into high - mass systems , which are observed experimentally . we speak of single or double high - mass diffractive dissociation . we introduce high - mass dissociation in the next section and give detailed discussion of these lrg events in section [ sec : lrg ] . before we do this , let us recall that the interaction radius ( i.e. the elastic slope ) expected at the lhc will already exceed 1 fm the distance where _ confinement _ may enter the game . thus it is possible that confinement will stop the growth of the radius ; that is , the growth of @xmath46 . another interesting possibility , proposed by v.v.anisovich @xcite , is that the partons ( the gluons and quarks ) in the proton will start to form a few colourless clusters ( like the nucleons in nucleus ) , rather than the normal homogeneous distribution . this may lead to a rich diffractive dip structure in the differential elastic cross section @xmath57 . it is an additional argument to study experimentally the @xmath16 behaviour of the elastic cross section at the lhc interval , can be found in @xcite . ] . so much for elastic scattering , which we may call elastic diffraction . now we turn to inelastic diffraction , which is a consequence of the _ internal structure _ of the protons . besides the pure elastic two - particle intermediate states shown in fig . [ fig : epip](a ) , there is the possibility of proton excitation , @xmath58 , shown in the small sketch in fig . [ fig : epip](b ) . at high energies , when the lifetime of the fluctuations of the fast proton is large , @xmath59 , the corresponding fock states can be considered to be ` frozen ' . each constituent of the proton can undergo scattering and thus destroy the coherence of the fluctuations . as a consequence , the outgoing superposition of states will be different from the incident particle , so we will have _ inelastic _ , as well as elastic , diffraction . ; and ( c ) the inclusion of the multi - pomeron - pomeron diagrams which allow for high - mass dissociation . , height=302 ] to discuss inelastic diffraction , it is convenient to follow good and walker @xcite , and to introduce states @xmath60 which diagonalize the @xmath56 matrix . these , so - called diffractive , eigenstates only undergo ` elastic ' scattering . to account for the internal structure of the proton we , therefore , have to enlarge the set of intermediate states , from just the single elastic channel , and to introduce a multi - channel eikonal . the situation is pictured in fig . [ fig : epip](b ) . what about proton dissociation into high - mass systems ? at first sight , it appears that we may account for it by simply enlarging the number of diffractive eigenstates @xmath60 . even if this were practical , we would still face the problem of double counting when partons originating from the dissociation of the beam and the ` target ' protons overlap in rapidities . instead , high - mass dissociation is described in terms of so - called `` enhanced '' multi - pomeron diagrams . the first , and simplest , is the triple - pomeron diagram , shown in fig . [ fig : opt ] and again in fig . [ fig : epip](c ) . in fact , high - mass dissociation is much larger than low - mass dissociation at very high energy , and it will play a central role in our discussion of soft interactions . for simplicity , therefore , we do not discuss low - mass dissociation further , but consider that it can be easily allowed for in terms of a multi - channel eikonal . so , from now on we should regard @xmath61 as an effective opacity embodying low - mass dissociation . why does high - mass dissociation become so important at high collider energies ? a simplified way to see this is to note that the cross section for dissociation of a proton into a high - mass ( @xmath62 ) system has the approximate form with @xmath63 . the final equality in ( [ eq : simple ] ) can be deduced by taking the ratio of the couplings indicated in the regge expressions in fig . [ fig : opt ] . ] _ sd = ~ s _ el , [ eq : simple ] with @xmath64 , where @xmath25 is the triple - pomeron coupling and @xmath24 is the coupling of the pomeron to the proton , see fig . [ fig : opt ] . the ln@xmath65 ` rapidity ' factor comes from the integration @xmath66 . note that here we have used @xmath67 and @xmath68 , see fig [ fig : opt ] . eq . ( [ eq : simple ] ) is grossly oversimplified . for one thing , the dissociation cross section is subject to even greater absorptive ( that is , multi - pomeron ) corrections than the elastic cross section . when these are taken into account it is found from analyses of the available high - energy data that g_3p / g_n 0.25 , see , for example , @xcite . the important point is that the relative size of the @xmath69 and @xmath70 cross sections is governed , not simply by the parameter @xmath71 , but rather by @xmath72 . for each fixed rapidity interval the probability of high - mass dissociation is relatively small , but the cumulative effect in the complete interaction amplitude is enhanced by the large phase available in rapidity at high collider energies . in the regge approach , we have seen that high - energy soft interactions are driven by pomeron exchange , together with its absorptive ( multi - pomeron ) corrections . we may call this the `` soft '' pomeron . note that the @xmath65-channel asymptotic behaviour arising from @xmath16-channel pomeron exchange corresponds to an interaction radius which grows with energy , @xmath73 . the pomeron must also describe multiparticle interactions . the sum of ladder diagrams of the type of fig . [ fig:2](a ) is the simplest multiparticle structure which reproduces the power - like @xmath74 behaviour of the pomeron pole . before the advent of qcd , ladder diagrams composed of @xmath16-channel mesons ( dominantly spinless pion exchange , to the @xmath65-channel physical domain , @xmath75 , due to the smallness of the pion mass . ] ) were summed , and it proved difficult to generate an intercept @xmath76 . shortly after the discovery of qcd it was noticed @xcite that ( colourless ) two - gluon exchange has the properties of pomeron exchange : vacuum quantum numbers , even signature and a singularity at @xmath77 . due to the polarisation vectors of the spin - one particle ( the gluon ) , the numerator of the propagator contains the @xmath78 tensor such that the 4-momenta of the incoming fast protons occur in the form @xmath79 . therefore the gluon - exchange amplitude contains an extra power of @xmath65 as compared with spinless particle exchange . later , by applying the regge approach to gluons , rather than to hadrons , it was possible , if the gluon transverse momenta @xmath6 were sufficiently large , to describe high energy ( low @xmath80 ) interactions in terms of perturbative qcd . then , on summing the leading contributions , where the smallness of the qcd coupling @xmath81 is compensated by large values of ln@xmath82 , we build up the `` hard '' pomeron . below , we briefly trace this history and then apply these ideas to gain insight into high - energy soft interactions . in terms of regge theory , the high - energy amplitude is described by the exchange of a pomeron ( or a few pomerons ) . as mentioned above , from a quantum field theory viewpoint , the pomeron may be regarded as the sum of ladder - type diagrams , fig . [ fig:2](a ) . in other words , it corresponds to a sum of completely inelastic @xmath83 processes ; that is , to the last term @xmath84 in the unitarity equation ( [ eq : elun ] ) . now let us ` cut ' the pomeron ; see section [ sec : cut ] for further discussion . that is , we treat the graph of fig . [ fig:2](a ) , not as a diagram for the amplitude , but rather as the diagram for the @xmath83 cross section ( 2n)=a^*(2n)a(2n ) , see fig . [ fig:2](b ) . then , we get the inelastic production of @xmath85 particles homogeneously distributed in the available rapidity , @xmath86 , interval . in the original ` soft ' regge approach it was assumed that the transverse momenta @xmath6 of these secondaries were limited , and bfkl equation predicts the ` diffusion ' of the gluons in @xmath87 space @xcite . ] . the original idea , proposed by amati , fubini and stanghellini in 1962 @xcite , was to explain a large high - energy cross section , which did not decrease with energy , as a sequence of interactions in which the energy of a pair of adjacent particles ( pions ) is relatively small and lies in the resonance region where the interaction is strong . the integration corresponding to each cell of such a ladder is over the rapidity @xmath86 and the transverse momentum @xmath6 , see fig . [ fig:2](a ) . if we denote the momentum transferred through the whole ladder by @xmath88 , then the contribution from a particular cell is @xmath89 where ( q_t^2 ) = , [ eq : s0 ] and @xmath90 is the coupling . the integration over the rapidities gives a(y ) = _ n ^n_i ^y_0 dy_i = _ n = e^y = s^ , where the total rapidity spanned by the ladder @xmath91 , and @xmath92 reflects the identity of the cells or , equivalently , the ordering of the rapidities @xmath93 . cross section , ( c ) two - pomeron exchange , and ( d , e ) the triple - pomeron contribution to high - mass @xmath62 diffractive dissociation . the particles that are ` cut ' by the vertical dashed lines in ( b ) and ( c ) are on - mass - shell.,height=143 ] a convenient alternative way to obtain this result is to consider the evolution equation in rapidity = a(y ) . [ eq : afs ] in such a form @xmath94 plays a role similar to a dglap splitting function . note that the value of the ` effective spin ' @xmath95 depends on the momentum transfer through the ladder , @xmath96 . the slope of the pomeron trajectory @xmath97 at @xmath98 is controlled by the mean value of the transverse momenta @xmath99 or , in the case of ( [ eq : s0 ] ) , by the mass @xmath100 of the @xmath16-channel particle . however , with this procedure it turns out to be practically impossible to obtain a large intercept @xmath101 whilst keeping the @xmath16-channel exchanges to have masses @xmath102 gev . the corresponding couplings @xmath90 are too small . on the other hand , if we include heavier states , then we go to larger @xmath6 ; that is to small distances where it looks reasonable to work in terms of perturbative qcd . with qcd , the pomeron , at least at small distances , is described by the ladder of fig . [ fig:2](a ) built of gluons , rather than of spinless particles ; with the coupling @xmath103 in ( [ eq : s0 ] ) replaced by the qcd coupling @xmath81 . the integral over @xmath6 now has a more complicated structure . it is convenient to consider the recursion relation between the amplitudes @xmath104 with @xmath85 and @xmath105 cells f_n(x , k_t ) = ^1_x k(k_t , k_t)f_n-1(x,k_t ) , where we keep just the leading ln@xmath106 term in the @xmath107 integration . note that now we have to account for the dependence of the amplitude on the transverse momenta , @xmath6 , of the @xmath16-channel gluons . the recursion relation may be rewritten in differential form , as the bfkl evolution equation @xcite = kf [ eq : bfkleq ] so that @xmath108 , where = n_c_s k / = ( t)-1 , where @xmath109 is the leading eigenvalue calculated using equation ( @xmath110 ) below . since we now have spin - one gluons in the ladder , the intercept @xmath111 already in the limit @xmath112 . in fact , if just the leading ln@xmath82 contributions are summed then , explicitly , @xmath1134ln2 , where @xmath114 @xcite . however , the next - to - leading logarithm ( nll ) contributions are numerically rather large . it was found @xcite that , accounting these @xmath81 corrections , @xmath115 . for the relevant , not too small , values of @xmath116 , a resummation is necessary . since the nll contribution is mainly of kinematic origin , it is possible , and looks reasonable , to resum all the major higher - order corrections . this leads to @xmath117 @xcite in a wide region of @xmath6 . for simplicity , we consider just the forward amplitude the kernel @xmath118 is a bit more complicated , but qualitatively similar in form . ] with @xmath119 . then the elastic forward amplitude is a = is f(x , k_t ) , and , at lo , the bfkl kernel @xmath118 acts as @xmath120 since in qcd we deal with a dimensionless coupling @xmath81 and a massless gluon , the mean value of @xmath6 in each cell is determined by the value of transverse momentum @xmath121 in the previous cell . this is an important property of the bfkl equation that is , the so - called diffusion in @xmath122 . the value of @xmath123 in the current cell may differ from that in the previous cell by some quantity @xmath124 . in other words , at each step of evolution , not only the impact parameter @xmath7 can be changed by @xmath125 , but also the value of @xmath122 can be changed by @xmath124 . there are phenomenological hints that at large distances the `` soft '' pomeron should have qualitatively similar structure as the `` hard '' ( qcd ) pomeron . indeed , first , no irregularity is observed in the hera data in the transition region , @xmath126 , between the ` soft ' and ` hard ' interaction domains ; the data are smooth throughout this region . second , a small slope @xmath127 of the pomeron trajectory , is obtained in global analyses of all available soft high - energy data , after accounting for absorptive corrections and secondary reggeon contributions . this indicates that the typical values of @xmath6 inside the pomeron amplitude are relatively large ( @xmath128 ) . finally , recent ` soft ' model data analyses @xcite which account for the enhanced absorptive effects find an intercept of the initial , bare pomeron @xmath129 close to the intercept of the bfkl pomeron after the nll corrections are resummed @xcite . thus it looks reasonable to assume that in the soft domain we deal with the same perturbative qcd pomeron ; at least , there is a smooth transition from the soft to the hard pomeron . in summary , the _ bare _ hard pomeron , with a trajectory with intercept @xmath130 and small slope @xmath131 , is subject to increasing absorptive effects as we go to smaller @xmath6 which allow it to smoothly match on to the attributes of the _ soft _ pomeron . in the _ limited _ energy interval up to the tevatron energy , some of these attributes ( specifically those related to the elastic amplitude ) can be mimicked or approximated by an _ effective _ pomeron pole with trajectory intercept @xmath132 and slope @xmath133 . for further discussion , we write the bfkl evolution equation ( [ eq : bfkleq ] ) in the oversimplified form = a(y ) . [ eq : evoleq ] where @xmath134 acts as the splitting function . this evolution equation is the analogue of ( [ eq : afs ] ) , but , now , the possibility of multiparticle production leads to an additional power growth of the amplitude , @xmath135 , in comparison with the born amplitude , @xmath136 . in other words , the evolution ( [ eq : evoleq ] ) describes the development of the parton cascade from the ` beam ' proton to the ` target ' proton , and the @xmath137 growth of the inelastic cross section reflects the power growth of the parton density in the cascade . after one of the partons interacts with the target the coherency of the beam proton wave function is destroyed and a number of secondaries is produced . note that in the ladder of fig . [ fig:2](b ) only one branch of the whole cascade is shown , which finally leads to a parton being absorbed by the target . the number @xmath138 of secondaries produced is equal to the number of steps ( partons ) in this particular branch , shown in fig . [ fig:2](b ) . according to ( [ eq : evoleq ] ) , where @xmath139 plays the role of the particle density per unit of rapidity , we have @xmath140 . on the other hand , each of the partons continue to develop its own cascade . one example is shown in fig . [ fig:2](b ) , where parton @xmath141 develops its own branch . however , the branches which were not affected by the target conserve their coherence , and in the final state appear as a single parton . thus we get the power growth of the amplitude , but only a logarithmic growth of the multiplicity @xmath138 . the studies of coherence phenomena in a partonic cascade were pioneered by v.n . gribov in 1972 , see @xcite . the fact that the coherence of the wave function of the beam proton was destroyed after the interaction with the target , leads not only to inelastic high - multiplicity production , but via unitarity relation ( [ eq : elun ] ) , also to elastic scattering . elastic scattering is due to the absorption of an initial ( coherent ) component in some domain of the impact parameter @xmath7 plane , and originates from the remaining part of the initial wave function which conserves the coherence between the partons . that is , elastic scattering occurs as the ` shadow ' of the inelastic interaction . if the inelastic contribution @xmath37 is represented by the single ladder of fig . [ fig:2](b ) , then the elastic contribution may be drawn as a ` cut ' between the two pomerons of fig . [ fig:2](c ) , see also fig . [ fig:4 ] . so much for elastic scattering of the incoming proton . however , an intermediate parton in a ladder may be scattered elastically as well . a unitarity equation , analogous to ( [ eq : elun ] ) , may , and should be , written for any intermediate parton . the elastic scattering of an intermediate parton @xmath141 can be represented by the diagram of fig . [ fig:2](d ) . here , both branches of the cascade which start to evolve from parton @xmath141 save their coherence . therefore in the final state we will have a large rapidity gap between the parton @xmath141 and the target . such a process is called the diffractive dissociation of the beam proton into a high - mass state @xmath62 . it is described by the triple - pomeron diagram of fig . [ fig:2](e ) , in which one pomeron is split into two pomerons , see also fig . [ fig : opt ] . the probability @xmath142 of this splitting within a unit rapidity interval is relatively small . first , due to the small parton @xmath141 density @xmath139 , where @xmath139 arises from ( [ eq : evoleq ] ) . next , due to the parton-(target ) proton cross section being smaller than the proton - proton cross section , since the incoming proton contains many partons . on the other hand , each intermediate parton in the ladder may generate a pomeron splitting . thus , the whole effect accumulated during the evolution ( [ eq : evoleq ] ) is _ enhanced _ by the parton multiplicity @xmath138 in other words , by the size of available rapidity space @xmath143 , see also the discussion around equation ( [ eq : simple ] ) . therefore a triple - pomeron diagram ( like that in fig . [ fig:2](e ) ) or diagrams with more complicated multi - pomeron vertices ( like those in fig . [ fig:3](b ) ) are called _ enhanced _ diagrams . : ( a ) neglecting the gap survival factor @xmath10 , and ( b ) including the @xmath10 suppression caused by both eikonal ( depicted here by the pomeron line connecting the two protons ) and enhanced rescattering involving intermediate partons . plot ( c ) shows the expected long - range correlations @xmath144 of ( [ eq : r2 ] ) , corresponding to diagram ( a ) , with @xmath145 , supplemented by eikonal screening.,height=264 ] already at the tevatron , the cross section of the events with a lrg , that is of single- and double - diffractive dissociation of one or both protons ( integrated over the mass @xmath62 of the final ` diffracted states ' ) , is comparable to the elastic @xmath146 cross section . if we were to take the contribution of the diagram of fig . [ fig:2](e ) literally and integrate over the rapidity of vertex @xmath147 , then , at sufficiently high energy , the cross section of events with a lrg would exceed the whole inelastic @xmath37 contribution described by a single pomeron , that is the contribution of fig . [ fig:2](b ) ( recall @xmath148 ) . the situation is even worse when we consider events containing a large number of lrg , arising , for example , from the multi - pomeron diagram of fig . [ fig:3](a ) . for simplicity , let us first account for gaps of limited size , say @xmath149 . we denote the probability to form such a gap as @xmath150 . then , integrating over the central position @xmath93 of each gap , we obtain for @xmath151 the contribution @xmath152 which grows as a power of @xmath65 . if we neglect the limit @xmath149 , and allow for the integration over the gap size @xmath153 , then the cross section increases faster than any power of @xmath65 , and clearly violates unitarity . this problem was discussed long ago in @xcite . two main scenarios were proposed to prevent such a growth and to restore unitarity . first , the so - called _ weak _ coupling scenario assumes that the triple - pomeron vertex @xmath154 ( and correspondingly the value of @xmath150 ) vanishes as @xmath155 @xcite . since the mean value of momentum transfer @xmath156 decreases as @xmath157 ( the shrinkage of the diffractive cone ) the decrease of @xmath147 compensates the increase of the available rapidity interval . however this hypothesis is not supported by experiment region to confirm this conclusion ; see @xcite . ] . in the favoured _ strong _ coupling scenario @xcite , the multi - gap cross section is suppressed by a small ` gap survival probability ' @xmath10 which decreases with energy . in this case the pomeron coupling @xmath158 as @xmath155 , while the small probability of lrg production is due to the large probability that the gaps are populated by secondaries produced in additional soft rescattering interactions between the protons ( and also the intermediate partons ) . in the simple eikonal model , ( [ eq : eik ] ) , the probability , not to have an additional inelastic interaction which will populate the gap , is given by a factor @xmath159 . recall that in the eikonal model @xmath84 . however , this factor accounts for the rescattering of the incoming _ protons _ only . since , now , we have also rescattering between the _ intermediate _ partons we have to consider more complicated multi - pomeron diagrams , like those shown in fig.3(c ) . we emphasize that each multi - pomeron exchange diagram describes simultaneously a few different processes . we have already seen that the first ladder diagram gives , on the one hand , the elastic @xmath8 scattering amplitude , fig . [ fig:2](a ) , while , on the other hand , it may be considered as the ` cut ' diagram for the cross section for multiparticle production , fig . [ fig:2](b ) . in the latter case cutting the pomeron of fig . [ fig:2](a ) , that is taking the discontinuity , disc@xmath56=2im@xmath56 , gives the contribution to multiparticle production , @xmath160 . with zero multiplicity in the central region ; to the process with single pomeron multiplicity ( @xmath161 ) ; and , finally , with double pomeron multiplicity ( @xmath162).,height=151 ] now consider the two - pomeron exchange diagram of fig . [ fig:2](c ) . strictly speaking , the diagram should be drawn more precisely , as has been done in fig . [ fig:4 ] . we have the possibility to cut both pomerons simultaneously , [ fig:4](d ) , to cut only one pomeron , [ fig:4](c ) , or to cut between the two pomerons , [ fig:4](b ) . the sum of all cuts gives the total contribution of two - pomeron exchange to the imaginary part of the elastic amplitude . this contribution is negative and describes the _ absorptive _ correction to the one - pomeron amplitude . on the other hand , the elastic cross section corresponding to fig . [ fig:4](b ) is positive ; as is the cross section of the events with a ` doubled ' particle density , which we obtain by cutting both pomerons as in fig . [ fig:4](d ) . the only negative contribution is the absorptive correction to the events with a single particle density ( i.e. the correction to single - pomeron exchange ) . the famous agk cutting rules @xcite gives the relation between the different subprocesses originating from the same reggeon diagram . for the respective cut diagrams of fig . [ fig:4 ] , the relation is _ 0 : _ 1 : _ 2 = 1 : -4 : 2 , [ eq:142 ] while the whole contribution to the elastic amplitude is @xmath163 . analogous relations hold for more complicated diagrams with many pomerons . note that , according to the agk rules , the multi - pomeron diagrams do not change the single particle inclusive cross section @xmath164 . indeed , the particle density arising from diagram [ fig:4](d ) is twice larger than that from [ fig:4](c ) , and these two contributions cancel each other . the same is true for more complicated multi - pomeron diagrams . if we start with the _ eikonal _ model , then the sum of all the multi - pomeron exchange diagrams gives the total probability of the interaction at fixed @xmath7 , _ tot=2(1-e^-/2)=2^_n=1 ( -1)^n-1=^_n=0 _ n ( b ) , [ eq : s1 ] where @xmath165 is given by one - pomeron exchange and _ 0=_el=(1-e^-/2)^2 , _ n = e^- for n 1 . [ eq : pois ] here , @xmath166 is the probability of the elastic interaction ; and @xmath167 is the probability to produce a particle density @xmath85 times larger than that arising from one - pomeron exchange , that is the probability to cut @xmath85 pomerons in the whole amplitude . it is easy to see that _ inel ^_n=1_n= ^_n=1e^-= 1-e^-=g_inel [ eq : s3 ] is the whole inelastic contribution at a fixed @xmath7 . the penultimate expression for @xmath168 represents the whole probability , 1 , minus the probability @xmath169 to have no inelastic interaction , whereas in the previous expression for @xmath170 each term @xmath171 represents the probability of @xmath85 inelastic interactions ( where @xmath92 accounts for the identity of the interactions ) multiplied by @xmath169 which guarantees that there are no further inelastic interactions . the multi - pomeron exchange structure of processes of ( [ eq : s1])-([eq : s3 ] ) is sketched in fig . [ fig : a ] . the dashed vertical lines represent the cut leading to the elastic contribution , @xmath172 , and the cuts of the @xmath85 pomerons in @xmath173 . the circle is to indicate that all @xmath85 pomerons are cut.,height=302 ] similar rules , with the same combinatorics , are valid in the presence of _ enhanced _ screening . at each multi - pomeron vertex , we can cut one or more pomerons , or cut between the pomerons placing the uncut pomerons either to the left or to the right of the cut . for illustration , we discuss the contributions arising from cutting a triple - pomeron diagram which is screened by a single eikonal pomeron , as shown in fig . [ fig : b](a ) . the pomeron ladders corresponding to this diagram , fig . [ fig : b](b ) , can be cut in 8 different ways . first , we have the elastic @xmath172 cut between all the ladders . next , the cut of the eikonal pomeron ladder gives the usual single - pomeron multiplicity . then , we have three possibilities to cut the triple - pomeron graph ; one with single multiplicity in the whole rapidity interval , one with a gap in the upper part , one with double multiplicity in the upper part . finally , we have the same three , but now simultaneously cutting the eikonal pomeron . thus , the different cuts , which lead to 8 different process , give the particle multiplicities as a function of rapidity that are sketched in fig . [ fig : b](c ) . the agk cutting rules may be used to determine the relative probabilities of these processes ; the result is given below the multiplicity plots . for the example of fig . [ fig : b ] we have to use relation ( [ eq:142 ] ) twice . for instance , the probability @xmath174 comes from @xmath175 , whereas @xmath176 comes from @xmath177 ( @xmath178 corresponds to the cut of the triple - pomeron diagram without cutting the single pomeron ; note that in ( [ eq:142 ] ) @xmath179 , where the two terms correspond to cutting one or the other pomeron in fig.[fig:4 ] ) . [ t ] , as a function of rapidity @xmath86 , expected for the processes obtained from these 8 different cuts , together with their relative probabilities , which may be obtained from the agk cutting rules.,height=415 ] moreover , agk rules show that , in general , these factors do not depend on exactly how the multi - pomeron vertices are cut . to be specific , the probability of a particular subprocess is given just by the combinatorial factors to choose @xmath85 cut pomerons from @xmath180 pomerons of the original amplitude . a factor 2 comes from disc@xmath56=2im@xmath56 of each cut pomeron , and another factor 2 comes from the possibility to place any uncut pomeron to the right or left of the cut . note that the eikonal model ( [ eq : eik ] ) predicts a long - range correlation between the secondaries produced in different rapidity intervals . indeed , we have possibility to cut any number of pomerons . cutting @xmath85 pomerons we get an event with multiplicity @xmath85 times larger than that generated by one pomeron . the probability to observe a particle from a diagram where @xmath85 pomerons are cut is @xmath85 times larger than that from the diagram with only one cut pomeron . the observation of a particle at rapidity @xmath181 , say , has the effect of enlarging the relative contribution of diagrams with a larger number of cut pomerons . for this reason the probability to observe another particle at quite a different rapidity @xmath182 becomes larger as well . this can be observed experimentally via the ratio of inclusive cross sections @xmath183 where @xmath184 is the particle density . without multi - pomeron effects the value of @xmath185 exceeds zero only when the two particles are close to each other , that is when the separation @xmath186 is not large . such short - range correlations arise from resonance or jet production . however , multi - pomeron exchange leads to a long - range correlation , @xmath187 , even for a large rapidity difference between the particles , @xmath188 . in fact , such a long - range correlation , with @xmath189 for @xmath190 , has even already been observed in the old cern - isr data @xcite . in the eikonal case , the value of @xmath185 may be calculated explicitly from the poisson distribution of @xmath173 given in ( [ eq : pois ] ) . at fixed @xmath7 it gives r_2 = ( 1+)(1-e^-)-1 . for a low optical density @xmath191 the correlation arises mainly from the two - pomeron exchange diagram , and increases as @xmath192 . at very large @xmath61 , the poisson distribution becomes narrow in the number @xmath85 of cut pomerons , and @xmath185 decreases as @xmath193 . the maximum value is @xmath194 for @xmath195 . it would be informative to study the pomeron loop structure by observing long - range correlations at the lhc energy . in the pure _ eikonal _ case , the correlation should be the same throughout the whole available rapidity interval ( except , of course , near the edges of rapidity interval , where multiparticle production is affected by energy conservation constraints ) . when we have _ enhanced _ diagrams , that is pomeron loops which produce lrg which occupy only a part @xmath196 of the rapidity space , the corresponding correlation takes place only within this @xmath196 interval . thus the existence of pomeron loops may be revealed _ both _ as lrg events ( from the cut with zero multiplicity inside the @xmath196 interval ) , and as long - range @xmath185 correlations in the @xmath153 intervals ( from the cuts generating large multiplicities of secondaries ) . the ` range ' of the correlation reflects the size of the pomeron loop in rapidity space . an example of this behaviour of @xmath185 is shown in fig . [ fig:3](c ) note that the simplest scenario to restore the unitarity at very high energies is not yet completely rejected . we mean the possibility that asymptotically the pomeron intercept becomes less than 1 , @xmath197 , and at very high energies the total cross section starts to decrease with energy . such a behaviour is expected in a theory with only the triple - pomeron coupling , and which neglects the more complicated multi - pomeron vertices , such as the @xmath198 pomeron coupling and so on @xcite . recall that the two - pomeron loop gives a positive probability for lrg production but its contribution to the amplitude is negative , see fig . [ fig:4 ] . if we now sum up the diagrams with many two - pomeron loop insertions ( fig . [ fig:3](a ) ) then we ` renormalise ' the pomeron propagator . in this way we obtain the ` dressed ' propagator in terms of a schwinger - dyson equation . the intercept of the final ` dressed ' propagator is @xmath199 . with increasing energy , the available rapidity interval increases , with the result that the loop renormalisation grows and hence the effective intercept @xmath200 decreases . asymptotically , the schwinger - dyson equation gives @xmath201 where @xmath147 is the triple - pomeron coupling . the loop contribution , @xmath202 , is calculated using the dressed pomeron propagator , and the factor @xmath203 comes from the integral over the loop size @xmath196 . for the reasonable values @xmath204 and @xmath117 , the naive estimate based on ( [ eq : sd ] ) gives @xmath205 . that is @xmath206 . this means that up to a rather large energy ( @xmath207 ) the total cross section will grow , but after this , that is just in the lhc region , it will start to decrease . for multiparticle production , such a regime will reveal itself as long - range rapidity fluctuations of the multiplicity arising from two - pomeron loops which may produce either lrg or a double multiplicity of secondaries in the rapidity domain occupied by the loop . let us return to the simplified eikonal model ( [ eq : eik ] ) , written for the partons in the incoming protons . we may ask the question - `` what happens to an incoming fast quark ? '' on the one hand , the factor @xmath208 describes its absorption . after the inelastic collision the incoming quark disappears from the initial wave function . on the other hand , the quark can not just _ disappear_. ] . it has baryon charge , etc . disappearance from the incoming beam ( wave function ) actually means _ migration _ after the collision the momentum of quark is changed . in the leading logarithmic approximation , the quark mainly changes its transverse momentum . each new pomeron - quark coupling may be considered as a new elastic scattering and it is known that after @xmath85 scatterings the mean transverse momentum squared , @xmath209 , increases @xmath85 times . thus , when we say that unitarity stops the growth of the parton density via an increase of the absorptive correction , actually we mean that the extra partons produced by the cascade are pushed out of the previous domain in @xmath210 configuration space into a region of larger @xmath6 and @xmath7 . accounting for the enhanced diagrams we get the same phenomena , not only for the incoming fast partons ( as in the eikonal model ) , but for any intermediate partons as well . thus finally we expect that the particle density will reach saturation at low @xmath6 , but will continue to grow with energy at larger @xmath6 . therefore , at higher energies a larger number of minijets with larger @xmath211 will be produced . due to saturation in the low @xmath6 region at very high energies , the majority of partons will have rather large @xmath6 . we already know that the diffusion in @xmath87 takes place in the perturbative qcd domain @xcite , and to provide a smooth matching to the larger @xmath6 domain we need to include some elements of a similar diffusion in the description of low ( and intermediate ) transverse momenta . at the tevatron and the lhc , there is a possibility to study just pomeron - pomeron collisions by selecting events with large rapidity gaps on either side of some centrally produced hadronic system of mass @xmath62 . this can be done , either , by detecting the two forward outgoing protons which each carry a large fraction , @xmath212 , of their incoming momenta and the size of the gaps in rapidity are @xmath213 . ] , say @xmath214 , or , just by using the `` gap - hadron - gap '' trigger . since the interaction across each gap is described by pomeron exchange such events may be treated as pomeron - pomeron collisions . usually it is called a double - pomeron - exchange ( dpe ) process . in a dpe process we may hope to observe the properties of the pomeron caused by the fact that ( i ) the pomeron consists mainly of gluons and ( ii ) the transverse size of the pomeron is small , that is the typical transverse momenta , @xmath6 , in the diagrams which describe pomeron exchange are relatively large . indeed , the value of @xmath6 increases due to bfkl diffusion and screening effects ( as explained in the previous section ) . another indication in favour of large @xmath6 is the rather small value of the slope @xmath215 of the pomeron trajectory , see section [ sec : trans ] and the discussion after eq.([eq : afs ] ) . for these reasons , dpe events have several characteristic features . first , since the pomeron consists of mainly of gluons , we expect there to be a larger fraction of the glueballs and/or @xmath216 mesons in the central system @xmath217 , than that in @xmath8-interactions . recall that su(3 ) iso - singlet @xmath216 mesons contain a sizeable gluon component . moreover , because dpe has no `` incoming '' valence quarks , we expect the baryon / hyperon content to be smaller than in @xmath8 interactions . further information can be obtained by comparing dpe processes with @xmath8 interactions at the lower energy @xmath218 . for example , the multiplicity of secondaries is expected to be close to that in the @xmath8 case interactions , since there the multiplicity , and other characteristics of the final distribution , are driven by specific double loagarithms which are different from those in @xmath8 collisions @xcite . ] . strictly speaking , we expect a bit higher multiplicity since in pomeron - pomeron collisions the initial energy goes mainly to mesons , while the fraction of baryons is smaller than in @xmath8 interactions . there are several consequences of the small size of the pomeron . first , the transverse momenta of secondaries are expected to be a bit larger for dpe than that in the @xmath8 case at @xmath219 , but smaller ( due to the kinematic constraints ) than in the original pure inelastic process which takes up the whole initial @xmath8 energy @xmath5 . the high @xmath220 dijets produced by dpe will be dominantly gluon jets . next , the relatively small size of the pomeron should result in a smaller interaction radius , as measured via bose - einstein correlations , than that found in the usual inelastic @xmath8 collision . finally , the probability of double parton scattering , in particular the probability to observe two pairs of high @xmath220 jets , will be larger than that in a @xmath8 collision at @xmath219 . accounting for the anticipated lhc luminosity , the rate of dpe events is rather large . for example , from ref . @xcite , we see that the dpe cross section integrated over the @xmath221 interval , corresponding to the central production of a system of mass @xmath222 100 gev , is about 10 @xmath223 . alternatively , the expected dpe cross section is @xmath224 @xcite ; here @xmath225 . so it should be possible to study the properties of dpe events in detail . we should discuss how the saturation of particle densities can be consistent with confinement . the growth of the transverse momenta @xmath6 tames the increase of the interaction radius with energy . one possibility is that the simultaneous action of confinement and the growth of @xmath6 will finally lead to the saturation of the interaction radius ; that is to the saturation of the elastic slope , @xmath226 as @xmath227 . however the idea , discussed at the end of section [ sec : tot ] , that when the interaction radius becomes large confinement leads to the formation of colourless clusters , allows an alternative possibility . in a large , but finite , interval of rapidity the evolution is driven by the gluon ladder . in this case the impact parameter is almost frozen ; that is , the variation of the interaction radius @xmath228 is small . the system then emits a pion ( or another meson ) and this ` colourless cluster ' provides a noticeable increase of the impact parameter @xmath7 . following this , we again have gluons in some interval of evolution , and so on , as shown in fig . [ fig : m](a ) . the hard qcd interactions will be responsible for the growth of @xmath6 , while the part of the evolution described in terms of colourless mesons will be responsible for the increase in the interaction radius , that is of @xmath46 , due to the diffusion in * b * plane see fig . [ fig : m](b ) . in such a ` mixed ' picture we expect lower @xmath6 in peripheral collisions ( since the available rapidity interval was used to enlarge the value of @xmath7 ) and maximal @xmath6 in central collisions . [ t ] it may be informative to study multiparticle production at the lhc for individual events , as well as in terms of inclusive cross sections . at the high lhc energy the multiplicity of secondaries is large . thus it might be possible to observe the fluctuation of particle densities in individual events like those shown in fig . [ fig : b ] , or the effects arising from a mixture of intervals of evolution via colourless meson and coloured gluon exchanges , like those shown in fig . [ fig : m ] . in the latter case , the rapidity correlation length ( measured via @xmath185 of ( [ eq : r2 ] ) ) should be rather small for peripheral collisions since the main part of the available rapidity interval was used for meson exchange ( to enlarge @xmath7 ) and so only small intervals are left for gluon exchange where we expect a large probability to have several ` cut ' pomerons . it is appropriate to say a few final words about models of high - energy soft interactions . we emphasize that the aim of experiment is not to reject or to confirm one or another model by saying that the data prefer monte carlo version x or y. rather , the objective should be to isolate , and to study , the main qualitative features of the interaction . what , therefore , are the requirements of a realistic model of high - energy soft interactions ? to describe all the qualitative features discussed in this paper , it is clear that a realistic model should * contain , not only the eikonal but also the enhanced multi - pomeron contributions ; * include , not just the triple - pomeron vertex , but more complicated multi - pomeron vertices ( otherwise we will get an asymptotically decreasing cross section ) ; * allow for ` diffusion ' of the partons both in impact parameter , @xmath7 , and in transverse momenta , or , more precisely , @xmath87 space . a recent attempt at building such a multi - pomeron model , tuned to describe all the available data for high - energy soft interactions , is presented in @xcite . for convenience , we call this the kmr model . [ t ] cross sections , and the cross sections of dissociation to a fixed @xmath229 state ( c ) ; ( d ) the parton multiplicity ( solid lines ) and the number of ` colour tubes ' ( dashed ) produced by pomeron components of different size . the figure is taken from the kmr model of @xcite.,height=680 ] finally , can the characteristic features of the multi - pomeron description of soft interactions be observed at the high tevatron and lhc energies ? we list some below . * the multi - pomeron absorptive effects tame the power growth of the @xmath8 total cross section leading to values smaller than predicted before . for example , the kmr model prefers a value close to the lower of the two tevatron measurements , and a value of about 90 mb at the lhc energy of @xmath230 14 tev , see fig . [ fig : kmr](a ) . at 100 tev the prediction of 108 mb is at the lower limit of cosmic ray expectations . * simultaneously , these absorptive effects shrink the elastic differential cross section peak as the energy increases through the lhc range more than the expectations arising from a naive effective pomeron - pole of a donnachie - landshoff type of parametrization @xcite . * the cross section for proton dissociation into high - mass systems should grow with energy and be comparable to the elastic scattering cross section . the reason is that although , for each fixed rapidity interval , the probability of high - mass dissociation is relatively small , the overall effect is enhanced by the large phase space that is available in rapidity at high collider energies . the predicted values of @xmath69 and @xmath70 for the kmr multi - pomeron model are shown in figs . [ fig : kmr](b , c ) for energies in the tevatron lhc energy range . * the main growth in multiplicity , as we go from tevatron to lhc energies , is due to the small size ( ` qcd ' ) pomeron component , which produces particles with typically @xmath231 gev . there is essentially no growth in multiplicity at small @xmath232 . this simply confirms the trend that has been observed through the cern - isr to tevatron energy range , see the data points in fig . [ fig : inclmult ] . + [ h ] + ranges correspond to the large- , intermediate- and small - size components of the pomeron , see also fig . [ fig : kmr](d),height=566 ] + in other words , starting with the same pomeron intercept ( @xmath117 of section [ sec : pqcd ] ) the contribution of the large - size component _ after the absorptive correction _ becomes practically flat in energy ( @xmath233 ) , while the small - size contribution , which is much less affected by the absorption , continues to grow with energy ( @xmath234 ) . such behaviour is consistent with the experiment ( see fig . [ fig : inclmult ] ) where the density of low @xmath232 secondaries is practically saturated while probability to produce a hadron with a large ( say , more than 5 gev ) transverse momentum grows with the initial energy . * multi - pomeron exchange diagrams have a characteristic pattern of ` enhanced ' multiplicities of secondary particles , as well as large rapidity gaps , governed by the agk cutting rules . these lead to long - range rapidity correlations , for example , which may be observed at the lhc via @xmath185 of ( [ eq : r2 ] ) . * the observed rate of large rapidity gap ( lrg ) events in various processes at the lhc will be particularly informative . the survival probability of a lrg may be calculated in terms of a multi - pomeron exchange model of soft interactions . it is necessary to calculate the survival to both _ eikonal _ soft rescattering ( between the colliding protons ) , and _ enhanced _ rescattering ( involving soft interactions with intermediate partons with different @xmath6 ) . to calculate the latter we need to include the @xmath6 dependence of pomeron exchange , see @xcite this was the third requirement listed for a realistic model . * a topical example of a lrg process is the _ exclusive _ production of a heavy mass system @xmath56 , that is @xmath235 where the @xmath236 signs denote lrgs . this process , with @xmath237higgs , is a novel and promising way to study the higgs sector at the lhc , which gives a strong motivation for the addition of forward proton detectors to enhance the discovery and physics potential of the atlas and cms detectors at the lhc @xcite . already , at the tevatron , the cdf collaboration have observed exclusive processes with @xmath238 , dijet and @xmath239 with rates consistent with the estimates of the lrg survival probabilities predicted by the multi - pomeron model of soft interactions , see @xcite and references therein . more data on exclusive processes , from both the tevatron and , especially , the lhc , will be illuminating . it is a pleasure to thank m.g . albrow , v.v . anisovich , a. de roeck and r. orava for interesting discussions . mgr would like to thank the ippp at the university of durham for hospitality . this work was supported by the grant rfbr 07 - 02 - 00023 , by the federal program of the russian state rsgss-3628.2008.2 and by the russia - israel grant 06 - 02 - 72041-mnti . khoze , a.d . martin and m.g . ryskin , eur . j. * c14 * , 525 ( 2000 ) ; _ ibid _ , * c18 * ( 2000 ) 167 ; _ ibid _ , * c23 * , 311 ( 2002 ) ; + e. gotsman , e. levin , u. maor , e. naftali and a. prygarin , _ proc . of the workshop hera and the lhc _ , hep - ph/0601012 , p. 221 ( 2006 ) . collins , _ regge theory and high energy physics _ , ( cambridge univ . press , 1977 ) ; + a.c . irving and r.p . worden , phys . rept . * 34 * , 117 ( 1977 ) . boreskov , a.b . kaidalov and o.v . kancheli , phys . of atomic nuclei , * 69 * , 1765 ( 2006 ) . a. donnachie and p.v . landshoff , nucl . b231 * , 189 ( 1984 ) . s.m . troshin and n.e . tyurin , int . j. mod a22 * , 4437 ( 2007 ) . m.g . ryskin , a.d . martin and v.a . khoze , eur . j. * c60 * , 249 ( 2009 ) . cudell et al . [ compete collaboration ] , phys . lett . * 89 * , 201801 ( 2002 ) ; + m.m . block , phys . rept . * 436 * , 71 ( 2006 ) ; + r. fiore , l.l . jenkovszky , r. orava , e. predazzi , a. prokudin and o. selyugin , arxiv:0810.2902 [ hep - ph ] ; + p.v . landshoff , arxiv:0903.1523 [ hep - ph ] . good and w.d . walker , phys . rev . * 120 * , 1857 ( 1960 ) . fadin , e.a . kuraev and l.n . lipatov , phys . lett . * b60 * , 50 ( 1975 ) ; + e.a . kuraev , l.n . lipatov and v.s . fadin , sov . jetp * 44 * , 443 ( 1976 ) ; _ ibid . _ * 45 * , 199 ( 1977 ) ; + i.i . balitsky and l.n . lipatov , sov . * 28 * , 822 ( 1978 ) . m. ciafaloni , d. colferai and g. salam , phys . rev . * d60 * , 114036 ( 1999 ) ; + g. salam , jhep * 9807 * , 019 ( 1998 ) ; act . phys . pol . * b30 * , 3679 ( 1999 ) ; + v.a . khoze , a.d . martin , m.g . ryskin and w.j . stirling , phys . rev . * d70 * , 074013 ( 2004 ) . i.a . verdiev et al . jetp * 19 * , 1148 ( 1964 ) ; + j. finkelstein and k. kajantie , phys . lett . * b26 * , 305 ( 1968 ) ; nuovo cim . * 56a * , 658 ( 1968 ) . gribov and a.a . migdal , sov . * 8 * , 583 ( 1969 ) . luna , v.a . khoze , a.d . martin and m.g . ryskin , eur . j. * c59 * , 1 ( 2009 ) . v.n . gribov and a.a . migdal , sov . jetp * 28 * , 784 ( 1969 ) . abramovsky , v.n . gribov and o.v . kancheli , sov . . phys . * 18 * , 308 ( 1973 ) . h. dibon et al . , phys . lett . * b44 * , 313 ( 1973 ) ; + s.r . amendolia et al . , phys . * b48 * , 359 ( 1974 ) .

we establish a quantum version of the classical isoperimetric inequality relating the fisher information and the entropy power of a quantum state . the key tool is a fisher information inequality for a state which results from a certain convolution operation : the latter maps a classical probability distribution on phase space and a quantum state to a quantum state . we show that this inequality also gives rise to several related inequalities whose counterparts are well - known in the classical setting : in particular , it implies an entropy power inequality for the mentioned convolution operation as well as the isoperimetric inequality , and establishes concavity of the entropy power along trajectories of the quantum heat diffusion semigroup . as an application , we derive a log - sobolev inequality for the quantum ornstein - uhlenbeck semigroup , and argue that it implies fast convergence towards the fixed point for a large class of initial states .

1606.08603

the convolution operation @xmath0 between two real- ( respectively vector- ) valued independent random variables @xmath1 and @xmath2 plays a central role in classical information theory . the operation is defined in terms of the action on probability density functions as @xmath3 the convolution models a general class of additive noise channels , and thus provides a natural framework for the study of information processing and associated capacities . the operation also is a central element in many functional analytic inequalities , most notably young s inequality @xcite , the brascamp - lieb inequalities @xcite , de bruijn s identity @xcite , the fisher information inequality @xcite , and the entropy power inequality @xcite . such inequalities have wide use in information theory , yielding bounds on communication capacities , as observed by shannon @xcite . they can also provide , for example , bounds on the convergence rate in the central limit theorem @xcite . beyond these applications , these results are appealing from a conceptual , geometric viewpoint : many inequalities can be regarded as information - theoretic counterparts of related statements about convex bodies . for example , identifying entropy power with volumes reveals a formal similarity between the brunn - minkowski inequality and the entropy power inequality @xcite . indeed , there is even a proof of the latter guided by this intuition @xcite . we refer to @xcite for detailed accounts of this wealth of inequalities and their interrelationships . our work is guided by the question of whether a similar array of inequalities exists in a quantum setting . a key first step in this direction one which is directly relevant to our work was taken by werner @xcite . he introduced the convolution operation @xmath4 which involves a probability density function @xmath5 on phase space as well as a state @xmath6 ( of a bosonic system ) . the operation results in the average state @xmath7 when displacing @xmath6 according to @xmath5 using the ( weyl ) displacement operators @xmath8 . here we introduce the parameter @xmath9 for convenience , the case @xmath10 was considered in @xcite . treating the convolutions and on the same algebraic footing , werner established ( among other results characterizing ) a form of young s inequality . it is worth mentioning that carlen and lieb have recently established generalizations of the latter in a fermionic context @xcite . more recent work @xcite has centered around a convolution operation of the form @xmath11 where @xmath12 is a ( @xmath13-mode ) beamsplitter of transmissivity @xmath14 $ ] . it is worth pointing out that the action of this operation is formally similar to when expressed in terms of the wigner functions describing the quantum states @xmath15 and @xmath16 . the map describes a process where two states interact . it captures , in particular , the situation where one of the states is transmitted through an additive ( bosonic ) noise channel . in @xcite , the authors established an entropy power inequality of the form @xmath17 for the convolution and for @xmath18 , where @xmath19 denotes the von neumann entropy . subsequent work @xcite managed to lift the restriction on @xmath20 , and generalized this result to more general ( gaussian ) unitaries in place of @xmath12 . a related inequality of the form @xmath21 for @xmath14 $ ] was also shown in @xcite , generalizing classical results @xcite ( see also @xcite for a discussion of the relationship between the two ) . a generalization to conditional entropies was proposed in @xcite , and an application to channel capacities was discussed in @xcite . a key tool in establishing these results is the quantum fisher information @xmath22 , defined for a state @xmath6 as the divergence - based fisher information of the family @xmath23 obtained by displacing @xmath6 along a phase space direction ( see section [ sec : quantumdef ] for a precise definition ) . it was shown in @xcite for @xmath18 and in @xcite for general @xmath14 $ ] that this quantity satisfies the fisher information inequality @xcite @xmath24 this identity is a consequence of the strong subadditivity ( data processing ) inequality for relative entropy , and lies at the heart of the proof of . following the theme of entropy power inequalities for quantum systems , audenaert , datta , and ozols @xcite have obtained strong majorization - type results for the finite - dimensional case . these center around an operation of the form , but with @xmath15 and @xmath16 being states on a finite - dimensional hilbert space , and a family @xmath25 of unitaries generated by a hamiltonian @xmath26 realizing the swap - operation of the two systems . as argued by audenaert et al . , these results imply several entropy - power - type inequalities . very recently , carlen , lieb , and loss @xcite have provided an elegant short proof of these statements . guha , shapiro , and garca - patrn have discussed alternative definitions of the quantum entropy power and corresponding entropy power inequalities @xcite . here we focus on the convolution operation . we find that this operation satisfies similar properties as the convolutions and . in particular , we establish a stam inequality for the fisher information of the form @xmath27 this classical - quantum version of stam s inequality has several immediate consequences , all of which follow the reasoning used in establishing classical results @xcite . for example , we use to establish an entropy power inequality of the form @xmath28 taking @xmath29 to be a unit - variance centered gaussian , we find a quantum isoperimetric inequality of the form @xmath30 note that @xmath31 is the result of applying a classical noise channel to @xmath6 , where the variance of the displacement goes to @xmath32 in the limit @xmath33 . this family of maps constitutes a semigroup , which we call the heat diffusion semigroup : it is generated by a liouvillian @xmath34 such that @xmath35 we find that the entropy power along trajectories generated by this liouvillian is concave , i.e. , @xmath36 eq . generalizes a celebrated result @xcite concerning the classical heat equation . the entropy power inequality for gaussian @xmath5 implies the lower bound @xmath37 , and establishes the isoperimetric inequality for the fisher information @xmath38 for states @xmath6 of @xmath13 bosonic modes . we find that for @xmath39 , this statement is tight : gaussian thermal states achieve equality in in the limit of large mean photon numbers . we apply our results , in particular eq . , to the quantum ornstein - uhlenbeck ( qou ) semigroup for a one - mode bosonic system . this is a one - parameter group of cptp maps @xmath40 generated by a linear combination of liouvillians of a quantum amplifier and an attenuator channel , respectively : @xmath41 where @xmath42 and @xmath43 are the creation and annihilation operators ( i.e. , @xmath44 = \mathsf{id}$ ] ) . the qou semigroup is a natural counterpart of the classical semigroup generated by the fokker - planck equation ( see appendix [ sec : classicalou ] ) . the unique fixed point of the semigroup @xmath40 is the state @xmath45 i.e. , it is diagonal in the number state basis @xmath46 with a geometric distribution , hence a gaussian thermal state . we conjecture that for an arbitrary initial state @xmath6 , this semigroup converges to the fixed point at an exponential rate given by the exponent @xmath47 , that is , @xmath48 where @xmath49 is the relative entropy . in appendix [ app : qou ] , we show that holds for all gaussian states @xmath6 , and the exponent @xmath47 is optimal . in other words , our conjecture amounts to the statement that certain gaussian thermal states converge `` most slowly '' to the fixed point , and the log - sobolev-@xmath50-constant , defined as the largest constant @xmath51 such that @xmath52 for any state @xmath6 and all @xmath9 , is given by @xmath53 . the quantum isoperimetric inequality provides evidence for this conjecture : taking as a specific example the case where @xmath54 and @xmath55 ( and thus @xmath56 ) , we can show ( see example [ ex:2 - 1_process_n ] in section [ sec : app ] ) that @xmath57 we also show ( see example [ ex:2 - 1_process_s ] in section [ sec : app ] ) that recent work by de palma , trevisan , and giovannetti @xcite similarly implies exponential convergence of the form for initial states @xmath6 having large entropy : the isoperimetric inequality of @xcite ( which we discuss in more detail below ) implies that @xmath58 while this does not establish the scaling for all states @xmath6 , it illustrates the use of the quantum isoperimetric inequality in a concrete context . we clarify the relationship between our case and the one in the classical setting ( see section [ sec : classicalcontext ] ) , where the log - sobolev-@xmath50 constant can be obtained from the isoperimetric inequality for the classical fisher information , following work by carlen @xcite . in fact , this is the main motivation for our conjecture : gaussian distributions give the optimal convergence rate in the classical problem . whether conjecture holds is a challenging open question . we remark that it would significantly strengthen known results about the qou semigroup . specifically , carbone et al . @xcite established that the qou semigroup is hypercontractive . in ( * proposition 4.2 ) , the following inequality was shown for the log - sobolev-@xmath59 constant in @xcite . ] @xmath60 of @xmath61 : @xmath62 in this expression , @xmath63 is the log - sobolev-@xmath59 constant of the associated classical birth - and - death process ( which is unknown ) , but for which the bounds @xmath64 were shown ( see ( * ? ? ? * proposition 4.1 ) where @xmath63 is denoted @xmath65 ) . following @xcite , this implies that for all states @xmath6 , we have @xmath66 ( respectively , we have @xmath67 ) , if the semigroup can be shown to be strongly ( respectively weakly ) @xmath68-regular . the derivation of our fast convergence results from the isoperimetric inequality follows , to some extent , a well - known line of reasoning considered in the classical context . indeed , carlen @xcite has shown that the isoperimetric inequality gives rise to a log - sobolev inequality , which in turns provides bounds on the convergence of the classical ornstein - uhlenbeck ( cou ) semigroup to the fixed point ( see appendix [ sec : classicalou ] ) . however , in our case we find that , while a log - sobolev inequality again easily follows from , the quantities appearing in it are not easily estimated . this concerns , in particular , the entropy rate of the attenuator @xmath69 given an arbitrary initial state @xmath6 ( the factor @xmath59 is chosen for convenience only ) . this is in sharp contrast to the classical case : here the trivial identity @xmath70 , for a scalar @xmath71 , satisfied by the differential entropy @xmath72 of a random variable @xmath1 , is sufficient for the purpose of establishing fast convergence of the cou semigroup to the fixed point ( see appendix [ sec : classicalou ] ) . de palma et al . @xcite showed that the infimum @xmath73 over all states @xmath6 with a given entropy is achieved by a gaussian thermal state . this statement , combined with a lower bound on the corresponding quantity @xmath74 for the amplifier from @xcite , valid for all states @xmath6 , immediately yields inequality . to establish , we prove another lower bound on @xmath75 : more precisely , we show that the infimum @xmath76 over all states @xmath6 with mean photon number bounded by @xmath77 is achieved by a gaussian thermal state . the proof proceeds by reduction to the classical case using recent majorization - type results @xcite . the latter can be treated using the results from @xcite . this , then , provides the required lower bound and yields statement . it is not , however , tight enough to establish conjecture . understanding the relationship between entropy production rates along trajectories of the qou semigroup , i.e. , different quantities of the form @xmath78 for @xmath79 , remains an open problem . we believe that progress in this direction could help provide further evidence for ( or indeed lead to a proof of ) the validity of conjecture . our work in the quantum setting follows a long sequence of well - known existing arguments applicable to classical probability distributions . all geometric inequalities established here have classical counterparts , and their proofs are inspired by ( and directly generalize ) corresponding classical proofs . this raises the question of whether other analogues of classical results exist : for example , one may conjecture that there is a quantum counterpart of young s inequality for the convolution operation . it is not our intention to provide a complete review of this assortment of classical results : it is hardly possible to do justice to the many important developments in this area . we refer to the article @xcite for a survey of many known connections . instead , we briefly review some of the basic definitions and seminal results which are directly relevant to our work . in the classical setting we assume to have @xmath80-valued random variables with absolutely continuous density functions . for such a random variable @xmath1 with density function @xmath5 ( which we often assume to be non - vanishing everywhere for simplicity ) , a fundamental information measure of interest is the _ shannon ( differential ) entropy _ @xmath81 also denoted as @xmath72 . the _ entropy power _ is given by @xmath82 and also written as @xmath83 . up to a factor , this quantity coincides with the variance of a gaussian distribution having the same entropy . the _ divergence _ , or _ relative entropy _ , of two density functions @xmath84 is defined as @xmath85 the _ fisher information _ of a random variable @xmath1 with density function @xmath5 is defined as the following quantity ] corresponding to . the fisher information matrix provides a lower bound on the variance of an unbiased estimate of the unknown parameter @xmath86 according to the cramer - rao inequality . for our purposes , it is sufficient to consider the family . ] @xmath87 which is associated with the family of translated probability density functions @xmath88 the quantity @xmath89 is also often denoted as @xmath90 , to emphasize that it is the fisher information of a random variable @xmath1 which has density function @xmath5 . for two densities @xmath5 and @xmath91 describing random variables @xmath1 and @xmath2 , the convolution operation of interest is given by , and describes the addition of the two random variables . of particular interest is the case where one of the random variables is a centered normal distribution with unit variance . such a random variable is denoted as @xmath92 below . key results in this setting are the classical de bruijn identity : : : @xmath93 + this result was established by de bruijn and gives an important relation between the fisher information and the entropy when a random variable @xmath1 is perturbed under an additive gaussian noise channel . it is a key ingredient in proofs of many information - theoretic inequalities . a simple proof can be found in @xcite . the fisher information inequality : : : @xmath94 , \label{eq : fisherinformationineq}\end{aligned}\ ] ] + as well as the related inequality @xmath95 proofs of these inequalities are given in @xcite , @xcite , and @xcite . zamir @xcite gives a particularly useful proof which relies on the information - processing inequality . this inequality states that the application of a channel can not increase the fisher information . zamir s proof of the fisher information inequality can be generalized to the quantum case . the fisher information isoperimetric inequality : : @xmath96^{-1}\geq 1\ . \end{aligned}\ ] ] + this inequality implies that for gaussian states , the inverse of the fisher information has minimal sensitivity to additive gaussian noise @xcite . the concavity of the entropy power : : @xmath97 + this celebrated result establishes that the entropy power is a concave function along trajectories of the heat flow semigroup . a proof is given in @xcite , and some shorter ones are presented in ( * ? ? ? * ; * ? ? ? the entropy power inequality : : @xmath98 + stam @xcite gave a proof of the entropy power inequality which relies on the de bruijn identity and the fisher information inequality . the proof was later simplified by blachman @xcite and others ( * ? ? ? * ; * ? ? ? * ) . the isoperimetric inequality for entropies : : @xmath99 + as given in @xcite . the isoperimetric inequality for entropies implies that gaussians have minimal entropy power among random variables with fixed fisher information , and can be used to derive log - sobolev inequalities for the ornstein - uhlenbeck semigroup @xcite , as we review in appendix [ sec : classicalou ] . we consider a @xmath13-mode bosonic system with `` position '' and `` momentum '' operators @xmath100 of the @xmath101-th mode satisfying the canonical commutation relations @xmath102=i\delta_{j , k}i$ ] . denoting the vector of position- and momentum - operators by @xmath103 , the weyl displacement operators are defined as @xmath104 here @xmath105 is the matrix defining the symplectic inner product . the factor @xmath106 in the definition is for convenience only . from the commutation relations of position and momentum operators and the campbell - baker - hausdorff formula , it is straightforward to check that the weyl operators satisfy @xmath107 consider a state @xmath6 on @xmath13 modes . quantities of interest are the von neumann entropy @xmath108 , as well as the relative entropy @xmath109 . the latter expression is defined for positive operators @xmath110 , and we will assume without further comments that the states @xmath6 , @xmath111 have full rank . for a multi - parameter family @xmath112 of states depending smoothly on the parameters @xmath86 , the divergence - based quantum fisher information is defined as the trace of the fisher information matrix @xmath113 this definition quantifies the dependence of the states on the parameter @xmath86 in the neighborhood of @xmath114 . in the following , we will apply this definition to the family @xmath115 of states obtained by translating a given @xmath13-mode state @xmath6 : analogously to , where we translated a given probability density function @xmath5 , we define the translated states @xmath116 here translation by the parameter @xmath86 on phase space is achieved by means of the weyl operators . the corresponding quantity @xmath117 will simply be called the fisher information of @xmath6 . note that this definition matches that of the classical fisher information ( cf . ) . we emphasize that the concept of quantum fisher information is non - unique ( see @xcite ) , but the use of is sufficient for our purposes . consider the liouvillian defined on @xmath13 modes as @xmath118}\right]}\ .\end{aligned}\ ] ] ( the factor @xmath119 differs from the convention used in @xcite , but turns out to be convenient in the proof of theorem [ thm : entropy_power ] , as explained later . ) the one - parameter semigroup @xmath120 of completely positive trace - preserving maps ( cptpms ) generated by @xmath34 will be called the quantum ( heat ) diffusion semigroup . it has various nice properties : for example , as shown in @xcite , a quantum version of the de bruijn identity reads @xmath121 we remark that the proof of , which has subsequently been applied , e.g. , in @xcite and generalized in @xcite , involves certain formal manipulations whose rigorous justification remains an interesting mathematical problem : as a quantum counterpart of partial integration , for example , arguments under the trace need to be cyclically permuted . it is clear that such manipulations should be valid for sufficiently regular families of states ( and indeed , are established for gaussian states ) , but corresponding conditions are currently unknown . we believe that the recent introduction of schwartz operators in @xcite provides an appropriate framework to shed light on this aspect . in the following , we will assume that our states under consideration satisfy the required regularity assumptions . we hope that this issue will eventually be resolved in a similar manner as in the classical setting , where initial work by shannon @xcite was followed by a long sequence of papers with increasing rigor . in the case of the classical de bruijn identity , barron @xcite , based on stam s work @xcite , has shown validity for all random variables @xmath1 with finite variance . the map @xmath122 can be explicitly written as ( see @xcite ) @xmath123 we may interpret this as the result of applying a certain convolution operation to a gaussian distribution and a quantum state . more precisely , for @xmath9 , we define a convolution operation @xmath124 between a probability density function @xmath125 and a @xmath13-mode state @xmath6 by @xmath126 in this terminology , eq . becomes @xmath127 where @xmath128 is a unit - variance centered gaussian distribution , that is , @xmath129 to close this section , we list two elementary properties of the convolution which can be checked by straightforward calculation . if @xmath130 is the probability density function of a random variable @xmath1 and @xmath131 , then @xmath132 where the probability density function of the rescaled random variable @xmath133 is given by @xmath134 . addition of random variables corresponds to convolution in the following sense : @xmath135 where @xmath136 is defined as in . in this section , we present several statements about the convolution operation and the quantum fisher information . the key idea in establishing these results is the fact that the convolution operation constitutes data processing , and hence provides an inequality because of the monotonicity of relative entropy . this is expressed in the following lemma . in the classical setting , the analogous argument for obtaining the fisher information inequality was first emphasized by zamir @xcite . ( data processing inequality for convolution)[lem : dataprocv ] let @xmath137 be probability density functions with full support . then @xmath138 for any states @xmath110 . the quantum relative entropy satisfies the following scaling property for scalars @xmath139 : @xmath140 defining @xmath141 , we obtain ( using the translated states defined in eq . ) @xmath142 here the inequality we used is the joint convexity of the relative entropy ( see ( * ? ? ? * theorem 1 ) ) . for the third equality we used property , and for the fourth equality we used unitary invariance of the relative entropy and the trace as well as the fact that @xmath5 is a probability distribution . the last equality follows from the definition of the divergence . to convert lemma [ lem : dataprocv ] into a statement about fisher information , we need the following covariance property of the convolution operation : it breaks down translations of the state @xmath143 into translations of the function @xmath5 and the state @xmath6 , respectively . [ lemma : equiv ] let @xmath144 and @xmath9 . then @xmath145 where @xmath146 . according to definition and we have @xmath147 on the other hand , we similarly have @xmath148 through a simple change of variables and recalling that @xmath146 , the claim follows from and . combining the data processing inequality of lemma [ lem : dataprocv ] with lemma [ lemma : equiv ] , we prove an inequality which may be seen as a classical - quantum version of the stam inequality . it relates the fisher information of the state @xmath7 to the fisher informations of @xmath5 and @xmath6 , respectively . [ thm : infoineq ] ( quantum stam inequality ) let @xmath149 and @xmath9 . then @xmath150 where @xmath146 . in particular , @xmath151 let @xmath152 . for @xmath153 and @xmath154 , introduce the vector @xmath155 define the functions @xmath156 and @xmath157 from the definition of the relative entropy and the data processing inequality , for every @xmath158 , we have @xmath159 the second derivative of @xmath91 can be written as the limit @xmath160 and , therefore , is it bounded @xmath161 since @xmath162 and @xmath163 we conclude that @xmath164 we remark that above inequality can also be derived as a matrix inequality without tracing both sides . however , this is not required for our purposes and our definition of the fisher information . using lemma [ lemma : equiv ] , the left - hand side of this equation for @xmath165 can be written as @xmath166 the right - hand side of eq . can be simplified by noticing that both the classical and quantum fisher information matrices satisfy reparametrization formulas ( see ( * ? ? ? * lemma iv.1 ) ) @xmath167 therefore , taking @xmath168 leads to @xmath169 with , , and , we arrive at the desired result . finally , setting @xmath170 and @xmath171 , we obtain . in the next lemma we show how the quantum stam inequality implies an isoperimetric inequality for the quantum fisher information . ( quantum fisher information isoperimetric inequality ) the following inequality holds : @xmath172^{-1 } \geq 1\ .\end{aligned}\ ] ] recall that we have @xmath173 for a gaussian random variable @xmath92 . in the quantum stam inequality , take @xmath29 , then @xmath174 taking the limit @xmath33 , we arrive at the desired inequality . the isoperimetric inequality is tight for one mode ( @xmath175 ) and saturated by the gaussian thermal state @xmath176 with mean - photon number @xmath77 : as shown in appendix [ app : fisher ] , we have @xmath177^{-1 } & = \frac{1}{{{\bf n}}({{\bf n}}+1)}\log^{-2}\left(1+\frac{1}{{{\bf n}}}\right)\rightarrow\ 1\qquad \text { as } { { \bf n}}\rightarrow\infty\ .\end{aligned}\ ] ] as in classical information theory , the isoperimetric inequality for the quantum fisher information implies concavity of the entropy power under diffusion as an immediate consequence . we define the entropy power as @xmath178 ( concavity of the quantum entropy power ) the entropy power along trajectories of the diffusion semigroup is concave , i.e. @xmath179 two applications of de bruijn identity yield @xmath180 ^ 2+\frac{1}{2d}\frac{d}{dt}\bigg|_{t=0}j(e^{t{{{\mathcal l}}_{\mathsf{heat}}}}(\rho ) ) \right)\ . \end{aligned}\ ] ] the quantum fisher information isoperimetric inequality is equivalent to @xmath181 this completes the proof . to proceed , we establish bounds on the asymptotic scaling of the entropy power for large times . the following lemma follows directly from @xcite and ( * ? ? ? * corollary iii.4 ) ( see also @xcite ) . [ lem : scaling ] let @xmath182 be the generator of the classical heat diffusion semigroup on @xmath183 . in the limit @xmath184 , we have @xmath185 independent of the probability density function @xmath5 on @xmath183 and the @xmath13-mode state @xmath6 , respectively . having the same scaling for classical and quantum heat flows motivated the choice of constants in and . note that if @xmath1 is a random variable with probability density function @xmath5 , then @xmath186 is the probability density function of the random variable @xmath187 obtained by adding a centered gaussian random variable with unit variance , see eq . . [ lem : compability ] we have @xmath188 whenever @xmath189 . observe that writing @xmath190 for @xmath191 , we get @xmath192 on the other hand , @xmath193 we have used properties and . in the penultimate step we have used that for independent unit - variance centered gaussian random variables @xmath194 and @xmath195 , we have @xmath196 = @xmath197 . hence the two expressions are equal and the statement follows . the next theorem presents the entropy power inequality for both the convolution operation and the heat diffusion semigroup . [ thm : entropy_power](entropy power inequality ) for @xmath9 , the following inequality holds @xmath198 in particular , choosing @xmath199 as the distribution of a unit - variance centered gaussian defined in eq . , we have @xmath200 the proof is inspired by the proof of the entropy power inequality in @xcite , which itself is inspired by the proof for classical random variables by blachman @xcite . here we provide all necessary details modified to the present situation . for @xmath201 , define functions @xmath202 where @xmath203 is the generator of the classical heat semigroup as defined in eq . . the initial value problems @xmath204 have solutions @xmath205 . fix a pair of such solutions @xmath206 and @xmath9 . define @xmath207 these functions diverge , i.e. @xmath208 because of and @xmath209 . consider the function @xmath210 with the initial conditions , it follows that the claim of the theorem is equivalent to @xmath211 this inequality follows from two facts : first , the fact that @xmath212 as follows from the asymptotic scaling shown in lemma [ lem : scaling ] , the divergence , and the choice of @xmath213 ; second , the fact that @xmath214 it remains to show identity . computing the derivative of @xmath215 leads to the following equality @xmath216 define the fisher informations @xmath217 for @xmath218 . from the quantum de bruijn identity and the classical de bruijn identity we obtain @xmath219 with these identities and eq . , eq . is equivalent to @xmath220 where we have used the shorthand notation @xmath221 , @xmath222 , and @xmath223 , and similarly for @xmath224 . recall that by lemma [ lem : compability ] , @xmath225 , and by the quantum stam inequality , we have the bound @xmath226 inserting this upper bound into , we obtain @xmath227 this proves . as the last statement in this section , we derive an isoperimetric inequality for entropies from the entropy power inequality . ( isoperimetric inequality for entropies ) we have @xmath228 applying the de bruijn identity to the definition of the entropy power @xmath229 , we obtain @xmath230 on the other hand , for @xmath9 , the entropy power inequality reduces to @xmath231\geq 2\pi e\ .\end{aligned}\ ] ] therefore , taking the limit @xmath33 , we obtain the desired bound . to conclude this section , we remark that the isoperimetric inequality for entropies is tight in the one - mode case , @xmath39 : for a gaussian thermal state @xmath232 with mean photon number @xmath77 we obtain @xmath233 detailed calculations are provided in appendix [ app : isoperim ] . in this section we show that gaussian thermal states minimize the entropy rate for the one - mode attenuator semigroup among states with bounded mean photon number . the semigroup is defined by its generator @xmath234 we prove the following theorem : [ thm : gaussian_optimal ] for any @xmath235 , the infimum @xmath236 over states @xmath6 with mean photon number @xmath77 is achieved by the gaussian thermal state @xmath232 defined in . in particular , @xmath237 the proof proceeds by reduction to a recent result by de palma , trevisan , and giovannetti @xcite , where it is shown that gaussian thermal states minimize the entropy rate of the quantum attenuator among all states with a given input entropy . our argument additionally uses the recently introduced concept of fock majorization and associated results by jabbour , garcia - patron and cerf @xcite , as well as the ( classical ) gaussian maximum entropy principle . in more detail , we first show that the problem of minimizing the entropy rate reduces to the study of properties of a classical semigroup describing a pure - death process . this connection was used previously in @xcite ( see section [ sec : appornstein ] ) and is also an essential first step in @xcite . more explicitly , in section [ sec : gaussian : conn_class ] ( theorem [ thm : correspondence_class ] ) we prove the identity @xmath238\leq{{\bf n } } } \frac{d}{dt}\big|_{t=0 } h(e^{t{{\mathcal c}}_-}(p))\ .\end{aligned}\ ] ] here the infimum is over all probability distributions @xmath239 on @xmath240 with expectation value @xmath241 $ ] bounded by @xmath77 , the quantity @xmath242 is the shannon entropy of the distribution @xmath239 , and @xmath243 is the generator of a semigroup describing a classical pure - death process ( see for a precise definition ) . using the results of @xcite , we then show that the entropy rate for the classical process is optimized by a geometric distribution : in section [ sec : gaussian : class ] ( theorem [ thm : correspondence_class ] ) we prove that for @xmath235 @xmath244\leq{{\bf n } } } \frac{d}{dt}\big|_{t=0 } h(e^{t{{\mathcal c}}_-}(p))=-{{\bf n}}\log\left(1+\frac{1}{{{\bf n}}}\right)\ .\end{aligned}\ ] ] finally , in section [ sec : gaussian : gaussian ] , we calculate the entropy rate for the gaussian thermal state @xmath232 with mean photon number at most @xmath77 and find that @xmath245 for an initial state @xmath246 which is diagonal in the number state basis @xmath247 ( where @xmath248 ) , the time - evolved state has the same form , i.e. , @xmath249 . thus the attenuator semigroup @xmath250 gives rise to a semigroup @xmath251 on classical probability distributions by @xmath252 . its generator @xmath243 describes the dynamics of a classical pure - death process . it can be obtained from by inserting a number state @xmath253 : it is straightforward to check that @xmath254 in particular , the coefficients @xmath255 satisfy the system of differential equations @xmath256 with initial condition @xmath257 for @xmath258 . the expression on the right - hand side of defines the generator @xmath243 , that is , we have @xmath259 the following theorem reduces the problem of minimizing the entropy rate for the quantum attenuator semigroup to the classical problem of minimizing the entropy rate for this pure - death process . [ thm : correspondence_class ] we have the identity @xmath260\leq{{\bf n } } } \frac{d}{dt}\big|_{t=0 } h(e^{t{{\mathcal c}}_-}(p))\ , \end{aligned}\ ] ] where @xmath241=\sum_{n=0}^\infty n p_n$ ] and @xmath261 are the expectation value and entropy of the distribution @xmath239 , respectively . the proof of theorem [ thm : correspondence_class ] relies on results obtained in @xcite and @xcite . here we review the necessary definitions and results , and specialize them to our situation . let @xmath239 and @xmath262 be decreasing summable sequences of positive numbers . then @xmath239 weakly sub - majorizes @xmath262 , @xmath263 , if and only if @xmath264 let @xmath265 and @xmath266 be positive trace - class operators with eigenvalues @xmath267 and @xmath268 , arranged in decreasing order . then @xmath265 weakly majorizes @xmath266 , i.e. , @xmath269 , if and only if @xmath263 . also , @xmath265 majorizes @xmath266 , i.e. , @xmath270 , if the traces of @xmath265 and @xmath266 are identical and @xmath271 . let @xmath1 be a positive trace - class operator with eigenvalues @xmath272 in decreasing order . the fock rearrangement ( or passive rearrangement ) is defined as @xmath273 our restriction on the mean photon number forces us to consider an additional majorization relation . in contrast , in @xcite , where the authors restrict the input entropy instead , fock majorization does not need to be considered . the fock majorization relation , denoted @xmath274 , was introduced in @xcite as follows : @xmath275 where @xmath276 ( @xcite)[thm : entropy_ineq_fock ] for any state @xmath6 and all @xmath9 we have @xmath277 this implies that @xmath278 the first statement is shown in ( * ? ? ? * eq . ( vi.10 ) ) , and the second statement then follows from ( * ? ? ? * theorem iii.3 ) . ( ( * ? ? ? * lemma iv.9))[lem : depalmamajorx ] suppose @xmath279 are positive trace - class operators with @xmath280 then @xmath281 we begin proving theorem [ thm : correspondence_class ] by investigating the change of the mean photon number under the fock rearrangement procedure . [ lem : photonnumberrearrange ] for any state @xmath6 , the fock rearrangement does not increase the mean photon number @xmath282 where @xmath283 . let @xmath258 be arbitrary , and set @xmath284 with @xmath285 defined in . then clearly @xmath286 . setting @xmath287 , @xmath288 , we have @xmath289 , and @xmath290 , according to for @xmath291 , i.e. @xmath292 thus , lemma [ lem : depalmamajorx ] leads to @xmath293 according to definition , this is equivalent to @xmath294 it was shown in ( ( * ? ? ? * eq . ( 5 ) ) ) that @xmath295 the claim follows by taking @xmath296 in the last statement . since @xmath297 , theorem [ thm : entropy_ineq_fock ] implies @xmath298 with lemma [ lem : photonnumberrearrange ] , we therefore obtain @xmath299 since @xmath300 is a passive , fock - rearranged state , the right - hand side is a classical problem related to the pure - death process , and , therefore , can be replaced with the infimum of the entropy rate of a probability distribution evolving under a pure - death process . thus the claim follows . for a probability distribution @xmath239 on @xmath240 , let @xmath301 denote the entropy rate when @xmath239 evolves under the classical death - process @xmath243 ( cf . ) . we are interested in distributions @xmath239 with a fixed expectation value @xmath241=\sum_{n=0}^\infty n p_n$ ] . the main result we use here is ( * ? ? ? * theorem 24 ) : it states that for any probability distribution @xmath239 on @xmath240 , the quantity @xmath302 is bounded by @xmath303 where @xmath304 and where @xmath305 is the entropy of a geometric distribution with expectation value @xmath77 ( or equivalently the entropy of a gaussian state with mean photon number @xmath77 ) . we use this to show the following : [ thm : alldistributionslowerbound ] the infimum @xmath306 \leq { { \bf n } } } j_-(p)$ ] is achieved by the geometric distribution @xmath307 with @xmath308 . in particular , for @xmath235 , @xmath309 \leq { { \bf n } } } j_-(p ) = - 2{{\bf n}}\log\left(1 + \frac{1}{{{\bf n}}}\right)\ .\end{aligned}\ ] ] we show that @xmath309 = { { \bf n } } } j_-(p ) = j(p^{\mathrm{geo},{{\bf n}}})= - 2{{\bf n}}\log\left(1 + \frac{1}{{{\bf n}}}\right)\ .\label{eq : toshownvf}\end{aligned}\ ] ] since the right - hand side is monotonically decreasing with @xmath77 , eq . implies the claim of the theorem . the geometric distribution @xmath310 with @xmath311 has expectation value @xmath312 = { { \bf n}}$ ] and entropy @xmath313 . by the maximum entropy principle ( * ? ? ? * chapter 12 ) , we know that geometric distributions are the distributions with maximal entropy among all distributions with a fixed expectation value @xmath241 $ ] . therefore we have @xmath314 = { { \bf n}}\qquad \rightarrow\qquad h(p ) \leq g({{\bf n}})\ .\end{aligned}\ ] ] combining this with we obtain @xmath315 = { { \bf n } } } j_-(p ) \geq \inf_{p : h(p ) \leq g({{\bf n } } ) } j_-(p ) \geq 2\inf_{p : h(p ) \leq g({{\bf n } } ) } f(h(p))\ , \end{aligned}\ ] ] since @xmath5 is decreasing ( * ? ? ? * lemma 5 ) , it follows that @xmath316 = { { \bf n } } } j_-(p ) \geq 2 f(g({{\bf n}}))= -2 { { \bf n}}\log\left(1 + \frac{1}{{{\bf n}}}\right)\ . \label{eq : jminusboundcl}\end{aligned}\ ] ] however , since @xmath312 = { { \bf n}}$ ] and @xmath317 , we have equality in . from theorem [ thm : correspondence_class ] and theorem [ thm : alldistributionslowerbound ] we have @xmath318 let @xmath232 be the gaussian thermal state with mean photon number @xmath77 as defined in . we have @xmath319 for @xmath320 under the map @xmath321 , the state @xmath232 evolves into the thermal state @xmath322 according to @xmath323 in particular , @xmath324 therefore the considered infimum is achieved by the gaussian thermal state @xmath232 . in this section we consider a one - parameter group of cptp maps @xmath325 generated by the linear combination @xmath326 where @xmath327 is defined by , and @xmath328 is defined by @xmath329 in the following , we use the entropy production rates @xmath330 the factor @xmath59 here is for convenience to match de bruijn identity for @xmath22 . these quantities are related to the fisher information @xmath22 by @xmath331 because of de bruijn s identity and the fact that @xmath332 . [ lem : rate_decay ] let @xmath333 . then @xmath334 for any state @xmath6 , where @xmath335 , @xmath336 , and @xmath337 is the fixed point of @xmath61 . as mentioned in section [ sec : appornstein ] , the unique fixed point of the semigroup @xmath40 is the state @xmath338 where @xmath339 . in particular , this implies that for any state @xmath6 @xmath340 straightforward calculations show that the mean photon number of @xmath6 converges to the mean photon number of the fixed point @xmath341 with an exponential rate @xmath342 where @xmath343 . therefore @xmath344[eq : photon_number_convergence ] combining the last equality with , we find that @xmath345 with and , and setting @xmath336 , we obtain the desired equality . the choice of @xmath336 in the lemma is motivated by gaussian states : in appendix [ app : qou ] we show that for any gaussian state @xmath6 the following inequality holds @xmath346 furthermore , for any @xmath347 there exists a gaussian state @xmath6 such that @xmath348 let us now consider a specific example of a quantum ornstein - uhlenbeck process . [ ex:2 - 1_process_s ] consider @xmath349 , @xmath350 . then @xmath351 for any state @xmath6 with @xmath352 . in comparison , the entropy of the fixed point is @xmath353 . in lemma [ lem : rate_decay ] we can bound @xmath354 ( * ? ? ? * eq . ( 43 ) ) , and the linear combination of @xmath75 and @xmath355 can be bounded by the result of de palma et al . that is , for any state @xmath6 with @xmath356 , @xmath357 where @xmath358 , and @xmath359 . substituting @xmath360 gives @xmath361 that is , for any state with @xmath356 , we have @xmath362 with the choice @xmath349 , @xmath350 , the function @xmath363 is monotonically increasing for @xmath364 . for @xmath365 , the right - hand side of the last inequality is non - negative . the isoperimetric inequality for entropies for one mode ( @xmath175 ) can be written as @xmath366 for any @xmath367 , using @xmath368 , we get @xmath369 therefore @xmath370 [ lem : logsobolev ] let @xmath333 and @xmath371 . then @xmath372 for all states @xmath6 , where @xmath373 for any @xmath367 . here @xmath335 , and @xmath337 is the fixed point of @xmath61 . in particular , choosing @xmath374 and @xmath375 , we have @xmath376 where @xmath377 . combining and leads to @xmath378 using this inequality in lemma [ lem : rate_decay ] we obtain the result . for the second part of the statement , theorem [ thm : gaussian_optimal ] shows that @xmath379 where @xmath232 is the gaussian thermal state with mean photon number @xmath77 . . then follows directly from the first part of the statement with the choice @xmath380 , inequality , and the choice of @xmath336 . considering the same specific qou process as in example [ ex:2 - 1_process_s ] we obtain the same rate of convergence to its fixed point , but now only for states with low mean photon number instead of large initial entropy as in example [ ex:2 - 1_process_s ] . [ ex:2 - 1_process_n ] consider @xmath349 and @xmath350 . then @xmath351 for any state @xmath6 with @xmath381 . in comparison , the mean photon number of the fixed point is @xmath382 . choosing @xmath349 and @xmath350 in lemma [ lem : logsobolev ] , we obtain @xmath383 the right - hand side is monotonically decreasing and for @xmath384 , it is non - negative . we have established new information - theoretic inequalities for bosonic quantum systems . our inequalities are motivated by and directly generalize well - known existing results concerning the sum of two real - valued random variables . they also complement recent results concerning the `` addition '' of two bosonic quantum states by means of a beamsplitter : we consider a hybrid operation which amounts to a certain way of combining a classical probability distribution on phase space with a quantum state . mathematically , our work thus makes progress towards a unified view of three types of convolution operations : the convolution of two classical probability density functions , of two quantum wigner functions , and of a pair consisting of a classical probability density function and a quantum wigner function . operationally , our results extend entropic characterizations of classical additive noise and quantum additive noise to so - called classical noise in bosonic systems . indeed , the proofs of our main inequalities , which include hybrid versions of the fisher information and entropy power inequalities , are formally very similar to existing proofs in the fully classical as well as fully quantum settings . the consideration of the hybrid classical - quantum convolution operation brings additional advantages , however : it allows for the study of infinitesimal gaussian perturbations to a given quantum state . in contrast , existing fully quantum entropy power inequalities are not directly amenable to such arguments ( at least not in an obvious way ) since basic uncertainty relations prevent us from making sense of e.g. , a gaussian state with infinitesimal variance . mirroring the derivation of the isoperimetric inequality from the brunn - minkowski inequality ( where a given set is perturbed by adding an infinitesimally small ball ) , we obtain a quantum isoperimetric inequality relating fisher information and entropy power . a striking simple consequence of the latter is the statement that the entropy power is a concave function of time for the quantum heat diffusion semigroup : again , this provides a quantum generalization of a fundamental result about the classical heat equation . let us conclude by mentioning a few potential directions for future work , as well as some open problems . for concreteness and simplicity , we have considered the simplest non - trivial definition of a hybrid convolution operation defined on multiple modes , as studied in @xcite . one may generalize our convolution and the associated results by considering additional ( linear ) operations along the lines of @xcite . from the point of view of information theory , it is also interesting to examine the implications of our results for the capacity of the classical noise channels similarly to @xcite . on a more speculative side , one may wonder whether alternative quantum generalizations of the results considered here exist , especially related to the conjectured photon number inequality by guha , erkmen and shapiro @xcite . these authors ( and subsequent work such as @xcite ) suggest replacing the quantum entropy power @xmath385 by the arguably more natural expression @xmath386 . this is the mean photon number of a gaussian state with identical entropy as @xmath6 . it appears that at least for our isoperimetric inequality , such a generalization would require more than a nave substitution as there is no meaningful lower bound on the product @xmath387 even for gaussian states . while this may be considered as additional mathematical justification for our formulation of these inequalities , we believe that progress in this direction could be helpful in resolving , e.g. , our conjecture concerning the convergence rate to the fixed point of the quantum ornstein - uhlenbeck ( qou ) semigroup . for the latter problem , apart from proving our conjecture , it would also be interesting to obtain multi - mode generalizations . this concerns , in particular , the entropy production rates for the qou semigroup . here , a resolution of the conjecture of @xcite for the multi - mode attenuator would likely provide important insights . finally , we mention some challenging mathematical problems resulting from our work . for example , while our isoperimetric inequality is tight for gaussian states , necessary and sufficient conditions for equality in most of our statements are currently unknown . finally , a rigorous discussion of the family of states for which the de bruijn identity holds , possibly using the framework of schwartz operators @xcite , would be an interesting task for future work . after posting our paper to the arxiv , we were made aware of concurrent related work by rouz , datta , and pautrat . their paper has now been made available @xcite . rk thanks the organizers of the workshop on hypercontractivity and log - sobolev inequalities in banff and r. f. werner for interesting discussions . rk and sh are supported by the technische universitt mnchen - institute for advanced study , funded by the german excellence initiative and the european union seventh framework programme under grant agreement no . they acknowledge additional support by dfp project no . k05430/1 - 1 . av is supported by the john templeton foundation grant no . 48322 , and by the basque government through the berc 2014 - 2017 program and by spanish ministry of economy and competitiveness mineco : bcam severo ochoa excellence accreditation sev-2013 - 0323 . in this appendix we discuss known classical results for the reader s convenience : we briefly review the relationship between the isoperimetric inequality for classical fisher information , the log - sobolev inequality , and the rate of convergence to the fixed point for the classical ornstein - uhlenbeck semigroup . these arguments were given by carlen @xcite for a particular element of the two - parameter family of ornstein - uhlenbeck processes . here we specialize to real - valued random variables , but allow arbitrary parameters in order to illustrate the parallels to the qou semigroup . we also explicitly discuss the convergence to the fixed point , which is only implicit in @xcite but appears to be folklore . let @xmath388 be a probability density on @xmath389 of a real - valued random variable @xmath1 . the classical ornstein - uhlenbeck ( cou ) process , for @xmath390 and @xmath391 is given by the fokker - planck equation @xmath392+\frac{\sigma^2}{2}\frac{\partial^2 f}{\partial x^2}={{\mathcal a}}_{\theta,\sigma^2}(f)\ .\end{aligned}\ ] ] ( carlen considers the case where @xmath393 and @xmath394 . ) the solution to this equation can be written ( in terms of random variables ) as @xmath395 where @xmath396 is an independent centered gaussian random variable with unit variance . the stationary solution @xmath397 therefore is a centered gaussian distribution with variance @xmath398 . in particular , implies that the second moments satisfy @xmath399&=e^{-2\theta t}\mathbb{e}[x_0 ^ 2]+ ( 1-e^{-2\theta t})\frac{\sigma^2}{2\theta}\ .\label{eq : momentsevolutionc}\end{aligned}\ ] ] we use that the relative entropy between a random variable @xmath1 and a centered normal variable @xmath400 is given by @xmath401 $ ] as can be verified easily . in particular , the relative entropy between the solution at time @xmath402 and the fixed point @xmath403 is given by @xmath404\ .\label{eq : relativeentropytofixed}\end{aligned}\ ] ] furthermore , we have the following de bruijn - type identity : let @xmath405 be of the form , i.e. , a solution to the cou process . let @xmath406 denote the fisher information of a random variable @xmath1 with distribution function @xmath5 . then @xmath407 observe that the second summand essentially stems from the fact that entropies transform very simply under rescaling of random variables , namely @xmath408 eq . significantly simplifies the analysis . such a property does not hold in the quantum case : as a consequence , we do not have a simple expression in terms of @xmath409 only . the derivative of the entropy along a semigroup with generator @xmath410 is given by the expression @xmath411 . using this fact gives @xmath412\right)\log f(x)dx-\frac{\sigma^2}{2}\int \frac{\partial^2 f(x)}{\partial x^2 } \log f(x)dx\ .\end{aligned}\ ] ] denoting @xmath413 , we obtain @xmath414 where we have used partial integration and the fact that boundary terms vanish twice . similarly , we have @xmath415 by partial integration . combining these statements gives the claim . following @xcite , we can write the isoperimetric inequality @xmath416 as @xmath417 in particular , using @xmath368 , we get @xmath418 for any @xmath367 . using inequality , it is straightforward to show the following . [ thm : carlengeneralized ] let @xmath405 be of the form , i.e. , a solution to the cou process with parameters @xmath419 . then @xmath420 note that because we are considering a semigroup , this result immediately implies that @xmath421 also , this result is tight : if @xmath422 is a centered gaussian random variable with variance @xmath423 , then implies @xmath424 where the variance of @xmath425 is @xmath426 inserting into , using @xmath427 yields @xmath428 in particular , we obtain @xmath429 according to eqs . , and , we have @xmath430 + 2\theta\ .\label{eq : ddterivativeclassical}\end{aligned}\ ] ] combining with yields @xmath431\ .\label{eq : dxzerozsigma}\end{aligned}\ ] ] combining with yields @xmath432\ .\label{eq : logsobolevclassical } \end{aligned}\ ] ] the claim then follows by choosing @xmath433 . consider a one - mode gaussian thermal state @xmath232 with mean photon number @xmath235 . its entropy is @xmath434 under the diffusion semigroup , the state @xmath232 evolves as @xmath435 in particular , by the de bruijn identity @xmath436 also , @xmath437 calculating the right - hand side of the quantum fisher information isoperimetric inequality , we obtain @xmath177^{-1 } & = \frac{1}{{{\bf n}}({{\bf n}}+1)}\log^{-2}\left(1+\frac{1}{{{\bf n}}}\right)\rightarrow\ 1\qquad \text { as } { { \bf n}}\rightarrow\infty\ .\end{aligned}\ ] ] consider a one - mode gaussian thermal state @xmath232 with mean photon number @xmath77 . from eq . we know that @xmath438 and that the entropy power of @xmath232 is given by ( cf . ) @xmath439 combining these two expressions , we see that the left - hand side of is @xmath440 in this section we show that for a one - mode gaussian state @xmath6 the following inequality holds : @xmath441 furthermore , this statement is optimal : for any @xmath347 there exists a gaussian state @xmath6 such that @xmath348 first , we note that the entropy of a gaussian state does not depend on its first moments . hence it follows that for a gaussian state the right - hand side of eq . does not depend on its first moments , and it is suffices to consider centered gaussian states . let @xmath6 be a one - mode centered gaussian state with mean - photon number @xmath77 and covariance matrix given by @xmath445 , where @xmath446 , and @xmath447 with @xmath448 and @xmath449 . denote @xmath450 then we have the covariance matrix of the time - evolved state @xmath452 is @xmath453 where @xmath454 therefore writing @xmath455 $ ] for the entropy of a gaussian state with covariance matrix @xmath443 , we have @xmath456 & = s\left[c_1^\pm(t ) \kappa s^t s + c_2^\pm(t ) { \mathop{\rm id}}\right ] = s\left[c_1^\pm(t ) \kappa o_2^t k^t o_1^t o_1 k o_2 + c_2^\pm(t ) o_2^t o_2 \right ] \\ & = s\left[c_1^\pm(t)\kappa k^t o_1^t o_1 k + c_2^\pm(t ) { \mathop{\rm id}}\right]\\ & = s\left[\begin{pmatrix } c_1^{\pm}(t ) \kappa z^2 + c_2^{\pm}(t ) & 0 \\ 0 & c_1^{\pm}(t)\kappa / z^2 + c_2^{\pm}(t ) \end{pmatrix}\right]\ . \label{eq : sympleval}\end{aligned}\ ] ] the symplectic eigenvalue of the matrix argument in is the square root of its determinant : @xmath457 the entropy of the time evolved state is @xmath458 where @xmath459 and @xmath460 . by the chain rule , we have that @xmath461 since @xmath462 , we only need to find @xmath463 . from lemma [ lem : rate_decay ] it is clear that we are interested in minimizing @xmath442 . both expressions in have a minimum at @xmath470 . therefore , with @xmath374 and @xmath471 , from lemma [ lem : rate_decay ] we obtain for fixed @xmath473 , the function @xmath474 satisfies @xmath475 . since @xmath474 is differentiable ( in fact smooth ) for @xmath235 , we can find the global minimum by finding the zeros of the derivative @xmath476 the only solution is @xmath477 , and since @xmath478 it is the minimum . the value of the minimum is @xmath479 , hence it follows that @xmath480 for all @xmath481 and we have

in biological and synthetic materials , many important processes involve charges that are present in a medium with spatially varying dielectric permittivity . to accurately understand the role of electrostatic interactions in such systems , it is important to take into account the spatial dependence of the permittivity of the medium . however , due to the ensuing theoretical and computational challenges , this inhomogeneous dielectric response of the medium is often ignored or excessively simplified . we develop a variational formulation of electrostatics to accurately investigate systems that exhibit this inhomogeneous dielectric response . our formulation is based on a true energy functional of the polarization charge density . the defining characteristic of a true energy functional is that at its minimum it evaluates to the actual value of the energy ; this is a feature not found in many commonly used electrostatic functionals . we explore in detail the charged systems that exhibit sharp discontinuous change in dielectric permittivity , and we show that for this case our functional reduces to a functional of only the surface polarization charge density . we apply this reduced functional to study model problems for which analytical solutions are well known . we demonstrate , in addition , that the functional has many properties that make it ideal for use in molecular dynamics simulations .

1309.7073

many biological systems involve mobile or fixed charges , the electrostatic response of which is key to our understanding of the physical behavior of such systems . nucleic acids and many proteins are charged in physiologically relevant conditions . the interactions that arise from the presence of these charges are crucial in the determination of the structure and function of these polymers @xcite . biological processes such as signaling in cells involve the creation of electrical potential differences and transport of ions across the cellular membrane @xcite . on the other hand , in the design and stabilization of many different synthetic structures , the electrostatic forces play a major role . examples include self - assembled colloidal dispersions @xcite , polynucleotide adsorption @xcite , dna precipitation in multivalent salts @xcite , overcharged surfaces @xcite , patterned surfaces @xcite , spontaneous adsorption of ions at liquid - liquid interfaces@xcite , faceted thin shells @xcite , viral assembly @xcite , and various dynamical processes including dna gel electrophoresis @xcite and related polyelectrolyte separation process @xcite . theoretical investigations of these interesting materials and biological systems must therefore accurately incorporate electrostatic interactions . under the conditions of high ionic concentration or in the presence of multivalent ions , when the finite size of ions and the inter - ionic correlations become significant , mean field theories are generally found inadequate to capture important electrostatic effects @xcite . for arbitrarily curved geometries or where the dielectric response of the medium is not homogeneous , the associated electrostatics problem gets too complicated even for the more sophisticated analytical treatments @xcite , and the use of numerical techniques becomes necessary . however , an accurate computer simulation involving electrostatic interactions presents its own challenges . the first challenge stems from the long range of the coulomb force which implies that every charge interacts with every other charge . thus , a system of @xmath0 charges requires an expensive @xmath1 force ( or energy ) calculation at every simulation step . attempts to ameliorate this scaling behavior have resulted in the development of several methods : e.g , ewald summation , particle - mesh methods , fast multipole methods @xcite , and local electrostatics algorithms @xcite . the other main challenge arises due to the presence of dielectric heterogeneities in the medium , and this constitutes the main focus of the present paper . free charges polarize their surrounding dielectric medium and the resulting net polarization and electric fields can have complex behavior . modeling of systems with electrostatic interactions should , ideally , incorporate this dielectric response of the medium . an explicit inclusion of the medium components ( molecules of the solvent , for example ) as a part of the model for the real system renders a prohibitively large number of degrees of freedom to simulate , such that even with the most efficient methods @xcite direct simulation becomes too computationally expensive . in many cases , the introduction of a spatially varying dielectric constant in the model is sufficient to capture the effects of polarizability and describe the dielectric response . in the simplest case of a uniform dielectric response , a single dielectric constant can describe a coarse - grained medium , and simulations can proceed as they would in free space , albeit with a scaled coulomb s law . however , most real situations involve regions with different dielectric response , as is the case for proteins within an aqueous cellular medium or for emulsions where oil and water are partitioned @xcite . in the presence of this varying dielectric response , the simplest form of coulomb s law breaks down and one has to accurately solve the poisson equation , at _ each _ simulation step , to obtain the necessary force ( energy ) information for the propagation of ionic coordinates . this adversely affects the stability and efficiency of the resulting numerical procedure . because of these computational challenges , the problem of treating dielectric heterogeneities in charge simulations continues to be a subject of intense research @xcite . a few previous attempts towards the solution of the problem of inhomogeneous dielectric response have involved a reformulation of electrostatics as a variational problem @xcite . here , the solution to the poisson equation is obtained not as a solution of a differential equation , but as the extremum of a suitably constructed functional @xcite . an important advantage of adopting a variational approach is that it offers the possibility of bypassing the effort to explicitly optimize the functional at each step by framing the problem in such a way that the very process of updating the simulation guarantees the optimization of the functional . in other words , since the optimization of the functional is equivalent to the solution of the poisson equation , we are offered the possibility of solving the poisson equation _ on - the - fly _ in tandem with the generation of the new charge configuration . however , this possibility arises only when the variational approach is based on an energy functional : a functional which _ minimizes _ to the _ true _ electrostatic energy . we note that in the literature , there is an abundance of functionals that are _ not _ energy functionals @xcite and therefore the numerical schemes associated with these functionals do not employ the ideas of dynamical optimization . in addition to the need for a true energy functional for an efficient numerical implementation of the variational procedure , it is also crucial to produce a functional with the appropriate function variable(s ) . in many cases a particular electrostatic variable offers distinct advantage over others . for example , a coarse - grained model often employed to study phenomena in both biological and synthetic settings is that of ions present in piecewise - uniform dielectrics separated by sharp interfaces ( see fig . [ fig1 ] ) . in this case , it is advantageous to choose the polarization charge density as the variable to solve for , rather than the electrostatic potential or the polarization vector . this is because when the dielectric response of the medium is piecewise uniform , the unknown polarization charge density resides only on the interface , and thus we are presented with the possibility to reduce the full three - dimensional electrostatic problem to a two - dimensional one . a uniform dielectric medium characterized by permittivity @xmath2 separated from another region of permittivity @xmath3 by a boundary that is treated as a thin interface . positive ( red ) and negative ( green ) ions are present in either media . coarse - grained models of this kind are often employed to study systems such as charged bio - macromolecules in aqueous solution or charged colloidal suspensions . ] in the light of the above discussion , it is clear that an energy functional of the polarization charge density would provide many advantages with regards to the numerical implementation of the variational method formulated to study problems involving dielectric heterogeneities . in the literature , one finds many functionals along with their associated numerical minimization procedures . though a functional of polarization density for a general system has been developed in ref . , it is not an energy functional . similarly , the functionals derived in refs . are energy functionals , but the basic field variables employed are vector fields such as the polarization vector @xcite or the electric field @xcite , not the polarization charge density . attard @xcite has provided an energy functional of the surface polarization charge density , but this functional is derived for a specific system that involves all free charges to be constrained in one uniform dielectric medium . in a recent paper ( ref . ) , we introduced a variational formulation of electrostatics that produced an energy functional of the polarization charge density . this functional reads : @xmath4&=\frac{1}{2}\iint \rho_{\mathbf{r}}g_{\mathbf{r},\mathbf{r ' } } \left(\rho_{\mathbf{r'}}+\omega_{\mathbf{r'}}[\omega]\right ) d^{3}r'd^{3}r\\ & - \frac{1}{2}\iint \omega_{\mathbf{r}}[\omega ] g_{\mathbf{r},\mathbf{r ' } } \left(\omega_{\mathbf{r ' } } - \omega_{\mathbf{r'}}[\omega ] \right ) d^{3}r'd^{3}r , \end{split}\ ] ] where @xmath5 is the polarization charge density and @xmath6 $ ] is both a functional of @xmath7 and a function of @xmath8 , and is defined as @xmath9 = \nabla\cdot\left ( \chi_{\mathbf{r}}\nabla\int g_{\mathbf{r},\mathbf{r ' } } \left ( \rho_{\mathbf{r ' } } + \omega_{\mathbf{r ' } } \right ) d^{3}r'\right).\ ] ] in the above equations @xmath10 is the susceptibility of the medium , @xmath11 is the free charge density , and @xmath12 is the green s function in free space . in our derivation we assumed that the medium polarization obeys linear response and also assumed the existence of dirichlet boundary conditions . both these assumptions are standard from the point of view of constructing electrostatic free energy functionals . @xmath13 $ ] is applicable to any configuration of free charges and works for arbitrary spatial variation in dielectric response . we also provided the expression of this functional for the important case of point charges present in piecewise - uniform dielectrics and developed a car - parrinello molecular dynamics scheme to study the equilibrium properties of such systems . as an application , we computed the density profiles for monovalent salt ions near a spherical emulsion droplet separating two liquids of different dielectric constants . in this paper , we present a detailed derivation of @xmath13 $ ] , showing steps that were omitted in ref . for the lack of space , and discuss the important features of our variational formulation that enable the production of an energy functional . also , we explore in detail , with several examples , the particular case of piecewise - uniform dielectric response . in addition , we provide the proofs that show that @xmath13 $ ] is an energy functional . in the supplementary information of ref . we proved the minimum property of this functional at its extremum . here we show that upon extremizing this functional one finds the usual electrostatic relation for the polarization charge density and the extremum value of the functional coincides with the true electrostatic energy of the system . for the sake of completeness we also include the proof of the functional becoming a minimum at its extremum . the paper is organized as follows . in sec . [ sec : vform ] we derive @xmath13 $ ] and discuss the key aspects of our variational formulation . in sec . [ sec : pud ] , we explicitly specialize @xmath13 $ ] to the case of sharp dielectric interfaces , and apply the resulting functional to some simple interfacial shapes . in sec . [ sec : numeric ] we demonstrate a grid - based numerical procedure to implement the functional minimization and some concluding remarks are made in sec . [ sec : conclusion ] . in appendix [ sec : ext ] we prove that @xmath13 $ ] is an energy functional . finally , appendix [ sec : app ] discusses the application of our variational principle to the simple case of a uniform dielectric . in the first half of this section we provide a detailed derivation of the functional given in eq . . the second half discusses the important features of the variational formulation that determine the extremal properties of the resulting functional . gaussian units are used throughout . we begin with the standard expression for the electrostatic energy written in its equivalent functional form : @xmath14=\frac{1}{8\pi}\int \epsilon_{\mathbf{r}}\left|\mathbf{e}_{\mathbf{r}}\right|^{2 } d^{3}r.\ ] ] here @xmath15 is the dielectric permittivity and @xmath16 is the electric field . following the formulation introduced in ref . , we include gauss s law as a constraint to the functional in via the lagrange multiplier @xmath17 , obtaining @xmath18=\frac{1}{8\pi}\int \epsilon_{\mathbf{r}}|\mathbf{e}_{\mathbf{r}}|^{2 } d^{3}r \nonumber\\ & & -\int \phi_{\mathbf{r}}\left(\nabla \cdot \left(\frac{\epsilon_{\mathbf{r}}\mathbf{e}_{\mathbf{r}}}{4\pi}\right ) -\rho_{\mathbf{r}}\right ) d^{3}r.\end{aligned}\ ] ] we note that @xmath17 can be shown to coincide with the electrostatic potential at equilibrium . also , we take @xmath19 to depend parametrically on the free charge density @xmath11 , implying that the latter will not be used as a variational field . we assume that the medium polarization @xmath20 obeys linear response : @xmath21 , where @xmath10 is the susceptibility connected to @xmath15 by the relation @xmath22 . employing this relation between @xmath15 and @xmath10 we now introduce the field variable @xmath20 in in the following way : @xmath23=\frac{1}{8\pi}\int |\mathbf{e}_{\mathbf{r}}|^{2 } d^{3}r + \int \frac{|\mathbf{p_{\mathbf{r}}}|^{2}}{2\chi_{\mathbf{r } } } d^{3}r \\ & - \int \phi_{\mathbf{r}}\left ( \nabla \cdot \frac{\mathbf{e}_{\mathbf{r}}}{4\pi } + \nabla \cdot \mathbf{p}_{\mathbf{r } } - \rho_{\mathbf{r}}\right ) d^{3}r . \end{split}\ ] ] variations of with respect to @xmath16 and @xmath17 give : @xmath24 in obtainig the above variations we make use of the dirichlet boundary condition ( dbc ) : @xmath25 where @xmath26 is a boundary invoked at infinity . all the surface integrals that appear as a consequence of taking the variations are rendered void by the use of dbc . from it is clear that @xmath17 must be the electrostatic potential . using we eliminate @xmath16 from and obtain @xmath27 eq . is the poisson equation satisfied by the potential @xmath17 when the charge density in _ free space _ is given by @xmath28 . the solution of the above equation can be written as @xmath29 where @xmath30 is the green s function in free space which satisfies the equation : @xmath31 and is given by @xmath32 note that @xmath30 also obeys dbc . substituting @xmath17 from in , we obtain @xmath16 in terms of @xmath20 : @xmath33 using and we eliminate @xmath17 and @xmath16 from to obtain a functional with @xmath20 as the sole variational field : @xmath34=\int\frac{|\mathbf{p}_{\mathbf{r}}|^{2}}{2\chi_{\mathbf{r}}}d^{3}r & + \frac{1}{2}\iint\left ( \rho_{\mathbf{r } } - \nabla\cdot\mathbf{p}_{\mathbf{r } } \right ) g_{\mathbf{r},\mathbf{r ' } } \\ & \times\left(\rho_{\mathbf{r ' } } - \nabla\cdot\mathbf{p}_{\mathbf{r ' } } \right ) d^{3}r'd^{3}r . \end{split}\ ] ] it can be shown that the correct constitutive relation between the polarization field and the electric field is obtained as a result of the extremization of the above functional @xcite . furthermore , one can prove that @xmath35 $ ] is an energy functional ; that is , its minimum computes the equilibrium electrostatic energy @xcite . the functional in has been obtained previously @xcite , but with different derivations than ours . we now show how to transform @xmath35 $ ] to an energy functional of the polarization charge density @xmath5 . this transition begins by inserting the definition of @xmath5 , namely , @xmath36 in by means of a lagrange multiplier @xmath37 : @xmath38=\int\frac{|\mathbf{p}_{\mathbf{r}}|^{2}}{2\chi_{\mathbf{r}}}d^{3}r + \frac{1}{2}\iint\left ( \rho_{\mathbf{r } } + \omega_{\mathbf{r } } \right ) g_{\mathbf{r},\mathbf{r'}}\\ & \times\left(\rho_{\mathbf{r ' } } + \omega_{\mathbf{r ' } } \right ) d^{3}r'd^{3}r -\int\psi_{\mathbf{r}}\left ( \omega_{\mathbf{r } } + \nabla\cdot\mathbf{p}_{\mathbf{r } } \right ) d^{3}r . \end{split}\ ] ] we note that @xmath37 will soon be shown to coincide with the electrostatic potential @xmath17 at equilibrium . taking variations of the above functional with respect to @xmath5 and @xmath20 gives the following relations : @xmath39 equation expresses @xmath37 in terms of @xmath5 . substituting @xmath37 from in expresses @xmath20 in terms of @xmath5 : @xmath40 at this point , by using and , we can eliminate @xmath37 and @xmath20 from in favor of @xmath5 and complete the desired transformation . however , while the functional that results from this procedure does single out the correct physical quantity upon extremization , it becomes a maximum , not a minimum , at equilibrium . we elaborate more on this observation in sec . [ sec : dis ] . to obtain the functional of @xmath5 with the desired extremal behavior , one must resist substitution at this stage and instead take the unutilized variation of @xmath41 $ ] with respect to @xmath37 which leads to @xmath42 substituting @xmath20 from in the above equation we obtain @xmath43 the above relation , as one can tell by inspection , is the correct physical relation that the polarization charge density must satisfy . equations and , when viewed together , and compared with imply that @xmath37 is indeed the electrostatic potential at equilibrium . at this point the substitution of @xmath37 , @xmath20 , and @xmath5 from , , and respectively , into the functional of leads to our central result : the functional in eq . . in appendix [ sec : ext ] we prove that @xmath13 $ ] is an energy functional and its minimization provides the correct induced charge density . we now analyze in some depth why some substitutions lead to an energy functional and why others do not . in this section we elaborate on some key observations made during the process of deriving @xmath13 $ ] . we noted in sec . [ sec : deriv ] that not all substitutions to eliminate field variables from in favor of @xmath5 lead to the desired result . we observed that @xmath37 and @xmath20 can be eliminated from using equations and , thus leading to a functional with @xmath5 as the sole variational field . one can show that this process results in a functional @xmath44 $ ] with the functional density : @xmath45\right ) / 2 - \omega_{\mathbf{r } } g_{\mathbf{r},\mathbf{r'}}\left(\omega_{\mathbf{r ' } } - \omega_{\mathbf{r'}}[\omega]\right ) / 2 $ ] . upon extremization , @xmath44 $ ] singles out the correct physical quantity , but becomes a maximum at equilibrium . in fact , @xmath44 $ ] is exactly the negative of the functional in ref . , neither of which are energy functionals . we note that functionals @xmath13 $ ] and @xmath44 $ ] share a common structure : the expression for the total electrostatic energy ( the first term in either functional ) is constrained by the correct physical relation that @xmath5 must satisfy , namely @xmath46 = 0 $ ] . the one but crucial difference between these functionals is in the choice of the constraint that is enforced by means of lagrange multipliers . while the constraints themselves might appear equivalent , their different explicit forms can endow the functional with different properties . previous functionals that lack some desirable properties can be understood as arising from deficient constraint expressions . our current formulation provides the appropriate constraint form . it is equally important to point out that the set of substitutions that we employed in sec . [ sec : deriv ] to arrive at the desired result are not the only ones that lead to an energy functional . due to the iterative nature of eq . , different sets of substitutions leading to different energy functionals are possible . for example , resisting substitutions post eq . , and instead employing to replace @xmath5 with @xmath47 $ ] in leads to a new relation between @xmath37 and @xmath5 . starting with this new relation , we can execute the same cycle of steps as before to obtain new expressions for @xmath20 and @xmath5 in terms of @xmath5 . we thus arrive at the following set of relations : @xmath48\right ) d^{3}r ' , \end{split } \\ \begin{split}\label{eq : newp } \mathbf{p}_{\mathbf{r } } = -\chi_{\mathbf{r}}\nabla\int g_{\mathbf{r},\mathbf{r ' } } \left(\rho_{\mathbf{r ' } } + \omega_{\mathbf{r'}}[\omega]\right ) d^{3}r ' , \end{split } \\ \begin{split}\label{eq : newomega } \omega_{\mathbf{r } } = \nabla\cdot\left ( \chi_{\mathbf{r}}\nabla\int g_{\mathbf{r},\mathbf{r ' } } \left ( \rho_{\mathbf{r ' } } + \omega_{\mathbf{r'}}[\omega ] \right ) d^{3}r'\right ) . \end{split}\end{aligned}\ ] ] we note that at equilibrium the above obtained relations for @xmath37 , @xmath20 , and @xmath5 coincide with the corresponding relations obtained in sec . [ sec : deriv ] : equations , , and . at this point , if we substitute @xmath37 , @xmath20 , and @xmath5 from , , and respectively into the functional of , one can show that the resulting functional ( @xmath49 $ ] ; see eq . below ) is also an energy functional . as should be evident , the above outlined cycle of steps can be repeated many times , yielding more energy functionals . specifically , we find a family of functionals @xmath50\}$ ] with @xmath51 , where the @xmath52 member has the form @xmath53=\frac{1}{2}\iint \rho_{\mathbf{r}}g_{\mathbf{r},\mathbf{r ' } } \left(\rho_{\mathbf{r'}}+\omega^{(n)}_{\mathbf{r'}}\right ) d^{3}r'd^{3}r\\ & - \frac{1}{2}\iint \omega^{(n)}_{\mathbf{r } } g_{\mathbf{r},\mathbf{r ' } } \left(\omega^{(n-1)}_{\mathbf{r ' } } - \omega^{(n)}_{\mathbf{r ' } } \right ) d^{3}r'd^{3}r . \end{split}\ ] ] in eq . , @xmath54 is both a function of @xmath8 and a functional of @xmath5 , but we have suppressed the functional part of the notation for brevity . @xmath54 is given by : @xmath55 with @xmath56 $ ] defined as @xmath56 = \int\delta(\mathbf{r } - \mathbf{r'})\omega(\mathbf{r'})d^{3}r ' = \omega$ ] . note that , by letting @xmath57 in we obtain @xmath58 = \omega[\omega]$ ] , where @xmath47 $ ] is given by eq . . it can be shown that for every @xmath59 the functional given by is an energy functional . although these functionals are different from one another , upon extremization each of them give the same iterative relation , eq . , and all minimize to the true electrostatic energy . proofs of these assertions are similar to the ones that appear in appendix [ sec : ext ] . by letting @xmath57 in it is easy to see that @xmath60 = \mathscr{f}[\omega]$ ] . @xmath13 $ ] thus represents the simplest member of a large family of energy functionals , offering the most ease with regards to use in analytical and numerical minimization procedures . we will only work with @xmath13 $ ] in the rest of this paper . each member of the family @xmath50\}$ ] possesses the same basic structure alluded to before : to a term representing the electrostatic energy ( the first double integral in ) , the iterative relation that @xmath5 must satisfy is included as a constraint . as is evident from , for each iterative relation that supplies the constraint equation , our variational formalism finds the appropriate lagrange multiplier required to enforce this constraint such that the resulting functional acquires the desired extremal properties . the functional @xmath13 $ ] derived in sec . [ sec : vform ] works for any medium with linear dielectric response , even for arbitrary spatial variations . in many instances , it is sufficient to represent the real system by a coarse - grained model where regions of uniform , but different , dielectric response are separated from each other by interfaces that can be assumed to be thin . for example , in the problem of colloids in a polar solvent , modelling the colloid as one uniform dielectric continuum and the surrounding solvent as another uniform dielectric of different permittivity provides a good representation of the real system . other examples where coarse - graining of this kind is often employed include : oil - water emulsions and biopolymers , such as lipid bilayers , in aqueous solution . in this light , we now consider the application of our functional to the problem of ions present in a system exhibiting this piecewise - uniform dielectric response . we show that for this specific dielectric response the functional @xmath13 $ ] reduces to a functional with only the interfacial induced charge density as the variational field . for the sake of brevity , we restrict ourselves to two uniform dielectrics separated by a single sharp interface @xmath61 , see fig . [ fig2 ] . extension to multiple dielectrics is straightforward . we assume that ions reside in the bulk of either dielectric . note that the interface can assume arbitrary geometry . let @xmath2 and @xmath3 denote the permittivities of the two media . we consider ions to be point particles with the @xmath62 ion having a charge @xmath63 . for a system with @xmath0 ions , the free charge density can be written as @xmath64 , where @xmath65 prescribes the position of the @xmath62 ion . it is useful to define the permittivity at the interface , taken to be the mean of permittivities on either side : @xmath66 ; and introduce @xmath67 = @xmath68 as a measure of the permittivity difference across @xmath61 . sketch showing two uniform dielectric regions of permittivities @xmath2 and @xmath3 separated by a sharp interface that may assume an arbitrary shape . the figure also shows the presence of point charges in either media and the induced charge on the interface . the effective interactions appearing in eq . between the various charged constituents of the system are also shown . ] clearly , the gradient of @xmath15 , or equivalently @xmath10 , vanishes everywhere except at the interface . mathematically we express this as : @xmath69 where @xmath70 is the unit normal vector at the interface , chosen to point in the direction of increasing permittivity , and @xmath71 is the position vector of an arbitrary point on the interface . when @xmath72 , @xmath73 , otherwise @xmath74 . because each medium offers a uniform dielectric response , the induced charge density in the bulk is known analytically from simple electrostatics principles @xcite , and it is : @xmath75 it is well known that the above bulk contributon leads to an effective charge density of @xmath76 . due to the discontinuity in the permittivity at the interface , induced charges also exist on the interface and their magnitude is in general unknown . thus , the overall induced charge density is expressed as the sum of two terms : @xmath77 where the first term on the right hand side in is given by , and the second term is the interfacial ( surface ) induced charge density . similar to the mathematical representation of @xmath78 , we have expressed the interfacial induced charge density as a surface integral , such that when @xmath79 the integral vanishes . substituting @xmath5 from in , and using eq . , we find that several volume integrals in reduce to surface integrals and @xmath13 $ ] is transformed to a functional of the surface induced charge density : @xmath80 & = \frac{1}{2}\iint\rho_{\mathbf{r}}k^{^{^{\negthickspace\negthickspace\negmedspace\negthickspace\circ\circ}}}_{\mathbf{r},\mathbf{r'}}\rho_{\mathbf{r'}}d^{3}rd^{3}r ' + \frac{1}{2}\iint_{\mathcal{i}}\rho_{\mathbf{r}}k^{^{^{\negthickspace\negthickspace\negmedspace\negthickspace\circ\bullet}}}_{\mathbf{r},\mathbf{s}}\omega_{\mathbf{s}}d^{3}rd^{2}s \nonumber\\ & + \frac{1}{2}\int_{\mathcal{i}}\int_{\mathcal{i}}\omega_{\mathbf{s}}k^{^{^{\negthickspace\negthickspace\negmedspace\negthickspace\bullet\bullet}}}_{\mathbf{s},\mathbf{s'}}\omega_{\mathbf{s'}}d^{2}sd^{2}s',\end{aligned}\ ] ] where @xmath81 is the induced charge density at the position @xmath71 on the interface , and @xmath82 , @xmath83 , and @xmath84 are , respectively , the effective potentials of interaction between two free charges , between a free charge and an induced charge , and between two induced charges ( see fig . [ fig2 ] ) . these effective interactions are given by : @xmath85 while the function @xmath86 in is the bare green s function given by , we find two new potentials of interaction in , @xmath87 and @xmath88 , which are defined as : @xmath89 where @xmath90 are arbitrary position vectors and @xmath91 are position vectors of arbitrary interfacial points . the functional in eq . can be compared with the functional of the surface polarization charge density obtained in ref . . the latter functional , as we noted earlier in sec . [ sec : dis ] , is not an energy functional . we find that the major difference between these two functionals is the _ absence _ of the interaction @xmath88 in the functional of ref . . we note that the presence of this particular interaction potential in @xmath92 $ ] is the direct result of employing the appropriate choice for the constraint to the electrostatic energy , an aspect of the variational formulation we highlighted in sec . [ sec : dis ] . it also appears that a functional of polarization charge density constructed with a combination of only @xmath86 and @xmath87 interaction potentials ( like the one in ref . ) ceases to remain an energy functional ; although we have not been able to rigorously show this . our attempts to construct an energy functional involving only @xmath86 and @xmath87 interactions via the variational formulation presented here , or otherwise , failed . we now employ @xmath92 $ ] to study some simple model systems exhibiting piecewise - uniform dielectric response . for these systems we can analytically carry out the integrals involved in eq . [ eq : sfnal ] by finding a suitable basis and expanding the green s function and induced charge density in terms of the associated basis functions ; eventually expressing the functional as a single integral involving the undetermined coefficients of expansion . we show that for these solvable models , @xmath92 $ ] can be minimized analytically which leads to the well known expressions for the associated induced charge density . we consider a planar interface separating two dielectrics of different permittivities @xmath2 and @xmath3 , with @xmath93 assumed . a point particle of charge @xmath94 is placed at a distance @xmath95 from the interface in the region with lower permittivity ( see fig . [ fig3 ] ) . we derive the induced charge density at the interface for this system using our variational formalism , in the process revealing the expression for the functional for this specific case . sketch showing a flat interface ( the @xmath96 plane ) separating the region @xmath97 of permittivity @xmath2 with the region @xmath98 characterized by a different dielectric constant @xmath3 . a point charge @xmath94 is present near the interface at a distance @xmath95 from the origin in the medium with permittivity @xmath3 . ] we adopt cylindrical coordinates @xmath99 , assume the interface to be the @xmath96 plane , and take the point charge to be on the positive @xmath100 axis at a distance @xmath95 from the origin . the @xmath98 domain then becomes the dielectric with @xmath3 permittivity . it is useful to choose bessel functions of integer order as the basis functions in this case . in this basis the bare green s function can be expanded as : @xmath101 we note that the position vector of the point charge is @xmath102 and the position vector of a point on the interface has the form @xmath103 . also , given the set up of this problem and recalling the definition of the normal vector ( it points from lower permittivity dielectric to the higher one ) , we have @xmath104 . let us now evaluate the functional given in for this particular different dielectric problem . for this purpose we would need the green s functions @xmath105 and @xmath106 , and the dot product of their gradients with the normal vector @xmath70 . these are readily evaluated from by employing , wherever necessary , the properties of bessel functions : @xmath107 , @xmath108 . using these functions the necessary renormalized green s functions @xmath87 and @xmath88 can be evaluated from , and employing them in , the effective interactions @xmath82 , @xmath83 , and @xmath84 are known . finally , just like the green s function in , the induced charge density , which due to symmetry is only a function of @xmath109 , can be written as an integral involving bessel function @xmath110 : @xmath111 where @xmath112 are as of now undetermined . submitting the evaluated interactions @xmath82 , @xmath83 , and @xmath84 , and @xmath113 from the above equation in the functional of , and remembering that the area element in our chosen coordinates is @xmath114 , we carry out most of the resulting integrals by employing the orthogonality relation @xmath115 and obtain the following functional : @xmath116= \frac{q^{2}}{8\epsilon_{2}^{2}}\left(\epsilon_{1}-\epsilon_{2}\right)\left(\epsilon_{1}-\epsilon_{2}-2\right)\int_{0}^{\infty}e^{-2\alpha d}d\alpha\\ & + \frac{q\pi}{2\epsilon_{2}}\left(\epsilon_{1}-\epsilon_{2}\right)\left(\epsilon_{1}+\epsilon_{2}-2\right)\int_{0}^{\infty}\frac{a\left(\alpha\right)}{\alpha}e^{-\alpha d}d\alpha\\ & + \frac{\pi^{2}}{2}\left(\epsilon_{1}+\epsilon_{2}\right)\left(\epsilon_{1}+\epsilon_{2}-2\right)\int_{0}^{\infty}\frac{a^{2}\left(\alpha\right)}{\alpha^{2}}d\alpha . \end{split}\ ] ] the subscript on @xmath19 in represents that the above is a functional for the planar interface case . clearly , @xmath117 $ ] is a functional of the lone function variable @xmath112 , which through represents the induced surface charge density . we now take the functional derivative of @xmath117 $ ] and set it to zero in order to determine @xmath112 . it is obvious that only the last two terms in contribute to this process and we obtain @xmath118}{\delta a(\alpha)}&= \frac{q\pi}{2\epsilon_{2}}\left(\epsilon_{1}-\epsilon_{2}\right)\left(\epsilon_{1}+\epsilon_{2}-2\right)\frac{e^{-\alpha d}}{\alpha}\\ & + \pi^{2}\left(\epsilon_{1}+\epsilon_{2}\right)\left(\epsilon_{1}+\epsilon_{2}-2\right)\frac{a\left(\alpha\right)}{\alpha^{2 } } , \end{split}\ ] ] which when set to zero reveals the coefficients @xmath112 to be the following : @xmath119 using the expression for @xmath112 from above in and carrying out the single integral involving the zeroth order bessel function , we get @xmath120 the above expression matches with the standard result for the surface induced charge density ( see ref . ) . we now derive the functional form and induced density for the case of a point charge near a spherical dielectric . the derivation is similar to the one just carried out for the planar case . we consider a dielectric sphere of radius @xmath121 and permittivity @xmath2 surrounded by a different dielectric of permittivity @xmath3 ( where @xmath122 is assumed ) . there is a free charge @xmath94 placed outside the sphere . in this example , the use of spherical coordinates @xmath123 is most convenient . we take the sphere to center at the origin @xmath124 and the point charge is assumed to lie on the positive @xmath100 axis at a distance @xmath95 from the origin ( see fig . [ fig4 ] ) , giving it the position vector @xmath125 . our conventions result in the the unit normal vector to be @xmath126 , i.e , pointing into the sphere . sketch of a dielectric sphere of radius @xmath121 and permittivity @xmath2 embedded in a region characterized by dielectric permittivity @xmath3 . the sphere is centered at the origin and is oriented such that the @xmath100 axis coincides with @xmath127 , where @xmath128 is the polar angle . a point charge @xmath94 is placed outside the dielectric sphere , on the @xmath100-axis at a distance @xmath95 from the origin . ] we start by choosing a suitable basis to expand the green s function . this basis turns out to be spherical harmonics and the expansion of @xmath30 in the latter is given by : @xmath129 where @xmath130 , @xmath131 , and the above expression holds for @xmath132 . as dictated by equation , in order to evaluate the interactions @xmath87 and @xmath88 for this problem , we need the following green s functions and its derivatives : @xmath133 , @xmath134 , @xmath106 , and @xmath135 , where @xmath71 and @xmath136 are the position vectors of arbitrary points on the sphere . these functions are readily evaluated from equation by employing the relations : @xmath137 for @xmath138 , and @xmath139 . once @xmath87 and @xmath88 are known , the effective interactions @xmath82 , @xmath83 , and @xmath84 for this problem are computed from . these interactions are then plugged into the functional in . also , for symmetry reasons the surface induced charge density @xmath81 is independent of the variable @xmath17 and just like any regular function of @xmath128 , it can be expanded in terms of spherical harmonics as : @xmath140 where the coefficients of expansion @xmath141 are as yet unknown . using in , and employing the orthonormality relation @xmath142 wherever necessary , the functional for this particular different dielectric problem is found to be : @xmath143 & = \frac{q^{2}}{2\epsilon_{2}^{2}}\sum_{l=0}^{\infty}\frac{\epsilon_{1}-\epsilon_{2}}{2l+1}l \left(\frac{\epsilon_{1}-\epsilon_{2}}{2l+1}l - 1\right)\frac{a^{2l+1}}{d^{2l+2}}\\ & + \frac{q}{2\epsilon_{2}}\sum_{l=0}^{\infty}\frac{4\pi}{2l+1}\frac{\epsilon_{1}-\epsilon_{2}}{2l+1}l(b-2)\frac{a^{l+2}}{d^{l+1}}a_{l}\\ & + \frac{a^{3}}{8}\sum_{l=0}^{\infty}\left(\frac{4\pi}{2l+1}\right)^{2}b(b-2)a^{2}_{l } , \end{split}\ ] ] where @xmath144 and the subscript on @xmath19 in denotes the spherical interface case under study . evaluating the functional derivative of @xmath145 $ ] and setting it to zero leads up to an equation for @xmath141 , which upon subsequent solving for , gives @xmath146 substituting @xmath147 from in the above equation leads to @xmath148 plugging @xmath141 from in gives the induced charge density on the interface to be @xmath149 which matches with the standard result available elsewhere @xcite . when many point charges are present near an arbitrarily shaped dielectric interface , one must resort to numerical methods to minimize @xmath92 $ ] in order to compute the induced charge density on the interface . therefore , we now turn towards discussing the numerical implementation of our variational method . to perform the minimization numerically we first partition the dielectric interface into @xmath150 finite elements . to each element @xmath151 we assign an average induced charge density @xmath152 , an area @xmath153 and a normal vector @xmath154 . under this discrete representation , @xmath92 $ ] becomes a functional of the set of discrete induced charge density values @xmath155 and eq . is transformed into : @xmath156 & = \frac{1}{2}\sum_{i=1}^{n}\sum_{j=1\atop j\neq i}^{n}q_{i } k^{^{^{\negthickspace\negthickspace\negmedspace\negthickspace\circ\circ}}}_{\mathbf{r}_{i},\mathbf{r}_{j}}q_{j } + \frac{1}{2}\sum_{i=1}^{n}\sum_{k=1}^{m}q_{i } k^{^{^{\negthickspace\negthickspace\negmedspace\negthickspace\circ\bullet}}}_{\mathbf{r}_{i},\mathbf{s}_{k}}\omega_{k}a_{k } \nonumber\\ & + \frac{1}{2}\sum_{k=1}^{m}\sum_{l=1}^{m}\omega_{k } k^{^{^{\negthickspace\negthickspace\negmedspace\negthickspace\bullet\bullet}}}_{\mathbf{s}_{k},\mathbf{s}_{l } } \omega_{l}a_{k}a_{l},\end{aligned}\ ] ] where @xmath157 is the position vector of the @xmath158 finite element and @xmath0 is the number of point charges . note that we represent the point charges with the density @xmath159 , where @xmath63 and @xmath65 are , respectively , the charge and position vector of the @xmath62 point charge . the effective interactions @xmath82 , @xmath83 , and @xmath84 in the above equation are the discretized version of their continuum counterparts in eq . . we note that the discretization process introduces divergences in , for example when @xmath160 , and to get around these divergences , we replace the sum at these points by an approximate integral which is evaluated analytically @xcite . @xmath161 $ ] can now be minimized by using steepest descent or simulated annealing methods . we choose the procedure of simulated annealing and implement it using a molecular dynamics ( md ) scheme @xcite . we include @xmath161 $ ] as the potential energy part of a lagrangian that contains a fictitious kinetic energy term : @xmath162 , where @xmath163 is a fictitious mass assigned to the @xmath158 induced charge value . the set @xmath155 represents a point in the ( fictitious ) configuration space and equations of motion of this point are derived from the lagrangian . a feature of the system that becomes important in simulations is that , as a result of gauss s law , the net induced charge at the interface is a constant . we directly enforce this constraint at each step of the simulation via the shake - rattle algorithm @xcite . the system for testing the numerical procedure designed to minimize @xmath92 $ ] : positive ( red ) and negative ( green ) ions inside and outside a spherically shaped dielectric region . the dielectric permittivity inside ( outside ) the sphere is @xmath164 ( @xmath165 ) . the ions have the charge of @xmath166 and are represented as spheres of diameter @xmath167 ; the radius of the dielectric sphere is @xmath168 . see text for the meaning of the symbols . ] polarization charge density ( in units of @xmath169 ) as a function of @xmath128 induced on the dielectric sphere from a point charge located outside the sphere ( see fig . [ fig4 ] ) . @xmath164 is the permittivity inside the sphere and @xmath165 is the permittivity outside . red triangles are the results from the numerical minimization of the functional @xmath92 $ ] and green circles are exact results given by eq . . the standard triangles and the hollow circles correspond to the case when @xmath170 and @xmath171 , while the inverted triangles and filled circles correspond to the inverse problem where @xmath172 and @xmath173 . the inset shows the details of the density around the point where it changes sign . ] the simulation begins at an arbitrarily chosen point in the fictitious configuration space and we choose @xmath163 and the simulation time step @xmath174 such that the ensuing dynamics is stable . the dynamics of this point is generated via standard md algorithm , using the force obtained as a result of computing the gradient of the functional in with respect to @xmath152 . the motion of the point is towards the minimum of the potential energy , resulting in the rise of the fictitious kinetic energy . after some time , a fraction of the kinetic energy is removed from the system , and the whole process of exploring the configuration space begins again . eventually , the system reaches its minimum potential energy and the set of induced charge values corresponding to this state is obtained as the solution . to demonstrate and test our numerical optimization strategy , we first applied it to the problem of a single positive charge outside a dielectric sphere as depicted in fig . [ fig4 ] , and then to the problem of many charges near a spherical dielectric interface as shown in fig . [ fig6 ] . for the single test charge problem , the exact result for the induced density is given by eq . . the exact results for the induced charge density in the many - charge system are obtained from a careful superposition of the induced densities generated by considering each charge separately . we test the accuracy of our numerical procedure against these exact results . we consider a spherical dielectric of permittivity @xmath164 surrounded by an exterior dielectric with permittivity @xmath165 . the radius of the sphere is @xmath168 , where @xmath167 is the diameter of the point charge taken to be @xmath175 and serves as the length unit . here , @xmath176 is the bjerrum length in water . the unit of charge is taken to be @xmath177 , the charge on a proton . the interface is discretized with roughly @xmath178 points and the fictitious md simulation parameters are : @xmath179 , @xmath180 , and @xmath181 . @xmath26 is the number of md steps and we quench the system every @xmath182 steps . in fig . [ fig5 ] we show the polarization charge density for the single test charge problem . we place the positive unit charge on the @xmath100-axis at a distance of @xmath183 from the center of the sphere . due to the inherent symmetry associated with this problem , the induced density is only a function of the @xmath128 variable . in addition , the density profile is symmetric around the @xmath184 point and hence the results are shown for @xmath185 $ ] . we compute the density for the case when @xmath186 , @xmath187 and also for the inverse problem where @xmath188 , @xmath189 . as is evident from fig . [ fig5 ] , our numerical results ( red triangles ) agree very well with the exact results ( green circles ) for both the cases studied . we observe that in the first case ( @xmath190 ) , the density induced on the portion of the interface that is nearest to the charge ( low @xmath128 values ) is positive , while in the latter case ( @xmath191 ) it is negative . also , in both cases , the sign of the induced density flips at some value of @xmath128 ( see inset in fig . [ fig5 ] ) . furthermore , in either case , the magnitude of the induced density falls rapidly in the beginning as the angle @xmath128 increases . all these observations , which are consistent with basic electrostatics principles , suggest that the test charge will repel away from the dielectric sphere when the dielctric constant of the latter is lower than the medium in which the test charge is embedded . otherwise , the test charge will be attracted towards the interface . we now present our findings for the case of many charges near the spherical interface . we consider 10 positive and 10 negative monovalent ions inside the sphere and the same outside . thus , the total number of ions equals 40 . the positions of the ions are chosen at random and the ions remain fixed at their locations . as the system is electroneutral in each dielectric , the net induced charge on the sphere is 0 . in fig . [ fig7](a ) we show our simulation results for the polarization charge density at the interface as a function of the angles @xmath128 and @xmath17 for the case : @xmath186 , @xmath187 . we refer to this graph as the polarization map . for this case , the point charge present inside ( outside ) the dielectric induces a charge of opposite ( same ) sign on the interface boundary closest to it . the regions of intense red ( highly positive ) or intense blue ( highly negative ) on the polarization map suggest the presence of an ion or many ions near the interface at the corresponding @xmath192 location . [ fig7](b ) shows the exact values of the induced density for this system . it is clear that the results from the numerical minimization of our functional are in excellent agreement with the exact results . in fig . [ fig7](c ) we show the polarization map for the same system as above , but with the dielectric media switched . thus , for this case @xmath188 and @xmath189 . we observe that by and large this map looks like the `` image '' of fig . [ fig7](a ) . regions with more positive ( red ) induced charge in fig . [ fig7](c ) are the ones that were highly negative ( blue ) in fig . [ fig7](a ) and vice versa . we indeed expect this as now , in direct contrast to before , an ion inside ( outside ) the dielectric induces a charge of the same ( opposite ) sign on the interface boundary closest to it . once again , comparison with exact results in fig . [ fig7](d ) confirms the accuracy of the numerical minimization procedure . we note that since our functional is an energy functional , the above described numerical minimization procedure for the case when the ions are static , can be suitably modified to incorporate the scenario when ions are moving , like in a conventional md simulation . results from such a dynamical optimization procedure were presented in ref . , along with a brief description of the method itself . we have presented a variational formulation of electrostatics specifically designed to treat the problem of dielectric heterogeneities in charged systems . assuming only the condition of linear response , we constructed an energy functional that employs the polarization charge density as its sole variational field . this functional is applicable for any configuration of free charges and arbitrary spatial dependence of the dielectric response . we discussed in some depth the basic structure of our functional , drawing comparisons with past functionals and revealing how more energy functionals can be constructed using our variational approach . next , we focused on the important case of uniform dielectrics separated by sharp interfaces . we showed that under this piecewise - uniform dielectric response , our functional reduces to a functional of only the surface polarization charge density . such a reduction of the 3-dimensional electrostatic problem to a 2-dimensional one has many advantages from a computational perspective . we then obtained the specific expressions for this reduced functional , and subsequently the induced charge density , for the case of a point charge near a planar interface and for a point charge near a spherical dielectric . finally , in the view of applying our approach to more complicated systems , we discussed the numerical implementation of our minimizing variational principle for a system exhibiting piecewise - uniform dielectric response . we illustrated this procedure for a system of monovalent ions near a spherical dielectric ; obtaining the polarization charge density induced on the interface and finding excellent agreement with exact results . due to the fact that @xmath13 $ ] is an energy functional , its minimization can be carried out in conjunction with the update of the ionic configuration . in ref . we demonstrated such a dynamical minimization method . this is of tremendous significance with regards to md simulations of ions in heterogeneous media , as the explicit solution of the poisson equation at each step is avoided . detailed explorations of the dynamical optimization of our functional and the associated md simulations investigating diverse systems such as charged colloidal dispersions and liquid - liquid emulsions will be the subject of a future study . thanks r. sknepnek for many useful discussions . v.j . was funded by the department of defense research and engineering ( ddr&e ) and the air force office of scientific research ( afosr ) under award no . fa9550 - 10 - 1 - 0167 and f.j.s . was funded by the nsf grant numbers dmr-0805330 and dmr-0907781 . in this appendix we investigate the extremal properties of @xmath13 $ ] . first , we derive the condition for which @xmath13 $ ] is an extremum . next , we prove that at its extremum the functional gives the true electrostatic energy . and finally , we show that the functional becomes a minimum at its extremum . the derivation of the extremum condition for @xmath13 $ ] begins by recording how much the functional changes when the function @xmath5 is changed by an arbitrary small amount @xmath193 . we use eq . to compute @xmath194 $ ] , retaining terms up to first order . the original functional @xmath13 $ ] is then subtracted from the result giving the first variation @xmath195-\mathscr{f}[\omega]$ ] . employing standard vector calculus identities @xcite wherever necessary and using dirichlet boundary condition to make the surface integrals vanish by invoking the boundary at infinity , we find the first variation @xmath196 to be @xmath197 where we have suppressed the functional part of the notation for @xmath198 for brevity . by definition , at the point of extremum , the first variation @xmath199 must vanish for an arbitrary @xmath193 . we see from that this is only true if the following condition holds : @xmath200 we now simplify eq . . operating on both sides of with the laplacian operator and using we obtain @xmath201 where towards the end we replaced the dummy variable @xmath202 with @xmath203 . it is useful to introduce @xmath204 using which , eq . can be written as @xmath205 multiplying both sides of by @xmath206 and integrating over whole space we obtain : @xmath207 integrating by parts and employing dbc , we transform the above integral into @xmath208 it is clear that since @xmath209 is always non - negative the integrand in the above equation is always non - negative . this means that the only way the integral is zero is if the integrand is identically zero at all points , which implies @xmath210 or @xmath206 is a constant . ( strictly speaking , the integrand can be zero without requiring that the gradient of @xmath211 vanishes : this happens when @xmath209 vanishes at _ all _ points . but this situation represents the presence of free space everywhere , and in that case our functional becomes independent of @xmath5 . ) using to expand @xmath211 , we thus obtain the equality : @xmath212 where @xmath213 is some constant . operating with the laplacian on both sides of and re - employing we get @xmath214 which , after expanding out the function @xmath198 using , becomes @xmath215 equation gives the extremum condition for @xmath13 $ ] . it is clear from the definition of @xmath5 , the condition of linear response , and basic laws of electrostatics that the right iterative relation for @xmath5 is obtained from the process of extremizing @xmath13 $ ] . as a first step towards proving that the functional @xmath13 $ ] is an energy functional , we investigate here its value at extremum . let @xmath216 be the function that extremizes the functional @xmath13 $ ] . the results of the last section show that @xmath217 must satisfy and so we obtain : @xmath218 = 0.\ ] ] to evaluate the value of @xmath13 $ ] at the extremum point we let @xmath219 in , thus obtaining @xmath220&=\frac{1}{2}\iint \rho_{\mathbf{r}}g_{\mathbf{r},\mathbf{r ' } } \left(\rho_{\mathbf{r'}}+\omega_{\mathbf{r'}}[\bar{\omega}]\right ) d^{3}r'd^{3}r\\ & - \frac{1}{2}\iint \omega_{\mathbf{r}}[\bar{\omega } ] g_{\mathbf{r},\mathbf{r ' } } \left(\bar{\omega}_{\mathbf{r ' } } - \omega_{\mathbf{r'}}[\bar{\omega } ] \right ) d^{3}r'd^{3}r . \end{split}\ ] ] using , the second double integral in the above equation vanishes and we obtain @xmath221&=\frac{1}{2}\iint \rho_{\mathbf{r}}g_{\mathbf{r},\mathbf{r ' } } \left(\rho_{\mathbf{r ' } } + \bar{\omega}\right ) d^{3}r'd^{3}r . \end{split}\ ] ] as we noted earlier , the function @xmath37 given by coincides with the electrostatic potential at the point of extremum . thus , using we obtain the following expression for the true electrostatic potential : @xmath222 using , the extremum value given in eq . becomes @xmath223&=\frac{1}{2}\int \rho_{\mathbf{r}}\phi_{\mathbf{r}}d^{3}r . \end{split}\ ] ] the expression on the right hand side of the above equation is the standard expression for the electrostatic energy , equivalent to @xmath224 . hence , the extremum value of @xmath13 $ ] gives the true electrostatic energy of the system . to complete the proof that @xmath13 $ ] is an energy functional we now show that @xmath13 $ ] becomes a minimum at its extremum . this begins by analyzing the terms in the variation of @xmath13 $ ] that are of second order in @xmath193 , terms which we ignored during the derivation of the extremum condition . if this second order change is shown to be positive then we would have proven that our functional becomes a minimum at the extremum point . it is clear from that the terms in @xmath225 - \mathscr{f}[\omega]$ ] that are quadratic in @xmath193 come only from the second double integral in , and we obtain @xmath226 where @xmath227 is given by @xmath228 and @xmath229 denotes the second order variation in @xmath13 $ ] . we focus on the first double integral in . integrating by parts and using the basic property of green s function , namely , eq . , the following identity can be derived : @xmath230 where @xmath231 is an arbitrary function . in deriving the above relation we invoked dbc for similar purposes as we have done before . using this identity with @xmath232 , the first term in transforms to @xmath233 we note that the right hand side of the above equation is always positive . we next probe the second term in . expanding @xmath227 using , the second term becomes @xmath234 integrating by parts and employing dbc we transform the right hand side of into a dot product of two gradients , as in @xmath235 which is equivalent to @xmath236 the two double integrals in can now be replaced with expressions obtained in eqs . and . doing so gives the following for the second variation : @xmath237 since @xmath209 is non - negative everywhere , it is clear that both the terms on the right hand side of eq . are always positive , implying @xmath238 , thus completing the proof . in this short appendix we apply our variational principle to the simplest case of a uniform dielectric . we derive the expression of our functional for this particular case and also obtain the induced density as a result of the minimization of the functional . for a uniform dielectric @xmath239 , where @xmath240 is a constant . employing this expression for @xmath209 in and using , we obtain @xmath241 = -4\pi\chi_{u}\left ( \rho_{\mathbf{r } } + \omega_{\mathbf{r } } \right).\ ] ] substituting @xmath6 $ ] from in transforms the latter equation into @xmath242= \frac{1}{2}&\left(\epsilon_{u}^{2 } - 3\epsilon_{u } + 3\right)\iint \rho_{\mathbf{r}}g_{\mathbf{r},\mathbf{r ' } } \rho_{\mathbf{r ' } } d^{3}r'd^{3}r \\ + & \left(\epsilon_{u } - 1\right)^{2}\iint \rho_{\mathbf{r}}g_{\mathbf{r},\mathbf{r'}}\omega_{\mathbf{r'}}d^{3}r'd^{3}r \\ + & \frac{1}{2}\epsilon_{u}\left(\epsilon_{u } - 1\right)\iint \omega_{\mathbf{r } } g_{\mathbf{r},\mathbf{r'}}\omega_{\mathbf{r ' } } d^{3}r'd^{3}r , \end{split}\ ] ] where we have expressed the resulting functional in terms of the uniform permittivity @xmath243 , which is connected to @xmath240 via the relation @xmath244 . equation gives the expression of our functional for the case of point charges in the presence of uniform dielectric response . let us now derive the expression for the density of induced charges in this case . from elementary electrostatics we expect that the induced charges are only to be found at the location of the free charges . the first variation of the functional in is : @xmath245 for @xmath246 to vanish for any @xmath193 , it is clear from that the following must be true : @xmath247 applying the laplacian on both sides of and employing , we obtain @xmath248 which simplifies to @xmath249 equation is indeed the standard expression for the induced charge density for the case of a uniform dielectric . 45ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ] + 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop _ _ , ed . ( , ) @noop _ _ , advanced book program ( , ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop _ _ ( , )

we describe radial velocity observations of a large sample of apparently single white dwarfs ( wds ) , obtained in a long - term effort to discover close , double - degenerate ( dd ) pairs which might comprise viable type ia supernova ( ) progenitors . we augment the wd sample with a previously observed sample of apparently single subdwarf b ( sdb ) stars , which are believed to evolve directly to the wd cooling sequence after the cessation of core helium burning . we have identified 18 new radial velocity variables , including five confirmed sdb+wd short - period pairs . our observations are in general agreement with the predictions of the theory of binary star evolution . we describe a numerical method to evaluate the detection efficiency of the survey and estimate the number of binary systems not detected due to the effects of varying orbital inclination , orbital phase at the epoch of the first observation , and the actual temporal sampling of each object in the sample . follow - up observations are in progress to solve for the orbital parameters of the candidate velocity variables .

astro-ph9802356

both theoretical and observational investigations into the origins of type ia supernovae ( ) have focused on binary star evolution . within this framework , scenarios are divided into `` single degenerate '' ( @xcite ) and `` double degenerate '' ( @xcite ; @xcite ) models . in the former models , a white dwarf ( wd ) accretes from a nondegenerate companion , while in the latter , two wds with a total mass exceeding the chandrasekhar limit ( ) are assumed to coalesce . the latter scenario predicts that there should exist a population of close , short - period ( @xmath0 10@xmath1 ) binary wd systems , or double degenerates ( dds ) , with total system masses in excess of . the binary components are initially brought to short - period orbits by common envelope ( ce ) evolution , with the energy necessary to expel the envelope taken from orbital shrinkage ( @xcite ) . subsequently , further loss of orbital angular momentum through emission of gravitational wave radiation provides the mechanism through which the merger can take place . it is important to note that at present it is not entirely clear if the coalescence of two wds with a total mass exceeding really produces a or rather an accretion induced collapse ( cf . @xcite ; @xcite ) . however , the existence of dd systems is a definite prediction of the theory of close binary star evolution . recently , several authors have made the suggestion that the properties of observed are not consistent with the detonation of progenitors having a single , well - defined mass . these include an apparently real dispersion in the peak absolute b magnitudes and a correlation between the peak magnitude and the decay rate of the b light curve ( @xcite ; @xcite ) . @xcite report matching the relationship only by requiring a variation in the amount of carbon burned to ni@xmath2 of nearly a factor of two . it has been suggested that accumulation of he atop an ordinary sub- co wd , or the merger of sub- he+co wds might provide the necessary range of masses and produce edge - lit detonations ( elds ) with characteristics ( e.g. , @xcite ; @xcite ; @xcite ) . the viability of this mechanism is controversial ( see , e.g. , @xcite ; @xcite ) , and existing estimates of the expected rate of elds in sub- systems fall short of the inferred galactic rate ( @xcite ) . another alternative is that some of these systems may evolve into ultra - short period interacting am cvn systems , which are believed by many authors to consist of lower - mass co white dwarfs accreting from very low - mass he wd secondaries ( e.g. , liebert et al . 1996 ) . @xcite discovered the prototype dd , l 8702 ( eg 11 ) , which has an orbital period @xmath3 . however , the period is too long for the system to merge on an astrophysically interesting time scale . when the merger does take place , a probably will not result since the total system mass is below ( @xcite ) . in recent years , three systematic searches for dds have been performed . two of these ( @xcite , hereafter rs ; @xcite , hereafter fwg ) failed to discover any candidates at all . the third ( @xcite , hereafter bgrd ) did discover one dd ( and 4 other candidates ) , but the low primary mass ( @xmath4 0.37 ; @xcite ) of the confirmed dd renders it an unlikely progenitor . we note that bgrd found a period @xmath0 118 for their confirmed dd ( 0957@xmath5666 ) , however , recent observations ( moran , marsh , & bragaglia 1997 ) reveal that the system actually has a period @xmath0 88 minutes and primary and secondary masses of 0.37 and 0.32 , respectively . recently , observations targeting the low - mass tail of the wd mass distribution of bergeron , saffer , & liebert ( 1992 ) spectacularly confirmed the prediction that wds with masses less than the canonical core - helium ignition mass are likely to be post - common envelope objects residing in short - period binary systems ( marsh , dhillon , & duck 1995 ; marsh 1995 ) . however , the low wd masses again make them unlikely progenitors . these negative results have cast some doubt on the merging dd scenario as the favored mechanism for producing . there does exist one example of a dd system where the combined mass is believed to exceed the chandrasekhar limit . @xcite discovered that lb 11146 consists of a normal da wd and a highly magnetic daxp wd . the combined mass of the system is between 1.71.9 , but the orbital period is unknown . @xcite failed to detect radial velocity variations in the core of to a limit 20 with periods 1 hour , and a recent _ hubble space telescope _ imaging program ( @xcite ) did not resolve the components to a separation @xmath6 0025 , corresponding to @xmath6 1 a.u . on the plane of the sky . if the progenitor stars were massive , as seems likely given the large masses of the remnants , close binary interaction is likely to have occurred in a previous agb phase . further work to establish the orbital period of this remarkable binary is urgently needed . another possible outcome of close binary evolution is a merger during the late stages of a ce phase leading to the formation of a single , possibly massive white dwarf ( e.g. , @xcite ; @xcite ) . large numbers of hot white dwarfs were detected in the rosat and euve surveys , and analyses of optical follow - up observations ( e.g. , @xcite ; @xcite ; @xcite ) determined stellar parameters , including masses . comparing the rosat and euve samples with the cooler , optically selected sample of @xcite revealed a statistically significant excess of hot , supermassive wds , which may represent a population of coalesced binary wds . the high - mass excess may partly arise from differences in the cooling rates of normal ( 0.6 ) and massive ( @xmath7 1.0 ) stars , but there may still be too many massive white dwarfs to be explained by the predictions of single star evolution , with a conventional initial mass final mass relation , initial mass function , and star formation history . recent theoretical work ( @xcite , hereafter ylts ; @xcite , hereafter ity ) reveals that consideration of the `` visibility function '' , a consequence of wd cooling along evolutionary sequences , is an important factor in calculations of the expected dd number distribution . systems born with orbital periods @xmath0 10@xmath1 do not evolve to shorter periods by emission of gravitational radiation on interesting ( i.e. , short ) timescales . below 10@xmath1 , an approximate equilibrium is established between systems born into short - period orbits and those that vanish either by merging or by cooling in some @xmath8 years beyond the detection limits of surveys with relatively bright limiting magnitudes , such as our own . only a small fraction of these pairs will actually merge in the @xmath8 year period of visibility , and only 1/20 of the pairs with @xmath9 are visible ( cf . ity , figure 2 ) . these results indicate that previous dd searches had sample sizes too small to reveal the massive , short - period systems of interest . on the other hand , it is clear that the close dds that _ were _ discovered ( saffer et al . 1988 ; bgrd ; @xcite ; @xcite ) fall within the peak of the expected dd number distribution ( cf . ylts , ity ) . ylts suggested that further searches with increased sample sizes are likely to reveal the existence of the short - period , massive dd population . in this paper , we report observations of an enlarged sample in order to place meaningful constraints both on dds as progenitors and , more generally , on the theory of close binary star evolution . our sample comprises 107 apparently single da ( hydrogen line ) and db ( helium line ) wds with @xmath10 drawn from the catalog of @xcite , which are largely not in common with the previous three systematic surveys . over two observing seasons , in a total of 21 nights in 4 observing runs from 19941996 , we have obtained radial velocity spectroscopy at the kitt peak national observatory s 2.1-m reflector equipped with goldcam . this instrumentation provided 3fwhm spectral resolution over a spectral range of 700centered on and including he i @xmath11 . exposure times were chosen to achieve a signal - to - noise ( s / n ) ratio of 30 per pixel . we have augmented the wd sample with a sample of 46 apparently single subdwarf b ( sdb ) stars previously observed by saffer ( 1991 ; see also @xcite ) . the radial velocity observations were obtained using the steward observatory s kitt peak station 2.3-m reflector , equipped with instrumentation similar to that used at the kitt peak 2.1-m and providing nearly identical spectral coverage and resolution . the inclusion of the sdb stars in the sample is appropriate since these core helium - burning stars are believed to evolve directly to the wd cooling sequence on a relatively short time scale following core helium exhaustion . theoretical calculations which predict the period and mass distributions of dds do take into account the contribution from the descendents of he stars . our choices of spectral features , resolution , and s / n ratio for our observations convey several advantages : 1 ) in all but the hottest da wds , and in all sdb stars , has a sharp ( fwhm 1 ) non - lte core , making it possible at our resolution and s / n ratio to measure radial velocities with 2030 errors despite the extreme width of the stark - broadened absorption line . the he i @xmath11 line in db wds provides similar precision , 3040 . the close , short - period dds of interest have peak - to - peak radial velocity amplitudes nearly an order of magnitude larger than these detection limits , and any existing systems would have a significant probability of detection ( for our sample , 70% in the period range 2@xmath1220@xmath13 ; see 7 for details ) . 2 ) our method detects objects with similar temperatures , like the prototype l 8702 , by revealing variations of both absorption cores in a double - lined spectrum . the technique of differential photometry in the wings of the absorption line does not detect this type of system . 3 ) spectroscopic observations in the core of offer the maximum velocity resolution for any wd spectral features observable from the ground at intermediate resolution . 4 ) the large throughput at intermediate spectral resolution makes it possible to observe large , relatively faint samples with modest telescope apertures . our targets were observed twice on any given night separated by 12 hr , followed by a third observation a day or two later . radial velocity variations of short - period systems are easily detected simply by comparing spectra by eye , although in follow - up observations to determine the orbital parameters , velocities will be measured formally by line profile fitting . in particular , derived periods and velocity semi - amplitudes will determine mass functions and provide lower bounds on total system masses . in table [ wdsample ] , we list all the wds in our sample . successive columns give the wd number , name , @xmath14 magnitude , ( @xmath15@xmath14 ) and ( @xmath16@xmath15 ) colors when available in the literature ( occasionally only greenstein multichannel or str " omgren colors are available ) , and references . we also list 7 objects ( note # 3 ) from samples of sdb stars ( saffer 1991 ; @xcite ) , which show significant velocity variability . the remaining non - variable sdb stars and their weighted mean velocities and errors are given in table [ sdbsample ] . in table [ ddvariables ] we list candidate dd systems which show velocity variability identified by visual examination of stacked spectra , the spectroscopic analog of blinking photographic plates . the last column indicates a weight of either 1 or 2 , indicating definite or marginal velocity variability , respectively . we identify definite variables as those for which the spectral line cores vary by 1.5 , a 2.5@xmath17 detection , on average . marginal variables did appear to vary , but at a lower level . the references are the same as in table [ wdsample ] . in the first 13 panels of figure [ multistack ] , we show spectra of our best candidates ( weight 1 ) , emphasizing the cores of the lines . the sdb velocity variables are excluded from the figure unfortunately , the spectra could not be recovered from an ancient archive tape . however , variability in the sdb stars was originally measured by formal line profile fitting ( see table [ sdbsample ] for the results ) . the last two panels at the bottom right in figure [ multistack ] show the detection of narrow , variable emission in the cores of the wd absorption lines , betraying otherwise hidden low - mass main sequence ( ms ) companions in detached , pre - cataclysmic binary ( pcb ) systems . the emission likely results from reprocessing of uv photons from the hotter wd in the cooler stars atmospheres ( cf . @xcite ) . the laboratory wavelength of is indicated with a vertical line in each panel . the figure illustrates the ease with which velocity variables may be identified , even in intermediate - dispersion spectra , and despite the highly stark - broadened absorption wings . we have detected a total of 23 definite variables , including 7 sdb+wd pairs and 2 white dwarf plus main sequence ( wd+ms ) pairs . of the total , 11 wd+wd pairs , all 7 sdb+wd pairs , and 1 wd+ms pair are new detections , most of which require follow - up observations to solve for the orbital parameters . some follow - up observations have already been accomplished the previously detected sdb velocity variables feige 36 and ton 245 , both pre - white dwarfs , have been shown to be close binary systems with periods of 85 and 25 , respectively ( saffer et al . 1998 , in preparation ) . in addition , the sdb stars 0101 + 039 , 1432 + 159 , and 2345 + 318 have recently been shown to have a periods of 161 ( with a possible alias of 137 ) , 539 , and 578 , respectively ( moran et al . 1998 , in preparation ) . low - mass ms companions are ruled out by large mass functions and the non - detection of infrared excesses . the companions are thus inferred to be compact , most likely white dwarfs . approximately 35% of the objects in our survey are in common with previous surveys . however , most of these objects have not been sampled in the period range in which our survey has the greatest sensitivity : 1 . the largest area of overlap with previous surveys is with rs ( who found no variables ) , with 26 of the objects listed in table [ wdsample ] being in common . however , the rs temporal sampling provided sensitivity to orbital periods between 30@xmath18 and 3@xmath1 , while our survey is most sensitive to orbital periods between 2@xmath1 and 1020@xmath13 . there is thus no substantive overlap between the two samples in their common orbital period ranges . table [ ddvariables ] shows that we detected 7 of these objects as radial velocity variables , 4 marginally and 3 definitely . there are 12 objects in table [ wdsample ] which overlap with fwg , who also found no variables . their temporal sampling provided sensitivity to orbital periods between 310@xmath1 . of the 12 fwg objects , 6 are also in common with rs . of the remaining 6 objects , we have detected 2 as velocity variables , one marginally and one definitely . the overlap with the survey of bgrd , much of which took place in the southern hemisphere , comprises only 10 objects , 4 of which are also in common with rs and fwg . of the remaining 6 objects , we definitely detect 2 as velocity variables . these two objects were not detected as variables by bgrd . _ 0553 + 053 . _ this object is a strongly magnetic da wd , whose absorption line is widely split by the zeeman effect ( @xcite ) . we detect possible line profile variations in our spectra which might be indicative of rotational or orbital modulation . the velocity is otherwise very difficult to measure due to the extremely broad and shallow absorption components . _ 0710@xmath19741 , 1101@xmath19364 , 1317@xmath19453 , 1713@xmath19332 , 2032@xmath19188 . _ these objects were detected as short - period binaries by marsh , dhillon , & duck ( 1995 ) and marsh ( 1995 ) , and as such do not represent new detections . our survey was performed between november 1994 and april 1996 , during which time we targeted these objects prior to having knowledge of their duplicity . we detected variability in all but one relatively faint object ( 2032 + 188 ; @xmath14 = 15.34 ) , for which our spectra had very poor s / n in the maximum exposure time of 30 minutes . this limit was imposed to avoid smearing spectra of very short - period systems . our observations of these objects are very useful , in that they indicate that our detection of known variables is very efficient , at least 80% in this case . also , since they were observed independently they may be retained in the statistical analysis . _ 0713 + 584 , 1756 + 827 . _ these two objects have large blueshifts , 300 , and are likely members of the galactic halo . _ 0900 + 554 , 0940 + 068 , 1247 + 553 , 2204 + 071 . _ the first three objects are classified as da wds in the catalog of @xcite , but our observations clearly identify them as sdb or blue horizontal branch ( bhb ) stars , as seen in the top three spectra in figure [ wd2sdb ] . for comparison , in the bottom spectrum we show a normal da wd . the lower - gravity stars have narrow , sharp absorption lines , and he i @xmath206678 is also detected . the strength of the latter feature in 0900 + 554 and 1247 + 553 implies that these two objects could be slightly cooler , lower - gravity bhb stars and not sdb stars , as the latter in general tend to have very weak he i @xmath206678 absorption ( cf . this is certain for 0900 + 554 , for which an existing blue optical spectrum shows balmer absorption extending to @xmath21 14 . however , all three objects will eventually evolve to low - mass ( 0.5 ) wds and rightfully remain included in the sample . the pg coordinates for 0900 + 554 ( ra 09:00:13.4 , dec + 55:26:42 ) were found to be in error the object s location is ra 09:00:05 , dec + 55:28:50 ( 1950 ) . 1247 + 533 is a visual double the sdb / bhb star is the southeast object . 0940 + 068 is barely resolved as a visual binary in the telescope guide star camera . it is not identified as composite in the pg survey and appears single on the pg finder plate . the eastern object is an sdb star , the western object a late - type main sequence star . the fourth object , 2204 + 071 , is ambiguously identified in the pg survey , as there are actually two objects marked on the pg finder plate . the southern object is a da white dwarf analyzed by @xcite . we serendipitously observed the northern object thinking it to be the da . instead , we obtained the fourth spectrum from the top in figure [ wd2sdb ] . the object is either a bhb star , in which case the extreme strength of he i @xmath206678 would imply a helium abundance at least 23 times solar , or it is a newly - discovered dab or dba wd . observations in the blue part of the optical spectrum will be required to resolve the ambiguity . _ 1101 + 364 , 1123 + 189_. the pg coordinates for these objects ( ra 11:01:35.3 , dec + 36:26:20 ; ra 11:23:46.6 , dec + 18:55:18 ) are in error . the objects are located at ra 11:01:46.3 , dec + 36:26:56 and ra 11:23:42 , dec + 18:55:50 , respectively ( 1950 ) . _ 1104 + 602_. this object ( eg 75 ) is mistakenly identified in the pg survey as pg 1100 + 604 . the correct coordinates are ra 11:04:43 , dec + 60:14:48 ( 1950 ) . the spectrum of the object found at the pg coodinates ( ra 11:00:43.5 , dec + 60:27:41 ) and agreeing in the telescope guide star camera with the position on the pg finder plate resembles that of a main sequence g star , based on the depth and narrowness of the line . _ 1204 + 451 , 1459 + 305_. these objects are visual binaries . the da wds are the western ( 1204 + 451 ) and southern ( 1459 + 305 ) stars . _ 1422 + 095_. this object is the well - studied pulsating da gd 165a ( @xcite ) , which has a very low - mass ( possibly sub - stellar ) , distant companion ( @xcite ) . we detect the wd as a velocity variable ( see table [ ddvariables ] ) . if the cooling age of the older wd secondary could be determined , an improved lower limit could be placed on the cooling age of the potential brown dwarf companion . analysis of delays in the wd pulsation arrival times would be another way to study the system s orbital properties . _ 1411 + 219_. this object , identified as composite in the pg survey , is a visual double , which in retrospect can be seen to be barely resolved upon close inspection of the pg finder plate . the db wd is the northern component of the pair and shows weak absorption in addition to strong he i @xmath11 . the southern component shows late k- or m - type spectral features . _ 1559 + 369 , 1600 + 369_. these objects are separate entries in the catalog of @xcite , but in fact are the same object . the pg coordinates ( ra 16:00:05.2 , dec + 36:55:29 ) are in error . the object is located at ra 15:59:33 , dec + 36:56:58 ( 1950 ) . _ 1713 + 332_. this high - proper motion star has 1996.3 coordinates ra 17:13:46.7 , dec + 33:16:22 . _ 2009 + 622_. this object was known to show an infrared excess and was spectroscopically observed by schultz , zuckerman , & becklin ( 1996 ) in an effort to detect emission from the inferred low - mass main sequence companion , with ambiguous results . we clearly detect strong , variable emission in the core of the da wd absorption line . in addition , the emission velocity is seen to vary by some 270 , indicative of a very short - period system ( @xmath0 1 day ) . _ 2126 + 734_. this object was found to have a cool wd companion at a separation of 14 by zuckerman et al . in table [ knowndd ] , we present all known dds in close ( short - period ) binary systems having measured orbital periods and primary masses , including the 5 recently discovered sdb+wd pairs with confirmed periods . we omit the ultra - short period am cvn systems , since they are likely not detached dds , and their primary stars are believed to be currently accreting from he secondaries . in any case , their primary masses can not be directly measured , and there is a considerable range in the masses implied by the theory of their evolution ( e.g. , @xcite ) . successive table columns give the object , its name , the period in days , logarithm of the period , an estimate of the mass of the primary component , total system mass when known , and the source of the data . we now wish to compare the results of the observations with the predictions of theory . population synthesis calculations of the binary population of the galaxy ( ylts ; ity ) predict the birthrate and observable number distributions of dd and wd+ms pairs . here , we present the results of a recent set of calculations , which like ity and ylts take into account the effects of observational selection due to white dwarf cooling and the relative brightness of the components , and which also allow for a more complete comparison with observations ( see below ) . we note that in the context of dd evolution , the primary component is taken to be the visible member of the binary , while in fact the secondary ( dimmer ) component was most often the originally more massive star . exceptions are algol - like systems where conservative mass transfer has reversed the original mass ratio . the observable ( younger and brighter ) component may be a helium ( he ) or carbon - oxygen ( co ) wd , while the older , dimmer component ( usually unobserved ) may be a he or co wd , or a low - mass main sequence star . in systems composed of two close wds , the theory predicts that 82% of the primaries are he wds , in good agreement with observations . in fact , table [ knowndd ] shows that of all known dds with confirmed periods and primary masses , every wd except l 8702 is inferred to be of the he variety based on its low mass ( @xmath22 0.5 ) . the five sdb primaries are expected to be he stars which eventually will evolve to lower - mass co wds after cessation of core helium - burning . the goal of this and previous surveys aimed at detecting dd systems , beyond the desire to identify a pre- system , is to confront the observations with the mass and period distributions predicted by the theory . previously , these distributions have been considered independently ( cf . figures 1 and 2 of ylts ; figure 2 of ity ) . here , we consider the bivariate distribution @xmath23(log @xmath0 , log @xmath4 ) , which may provide insight into different evolutionary channels that , while producing dd systems with similar periods , might produce systems with significantly different masses . using the same population synthesis code as ity , we have recalculated the numbers of dd and wd+ms pairs on a fine log @xmath0 vs. log @xmath4 grid . the grid range and spacings are from 0.8 to 0.15 in steps of 0.025 , and log @xmath0 ( in days ) from 1.5 to 4.5 in steps of 0.1 . the masses are those of the primary ( visible ) wd . in figure [ bivary ] , we display a halftone plot of the theoretical dd number distribution , with the greyscale spanning values between 0 and 1.25 @xmath24 10@xmath25 . the filled symbols represent all known dds in table 3 , with circles for wd primaries and diamonds for sdb primaries . masses for the wd primaries have been taken from the literature . for the sdb stars , a mass of 0.5 has been assumed , based on their location in the temperature gravity plane ( @xcite ; @xcite ) . remarkable structure is evident in the bivariate distribution which is obscured in the projected univariate distributions @xmath23(log @xmath0 ) and @xmath23(log @xmath4 ) . a strong ridgeline extends from the low - mass , short - period region toward higher masses and longer periods . the structure is understood as the dd `` birthline '' at a given mass there is a period threshold beyond which systems do not interact and never evolve through a common envelope phase . the numbers of systems are largest at low mass , where the initial mass function produces the largest numbers of progenitor stars . the `` island '' at = 0.2 represents the wds with co cores . we further discuss the comparison between the theory and observations in section 8 . our survey method fails to detect all bonafide short - period binaries for several reasons : 1 . random inclination of the orbital planes to the line of sight masks radial velocity variations for near plane - on systems . 2 . our temporal sampling imposes two different constraints : @xmath26 ) exposure times for most objects were between 15 and 30 minutes , which would cause smearing of the observed line profiles for orbital periods shorter than 12 hours . this affects all objects in the same way . we adopt a period of 2 hours as a conservative lower limit to our sensitivity bandpass . @xmath27 ) some orbital periods will have low sensitivity due to period aliasing , affecting objects differently depending on the exact sequence of observation times . 3 . in systems with low - mass secondaries , the orbital velocity of the primary will be lower than for high - mass secondaries . wd+ms pairs are especially afflicted with this bias since the ms companion typically must have a mass @xmath6 0.3 in order to have remained hidden until detected spectroscopically . however , this bias is mitigated by the frequent presence of reprocessed emission in the ms secondary s photosphere , which is easily detected . 4 . given our modest radial velocity precision ( 2030 at our intermediate spectral resolution and s / n ratio ) , very long period systems will have velocities too small to detect variations . 5 . in very hot and/or strongly magnetic da and dao wds , the line profiles are usually shallow and very broad , and they lack the sharp , non - lte absorption core of line profiles of cooler or non - magnetic wds . velocity differences are therefore very difficult to discern , except for the very hottest da or dao wds , like feige 55 , which often show narrow photospheric emission in the absorption line ( cf . holberg et al . 1995 ) . fortunately , few wds in the sample fall into this category . previous investigators ( rs , fwg , @xcite , @xcite ) have used a number of analytical methods for estimating detection probabilities . each approach attempted to address a particular statistical question . for example , fwg calculated the probability that no binaries at all were found in their sample , while marsh et al . ( 1995 ) calculated the ratio of the probabilities that individual objects were binaries vs. single stars . our own concern is to estimate the fraction of bonafide dd systems our survey fails to detect due to selection effects . we choose a monte carlo approach , beginning with our calculated mass / period distribution ( figure [ bivary ] ) . for each object in table [ wdsample ] , we first randomly select an orbital period and primary mass from the theoretical distribution using the rejection method ( @xcite ) . we deem the experiment to have failed and continue with the next object if the orbital period is less than 2 hours . we then randomly select a secondary mass , using the same procedure , from one of the two distributions for wd+wd given in the top panel of figure 4 of ity , according to whether the primary is a he or co wd . the orbital separation and primary wd velocity are computed , an orbital plane inclination is chosen from the probability distribution @xmath28 , and an initial orbital phase is chosen from a uniform deviate . orbital phase zero is defined as the positive - going crossing of the systemic velocity by the primary s velocity curve . with all relevant orbital parameters fixed , radial velocities are then computed for each mid - exposure time of observation , and the maximum radial velocity difference is determined . we consider the simulated system to have been detected if the maximum velocity difference exceeds 65 , or 2.5 times an assumed 1@xmath17 precision of 25 ( consistent with our definition of definite detection when the observed line core variation exceeds 1.5 ) . a `` hit '' is recorded for each detected object , and the next object is then tested . the experiment was repeated for all wds in table [ wdsample ] , and for each complete pass through the sample the detection probability was taken to be the fraction of hits . this is specific to our sample , as it depends on the sequence of observation times for each object , and it assumes a parent population of 100% wd+wd binaries with the theoretical log @xmath0log @xmath4 distribution of figure [ bivary ] . the procedure was repeated 1000 times each for two cases of orbital plane inclination : @xmath29 ( edge - on ) and @xmath28 , and for three cases of allowed periods : all periods in our calculated grid , @xmath0 between 2@xmath120@xmath13 , and @xmath0 between 2@xmath1210@xmath13 . the result for the most favorable condition , with @xmath29 and @xmath0 between 2@xmath110@xmath13 , is shown in figure [ ddprob ] , where the distribution of detection efficiencies is nearly gaussian with a mean of 83% and standard deviation of 3% . the remaining results are given in table [ allddprob ] , which shows that the effects of random inclination of orbital planes and low velocities associated with longer orbital periods are both important contributors to decreased detection efficiency . we adopt an upper limit of 1020 days to our sensitivity bandpass ( with a lower limit of 2 hours ) , corresponding to a detection efficiency of 70% over our entire sample . we do not consider the probability of duplicity for individual objects , as theoretical predictions of the fraction of dds and the mass and period distributions pertain to the entire population of wds and sdbs in the galaxy . our method of estimating detection efficiencies allows an evaluation of selection effects on the raw observed fraction of dds detected in the sample as a whole . comparison of the existing observations with the bivariate distribution of figure [ bivary ] is not straightforward for two reasons : 1 ) the number of dd systems discovered to date which have orbital periods and primary ( visible ) masses determined is still quite small . 2 ) the measurement errors for the wd and sdb masses , based primarily on surface gravity errors , are generally 0.050.10 in and do not permit an unambiguous comparision . of the six systems near = 0.3 and lying just below the co peak at 0.2 ( 0.6 ) in figure [ bivary ] , five have sdb primaries with masses near 0.5 . as such , they sensibly ought to lie slightly below the co peak , and we do not consider this a serious disagreement . the sdb periods do agree very well with the predicted co wd period distribution . four objects lie comfortably near either the dd birthline or the two peaks at = 0.5 and 0.65 . the remaining objects do not appear to lie near peaks in the distribution . the agreement with theory would be improved if the derived wd primary masses were systematically too large , or if the theoretically predicted masses were systematically too small . we note that the effect of varying the common envelope efficiency parameter ( which was taken to be unity in our calculations ) is largely to shift the predicted periods , not the primary masses ( cf . @xcite ) . it is possible that at least some observed dd primary masses are overestimated due to the presence of a fainter and usually more massive wd companion . line profile analysis of the composite optical spectrum of the brighter , lower - gravity visible wd and a dimmer , but higher - gravity wd might result in a derived surface gravity higher than that of the primary alone and would imply a too - large primary mass . on the other hand , in cases where the mass ratios have been measured ( l 8702 , 1101 + 364 , 0957666 ) , the wds have similar masses , and it is unlikely that companions contribute more than 20% of the light in 1241010 , 1317 + 453 and 1713 + 332 ( t. marsh , private communication ) . also , the radial velocity data for 0957666 almost certainly preclude reducing the wd masses . there may be another effect which distorts the bivariate number distribution in the @xmath30@xmath31 diagram . the theoretical grid was constructed assuming that all wds follow the same cooling sequence computed by @xcite for 0.6 @xmath32 wds . following their calculations , we assumed that all wds are detectable for a time @xmath33 in other words , we assumed that the luminosity of the white dwarf decreases below a certain detection limit in @xmath34yr , irrespective of its mass . until very recently , cooling sequences for low mass wds were not available . however , recent calculations ( @xcite ) for @xmath4 = 0.150.45 show that the more massive wds ( 0.45 ) cool to the same luminosity several times slower than the lower - mass ones ( 0.15 ) . this may be understood as an effect of the larger surface areas of low - mass wds . an attempt to quantify the effect of this mass dependence of cooling curves on the number distributions of wds is currently underway ( tutukov & yungelson , in preparation ) . the difference in cooling curves may shift the theoretical distribution in figure 3 more in favor of high - mass systems , and the strong concentration of wds along the `` birthline '' below = 0.6 may diminish . in light of the difficulty still associated with interpreting the theoretical bivariate distribution , the total predicted number of dd systems and the projected log @xmath0 and log @xmath4 distributions remain at present the best means of comparing the theory and observations . in table [ ddtheory ] , we present the predicted fractions of dd and wd+ms systems for three values of . successive columns give , the number of close dds vs. all wds ( i.e. , the sum of all single wds plus all wide and close pairs ) , the number of pre- systems vs. all wds , and the number of close wd+ms pairs vs. all wds . the pre- column describes dds which have a total system mass greater than and may merge within a hubble time . the table also illustrates how many wds have to be surveyed for a detection of a pre- , and how much more effort has to be invested before observations will better constrain the theory . we now compare the predicted numbers of dds with two observed samples , the one reported here and that of bgrd , as these have similar magnitude limits and are sensitive to the same range of orbital periods ( 2 hr to a few days ) . three comparisons are possible : 1 . bgrd found 1 dd and 4 candidate dds in their sample of 54 wds ( 1/11 ) , and 4 wd+ms pairs , 2 which were unresolved , and 2 which were resolved as visual binaries ( 1/13 ) . in our combined sample of 107 wds and 46 sdbs we find 6 confirmed dds and 15 candidate dds , not all of which are new detections . we note that seven of these have sdb primaries , which have core helium - burning lifetimes of 12 @xmath24 10@xmath35 years , comparable to the interval during which wds are the brightest and are most likely to be discovered ( cf . however , the field sdb stars have a birthrate 0.5% of that of field wds ( @xcite ) , and they are thus quite heavily over - represented in our sample . considered separately , the fractions of dd systems with wd and sdb primaries among all stars in our sample are 1/8 and 1/7 , respectively . including the visual doubles and pg composites noted in tables [ wdsample ] and [ sdbsample ] , 12 wd+ms pairs are found in our sample ( 1/13 ) . excluding duplication of objects in common , the fractions of dds and wd+ms pairs found among all wds in our and bgrd s combined samples are 1/9 and 1/12 , respectively . these results are in good agreement with columns 2 and 4 of table [ ddtheory ] for values of near 1.0 and tend to exclude much lower values of . the population synthesis code also predicts the expected projected period and mass distributions of close dds . we first compare the observations with the predicted period distribution in figure [ ddnlogp ] . the theoretical prediction is the solid histogram ( taken as a projection onto the log @xmath0 axis from our most recent calculation of the bivariate distribution ) , and the dot - dashed histogram is the observed period distribution of the dds in table [ knowndd ] , including the newly - discovered sdb+wd pairs . the observed distribution ( arbitrarily normalized ) is generally in agreement with the predicted distribution , albeit possibly offset to shorter periods . part of the apparent deficit of systems at longer periods is due to the decline in our detection efficiency at orbital periods longer than 1020 days . 3 . in figure [ ddnlogm ] , the predicted distribution of primary masses of close dds is given as the solid histogram . there is a broad peak near 0.5 , corresponding to he wd primaries , and a second strong , narrow peak at 0.2 , corresponding to co wd primaries , with a tail extending to masses in excess of 1.2 . the dot - dashed histogram is the observed mass distribution of the dds in table [ knowndd ] . the observed distribution is generally in agreement with the predictions of theory , except for the apparent excess of stars in the bin at = 0.3 . all but one of these ( l 8702 ) are sdb stars , which as we have previously noted , are over - represented in our sample . + the theory also predicts that some 18% of primary stars in dds have co cores . we therefore expect to find 2 co wds in the confirmed short - period dd sample ; instead , we find none . we note that of the radial velocity variables in table [ ddvariables ] , 10 are in common with the sample of @xcite , among which only three objects have derived masses @xmath36 0.5 : 1232 + 479 ( 0.53 , weight 1 ) , 1446 + 286 ( 0.82 , weight 2 ) , and 2117 + 539 ( 0.50 , weight 1 ) . four other objects are in common with the sample of bragaglia et al . ( 1995 ) , but only one has a derived mass @xmath36 0.5 : 1422 + 095 ( 0.51 ; weight 1 ) . it is crucial to obtain accurate masses for the remaining objects in table [ ddvariables ] to identify any co primaries which may reside in potential pre- systems . the possibility that the primary stars optical spectra may be diluted with that of dimmer , possibly higher - gravity companions must be taken into account . when confirmed , the 18 new candidate dds ( including the sdb variables ) from our survey will increase the number of known short - period dds to 29 systems , approaching the number predicted to include a super - system with a period short enough to merge in a hubble time ( 1/40 for = 1.0 ; see table [ ddtheory ] ) . clearly , in addition to the follow - up observations to determine orbital periods and masses of the dd candidates in the current sample , further survey work is required in order to expand the wd sample by approximately a factor of 2 ( for of 1.0 ) . the previous discussions have not taken into account dd systems missed in our survey due to selection effects . we first consider the systems with wd primaries . assuming an efficiency of 70% , our raw number of 14 confirmed and candidate dd systems would imply the existence of some 20 bonafide dd systems with wd primaries in our sample of 107 wds , or a fraction just under 1/5 of all wds . the same result is obtained from the sample of dds with sdb primaries . table [ ddtheory ] shows that this substantially exceeds the predicted fraction for 1 . such a high implied fraction of close binaries among all wds suggests that , formally , @xmath7 12 . we note that the condition that @xmath7 1 has been suggested as a requirment for an explanation of the correlation of orbital periods and eccentricities of high mass binary pulsars ( e.g. , @xcite ) and millisecond binary pulsars ( e.g. , @xcite ; @xcite ; @xcite ) . however , since detailed numerical models of the ce process , which involve energy sources other than gravitational ( e.g. , @xcite ) are not available , any treatment of the evolution of orbital separations inside the ce , which ultimately determines the predicted number distributions , is at best an approximate estimate . the dao wds are a small class of hot , hydrogen - rich stars that show weak , narrow @xmath37 absorption in addition to the broad balmer lines . @xcite analyzed 12 daos and found that half of them have such low surface gravities ( and by implication low masses ) , that unlike most wds they can not be products of post - agb evolution . @xcite suggested that the low - mass daos might be the product of post - extended horizontal branch ( post - ehb ) evolution . another alternative was raised by @xcite , who noted that the very low mass of the dao feige 55 ( @xmath22 0.4 ) might be related to its double degenerate nature . the ehb stars ( sdbs and the hottest bhb stars ) are thought to have core masses very near 0.5 , and post - ehb evolution alone is unlikely to have produced feige 55 s low core mass . in contrast , close binary evolution is predicted to produce a wealth of low - mass remnants , as observations confirm . feige 55 was detected as a radial velocity variable in iue observations , and its orbital parameters were subsequently determined from data including observations obtained during our 1995 march observing run ( holberg et al . we also observed two other dao wds , 1210 + 533 and 2342 + 806 . the former star has relatively weak absorption and has neither a sharp non - lte absorption core nor a central reversal . it does not appear to be a velocity variable , but we can not rule it out conclusively . the latter star ( gd 561 ) is a near - twin of feige 55 , with a somewhat hotter temperature and an even lower surface gravity and mass . our observations do show a distinct emission reversal like that of feige 55 , but the star is constant in velocity at a level 15 over a span of 6 nights . these results point out the need for further investigation into the possible connection between close binary evolution and the very low masses of approximately half of all doa wds . we have performed a radial velocity survey of a total of 153 field wds and sdbs , most of which were not previously known to be binary . in the combined sample , we have discovered 18 new dd candidates with wd or sdb primaries , and 1 new confirmed wd+ms pair . among the 7 sdb variables , we have obtained orbital solutions for the short - period sdb+wd pairs feige 36 ( @xmath0 = 85 ) and ton 245 ( @xmath0 = 25 ) , and moran et al . ( 1998 , in preparation ) have found orbital periods for 0101 + 039 ( @xmath0 = 161 or 137 ) , 1432 + 159 ( @xmath0 = 539 ) , and 2345 + 318 ( @xmath0 = 578 ) . our conclusions are : 1 . the population of close , short - period dd systems predicted by the theory of close binary star evolution does in fact exist in significant numbers . the raw observed fractions of confirmed short - period dds and wd+ms pairs among all wds are in satisfactory agreement with the predictions of theory for values of the common envelope efficiency parameter near 1.0 . however , applying a correction for selection effects implies an excess of existing dds over the predicted number . a resolution of this apparent discrepancy will have to await a more complete understanding of the ce process . 2 . the observed dd orbital period and primary mass distributions do fall very nearly in the peak of the predicted distributions for = 1.0 , although possibly shifted slightly . we note that the discrepancy of the predicted and observed period distributions could be made to vanish by a small decrease in the value of , a free parameter in the theory ( `` the parameter of ignorance '' , i. iben ) . the discrepancy with the predicted mass distribution is less amenable to fine - tuning of the theory , and we suggest that some previous primary wd mass determinations may have been systematically biased to higher masses due to dilution of the optical spectra by the presence of unseen , higher - gravity companions . over - representation of sdb primaries in the combined sample is responsible for the apparent excess of stars with masses near 0.5 . a refined analysis of the effect of the mass - dependence of wd cooling sequences upon the @xmath30@xmath31 number distribution is also required . 3 . the low - mass systems observed by marsh et al . ( 1995 ) and marsh ( 1995 ) tend to have short orbital periods , in agreement with theoretical predictions . however , their total system masses likely are too small to make them viable progenitors . furthermore , the primary stars must have he cores , while co white dwarfs are thought to be the best candidates to undergo the central ignition and detonation of carbon required to produce a . the one known super- dd system , lb 11146 , has an unknown orbital period , although its projected orbital separation is 1 a.u . the large remnant masses suggest that the progenitor stars also were massive , had large radii in the agb phase , and that the system might have gone through at least one ce phase leading to a short orbital period . however , no `` loaded gun '' has yet been found , i.e. , a super- system with @xmath0 10@xmath1 . on the other hand , the sample of confirmed short - period dds is still smaller than the number predicted to contain a massive , short - period system . when confirmed , the 18 new dd candidates discovered in our survey will increase the number of known short - period dds among all wds to a fraction approaching that which theory predicts will include a pre - system . + we shall continue with follow - up observations to solve for the orbital parameters of the new wd+wd and sdb+wd candidates found in our survey , and with the survey work itself to further expand the sample . additional work on the dao wds is needed to elucidate the connection between close binary evolution and the very low masses of half of daos . infrared observations , especially for the higher - luminosity sdb stars , will be required to distinguish main sequence companions having masses as low as 0.1 from compact companions . our survey has revealed a significant new population of close dd systems . even if no convincing potential progenitor is found , the distributions of dd orbital periods and primary masses ( and secondary masses and mass ratios , where observeable ) constitute powerful probes of the ce process , and a meaningful test of the predictions of the theory of close binary evolution now seems within reach . acknowledgements : this work has been supported by nasa grants nagw-2678 , g005.44000 , and by the director s discretionary research fund at the space telescope science institute . lry acknowledges support through rfbr grant 960216351 , and is grateful for the hospitality of meudon observatory . ras acknowledges fruitful conversations with tom marsh , jim liebert , jay holberg , and gary schmidt . lry acknowledges day - to - day discussions of stellar evolution with alexander tutukov . the authors are grateful for the long - term committment of the kpno tac and awards of valuable kitt peak 2.1-m time , crucial to the success of our large - scale survey . we thank the anonymous referee for a critical review of the manuscript . cllccl 0002@xmath19729 & gd 408 & 14.35 & @xmath50.08 & @xmath50.91 & 1,13 0004@xmath19330 & gd 2 & 13.82 & @xmath50.29 & @xmath51.21 & 1,2 0101@xmath19039 & feige 11 & 12.06 & @xmath50.24 & @xmath50.98 & 3 0101@xmath19048 & g 1@xmath545 & 13.96 & @xmath190.31 & @xmath50.57 & 0112@xmath19142 & pg & 15.0 : & @xmath50.59 & @xmath50.47 & 3,4 0134@xmath19833 & gd 419 & 13.08 & @xmath50.64 & @xmath50.06 & 4 0148@xmath19467 & gd 279 & 12.44 & @xmath190.06 & @xmath50.63 & 1,2 0151@xmath19017 & g 73@xmath54 & 15.00 & @xmath190.16 & @xmath50.59 & 0205@xmath19250 & g 35@xmath529 & 13.22 & @xmath50.04 & @xmath50.85 & 2 0208@xmath19396 & g 74@xmath57 & 14.50 & @xmath190.30 & @xmath50.47 & 0232@xmath19525 & g 174@xmath55 & 13.75 & @xmath50.55 & @xmath190.08 & 4 0401@xmath19250 & g 8@xmath58 & 13.81 & @xmath190.11 & @xmath50.58 & 1 0425@xmath19168 & lp 415@xmath5415 & 14.02 & @xmath50.09 & @xmath50.97 & 0437@xmath19138 & lp 475@xmath5242 & 14.86 & @xmath50.43 & @xmath50.25 & 4,13 0439@xmath19466 & lsv @xmath1946 21 & 12.67 & @xmath50.30 & @xmath51.24 & 5 0453@xmath19418 & gd 64 & 13.77 & @xmath190.23 & @xmath50.61 & 1 0532@xmath19414 & gd 69 & 14.75 & @xmath190.32 & @xmath50.54 & 0549@xmath19158 & gd 71 & 13.06 & @xmath50.25 & @xmath51.16 & 1 0553@xmath19053 & g 99@xmath547 & 14.08 & @xmath190.61 & @xmath50.11 & 6 0612@xmath19177 & g 104@xmath527 & 13.40 & @xmath50.15 & @xmath50.98 & 1,2,7,8 0637@xmath19478 & gd 77 & 14.80 & @xmath190.13 & @xmath50.64 & 6 0644@xmath19376 & g 87@xmath57 & 12.10 & @xmath50.08 & @xmath50.92 & 1,2 0651@xmath5020 & gd 80 & 14.82 & @xmath50.21 & @xmath51.19 & 7 0706@xmath19295 & kuv & 15.45 & @xmath190.09 & @xmath50.66 & 0710@xmath19741 & gd 448 & 14.97 & @xmath50.06 & @xmath50.75 & 9 0713@xmath19584 & gd 294 & 12.04 & @xmath190.00 & @xmath191.57 & 1,10 0743@xmath19442 & gd 89 & 14.92 & @xmath190.07 & @xmath50.73 & 0834@xmath19501 & pg & 15.82 & @xmath50.36 & @xmath51.20 & 5 0839@xmath19232 & pg & 14.42 & @xmath50.14 & @xmath50.98 & 1 0854@xmath19404 & gd 98 & 14.87 & @xmath50.13 & @xmath50.94 & 0900@xmath19554 & pg & 13.84 & @xmath50.05 & @xmath190.49 & 1,10,11 0938@xmath19550 & pg & 14.77 & @xmath50.52 & @xmath190.05 & 4 0939@xmath19262 & ton 21 & 14.53 & @xmath50.35 & @xmath51.16 & 0940@xmath19068 & pg & 13.70 & @xmath50.19 & @xmath51.03 & 7,11,20 0943@xmath19441 & g 116@xmath552 & 13.32 & @xmath190.07 & @xmath50.54 & 1 0945@xmath19246 & lb 11146 & 14.32 & @xmath50.47 & @xmath50.19 & 4,6 1005@xmath19642 & gd 462 & 14.03 & & & 12 1011@xmath19570 & gd 303 & 14.57 & @xmath50.49 & @xmath50.32 & 4,13 1012@xmath19007 & pg & 14.90 & @xmath50.12 & @xmath50.30 & 3,10 1019@xmath19637 & g 235@xmath567 & 14.71 & @xmath190.36 & @xmath50.49 & 1022@xmath19050 & lp 550@xmath552 & 14.18 & @xmath190.19 & @xmath50.51 & 7 1026@xmath19023 & lp 550@xmath5292 & 14.20 & @xmath50.41 & @xmath190.42 & 4 1057@xmath19719 & pg & 14.95 & & & 12 1101@xmath19249 & feige 36 & 12.70 & @xmath50.23 & @xmath50.89 & 3 1101@xmath19364 & ton 1323 & 14.44 & @xmath50.46 & @xmath190.40 & 4,14 1104@xmath19602 & g 197@xmath54 & 13.80 & @xmath50.02 & @xmath50.81 & 1 1121@xmath19216 & g 120@xmath545 & 14.24 & @xmath190.31 & @xmath50.52 & 1123@xmath19189 & pg & 14.01 & @xmath50.41 & @xmath50.59 & 4,15 1126@xmath19384 & gd 310 & 14.89 & @xmath50.12 & @xmath51.01 & 1129@xmath19156 & pg & 14.04 & @xmath50.52 & @xmath190.11 & 4 1204@xmath19451 & pg & 14.84 & & & 12,20 1210@xmath19533 & pg & 14.07 & @xmath50.70 & @xmath50.73 & 4,5,22 1229@xmath5013 & pg & 14.24 & @xmath50.35 & @xmath190.05 & 4 1232@xmath19479 & gd 148 & 14.52 & @xmath190.06 & @xmath50.68 & 1 1247@xmath19553 & gd 319 & 12.31 & @xmath190.03 & @xmath50.93 & 11,20 1257@xmath19048 & gd 267 & 15.05 & @xmath50.09 & @xmath50.88 & 1310@xmath19583 & pg & 13.89 & & & 12 1317@xmath19453 & g 177@xmath531 & 14.13 & @xmath190.03 & @xmath50.56 & 1,16 1319@xmath19466 & g 177@xmath534 & 14.55 & @xmath190.00 & @xmath50.66 & 1408@xmath19323 & gd 163 & 14.06 & @xmath190.03 & @xmath50.81 & 1 1411@xmath19219 & pg & 14.38 & @xmath50.36 & @xmath50.21 & 4,13?,15,20 1422@xmath19095 & gd 165 & 14.32 & @xmath190.14 & @xmath50.59 & 7 1432@xmath19159 & pg & 13.90 & @xmath50.12 & @xmath190.12 & 3,10 1446@xmath19286 & ton 214 & 14.54 & @xmath50.09 & @xmath190.01 & 10 1459@xmath19305 & pg & 13.98 & @xmath190.30 & @xmath50.84 & 20 1537@xmath19652 & gd 348 & 14.64 & @xmath190.18 & @xmath50.49 & 1538@xmath19269 & ton 245 & 13.89 & @xmath50.24 & @xmath50.90 & 11 1559@xmath19369 & g 180@xmath523 & 14.36 & @xmath190.17 & @xmath50.56 & 1601@xmath19581 & pg & 14.30 & & & 12 1615@xmath5154 & g 153@xmath541 & 13.41 & @xmath50.22 & @xmath51.08 & 2 1630@xmath19618 & gd 354 & 15.5 & cc@xmath51 & & 11 1632@xmath19177 & pg & 13.08 & @xmath50.13 & @xmath190.47 & 4 1633@xmath19433 & g 180@xmath563 & 14.82 & @xmath190.42 & @xmath50.42 & 1637@xmath19335 & g 180@xmath565 & 14.65 & @xmath190.18 & @xmath50.57 & 1647@xmath19591 & g 226@xmath529 & 12.24 & @xmath190.16 & @xmath50.62 & 2 1713@xmath19332 & gd 360 & 14.46 & @xmath50.11 & @xmath50.90 & 1,16 1713@xmath19695 & g 258@xmath53 & 13.27 & @xmath190.07 & @xmath50.70 & 1 1716@xmath19426 & pg & 13.85 & @xmath50.66 & @xmath50.35 & 3,4 1756@xmath19827 & g 259@xmath521 & 14.30 & @xmath190.35 & @xmath50.52 & 1822@xmath19410 & gd 378 & 14.39 & @xmath50.19 & @xmath50.96 & 1,13?,19 ? 1824@xmath19040 & g 21@xmath515 & 13.90 & @xmath190.05 & @xmath50.55 & 1,7 1826@xmath5045 & g 21@xmath516 & 14.54 & @xmath190.24 & @xmath50.56 & 1840@xmath19042 & gd 215 & 14.73 & @xmath190.25 & @xmath50.62 & 1845@xmath19019 & kpd & 12.96 & @xmath50.23 & @xmath51.09 & 7 1919@xmath19145 & gd 219 & 13.01 & @xmath190.06 & @xmath50.66 & 1 1935@xmath19276 & g 185@xmath532 & 12.97 & @xmath190.17 & @xmath50.56 & 1936@xmath19327 & gd 222 & 13.58 & @xmath50.12 & @xmath50.88 & 1943@xmath19163 & g 142@xmath550 & 14.08 & @xmath50.06 & @xmath50.86 & 1953@xmath5011 & g 92@xmath540 & 13.71 & @xmath190.30 & @xmath50.61 & 7 2009@xmath19622 & gd 543 & 15.15 & @xmath50.69 & @xmath50.29 & 4,17 2028@xmath19390 & gd 391 & 13.37 & @xmath50.15 & @xmath50.98 & 2032@xmath19188 & gd 231 & 15.34 & @xmath50.04 & @xmath50.80 & 16 2032@xmath19248 & wolf 1346 & 11.54 & @xmath50.07 & @xmath50.87 & 1,2 2047@xmath19372 & g 210@xmath536 & 12.93 & @xmath190.14 & @xmath50.66 & 2111@xmath19498 & gd 394 & 13.09 & @xmath50.24 & @xmath51.15 & 2117@xmath19539 & g 231@xmath540 & 12.33 & @xmath190.07 & @xmath50.67 & 2 2124@xmath19550 & g 231@xmath543 & 14.66 & @xmath190.15 & @xmath50.68 & 2126@xmath19734 & g 261@xmath543 & 12.88 & @xmath190.01 & @xmath50.66 & 2,18 2134@xmath19218 & gd 234 & 14.45 & @xmath50.04 & @xmath50.83 & 2136@xmath19828 & g 261@xmath545 & 13.02 & @xmath50.02 & @xmath50.72 & 1 2149@xmath19021 & g 93@xmath548 & 12.77 & @xmath190.00 & @xmath50.78 & 1,2,7 2150@xmath19338 & gd 398 & 15.11 & @xmath50.03 & @xmath190.25 & 10 2204@xmath19071 & pg & 14.74 & @xmath50.05 & @xmath50.30 & 19 2207@xmath19142 & g 18@xmath534 & 15.61 & @xmath190.32 & @xmath50.58 & 2246@xmath19223 & g 67@xmath523 & 14.31 & @xmath190.21 & @xmath50.67 & 2307@xmath19636 & g 241@xmath546 & 14.16 & @xmath50.70 & @xmath50.09 & 4 2309@xmath19105 & gd 246 & 13.11 & @xmath50.32 & @xmath51.23 & 1,2,7 2319@xmath19691 & gd 559 & 14.75 & @xmath50.07 & @xmath190.23 & 10 2326@xmath19049 & g 29@xmath538 & 13.10 & @xmath190.20 & @xmath50.65 & 2 2328@xmath19108 & pg , kuv & 15.53 & @xmath50.03 & @xmath50.90 & 2329@xmath19407 & g 171@xmath52 & 13.82 & @xmath190.03 & @xmath50.72 & 1 2341@xmath19322 & g 130@xmath55 & 12.92 & @xmath190.14 & @xmath50.61 & 1 2342@xmath19806 & gd 561 & 14.53 & @xmath50.25 & @xmath51.21 & 5,22 2345@xmath19318 & pg & 14.18 & @xmath50.05 & @xmath190.01 & 3,10 clrrl 0154@xmath19204 & & @xmath1913.0 & 30.9 & 0154@xmath19182 & & @xmath1986.5 & 25.8 & 0212@xmath19230 & & @xmath578.2 & 15.4 & 0250@xmath19189 & & @xmath194.4 & 11.4 & 0322@xmath19114 & & @xmath1928.5 & 9.3 & 0342@xmath19026 & & @xmath1912.6 & 7.0 & 0349@xmath19094 & & @xmath1930.4 & 10.8 & 0749@xmath19658 & & @xmath527.3 & 6.9 & 0806@xmath19516 & & @xmath195.6 & 14.4 & 0816@xmath19313 & ton 313 & @xmath552.7 & 14.9 & 0823@xmath19465 & & @xmath1923.7 & 9.9 & 0839@xmath19399 & k345@xmath530 & @xmath59.3 & 13.2 & visual double 0856@xmath19121 & & @xmath1997.0 & 10.2 & 0918@xmath19029 & & @xmath1979.4 & 16.3 & 0919@xmath19272 & ton 13 & @xmath533.2 & 16.9 & 1114@xmath19072 & feige 38 & @xmath193.6 & 8.3 & 1154@xmath5070 & & @xmath54.7 & 13.2 & 1223@xmath19058 & & @xmath19131.4 & 21.5 & 1224@xmath19671 & & @xmath541.1 & 7.0 & 1230@xmath19052 & & @xmath557.1 & 12.6 & 1233@xmath19426 & feige 65 & @xmath1961.2 & 13.4 & 1303@xmath5114 & & @xmath53.2 & 13.5 & 1323@xmath5085 & & @xmath553.2 & 7.8 & 1325@xmath19101 & & @xmath1911.5 & 22.3 & 1343@xmath5101 & & @xmath1948.9 & 17.4 & 1442@xmath19342 & & @xmath519.0 & 5.2 & 1623@xmath19386 & kuv & @xmath590.9 & 41.7 & visual double 1643@xmath19209 & & @xmath1952.3 & 12.8 & 1704@xmath19221 & & @xmath562.1 & 9.2 & 1708@xmath19409 & & @xmath5197.4 & 13.1 & 2059@xmath19013 & & @xmath57.6 & 14.3 & 2110@xmath19127 & & @xmath196.8 & 5.2 & pg comp 2128@xmath19096 & & @xmath1923.0 & 19.4 & 2135@xmath19044 & & @xmath539.0 & 7.3 & 2204@xmath19034 & & @xmath510.1 & 13.7 & clccccc 0101@xmath19039 & feige 11 & 0.67 & @xmath50.173 & 0.50 & & 1 & & 0.57 & @xmath50.243 & & & alias ? 0135@xmath5052 & l 8702 & 1.56 & @xmath190.193 & 0.47 & 0.99 & 2 0957@xmath5666 & l 10126 & 0.06 & @xmath51.234 & 0.37 & 0.69 & 3,4 1101@xmath19249 & feige 36 & 0.35 & @xmath50.456 & 0.50 & & 5 1101@xmath19364 & ton 1323 & 0.15 & @xmath50.824 & 0.31 & 0.58 & 6 1202@xmath19608 & feige 55 & 1.49 & @xmath190.173 & 0.40 & & 7 1241@xmath5010 & & 3.35 & @xmath190.525 & 0.31 & & 8 1317@xmath19453 & & 4.80 & @xmath190.681 & 0.33 & & 8 1432@xmath19159 & pg & 0.23 & @xmath50.649 & 0.50 & & 1 1538@xmath19269 & ton 245 & 2.50 & @xmath190.389 & 0.50 & & 5 1713@xmath19332 & & 1.12 & @xmath190.049 & 0.35 & & 8 2331@xmath19290 & & 0.17 & @xmath50.770 & 0.39 & & 8 2345@xmath19318 & pg & 0.24 & @xmath50.618 & 0.50 & & 1

self - propelled point - like particles move along circular trajectories when their translocation velocity is constant and the angular velocity related to their orientation vector is also constant . we investigate the collective behavior of ensembles of such circle swimmers by brownian dynamics simulations . if the particles interact via a velocity - trajectory coordination " rule within neighboring particles , a self - organized vortex pattern emerges . this vortex pattern is characterized by its particle - density correlation function , the density correlation function of trajectory centers , and an order parameter representing the degree of the aggregation of the particles . here , we systematically vary the system parameters , such as the particle density and the interaction range , in order to reveal the transition of the system from a light - vortex - dominated to heavy - vortex - dominated state , where vortices contain mainly a single and many self - propelled particles , respectively . we also study a semi - dilute solution of curved , sinusoidal - beating flagella , as an example of circling self - propelled particles with explicit propulsion mechanism and excluded - volume interactions . our simulation results are compared with previous experimental results for the vortices in sea - urchin sperm solutions near a wall . the properties of the vortices in simulations and experiments are found to agree quantitatively .

1309.3021

systems of self - propelled particles ( spps ) , which consume energy to maintain a persistent non - brownian motion , exhibit an abundance of fascinating non - equilibrium collective behaviors such as swarming , swirling , and clustering @xcite . examples are found in very different areas of physics and biology , and range from actin filaments and microtubules in motility assays @xcite through metallic nano - rods @xcite to flocking of birds @xcite and groups of people @xcite . in all of these examples , spps employ different propulsion mechanisms , and also interact with each other in different ways , including direct physical contact @xcite , chemotaxis @xcite , hydrodynamic interactions @xcite , and restricted visual contact @xcite . however , the collective behavior of spps systems is surprisingly similar , displaying phenomena such as giant density fluctuations with anomalously slow relaxation @xcite , order - disorder phase transition with increasing noise and decreasing density @xcite , etc . since the simplest model of interacting spps was introduced by vicsek et al . @xcite , nowadays widely referred as the vicsek model " @xcite , collective behavior of spp systems has attracted much interest @xcite . in the vicsek model , point particles moving with constant velocity align their directions of motion with the average direction of other particles in a prescribed interaction range , while internal or external noise is taken into account by adding a random increment to their orientation vectors . by a variation of the system parameters , e.g. the particle density and the strength of the perturbation , such a simple system has been found to undergo an order - disorder phase transition , whose nature ( first or second order ) is related to the particle velocity and the way the perturbation is introduced @xcite . according to these studies , the initial conditions and settings of the simulation , the type of interaction and the boundary conditions play an important role in the formation of certain collective patterns of motion @xcite . a special , but not rare class of patterns of collective motion in spps systems are swirls or vortices , in which a group of particles circle around a common center . in experiments , swirls and vortices emerge both in non - living particle systems ( vertically vibrated granular rods @xcite , motility assays of actin filaments @xcite or microtubules @xcite ) and in systems of living micro - organisms ( bacteria colonies @xcite , zoo - plankton under optical stimulus @xcite , sea - urchin sperm trapped near a substrate @xcite ) . in simulations , the emergence of swirls and vortices was found to depend on the initial conditions and the model settings , such as a circular boundary @xcite , alignment with a blind angle " of interaction behind each agent @xcite , a harmonic attractive pair potential with a noise above a critical value @xcite , and hydrodynamic interactions @xcite . in most of the computational models exhibiting vortices , a single spp is either assumed to perform a random walk @xcite , or to move with constant magnitude of velocity @xcite . in this paper , we consider a class of spps which move in the absence of noise along _ curved trajectories _ rather than straight lines , which we call circle spps . the driving force of a circle spp does not coincides with its propagation direction , so that its trajectory is a cycloid in three spatial dimensions or a circle in two dimensions @xcite . artificial circle swimmers can be constructed by introducing a tilted or bend structure to catalytically driven colloidal rods @xcite , giving thermophoretic colloidal swimmers a l - shape @xcite , or by designing micro - machines of connected beads which are moved relative to each other in a time - irreversible multi - step cycle @xcite . examples of circle swimmers in living systems include certain bacteria @xcite and spermatozoa @xcite when the micro - organisms are attracted or confined in their motion to a surface or wall . the most carefully analyzed experiment on the collective motion of circle spps might be the self - organized vortices of sea - urchin sperm @xcite . the 50-@xmath3m - long sperm circle clockwise when they gather at a substrate . with increasing surface density of sperm , the system exhibits a transition from a disordered state of randomly distributed sperm to a self - organized vortex - array state with local hexagonal order , in which each vortex consists of several sperm . in ref . , the emergence of this structure is attributed to the hydrodynamic interactions between individual sperm and between sperm vortices . however , simulations suggest that the steric interactions between rod - like spps @xcite and sinusoidal beating flagella @xcite strongly contribute to the collective motion and aggregation . while the morphology and properties of swirls and vortices in straight - trajectory spps systems have been studied intensively in recent years @xcite , much less is known theoretically about the collective motion of circle spps @xcite . the understanding of the effect of the spontaneous circular trajectory on the collective behavior of spps is essential for the understanding of many biological phenomena and for the design of microscopic machines . in addition to the parameters of spps ( like velocity , particle density , interaction range , and noise amplitude ) , the circle spp systems have at least the spontaneous curvature of the particle trajectory as an extra parameter . the larger parameter space suggests a more complex behavior of these systems . in order to study such systems , highly simplified models are often very useful , as has been demonstrated by the success of the vicsek model . an example for circle swimmers is the simplified model of ref . , designed to interpret the formation of vortex arrays of sea urchin sperm . in this model , each sperm is described by a point particle at its trajectory center , with a pairwise short - range attraction arising from hydrodynamic forces , and a longer range repulsion due to steric or hydrodynamic origin . inspired by this model , we construct a model of circle spps by point - like particles with a constant propagation velocity , where particles interact through a `` velocity - trajectory coordination '' , which takes the trajectory centers of neighbor particles into account . this algorithm of interaction is different from both the trajectory - center coordination " of ref . and the velocity coordination " of the vicsek model @xcite ( which averages the velocity of neighboring particles ) . we will analyze this model in detail , and show that it describes vortex formation and the evolution of a stationary vortex pattern . in addition to the point - like circle spps model , we also study curved , sinusoidal - beating flagella . sinusoidal beating of a filament or elongated body is a common self - propulsion mechanism in biological systems with low - reynolds - number hydrodynamics , e.g. nematodes @xcite and sperm of higher animals @xcite . the sinusoidal wave propagates from one end to the other on the filament - like body and pushes the surrounding fluid backwards to generate a forward force . thus , the cell or organism gains velocity in opposite direction of wave propagation . as indicated by experiments of biological systems @xcite and simulations of model systems @xcite , the sinusoidally undulating motion of the body does not destroy the general collective behaviors of rod - like spps . our flagellum model is coarse - grained as particles connected by harmonic springs , and the hydrodynamics is either approximated by anisotropic friction , or calculated by using multi - particle collision dynamics ( mpc ) , a mesoscopic particle - based simulation approach @xcite . by introducing a non - zero average curvature in the beating plane , the undulating flagellum traces out a circular trajectory in two dimensions , reminiscent the trajectory of sea urchin sperm at a substrate @xcite . the simplicity of the model allows us to analyze the contribution of volume exclusion and hydrodynamic interaction separately . the study provides insight into the effect of flagellar properties , such as frequency distribution and spontaneous curvature , on the collective motion and the emerging vortex patterns . the paper is organized as follows . section [ sec : models ] gives a brief description of our models and simulation methods . in sec . [ sec : spp ] , we analyze the collective motion of point - like circle spps systems . then , we study the collective motion of curved , sinusoidally beating flagella , and compare the results with circle spps models and sea - urchin sperm experiments in sec . [ sec : flagella ] . the results are summarized in sec . [ sec : summary ] . we consider @xmath4 point - like particles moving in a two - dimensional system of size @xmath5 . the number density of the particles is @xmath6 . each particle has a spontaneous circular trajectory of diameter @xmath7 , which is traversed in counter - clockwise direction , and a circular interaction region of diameter @xmath8 , as illustrated in fig . [ fig : circularswimmermodel ] . at time @xmath9 , the @xmath10-th particle has position @xmath11 , velocity @xmath12 , and trajectory center position @xmath13 of its spontaneous circle trajectory determined by @xmath14 where @xmath15 is a rotation matrix which rotates a vector in two dimensions counter - clockwise through an angle @xmath16 . note that @xmath17 is obtained from the instantaneous velocity and position , rather than from the full trajectory . the equation of motion for the @xmath10-th particles is then given by @xmath18 motivated by the nematic alignment of colliding rod - like particles @xcite , we define a majority rule for the the re - direction of the velocity as @xmath19 where @xmath20 is the simulation time step , @xmath21 is constant magnitude of the velocity , @xmath22 is the angular velocity of the orientation of the particle velocity which enforces a circular trajectory of diameter @xmath7 in dilute suspension , and @xmath23 is the sum of all particles within the interaction region of @xmath10-th particle including @xmath10-th particle itself . the tangential unit vector @xmath24 is obtained at the intersection of the spontaneous trajectory of @xmath25-th particle and the connect line between @xmath11 and @xmath26 , as illustrated in fig . [ fig : circularswimmermodel ] . the direction of @xmath24 is chosen to make @xmath27 . with such an interaction rule , a particle tends to move parallel or anti - parallel to the trajectories of its neighbors . therefore , the result of the interaction is a local nematic alignment of trajectories . in contrast to the vicsek model @xcite , the change of the velocity direction in our model is determined not by a polar alignment of velocities ( velocity coordination ) , but rather by a nematic alignment of local trajectories , which are determined by the instantaneous trajectory centers @xmath26 of the neighbor particles ( velocity - trajectory coordination ) . here , the choice of the alignment rule strongly depends on the interaction range relative to the preferred swimming radius . clearly , a majority rule can only generate alignment if the number of particles in the interaction range is sufficiently large . this is very difficult to achieve for @xmath28 when particles move collective in circular vortices . therefore , we choose @xmath29 . our simulations show that in this case velocity coordination as in the vicsek model does not give rise to vortex patterns in a circle spp system , because neighboring particles align their directions of motion but do not tend to generate vortices with several particles circling around a common center . note that we omit noise terms in eqs . ( [ eq : cseq1 ] ) and ( [ eq : cseq2 ] ) , so that our model is deterministic once the initial state is given . this is a reasonable description of microswimmers which are not too small , such as sperm . and @xmath25 . the dashed lines denote the spontaneous circular trajectories of the @xmath10-th and @xmath25-th particles . the gray area displays the neighbor region of particle @xmath10 . @xmath24 is the tangent unit vector to the trajectory of particle @xmath25 with @xmath30.,width=264 ] in the simulation , we use the diameter of spontaneous trajectory @xmath7 as length unit , @xmath20 as time unit , and @xmath31 . this implies a period of rotation of an interaction - free particle of @xmath32 . the interaction range @xmath8 is varied below , but is typically on the order of @xmath7 ( as explained above ) . the simulation box size is @xmath33 or @xmath34 . the simulation time for each system is @xmath35 . we construct a two - dimensional flagellum model @xcite from a linear sequence of @xmath36 beads , which are connected by harmonic springs with rest length @xmath37 . the local spontaneous curvature of such a filament is a function of time @xmath9 and position @xmath38 along the flagellar contour , measured from the first bead , @xmath39 where @xmath40 is the beating frequency , @xmath41 is the wave number , @xmath42 is the initial phase chosen independently from a uniform distribution on the interval @xmath43 for each flagellum , and @xmath44 is the average curvature . we denote a flagellum with non - zero @xmath44 a curved flagellum . the constant @xmath45 controls the strength of beating , which is related to the amplitude of the shape undulations . equation ( [ eq : beat ] ) generates a traveling wave , which propagates from the front to the end of the flagellum . the curvature elasticity of the flagellum is determined by a bending potential , which depends on the deviations of the angles between neighbor bonds from their preferred value @xmath46 . therefore , the elastic energy of flagellum @xmath25 is given by @xmath47 ^ 2 + \sum_{i=1}^{n_f-2}\frac{1}{2}\frac{\kappa}{l_0 ^ 3}\biggl[{\bf r}_{i+1,j } - \mathfrak r(c_j(x_i , t)l_0){\bf r}_{i , j}\biggr]^2 . \label{eq : potential}\ ] ] here , @xmath48 are bond vectors , and @xmath49 denotes the position of the @xmath10-th bead of the @xmath25th flagellum . @xmath50 is an operator rotating a two - dimensional vector counter - clockwise by an angle @xmath51 . @xmath52 is the spring constant and @xmath53 the bending rigidity @xcite . the volume exclusion of beads on different flagella is taken into account by a purely repulsive , truncated and shifted lennard - jones potential @xmath54 , & r\leq2^{1/6}l_0\\ 0 , & \text{otherwise } \\ \end{array}\right . \ , \label{eq : ljpotenttial}\ ] ] where @xmath55 is the distance between two beads belonging to different flagella . the swimming of both nematodes and sperm is determined by low - reynolds - number hydrodynamics , where viscous forces dominate over inertial forces . in this regime , the dynamics of a rod can often be well described by resistive - force theory @xcite , in which hydrodynamic interactions between different segments of a rod are approximated by an anisotropic friction ( af ) , so that @xmath56 , @xmath57 , where @xmath58 and @xmath59 are the velocity components of a flagellum segment on the directions parallel and perpendicular to the local tangent vector , respectively . @xmath60 is the friction coefficient for a segment of unit length , with @xmath61 or @xmath62 @xcite . hydrodynamic interactions are not included in simulation employing af , so that volume exclusion is the only interaction between the flagella in this case . considering the relatively large sizes of flagella and nematodes of several 10 @xmath3 m and a few millimeters , respectively , we neglect thermal fluctuations in our af simulations . in simulations with full hydrodynamics , the flagellum model is embedded in a two - dimensional fluid , and the time evolution of the system is carried out by employing a hybrid molecular dynamics approach multi - particle collision dynamics ( mpc ) for the fluid @xcite in combination with molecular dynamics ( md ) for the flagellum as follows . during the streaming step , the point - like particles of the mpc fluid move ballistically during a time interval @xmath63 , while newton s equations of motion for the flagellum particles are integrated by the verlet algorithm with a time step of @xmath64 . in the subsequent collision step , the flagella exchange momentum with the neighboring fluid particles . the collisions are performed by sorting all fluid and flagellum particles into the cells of a cubic lattice ( with lattice constant @xmath65 ) , which are labeled by an index @xmath66 . a rotation operator @xmath67 is assigned to each box . if @xmath68 is the pre - collisional center - of - mass velocity of all particles in box @xmath66 , the post - collisional velocity @xmath69 of a particle @xmath10 in the box is given by @xmath70 , where @xmath12 is its pre - collision velocity . this simple collision rule conserves mass , momentum and energy , which guarantees the emergence of navier - stokes hydrodynamics on length scales larger than @xmath65 . since the beating flagella perpetually inject energy into the fluid , we apply a local thermostat after each collision step . ( black solid squares ) of the circular trajectory of a single flagellum as a function of @xmath71 in af simulations . the red line is a linear fit to the data with slope of 80 . the inset is the dimensionless velocity @xmath72 of a single flagellum ( black solid squares ) as a function of @xmath73 . the red line is the linear fit to the data . in both ( a ) and ( b ) , the beating frequency of the flagellum is @xmath74 . , width=264 ] the simulation system contains @xmath4 flagella of length @xmath75 in a simulation box of size @xmath76 , where @xmath65 is size of a mpc collision box . the wave number @xmath77 implies the presence of two complete sinusoidal waves on the flagellum . the number density of flagellum is @xmath78 . periodic boundary conditions are employed . during the simulation , each flagellum has a constant frequency @xmath40 chosen from a gaussian distribution with the average @xmath79 and the variance @xmath80 , where @xmath81 is a dimensionless number characterizing the width of the frequency distribution . when the average spontaneous curvature @xmath44 is positive , the flagellum is curved and prefers a clockwise circular trajectory . the trajectory diameter @xmath7 depends approximately linearly on @xmath71 in af simulations , and the center - of - mass velocity @xmath82 of the flagellum decreases approximately linearly with @xmath73 from its value for @xmath83 , as shown in fig . [ fig : graph_bend_basic ] . therefore , for a flagellum beating with the average frequency @xmath84 , it takes @xmath85 beats to move a one body length . equivalently , the swimming velocity is @xmath86 . the beats number for a flagellum to complete a full circle trajectory is @xmath87/[0.066- 0.0057 c_0l_{fl}]$ ] . for example , in our af simulations , the most strongly curved flagellum , with @xmath88 , takes @xmath89 beats to complete a circle , while the least curved flagellum , with @xmath90 , takes @xmath91 beats . each simulation runs for at least @xmath92 . therefore , the flagella beat at least 2500 times on average in a simulation run , and complete about 35 to 110 full circles . the other parameters of our flagellum simulations are listed in ref . . the collision of two particles with elongated structure in a viscous fluid environment results in a cooperated motion that the particles move together with close packing . if the elongated particles are straight , the nematic interaction results in velocity alignment of neighbor particles and cluster formation @xcite . when two elongated and curved flagella encounter each other , they tend to get close and synchronize their configuration via hydrodynamic interactions as well as volume exclusion @xcite and form an effective extended flagellum " with the same average curvature . the circular motion of such an extended flagellum around a center is equivalent to the motion of two flagella around the same center , the first step towards the formation of a vortex . each black dash is the trajectory of a particle in a time interval @xmath93 , and the red dots are the corresponding trajectory centers . the inset shows a magnification of the area at the bottom left corner where two vortices consists of groups of circle spps with equal distance between neighboring groups . the number of particle groups of them are five and six , respectively . the parameters are @xmath94 , @xmath95 , and @xmath96 . , width=264 ] @xmath4 circle spps are distributed in the simulation box with random positions and random velocity directions at the initial time @xmath97 . shortly after the start , vortices emerge , which are formed by one or several particles circling around a common center , as illustrated in fig . [ fig : particle_snapshot ] and movies in the supplemental material @xcite . the vortices distribute in space with no obvious long - range order , and displace vigorously if the trajectory centers of the particles belonging to the same vortex do not coincide precisely at the same point . when two vortices overlap or collide , particles are exchanged between the vortices until they either fuse or separate from each other far enough . therefore , the vortex mass , defined as the number of circle spps forming it , changes during collisions . note that the system is deterministic due to the absence of noise , and a noise - induced fission of vortices does not occur in the simulation . we characterize the emergent vortex pattern of circle spps by the particle density - correlation functions , the trajectory - center density - correlation functions , and the order parameter reflecting the degree of the particle aggregation . by systematically changing the particle density @xmath98 and the size of the interaction region @xmath8 , we then determine the dynamics state diagram of this system . the correlation function of particle densities at time @xmath9 is given by @xmath99 and the correlation function of the trajectory - center density at time @xmath9 is @xmath100 where @xmath101 and @xmath102 are the number densities of the particles and the trajectory centers at position @xmath103 and time @xmath9 , respectively . we consider times @xmath9 in the interval @xmath104 $ ] , where @xmath105 and @xmath106 . most of the systems have not yet reached a stationary state at @xmath107 , so that the correlation functions @xmath108 and @xmath109 still contain some averaging over different structures and spatial arrangements of the vortices . . the inset is the density correlation function of an annular region of evenly distributed particle density . the diameter of the inner circle of the annular region is @xmath110 , and the diameter of the outer circle is @xmath111 . , width=264 ] as a function of the particle density @xmath112 . the red dashed line is the guide to the eye at value @xmath113 . ( b ) @xmath114 as a function of the diameter of the interaction region @xmath8 . the open symbols and the solid symbols represent the light - vortex - dominated systems and the heavy - vortex - dominated systems , respectively . the error bars " in ( a ) and ( b ) indicate the half width of peak ii as defined in fig . [ fig : graph_particle_grho ] , and therefore characterize the width of the band of particle trajectories within a vortex . , title="fig:",width=264 ] + as a function of the particle density @xmath112 . the red dashed line is the guide to the eye at value @xmath113 . ( b ) @xmath114 as a function of the diameter of the interaction region @xmath8 . the open symbols and the solid symbols represent the light - vortex - dominated systems and the heavy - vortex - dominated systems , respectively . the error bars " in ( a ) and ( b ) indicate the half width of peak ii as defined in fig . [ fig : graph_particle_grho ] , and therefore characterize the width of the band of particle trajectories within a vortex . , title="fig:",width=275 ] the particle density - correlation function @xmath115 mainly characterizes particle arrangement in a vortex , in particular for @xmath116 , as shown in fig . [ fig : graph_particle_grho ] . in the single - vortex region , @xmath115 displays a pronounced spatial dependence . there is a pronounced peak at @xmath117 , denoted as peak i , which arises from the interaction range of our model . the displacement of a particle in a simulation time step , @xmath118 , is much smaller than @xmath119 ; therefore , during the very short time interval for two particles to adjust their motion and form a vortex , their relative distance remains essentially constant . this distance only changes when a third particle comes into play . if the mass of a vortex is high , the multi - particle interaction affects and disturbs the distances of other particles in a vortex . thus , a higher density @xmath98 depresses peak i in fig . [ fig : graph_particle_grho ] . for example , when @xmath120 , doubling @xmath98 strongly depresses the height of peak i. similarly , increasing @xmath8 enhances peak i when @xmath98 is kept constant . therefore , a vortex is composed of several groups of particles with the constant distance @xmath119 between neighbor groups , as illustrated in fig . [ fig : particle_snapshot ] . if the tendency to keep the distance @xmath119 is strong for the particle groups , a second and third peak become evident , see fig . [ fig : graph_particle_grho ] . the position of the @xmath121-th peak is approximately @xmath122 where @xmath123 and @xmath114 is the diameter of the vortex . note that such periodic particle density modulation is a dynamic temporary structure which breaks and reconstructs with time . the peak of the vortex structure at @xmath124 , denoted as peak ii , indicates the diameter @xmath114 of a vortex . according to the definition of the density correlation function @xmath0 , this diameter is weighted by the mass of the vortices , so that the vortices consisting of more particles contribute more to @xmath114 . thus , @xmath114 mainly reflects the diameter of the heavy " vortices , containing a large number of particles in the vortex . we define the width of peak ii as the distance between the peak position and the position of the subsequent minimum . a larger peak width indicates a wider band of particles in the vortex . [ fig : particle_dv ] shows @xmath114 changing with the particle density @xmath112 and the diameter of the interaction region @xmath8 . the error bars " in fig . [ fig : particle_dv ] indicates the width of peak ii , _ i.e. _ the width of the particle band of the vortices . in the low - density limit , the particles hardly meet each other , and the vortices mainly contain a single spp , which is defined as a light - vortex - dominated state . in this state , @xmath125 and the width of peak ii is narrow , as illustrated in fig . [ fig : particle_dv]a for @xmath126 . at high particle density , e.g. for @xmath127 , a majority of the particles belongs to vortices of mass larger than 10 , which is defined as a heavy - vortex - dominated state . in this state , @xmath114 still approximately equals @xmath7 , but the width of peak ii is about twice as large as for @xmath126 . an extraordinarily large width at @xmath128 in fig . [ fig : particle_dv]a indicates a transition of the system from a light - vortex - dominated state to a heavy - vortices - dominated state . the same transition from light to heavy vortices also happens with increasing @xmath8 at constant density @xmath98 . the vortex diameters @xmath114 for different @xmath112 coincide in the heavy - vortex - dominated state , but are slightly smaller in the light - vortex - dominated state , see fig . [ fig : particle_dv]b . increasing @xmath8 causes a systematic increase of @xmath114 . the small dip at @xmath96 , can be interpreted as follows . @xmath114 is not only determined by the spontaneous trajectory diameter @xmath7 of circle spps , but also influenced by the interaction range @xmath8 , because the vortex prefers a dynamic temporary state with an integral number of particle groups with a constant neighboring distance @xmath119 , as shown in fig . [ fig : particle_snapshot ] . therefore , @xmath114 can be estimated by @xmath129 $ ] , where @xmath121 is the number of particle groups in a vortex ; this estimate provides qualitatively the correct trend for @xmath130 , but deviates a little quantitatively because @xmath121 differs for different vortices in the system . for example , heavy vortices with @xmath131 are dominant when @xmath132 and @xmath133 , while there are many heavy vortices of both @xmath131 and @xmath134 when @xmath132 and @xmath96 . for @xmath135 , @xmath0 approaches unity due to the loss of spatial correlations at large distances , which indicates a liquid - like order of the vortices . , with @xmath120 and @xmath128.,width=264 ] , as a function of the particle density @xmath112 . ( b ) @xmath136 as a function of the diameter of the interaction region @xmath8 with different particle densities . the red dashed line is @xmath137 . the inset is the order parameter @xmath2 as a function of time @xmath138 when @xmath133 . the red , blue and black lines represent the particle densities @xmath139 , @xmath140 , and @xmath141 , respectively.,title="fig:",width=302 ] + , as a function of the particle density @xmath112 . ( b ) @xmath136 as a function of the diameter of the interaction region @xmath8 with different particle densities . the red dashed line is @xmath137 . the inset is the order parameter @xmath2 as a function of time @xmath138 when @xmath133 . the red , blue and black lines represent the particle densities @xmath139 , @xmath140 , and @xmath141 , respectively.,title="fig:",width=302 ] the correlation function of the trajectory - center density , @xmath1 , shown in fig . [ fig : particle_gc ] , has a large peak at @xmath142 because of the aggregation of the trajectory centers . @xmath1 decays rapidly to zero for @xmath143 , indicating that the trajectory centers aggregate in a small spot at the center of the vortex . two heavy vortices can not overlap , otherwise they will fuse into one or separate into several distinct vortices . therefore , @xmath1 displays a depletion zone from @xmath7 to @xmath144 . for @xmath145 , @xmath1 increases rapidly and displays a peak corresponding to the the closest possible distance @xmath136 of two neighboring heavy vortices . for larger distances @xmath38 , @xmath1 approaches a constant with decaying oscillations , which reveals the absence of long - range order and corresponds to a liquid - like spatial arrangement of vortices . the dependence of @xmath136 on the particle density is show in fig . [ fig : particle_dvv]a . @xmath136 is nearly independent of density , with @xmath146 for @xmath120 and @xmath147 , when the system is heavy - vortices dominated . similarly , @xmath148 for @xmath149 , when the system is light - vortices dominated . the transition between light - vortices - dominated and heavy - vortices - dominated states occurs when @xmath150 , which agrees with the transition region extracted from @xmath0 ( compare fig . [ fig : particle_dv]a ) . as shown in fig . [ fig : particle_dvv]b , the increase of @xmath136 with @xmath8 for constant density @xmath112 is somewhat faster than the linear relation @xmath151 . for @xmath139 , @xmath136 displays a pronounced increase by more than @xmath152 when @xmath8 is increased from @xmath153 to @xmath152 . when @xmath154 , @xmath136 exhibits an unexpected decrease for @xmath133 ; we interpret this as the absence of a second vortex - mass - increasing time period due to the inadequate simulation time , as will be explained in sec . [ sec : op ] below . in the initial state , all particles are randomly distributed in space , so that the number of the particles and the number of the trajectory centers in a defined area assumes a binomial distribution . when the particles form vortices , the aggregation of the trajectory centers cause a large deviation of the number of trajectory centers in a defined area from its average value . a stronger aggregation leads to a larger center - density deviation . therefore , we define an order parameter @xmath2 as @xmath155 where @xmath156 is the variance of a binomial distribution of point particles , and @xmath157 is the variance of the trajectory center numbers in an area @xmath158 . if the center positions are randomly distributed in the space , @xmath2 vanishes . @xmath2 increases with the degree of the aggregation of centers , which indicates the formation of vortices . if all particles circle around a common center , @xmath159 . suppose @xmath4 particles form @xmath160 vortices in the two - dimensional space @xmath76 . the number of the vortices consisting of @xmath121 particles is @xmath161 , and @xmath162 in order to elucidate the relation between the average mass @xmath163 of the vortices and the order parameter , we assume that the trajectory centers of all particles in a vortex collapse into one point , the vortex center , although they always have some narrow distribution around the center in the simulations . according to the trajectory - center density correlation function @xmath1 ( fig . [ fig : particle_gc]a ) , the distance between two vortices in our systems is usually larger than @xmath152 in the systems dominated by heavy vortices , except when two vortices are colliding and merging . therefore , the possibility to find two heavy vortex center in an area @xmath158 is low . thus , we assume that the probability to find two vortex centers in @xmath158 is zero . we divide our system of size @xmath76 into boxes of size @xmath158 , and fill the boxes with at most one vortex . then , the variance of the center - number distribution in @xmath158 is @xmath164 note that the weight average of the vortex mass is @xmath165 thus , @xmath157 can be written in terms of @xmath166 as @xmath167 substituting eqs . ( [ eq : delta_0 ] ) and ( [ eq : delta^2 ] ) into eq . ( [ eq : orderparameter_eq ] ) , we find @xmath168 in our simulations , @xmath169 is negligible , so that @xmath170 this result reveals a simple linear relation between our order parameter @xmath2 and the weight average of vortex mass . thus , @xmath2 is a proper order parameter to characterize the degree of vortex formation . , for ( a ) various diameters @xmath8 of the interaction region , and ( b ) various particle densities @xmath98 . the dashed lines are power laws with different exponents , as indicated.,title="fig:",width=264 ] + , for ( a ) various diameters @xmath8 of the interaction region , and ( b ) various particle densities @xmath98 . the dashed lines are power laws with different exponents , as indicated.,title="fig:",width=264 ] the order parameter @xmath2 increases with time once the particle start to move and the trajectory centers aggregate , as shown in fig . [ fig : particle_order ] . @xmath2 first undergoes a fast increase during a time interval which we call period i. the increase of @xmath2 with time during period i occurs faster for larger @xmath98 , but is not sensitive to @xmath8 , as indicated by the dashed lines in fig . [ fig : particle_order ] . during this period , the vortices are formed mainly by particles which are close to each other in the initial state ; therefore , @xmath2 depends roughly linearly on the particle density , as shown in fig . [ fig : particle_order_time]b . period ii is defined as the time period after period i , during which @xmath2 increases only very slowly and approaches a plateau , as shown in fig . [ fig : particle_order ] . the plateau has a higher value for larger @xmath98 and @xmath8 . for high densities ( @xmath171 when @xmath120 in fig . [ fig : particle_order]b ) or large @xmath8 ( @xmath172 when @xmath139 in fig . [ fig : particle_order]a ) , there is another dynamical evolution after period ii , denoted period iii , in which @xmath2 displays another pronounced increase . the increase of @xmath2 in period iii indicates the formation of heavy vortices by vortex fusion not seen for some low @xmath173 or small @xmath8 systems . however , in our simulation time scale , some systems with high @xmath173 and large @xmath8 also do not display period iii ( see the inset of fig . [ fig : particle_dvv]b ) , in this case because the simulation time is not long enough for vortex fusion to occur . at times @xmath174 , @xmath175 , @xmath176 , @xmath177 , and @xmath178 , ( a ) as a function of @xmath8 when @xmath179 , and ( b ) as a function of @xmath112 when @xmath120 . the dashed lines in ( b ) are linear fits of @xmath2 at @xmath174 , @xmath175 , and @xmath176.,title="fig:",width=264 ] + at times @xmath174 , @xmath175 , @xmath176 , @xmath177 , and @xmath178 , ( a ) as a function of @xmath8 when @xmath179 , and ( b ) as a function of @xmath112 when @xmath120 . the dashed lines in ( b ) are linear fits of @xmath2 at @xmath174 , @xmath175 , and @xmath176.,title="fig:",width=264 ] by analyzing the order parameter @xmath2 at a given time @xmath9 for different systems , we can also determine the transition from the light - vortex - dominated state to the heavy - vortex - dominated state , as shown in fig . [ fig : particle_order_time ] . the transition happens when @xmath180 , for @xmath139 , and when @xmath150 for @xmath120 . @xmath2 assumes a small value near zero in the light - vortex - dominated systems , then increases rapidly with increasing @xmath8 and @xmath112 in the heavy - vortex - dominated systems . at @xmath181 from @xmath182 to @xmath176 the white lines divide the state diagram into four regimes according to the shape of @xmath183 . for the definition of regimes @xmath45 to @xmath184 see main text.,width=321 ] according to the different shapes of @xmath183 for different systems , compare fig . [ fig : particle_order_time ] , we obtain the state diagram of circle spps at @xmath185 , as shown in fig . [ fig : phasediagram ] . the state diagram is divided into four regimes . for systems locating in regime @xmath45 , @xmath2 has completed the first growth period i , and reached the plateau of period ii , i.e. the final state of these systems ( at @xmath186 ) is dominated by light vortices . in regime @xmath187 , the systems have passed the stationary period ii , and at the end of the simulation time are just undergoing the fusion of light vortices ( period iii ) ; thus , these systems are in the crossover between regimes @xmath45 and @xmath188 . in regime @xmath188 , the systems have come to the second stationary state after the pronounced increase of @xmath2 in period iii . finally , in regime @xmath184 , which occupies the high @xmath173 and high @xmath8 section of the state diagram , the evolution of the fusion of heavy vortices is still in progress at @xmath186 ; the end state is characterized in this case by heavy vortices of similar weight . note that the division of the regimes according to the temporal evolution of @xmath183 resembles but does not coincide with a classification according to the value of @xmath2 in the final state state . for example , @xmath2 does not have the highest value in regime @xmath184 . the reason is that the characteristic time scale of the evolution strongly depends on particle density , so that at the chosen final simulation time , the systems in regime @xmath184 has only been able to complete the initial collection of neighbor particles . ( see inset of fig . [ fig : particle_dvv]b ) . our simulations are performed without noise . therefore , the vortex mobility only depends on multi - particle interactions . increasing vortex mass leads to an increasing complexity of these interactions and can result in a high mobility of vortices . the light - vortex - dominated systems can not increase the vortex mass via collision because all vortices do not move after formation . in contrast , systems which form heavier vortices during period i can continue to increase the vortex mass via the collision of vigorous heavy vortices until a limit is reached , which is determined by the particle density . at the boundary between regimes @xmath45 and @xmath187 in fig . [ fig : phasediagram ] , the value of @xmath2 is between 1 and 3 , indicating the weighted vortex mass less than 5 . moreover , the position of the line agrees well with the transition between light - vortex - dominated state and heavy - vortex - dominated state indicated by the variation of @xmath114 and @xmath136 shown in fig . [ fig : particle_dv ] and fig . [ fig : particle_dvv ] . therefore , the boundary between regimes @xmath45 and @xmath187 can be identified with the transition from the light - vortex - dominated to the heavy - vortex - dominated state . , @xmath189 , and @xmath190 , ( b ) in af with @xmath191 , @xmath189 , and @xmath190 , and ( c ) in af with @xmath192 , @xmath193 , and @xmath190 . the dashed lines in ( c ) shows the local hexagonal order . ( d ) normalized flagellum density @xmath194 averaged over a time interval of @xmath195 of the system in ( b ) . see also movie s1 in the supplemental material @xcite . , title="fig:",width=151 ] , @xmath189 , and @xmath190 , ( b ) in af with @xmath191 , @xmath189 , and @xmath190 , and ( c ) in af with @xmath192 , @xmath193 , and @xmath190 . the dashed lines in ( c ) shows the local hexagonal order . ( d ) normalized flagellum density @xmath194 averaged over a time interval of @xmath195 of the system in ( b ) . see also movie s1 in the supplemental material @xcite . , title="fig:",width=151 ] + , @xmath189 , and @xmath190 , ( b ) in af with @xmath191 , @xmath189 , and @xmath190 , and ( c ) in af with @xmath192 , @xmath193 , and @xmath190 . the dashed lines in ( c ) shows the local hexagonal order . ( d ) normalized flagellum density @xmath194 averaged over a time interval of @xmath195 of the system in ( b ) . see also movie s1 in the supplemental material @xcite . , title="fig:",width=151 ] , @xmath189 , and @xmath190 , ( b ) in af with @xmath191 , @xmath189 , and @xmath190 , and ( c ) in af with @xmath192 , @xmath193 , and @xmath190 . the dashed lines in ( c ) shows the local hexagonal order . ( d ) normalized flagellum density @xmath194 averaged over a time interval of @xmath195 of the system in ( b ) . see also movie s1 in the supplemental material @xcite . , title="fig:",width=151 ] of the vortex as a function of the radius position in af simulations with @xmath196 and @xmath190 . the dashed lines are the functions @xmath197 . the inset shows @xmath198 as a function of @xmath199 in a mpc fluid with @xmath189 and @xmath191 ; the dashed line is a guide to the eye to indicate a plateau . , width=264 ] the flagella are initially distributed randomly in space , with random orientations . after an initial relaxation time " , which coresponds to the time a single flagellum needs to move several circles ( @xmath200 ) , the system reaches a stationary state in which the curved flagella spontaneously organize into rotating vortices @xcite , as illustrated in fig . [ fig : snapshot]ab . we start to gather data for averaging at @xmath201 , when the flagella have completed more than ten full circles even in the system with largest @xmath7 . the flagella are moving clockwise and the waves on the flagella are propagating counter - clockwise . in af simulations , the flagella only have hard - core interactions , while in a mpc fluid , the hydrodynamic interactions synchronize the flagellar beat in the same vortex and packs flagella tightly due to hydrodynamic attraction @xcite . the mass of each vortex changes dynamically due to the collision with flagella in neighboring vortices @xcite . figure [ fig : omega ] shows the angular velocity @xmath202 , where @xmath203 is the center - of - mass velocity of a flagellum and @xmath199 is the distance between the mass center of the flagellum and the vortex center to which the flagellum belongs , as a function of @xmath199 . in af simulations , @xmath198 approaches @xmath204 for large @xmath199 , where @xmath82 is the velocity of a freely swimming flagellum . at small radii , the volume exclusion between the propagating sinusoidal configurations of neighboring flagella reduces the angular velocity . thus , although @xmath198 still slowly increases with decreasing @xmath199 for @xmath205 , it is much lower than expected from the relation @xmath206 . on the other hand , at large radius ( @xmath207 ) , in some af systems ( for example , @xmath191 and @xmath192 in fig . [ fig : omega ] ) , @xmath198 is larger than @xmath204 due to the repulsive interaction between two flagella belonging to different vortices . in conclusion , the volume exclusion between the flagella depresses @xmath198 due to interactions between flagella in the same vortex , but enhances @xmath198 due to interactions between neighboring vortices . in a mpc fluid , @xmath198 also approximately obeys the power law @xmath208 at large @xmath199 . however , at radii @xmath209 , @xmath198 is nearly independent of @xmath199 , as shown in fig . [ fig : omega]b . here , the synchronized flagella form a closed ring with integer numbers of waves and rotate with same angular velocity , as shown in movie s2 in the supplementary material @xcite . although @xmath210 is not exactly equal to @xmath82 in systems with mpc fluid or with af , an assumption of a unique swimming velocity is still a good approximation for comparison of the flagella system with the circle spps system . we define a normalized density of flagellar segments as @xmath211 where @xmath212 is the number density of monomer beads averaged in a square box of area @xmath213 at the position @xmath103 at time @xmath9 . in order to gain better statistics of the vortex structures , @xmath214 is the average over time @xmath215 , with @xmath215 is chosen to be @xmath216 . an example of an image of @xmath217 is shown in fig . [ fig : snapshot]d . and ( b ) the normalized trajectory center density @xmath218 . both densities @xmath194 and @xmath218 are averaged over a time interval @xmath195 . the parameters are @xmath191 , @xmath189 , and @xmath190 . the symbols indicate the results of mpc - fluid ( red bullets ) and of af ( black squares ) simulations . , title="fig:",width=264 ] + and ( b ) the normalized trajectory center density @xmath218 . both densities @xmath194 and @xmath218 are averaged over a time interval @xmath195 . the parameters are @xmath191 , @xmath189 , and @xmath190 . the symbols indicate the results of mpc - fluid ( red bullets ) and of af ( black squares ) simulations . , title="fig:",width=264 ] the correlation function of flagellum density is then defined as @xmath219 similarly , the correlation function of flagellum trajectory - center density is @xmath220 where @xmath221 and @xmath222 is the number density of the centers of flagellum trajectories at @xmath103 and time @xmath9 . figures [ fig : correlation]a , b show examples of @xmath223 and @xmath224 , respectively . both correlation functions approach a constant at large distances , indicating the absence of long - range order and a liquid - like arrangement of vortices , similarly as discussed for circle spps in sec . [ sec : spp_corr ] . long - range order of vortices , such as a hexagonal arrangement , is not observed even for large flagellum densities . however , a local hexagonal order is still possible due to the volume exclusion between vortices , as indicated by the higher - order peaks of @xmath225 and shown in real - space snapshots in fig . [ fig : snapshot]c . the interpretation of the correlation function is of course very similar as for correlations functions of circle spps in sec . [ sec : spp_corr ] . the first local maximum of @xmath223 corresponds to the average vortex diameter , the first local maximum of @xmath224 to the average distance between neighbor vortices . a comparison of af and mpc simulations clearly shows that structures in the vortices in the af model are considerably more ordered , which leads to pronounced oscillations of the correlations functions . in the af simulations , @xmath224 nearly vanishes at @xmath124 , indicating that the area occupied by the trajectory centers in a vortex is significantly smaller than @xmath7 , in agreement with our circle spps observations . in the mpc - fluid simulations , the weaker correlations indicate a larger diversity of vortex sizes but also a larger mobility of the vortices . a comparison of the snapshots and correlation functions in figs . [ fig : snapshot ] and [ fig : correlation]b with those of sea - urchin sperm vortices in ref . shows that the phenomena observed in af simulations are more similar to the experimental behavior than those in mpc simulations . furthermore , the results of our af simulations agree much better with those of our circle spps simulations . we attribute the deviations of the mpc - simulation results from those of the experiments to the very strong hydrodynamic interactions in the two - dimensional model system . indeed , the flow field around a dragged point particle in two spatial dimensions decays only logarithmically with distance @xmath55 , while in three dimensions it decays much faster , like @xmath226 , and even faster in the presence of a wall . in the experiments , the motion of sperm near a wall is governed by three - dimensional hydrodynamics . thus , we conclude that a detailed numerical investigation of the importance of hydrodynamic interactions for the formation of sperm vortices requires full three - dimensional hydrodynamic simulations . the two - dimensional mpc simulation still provide the important information that the synchronized flagellum beating of the sea - urchin sperms in a vortex is the result of hydrodynamic interactions between the beating tails , as indicated by the snapshot in fig . [ fig : snapshot]a ( see also refs . ) . in the remainder of this section , we focus on the analysis of the systems with anisotropic friction . figure [ fig : d1d2 ] shows the average vortex diameter @xmath114 and the average vortex distance @xmath136 as functions of @xmath98 , @xmath7 , and @xmath81 . in the low - density limit , the flagella move without touching each other . thus @xmath114 must approach @xmath7 and @xmath136 must approach the average distance between flagella . as the density increases , @xmath114 and @xmath136 increase slowly , as shown in fig . [ fig : d1d2]a . as in the circle spps systems , compare figs . [ fig : graph_particle_grho ] and [ fig : particle_dvv ] , @xmath114 and @xmath136 also increase and level off when @xmath227 . however , the drop of @xmath114 and @xmath136 near @xmath128 for circle spps systems , which indicates the transition from the light - vortex - dominated to the heavy - vortex - dominated state , is not seen in the curved - flagella system . on the other hand , the increase of @xmath114 can also partially be attributed to the volume exclusion between flagella , so that the orbit at radius @xmath199 in the vortex can be occupied only by a limited number of flagella . similarly , @xmath136 increases with @xmath112 due to volume exclusion , which generates an effective repulsion between the neighboring vortices . when we change the preferred trajectory diameter @xmath7 of the curved flagella , the diameter of a vortex varies as @xmath125 , as shown in fig . [ fig : d1d2]b , in good agreement with the behavior of the circle spps systems . however , over a wide range of @xmath98 or @xmath7 , @xmath228 remains nearly independent of these parameters and fluctuates around a value @xmath229 . this value of @xmath228 is found for the interaction range @xmath230 to @xmath231 in circle spps systems ( fig . [ fig : particle_dvv]b ) . thus , we conclude that the effective size of the interaction region of a curved flagellum is approximately the same as the diameter of its circular trajectory . in the sea - urchin sperm vortex experiment @xcite , the radius of the sperm vortex is @xmath232 m , and the average vortex distance is @xmath233 m . therefore , @xmath234 in the experiment , in excellent agreement with @xmath235 in our af simulations . in our circle spps simulations , @xmath236 for systems with @xmath237 in the heavy - vortex - dominated state . the vortex formation is not sensitive to the width of the beating - frequency distribution of the flagella , as shown in fig . [ fig : d1d2]c , although it leads to a range of flagellar velocities . this insensitivity explains the emergency of the vortices in the sea - urchin sperm experiment even though there is a spread of beating frequency of about 9 percent @xcite . figure [ fig : orderparameter ] shows that the order parameter @xmath2 , which represents the degree of aggregation and is closely related to the weight average of vortex mass , grows with increasing @xmath98 and @xmath7 . interestingly , in the stationary state of the flagella system , @xmath2 increases linearly with flagellum density when @xmath238 , which is reminiscent of the linear relation between @xmath2 and @xmath112 for @xmath239 in circle spps systems ( fig . [ fig : particle_order_time]b ) . however , in circle spps systems , @xmath2 continues to increase after @xmath240 . we attribute the linear relation of @xmath2 and @xmath98 in the flagella system to the effect of volume exclusion . the linear dependence of @xmath2 on the flagellum density agrees very well with the experimental observations for sea - urchin sperm . on the other hand , @xmath2 is found to increase as @xmath241 , see fig . [ fig : orderparameter]b , indicating the importance of a second length scale , which should be related to the flagellum length . the effect of volume exclusion and synchronization of the flagellar beat manifests itself in the following three main aspects . first , as shown in fig . [ fig : snapshot]bc , the flagellum vortices are closed rings composed of synchronized flagella . this closed structure makes the fusion of large vortices to happen very infrequently , because the necessary force to open such a structure to fuse two neighboring vortices is very large . therefore , as suggested by our circle spps simulations , the flagellum system experiences vortex formation and reorganization only during the early stages of structure formation , corresponding to period i in the circle spp system , which leads to vortices described by an order parameter @xmath2 which depends linearly on the particle density . the next step to raise the aggregation number , corresponding to period iii in the circle spp system , which requires the fusion of heavy vortices , is prevented by their nearly impenetrable closed structure . second , an extraordinary heavy vortex can not exist for long , because the maximum mass of a vortex is determined by the balance of forces between flagella . for a flagellum swimming at a distance @xmath199 from the center , the prevalent trajectory curvature is @xmath242 , while the preferred trajectory curvature of a flagellum is @xmath243 . when @xmath244 , the flagellum at @xmath199 pushes outwards and exert an outward force on other flagella in outer layers ; similarly , when @xmath245 , a flagellum at @xmath199 pushes inwards and exerts an inward force on other flagella in inner layers . when @xmath199 exceeds @xmath7 , an instability should develop , which leads to a breakup into smaller vortices of radius @xmath246 . for systems with larger @xmath7 , the region @xmath244 is larger and the typical difference between @xmath242 and @xmath243 is smaller , so that the maximum vortex mass is larger than for systems with smaller @xmath7 . therefore , @xmath2 increases with increasing @xmath7 for fixed @xmath98 , as shown in fig . [ fig : orderparameter]ab . third , consider now a large flagellum density , for which the system is already full of vortices of similar mass , frequently colliding with each other ( see movie s3 in supplemental material @xcite ) . a further increase of @xmath98 starts to destroy the vortex structure and the order parameter @xmath2 decreases , as shown in fig . [ fig : orderparameter]a for @xmath247 . such a decrease of @xmath2 was not observed in the sea - urchin sperm experiment @xcite because the experimental system was not strictly two - dimensional . we conjecture that at a certain surface density of sperm , the substrate is completely packed with vortices and can not absorb any more cells . the local hexagonal order of the sperm vortex array @xcite is a clue for this close packing . a further increase of the surface density of sperm is not possible because higher density will cause more frequent collision and consequently expel some sperm from the near - substrate layer . it seems not possible @xcite to obtain a higher surface density of sperm than 6000/mm@xmath248 , the largest density investigated in ref . . for higher densities in the experiments , multiple layers of sperm developed , and the layers on top were neither ordered nor destroying the pattern below @xcite . a possibility to increase the density further in experiments might be to restrict the sperm in a narrow slit of one layer thickness between two flat substrates . the variance @xmath81 of the frequency distribution also influence the order parameter , as shown in fig . [ fig : orderparameter]b . @xmath2 is not sensitive to @xmath81 when @xmath249 , but decreases with increasing @xmath81 for @xmath250 . when there are large differences between the frequencies of flagella in a vortex , the collisions between the undulating shapes increase the short - range repulsion , separate the flagella and cause a looser vortex structure , so that vortex break up more easily in collisions with other vortices . at small @xmath251 , this effect is small . note that even for @xmath252 , vortices still exist , as indicated by the density correlation functions . however , stable structures hardly exist for a long time . the frequent fission and fusion of vortices make @xmath2 small , although the system is not completely disordered . we have simulated systems of self - propelled particles with preferred circular trajectories ( circle spps ) interacting via a velocity - trajectory coordination rule , and systems of curved flagella propelled by a sinusoidal beating motion . in both systems , we observe the formation of vortex arrays , controlled by particle density , interaction range , and diameter of the preferred circular trajectory . for the circle - spp systems , the vortex array shows liquid - like rather than hexagonal spatial order . the diameter of the vortices , @xmath114 , is about the diameter of a single particle trajectory @xmath7 with a slight increase with the diameter of the interaction region @xmath8 , but is not sensitive to the particle density @xmath98 . the average distance between neighbor vortices , @xmath136 , is also not sensitive to @xmath98 , but increases quickly with increasing @xmath8 . a transition from a light - vortex - dominated state ( at low @xmath98 and small @xmath8 ) to a heavy - vortex - dominated state ( at high @xmath98 or large @xmath8 ) is observed . we use an order parameter @xmath2 to characterize the degree of the vortex formation . by comparing the time evolution of @xmath2 , we find that the vortex formation can be divided into three time periods . during period i , the particles collect neighbor ones to form vortices and @xmath2 increase quickly with time . the increase of @xmath183 during period i is slightly elevated with @xmath98 , but is not sensitive to @xmath8 . in the subsequent period ii , @xmath183 increases very slowly . in period iii , the vortex mass increases again more rapidly through vortex collision and fusion . note that environmental noise is not described in our model . therefore the fission , fusion and displacement of vortices is purely the result of multi - particle interactions of circle spps . in order to compare with the experiments of sperm cells near surfaces , we have also studied a more detailed model of curved , sinusoidally beating flagella . vortex patterns in this system emerge from the hard - core repulsion of the curved body of the elongated self - propelled particles moving in a viscous environment . in the simulations with anisotropic frictions , the collective motion of the curved flagella system agrees very well with the behavior of the circle spps system , as well as the phenomenon observed in the sea - urchin sperm experiments @xcite . as in the circle spps system , the average size of the vortices @xmath114 equals approximately @xmath7 , and slightly increases with the flagellum density . by comparing @xmath136 with the circle spps systems , we find that the size of the effective interaction region of a curved flagellum can be approximately identified with @xmath253 . the order parameter @xmath2 increases with @xmath98 as well as @xmath7 . the fraction @xmath235 coincides the value @xmath254 calculated by using the data from @xcite for sea - urchin sperm system . in conclusion , the collective motion of self - propelled particles , which leads to the formation of vortex arrays , can be well reproduced by circle spps with a velocity - trajectory coupling interaction . the velocity - trajectory coordination rule is a different interaction type than the velocity coordination rules employed since the vicsek model @xcite for the simulations of collective motion . such an interaction mimics , for example , the hard - core interaction of curved , sinusoidal beating flagella . the analysis of a more specific model of beating flagella allows to elucidate the features related to an explicit propulsion mechanism and physical interactions . y.y . gratefully acknowledges support from the national natural science foundation of china ( grants 21304020 ) and the shanghai postdoctoral scientific program ( program no.11r21411200 ) . g.g . gratefully acknowledges support from the vw foundation ( volkswagenstiftung ) through the program _ computer simulation of molecular and cellular bio - systems as well as complex soft matter_. for the flagellum model , we chose the parameters @xmath255 , @xmath256 , monomer mass @xmath257 , @xmath258 , @xmath259 , @xmath260 , and @xmath261 . volume exclusion is obtained by a repulsive lennard - jones potential with @xmath262 and @xmath263 . for brownian dynamics simulations with anisotropic friction ( af ) , the friction coefficients are chosen to be @xmath264 and @xmath265 , in order to reproduce the single - flagellum velocity @xmath266 in mpc fluid when @xmath267 . the mpc fluid is governed by the side length @xmath65 of the collision boxes , the mass @xmath268 of the fluid particles , the thermal energy @xmath269 , the rotation angle @xmath270 , and the average density of fluid particles @xmath271 . the natural time scale in mpc simulations is @xmath272 . the collision time is @xmath273 , with these parameters , the dynamic viscosity of the mpc fluid is given by @xmath274 @xcite .

we discuss 3-manifolds which are cyclic coverings of the 3-sphere , branched over 2-bridge knots and links . different descriptions of these manifolds are presented : polyhedral , heegaard diagram , dehn surgery and coloured graph constructions . using these descriptions , we give presentations for their fundamental groups , which are cyclic presentations in the case of 2-bridge knots . the homology groups are given for a wide class of cases . moreover , we prove that each singly - cyclic branched covering of a 2-bridge link is the composition of a meridian - cyclic branched covering of a determined link and a cyclic branched covering of a trivial knot . + + _ mathematics subject classification 2000 : _ primary 57m12 , 57r65 ; secondary 20f05 , 57m05 , 57m25 . + _ keywords : _ 3-manifolds , cyclic branched coverings , 2-bridge knots and links , dehn surgery , heegaard diagrams , heegaard genus , gems , crystallizations , cyclically presented groups , fundamental groups , homology groups . contents \1 . introduction and preliminaries . \2 . cyclic branched coverings . \3 . polyhedral construction . \4 . heegaard diagram construction and genus . \5 . surgery construction . \6 . coloured graph construction . \7 . fundamental groups . \8 . homology groups . \9 . decomposition of singly - cyclic branched coverings .

math0106164

the family of 2-bridge knots / links is a well - studied subject , starting from the classical papers of listing , dehn , alexander , reidemeister , schubert and others . almost everything is known about the symmetries , invariants , and geometric properties of these links , of their complements , and of manifolds connected with them . the aim of the present paper is to provide a review of the properties of 3-manifolds which are cyclic branched coverings of 2-bridge knots / links . we recall that there are a lot of ways of describing a 3-manifold : by fundamental polyhedra @xcite , by heegaard diagrams @xcite , by surgery on links in @xmath0 @xcite , by branched coverings of @xmath0 @xcite and by gems / crystallizations @xcite . so , each manifold has many `` faces '' , which depend on the type of the description we choose . this choice depends on the nature of the problem one studies . in this paper we are going to present these aspects of cyclic branched coverings of 2-bridge knots / links . as an example , we will describe the different faces of the classical hantzsche wendt manifold @xcite . by the term _ manifold _ we always mean a compact , connected , orientable without boundary . let @xmath1 be triangulated @xmath2-dimensional manifolds and let @xmath3 be an @xmath4-subcomplex of @xmath5 . a non - degenerate map @xmath6 is an _ @xmath7-fold covering map branched over @xmath3 _ , with @xmath8 , if : ( i ) @xmath9 is an ordinary @xmath7-fold covering , and ( ii ) @xmath10 . the manifold @xmath11 is said to be a _ branched covering _ of @xmath5 , and @xmath3 is called the _ branching set _ of the covering . two branched coverings @xmath12 and @xmath13 are equivalent if there exist two homeomorphisms @xmath14 and @xmath15 such that @xmath16 . a remarkable result by r. h. fox @xcite states that a branched covering is uniquely determined by the ordinary covering induced by restriction . this proves the existence of a one - to - one correspondence between the @xmath7-fold coverings of @xmath5 branched over @xmath3 and the equivalence classes of monodromies ( i.e. transitive representations ) @xmath17 , where @xmath18 is the symmetric group on @xmath7 elements and @xmath19 is an arbitrary base point . if @xmath5 is an @xmath2-sphere , then a covering of @xmath5 branched over @xmath3 is simply called a _ branched covering of @xmath3_. two @xmath7-fold branched coverings @xmath12 , @xmath13 , with branching sets @xmath20 , @xmath21 and monodromy maps @xmath22 , @xmath23 respectively , are equivalent if and only if there exists an inner automorphism @xmath24 of @xmath18 and a homeomorphism @xmath15 , such that @xmath25 and @xmath26 , where @xmath27 is the homomorphism induced by @xmath28 on the fundamental groups . branched coverings of spheres are of great interest , in particular as a method for representing manifolds . a classical result concerning this point was obtained by j. alexander @xcite and states that every @xmath2-manifold is a covering of @xmath29 , branched over the @xmath4-skeleton of a standard @xmath2-simplex . fox s result gives the possibility of extending the concept of cyclic coverings from ordinary coverings to branched coverings . thus , a branched covering is called _ cyclic _ if its associated ordinary covering is cyclic . similar extension also applies to regular and abelian coverings . since a cyclic covering is abelian , it is determined up to equivalence by an epimorphism @xmath30 where @xmath31 is the cyclic group of order @xmath7 , embedded into @xmath18 through the monomorphism sending @xmath32 to the standard cyclic permutation @xmath33 . if @xmath34 and @xmath35 is a @xmath36-component link in @xmath0 , then @xmath37 and a basis is given by any set of homology classes of meridian loops around the components of @xmath3 . therefore , an @xmath7-fold cyclic branched covering @xmath38 of @xmath3 is defined by orienting @xmath3 and assigning an integer @xmath39 to each component @xmath40 , such that the set @xmath41 generates @xmath31 . if @xmath42 is a meridian around @xmath40 , coherently oriented with the chosen orientations of @xmath3 and @xmath0 , we define @xmath43=k_j\in\z_n$ ] and therefore @xmath44=(1\,2\,\cdots\,n)^{k_j}$ ] . we will denote this manifold by @xmath45 . by multiplying each @xmath46 by the same invertible element @xmath47 of @xmath31 , we obtain an equivalent covering . more precisely , two @xmath7-fold cyclic branched coverings @xmath12 and @xmath13 , with associated maps @xmath48 and @xmath49 respectively , are equivalent if and only if there exist @xmath50 and a homeomorphism @xmath51 , such that @xmath52 , @xmath53 and @xmath54 , where @xmath55 is the homomorphism induced by @xmath56 on the first homology groups and @xmath57 is the multiplication of @xmath58 by @xmath47 . general references on cyclic branched coverings of knots / links are the interesting books @xcite , @xcite , and @xcite . following @xcite we shall say that a cyclic branched covering @xmath45 is : ( i ) : : _ strictly - cyclic _ if @xmath59 , for every @xmath60 ; ( ii ) : : _ almost - strictly - cyclic _ if @xmath61 , for every @xmath60 ; ( iii ) : : _ meridian - cyclic _ if @xmath62 , for every @xmath63 ; ( iv ) : : _ singly - cyclic _ if there exists @xmath63 such that @xmath62 ; ( v ) : : _ monodromy - cyclic _ if it is cyclic . the following implications are straightforward : @xmath64 moreover , all five definitions are equivalent when either @xmath3 is a knot or @xmath65 . note that strictly - cyclic coverings are also called _ uniform _ in @xcite and meridian - cyclic coverings are also called _ strongly cyclic _ in @xcite . by a suitable reorientation of the link , an almost - strictly - cyclic covering becomes a strictly - cyclic one . for a singly - cyclic covering we can always assume @xmath66 , up to equivalence and possible reordering of the components of @xmath3 . therefore , when @xmath67 the covering is completely determined by an integer @xmath68 . in order to simplify the notations , the @xmath7-fold strictly - cyclic branched covering @xmath69 will be denoted by @xmath70 and the @xmath7-fold singly - cyclic branched covering of a 2-component link @xmath71 will be denoted by @xmath72 . in particular , @xmath72 is meridian - cyclic when @xmath73 and strictly - cyclic when @xmath74 . the class of 2-bridge knots / links ( also called `` rational knots / links '' ) is of great interest and the main properties of its elements can be found in @xcite . we denote by @xmath75 the 2-bridge knot or link of type @xmath76 , with integers @xmath77 and @xmath78 such that @xmath79 . it is well known that @xmath80 is a knot when @xmath81 is odd and a two - component link when @xmath81 is even . in our discussion we will skip the `` singular '' case of the 2-component trivial link , which is a 2-bridge link and is often indicated by @xmath82 . a standard diagram presentation for 2-bridge knots / links is schubert s normal form ( * ? ? ? actually , schubert s normal form is only defined when @xmath83 is odd , but this is not a restriction , up to equivalence ( see classification of 2-bridge knots / links below ) . figure 1 gives two examples of such diagrams for the figure - eight knot @xmath84 and for the whitehead link @xmath85 . and @xmath85 . ] another type of diagram for 2-bridge knots / links is introduced by j. conway @xcite ( see figure 2 ) . it comes from the representation of a rational number by a continued fraction . if @xmath86 we will say that @xmath87 $ ] are conway parameters for @xmath88 and the notation @xmath89 $ ] will be used . 2-bridge knots / links are assumed to be oriented in a standard way , as indicated in figures 1 and 2 . by the schubert construction of @xmath90 , there is an obvious orientation preserving involution @xmath91 exchanging the two bridges of @xmath90 and preserving the orientation of the component(s ) . recall that two oriented links @xmath3 and @xmath92 with @xmath36 components with a given order @xmath93 and @xmath94 are said to be _ equivalent _ if there exists an orientation preserving homeomorphism @xmath95 sending @xmath96 to @xmath97 , for @xmath98 , and preserving their orientations . for 2-bridge links the equivalence does not depend on the ordering of the components because of the homeomorphism @xmath99 previously described . h. schubert in @xcite classified 2-bridge knots / links ( see also ( * ? ? ? * theorem 2.1.3 ) ) : ( i ) : : two oriented two - bridge knots @xmath90 , @xmath100 are equivalent if and only if @xmath101 and @xmath102 mod @xmath81 ; ( ii ) : : two oriented two - bridge links @xmath90 , @xmath100 are equivalent if and only if @xmath101 and @xmath103 . will be considered mod @xmath104 , except when otherwise specified . ] if we consider unoriented links , then the condition of ( ii ) is reduced to that of ( i ) . below we will use the following properties : @xmath105 is equivalent to the mirror image of @xmath90 and the link @xmath106 is equivalent to the link @xmath90 with the opposite orientation on one of the two components @xcite . from now on , we use the notation @xmath107 for @xmath7-fold cyclic branched coverings of 2-bridge links ( i.e. , @xmath81 is even ) . in particular , @xmath108 will denote a singly - cyclic branched covering and @xmath109 will denote a strictly - cyclic one . the last notation will always be used even in the case when @xmath75 is a knot ( i.e. , @xmath81 is odd ) . the class of cyclic branched coverings of 2-bridge knots / links has been intensively studied by many authors , and lot of their properties were discovered . in particular , the two - fold coverings are homeomorphic to lens spaces . in section 2 we introduce our class of manifolds , discuss homeomorphisms , geometric structures , and present them as 2-fold branched coverings of @xmath0 . in section 3 we will describe the polyhedral construction of these manifolds , according to j. minkus @xcite . in section 4 we will give symmetric heegaard diagrams for cyclic branched coverings of 2-bridge knots . following from a result obtained in @xcite , this description arises from a construction by m. dunwoody @xcite . in the same section , some estimates for the genus are given . in section 5 we will give a surgery description of these manifolds , for the case of knots . it was pointed out in @xcite and @xcite that some cyclic branched coverings of 2-bridge knots are takahashi manifolds @xcite and therefore they can be obtained by dehn surgery on a certain chain of circles in @xmath0 . the takahashi construction has been generalized in @xcite , in order to obtain all cyclic branched coverings of 2-bridge knots . in section 6 we will deal with the coloured graph representation of manifolds introduced by m. pezzana @xcite and his school . a wide class of coloured graph encoding 3-dimensional ( possibly singular ) manifolds was defined by s. lins and a. mandel in @xcite , and intensively studied by many authors . it was shown in @xcite that all lins - mandel manifolds are cyclic branched coverings of 2-bridge knots / links . in section 7 we will describe the presentation of the fundamental group of these manifolds obtained in @xcite and @xcite . moreover , in the case of coverings of 2-bridge knots , we also describe the two different cyclic presentations obtained in @xcite and @xcite . in section 8 we give the first integer homology groups for a large class of cases . in section 9 we will prove that each singly - cyclic branched covering of a 2-bridge link is the composition of a meridian - cyclic branched covering of a determined link and a standard cyclic covering of @xmath0 branched over a trivial knot . the particular case of singly - cyclic coverings of the whitehead link was studied in @xcite . cyclic branched coverings of 2-bridge knots / links form a very important class of 3-manifolds , that is a natural generalization of the class of lens spaces . indeed @xmath110 is the lens space @xmath111 and , when @xmath73 , @xmath112 is the lens space @xmath113 @xcite . this class appears to be fairly rich , since it contains several interesting 3-manifolds , such as the poincar homology sphere , the seifert weber hyperbolic dodecahedron space , the euclidean hantzsche wendt manifold , the hyperbolic fomenko matveev weeks manifold and also an infinite family of brieskorn manifolds . more precisely , the poincar homology sphere @xcite is the @xmath114-fold cyclic branched covering of the trefoil knot @xmath115 and the @xmath116-fold cyclic branched covering of @xmath117 , i.e. , @xmath118 ; the seifert weber hyperbolic dodecahedron space @xcite is the @xmath114-fold singly - cyclic branched covering of the whitehead link @xmath119 defined by @xmath66 and @xmath120 , i.e. , @xmath121 ( its generalizations @xmath122 were discussed in @xcite ) ; the euclidean hantzsche wendt manifold @xcite is the @xmath116-fold cyclic branched covering of the figure - eight knot @xmath123 , i.e. , @xmath124 ( its generalizations @xmath125 , known as the fibonacci manifolds , were studied in @xcite ) ; the hyperbolic fomenko matveev weeks manifold @xcite , which is the hyperbolic 3-manifold with the smallest known volume , is the @xmath116-fold cyclic branched covering of the knot @xmath126 , i.e. , @xmath127 @xcite ( its generalizations @xmath128 were studied in @xcite ) ; the brieskorn manifold @xmath129 @xcite is the @xmath7-fold strictly - cyclic branched coverings of the torus knots or links @xmath130 , i.e. , @xmath131 ( see also @xcite ) . note that , as a consequence of the positive solution of the smith conjecture @xcite , this family of manifolds contains no sphere , since @xmath90 is never a trivial knot . now we list some sufficient homeomorphism conditions . let @xmath90 be a 2-bridge link and @xmath132 . then : ( i ) : : if @xmath133 , then @xmath108 is homeomorphic to @xmath134 ; ( ii ) : : @xmath108 is homeomorphic to @xmath135 ; ( iii ) : : @xmath108 is homeomorphic to @xmath136 ; ( iv ) : : if the links @xmath90 and @xmath100 are equivalent , then @xmath108 is homeomorphic to @xmath137 . let @xmath138 and @xmath139 be meridians corresponding to the components of @xmath140 . ( i ) let @xmath141 and @xmath142 the corresponding cyclic branched coverings of @xmath90 . if @xmath143 and @xmath144 are the associated monodromy maps , we have @xmath145=1 $ ] , @xmath146=k$ ] , @xmath147=1 $ ] and @xmath148=k'$ ] . therefore , @xmath149 and the two coverings are equivalent . ( ii ) the link @xmath105 is equivalent to the mirror image of @xmath90 . let @xmath150 be an orientation reversing homeomorphism of @xmath0 sending @xmath90 to @xmath105 . since @xmath150 preserve both the orientations of the two components of @xmath90 , we have @xmath151=-[m'_1]$ ] and @xmath152=-[m'_2]$ ] , where @xmath153 and @xmath154 are the generator meridians associated to @xmath105 . therefore , @xmath155 and the two coverings are equivalent . ( iii ) the link @xmath106 is equivalent to the link @xmath90 with the opposite orientation in the second component . let @xmath150 be the identity map on @xmath0 , then @xmath151=[m'_1]$ ] and @xmath152=-[m'_2]$ ] . therefore , @xmath156 and the two coverings are equivalent . ( iv ) let @xmath150 be the homeomorphism of @xmath0 realizing the equivalence . without loss of generality we can assume that @xmath157 and @xmath158 . then @xmath151=[m'_1]$ ] and @xmath152=[m'_2]$ ] . therefore , @xmath156 and the two coverings are equivalent . let @xmath90 be an arbitrary 2-bridge link and @xmath159 such that @xmath73 , then : ( i ) : : @xmath108 is homeomorphic to @xmath160 ; ( ii ) : : if @xmath161 , then @xmath108 is homeomorphic to @xmath162 . \(i ) see item ( i ) of the previous proposition . ( ii ) assume @xmath163 . since @xmath83 is odd , we have @xmath164 . therefore , @xmath90 is equivalent to @xmath106 and from the previous proposition we obtain @xmath165 . now , let @xmath166 . since @xmath83 is odd , we have @xmath167 . therefore , @xmath90 is equivalent to @xmath168 and from the previous proposition we obtain @xmath169 . when our manifolds have hyperbolic geometric structure , a partial converse to the previous results can be given ( see theorem [ theorem iff ] below ) . recall that a 2-bridge knot or link @xmath90 is hyperbolic ( i.e. , the complement @xmath170 has hyperbolic structure ) if and only if @xmath90 is non - toroidal , that is @xmath171 mod @xmath81 . thus , the geometric structure of @xmath108 can be obtained in these cases from w. thurston @xcite and w. dunbar @xcite results . moreover , when the branching set is toroidal , then @xmath109 turns out to be the brieskorn manifold @xmath129 . thus , we have the following result : [ proposition geom ] ( i ) : : @xcite if @xmath73 and @xmath172 mod @xmath81 , then @xmath108 ( or @xmath173 in the case of a knot ) is hyperbolic for ( i ) @xmath174 , @xmath175 and ( ii ) @xmath176 , @xmath177 . moreover , @xmath178 is euclidean and @xmath179 is spherical for all @xmath180 . ( ii ) : : @xcite if @xmath181 mod @xmath81 , then @xmath109 is a spherical manifold for @xmath182 , a nil - manifold for @xmath183 , and a @xmath184-manifold for @xmath185 . from theorem 1 of @xcite ( including the note ( a ) of page 293 ) and theorem 4.1 of @xcite ( see tables of page 184 ) , the next result holds . [ theorem 2 ] [ theorem iff ] let @xmath90 be a 2-bridge hyperbolic link . if @xmath73 , then @xmath134 and @xmath108 are homeomorphic if and only if ( i ) : : @xmath186 , when @xmath187 ; ( ii ) : : @xmath188 , when @xmath189 . volumes and chern simons invariants of hyperbolic cyclic branched coverings of 2-bridge knots were obtained in @xcite , where the table for small values of parameters is given . now we will show that every @xmath7-fold strictly - cyclic covering of a @xmath190-bridge knot or link can be obtained as the 2-fold branched covering of a certain @xmath7-periodic knot or link @xcite . if @xmath191 is the involution previously described , which leaves @xmath90 invariant and exchanges its bridges , we denote the quotient @xmath192 by @xmath193 , that is the trivial knot pictured in figure 3 . so @xmath194 is a 2-fold covering of @xmath195 branched over a trivial knot @xmath196 corresponding to the axis of the involution @xmath197 . this covering gives us a natural way to construct a periodic generalization of 2-bridge knots / links . denote by @xmath198 the preimage of @xmath199 under the n - fold covering of @xmath0 branched over @xmath196 . this knot or link admits a natural @xmath7-bridge presentation ( see an example in figure 3 ) and will be called the _ @xmath7-cyclic extension _ of @xmath88 . in particular , @xmath200 and @xmath201 is the torus knot / link of type @xmath202 . and @xmath203 . ] @xcite the manifold @xmath204 is the @xmath190-fold branched covering of @xmath205 . thus , as an example , the hantzsche wendt manifold is the @xmath116-fold branched covering of @xmath206 , pictured in figure 1 , and the @xmath190-fold branched covering of @xmath207 , pictured in figure 3 , that is a @xmath116-component link known as the _ borromean rings _ ( see also @xcite ) . independently , a @xmath190-fold covering description of @xmath208 , when @xmath81 is odd , arising from the generalized takahashi manifolds ( see section 5 ) , was obtained in @xcite . a classical method for constructing ( orientable ) closed @xmath116-manifolds consists in the pairwise identification of ( oppositely oriented ) boundary faces of a triangulated @xmath116-ball ( see @xcite ) . the resulting quotient complex triangulates a closed pseudomanifold , which is a manifold if and only if its euler characteristic vanishes . for strictly - cyclic branched coverings of 2-bridge knots / links @xmath209 this construction was realized by j. minkus @xcite . the idea is based on the method of realizing lens spaces . here we will briefly describe the construction . consider the boundary 2-sphere @xmath210 of the 3-ball @xmath211 . draw @xmath7 equally spaced great semicircles joining the north pole @xmath212 to the south pole @xmath213 . this decomposes @xmath210 into @xmath7 congruent lunes . subdivide each semicircle into @xmath214 equal segments by drawing @xmath215 equally spaced vertices on each semicircle . in this way , each lune can then be viewed as a curvilinear @xmath216 sided polygon on @xmath210 . now bisect each lune by drawing a great circle arc inside the lune , joining the vertex which is @xmath217 segments down from @xmath5 ( the point @xmath218 of figure 4 ) on each semicircle with the vertex @xmath217 segments up from @xmath219 to the next clockwise semicircle . figure 4 shows this decomposition of @xmath220 , where @xmath221 . . ] the result is the decomposition of @xmath222 into @xmath223 regions @xmath224 , @xmath225 , @xmath226 . the regions @xmath224 are around @xmath5 and the regions @xmath225 are around @xmath219 , and each @xmath225 can be reached from the corresponding @xmath224 by moving @xmath224 counterclockwise to the adjacent lune and then shifting from the northern to the southern hemisphere of @xmath222 . the cell 3-complex @xmath227 is obtained from @xmath228 by identifying @xmath224 with @xmath225 on @xmath222 for each @xmath229 by an orientation reversing homeomorphism which matches the vertex @xmath218 of @xmath224 with the vertex @xmath230 of @xmath225 . [ theorem minkus ] ( * ? ? ? * theorem 7 ) the manifold @xmath231 is the @xmath7-fold strictly - cyclic branched covering of the 2-bridge knot or link @xmath232 . therefore the manifolds @xmath231 are homeomorphic to the above defined manifolds @xmath233 and provide a polyhedral construction for them . as an example , figure 5 gives the polyhedral construction of the hantzsche wendt manifold , that is @xmath124 in our notations . the minkus polyhedral construction only produces strictly - cyclic branched coverings of 2-bridge knots / links . the generalization to singly - cyclic coverings @xmath234 is straightforward and is illustrated in figure 6 , which is the same as figure 5 of @xcite . in this case , the cell 3-complex @xmath234 is obtained from @xmath228 by identifying @xmath224 with @xmath225 on @xmath222 for each @xmath235 by an orientation reversing homeomorphism which matches the vertex @xmath218 of @xmath224 with the vertex @xmath236 of @xmath225 . . ] we remark that another polyhedral construction for @xmath237 is presented in @xcite . in this section we will discuss heegaard diagrams as well as cyclically symmetric heegaard diagrams for our manifolds . heegaard diagrams for meridian - cyclic branched coverings of 2-bridge knots / links can be directly obtained from the coloured graph construction that will be discussed in section 6 ( see @xcite ) . more precisely , a heegaard diagram of genus @xmath238 of @xmath239 , with @xmath73 , can be obtained from figure 7 by identifying the disk @xmath240 with the disk @xmath241 , for @xmath242 , according to the numeration of the vertices , and removing one of the @xmath7 closed curves arising from these identifications , illustrated by the dashed lines in figure 7 . figure 8 presents a heegaard diagram of @xmath124 obtained in this way . at the same time , it is natural to ask about heegaard diagrams for our manifolds corresponding to cyclic coverings . the connections between cyclic branched coverings of knots and cyclic presentations of groups induced by suitable heegaard diagrams have recently been discussed in several papers @xcite . in order to investigate these relations , m.j . dunwoody introduced in @xcite a class of heegaard diagrams , depending on six integers , having a cyclic symmetry and encoding a cyclic presentation for the fundamental group of the represented manifold @xmath243 . in @xcite it has been shown that the 3-manifolds represented by these diagrams ( called _ dunwoody manifolds _ ) are cyclic coverings of lens spaces branched over genus one 1-knots ( also called @xmath244-knots ) . as a corollary , it has been demonstrated that , for particular values of the parameters , the dunwoody manifolds turn out to be cyclic coverings of @xmath245 branched over some knots . this gives a positive answer to a conjecture made by dunwoody , which has also been independently proved in @xcite . it is interesting to note that the class of dunwoody manifolds properly contains the class of cyclic branched coverings of 2-bridge knots , as stated in the following theorem . note that in the statement @xmath246 is an integer only depending on @xmath247 and @xmath248 ( see details in section 3 of @xcite ) . [ theorem dunwoody ] @xcite for all @xmath249 and @xmath8 , the @xmath7-fold cyclic covering of @xmath0 branched over @xmath250 is the dunwoody manifold @xmath251 . thus , all branched cyclic coverings of 2-bridge knots are dunwoody manifolds . , the hantzsche wendt manifold . ] as a consequence of the previous theorem , each @xmath7-fold cyclic branched covering of a 2-bridge knot admits a heegaard diagram of genus @xmath7 with a cyclic symmetry of order @xmath7 . these diagrams induce `` geometric '' cyclic presentations for the corresponding fundamental groups ( see section 7 ) . now we discuss some results on genus ( classical , @xmath214-symmetric and equivariant ) of our manifolds . from the standard representation of a 2-bridge knot / link as a 4-plat @xmath252 , where @xmath253 , @xmath254 are 3-balls and @xmath255 , @xmath256 are pairs of trivially properly embedded arcs , we get a heegaard splitting of @xmath257 , where the splitting surface is an @xmath7-fold cyclic covering of @xmath210 branched over 4 points . from the riemann - hurwitz formula , the surface has genus @xmath258 ( see details in @xcite ) . therefore we have the following : [ proposition genus ] let @xmath259 be the genus of a 3-manifold @xmath11 . then @xmath260 in particular , for singly - cyclic coverings @xmath261 and for meridian - cyclic coverings @xmath262 . thus , for strictly - cyclic coverings @xmath263 . from theorem 4 of @xcite , for the case of strictly - cyclic branched coverings the same estimation also holds for the @xmath7-symmetric genus introduced by j. birman and h. hilden in @xcite . [ proposition symm - genus ] let @xmath264 be the @xmath7-symmetric genus of a 3-manifold @xmath11 . then @xmath265 in many interesting cases the result of proposition [ proposition genus ] can be improved . [ proposition braid ] ( i ) : : for all @xmath266 we have @xmath267 ( ii ) : : for all @xmath268 we have @xmath269 \(i ) the @xmath7-fold strictly - cyclic covering of @xmath130 is the 2-fold covering of the torus knot / link of type @xmath270 @xcite , which can be presented as the closure of a @xmath81-string braid . so it has bridge number @xmath271 and then @xmath272 @xcite . ( ii ) the @xmath7-fold strictly - cyclic covering of @xmath273 is the 2-fold covering of @xmath274 , which can be presented as the closure of a @xmath275-string braid . so it has bridge number @xmath276 and then @xmath277 @xcite . in particular , from ( ii ) we see that the @xmath7-fold cyclic branched covering of the figure - eight knot ( @xmath278 ) has genus @xmath279 if @xmath280 and @xmath281 if @xmath65 ( see also @xcite ) . similarly , the @xmath7-fold strictly - cyclic branched coverings of the whitehead link ( @xmath282 ) have genus @xmath283 . recall that the maximally possible order of a finite group of orientation - preserving homeomorphisms of a handlebody @xmath284 of genus @xmath285 equals @xmath286 ( see @xcite ) , analogously to the classical hurwitz @xmath287-bound for a closed riemann surfaces of genus @xmath288 . according to @xcite , a closed 3-manifold @xmath11 is said to be a _ @xmath289-manifold of genus @xmath290 _ if it admits an action of the finite group @xmath289 and @xmath290 is the minimal genus of a heegaard splitting of @xmath11 for which both handlebodies are invariant under the @xmath289-action . a @xmath289-manifold of genus @xmath291 is called _ minimal _ if the induced @xmath289-action on each of the two handlebodies of an invariant heegaard splitting of genus @xmath290 is a strong genus action ( i.e. , there is no action of @xmath289 on a handlebody of genus @xmath292 , with @xmath293 ) . moreover , if @xmath289 has maximal positive order @xmath286 then the @xmath289-manifold @xmath11 and the @xmath289-action are called _ maximally symmetric_. for @xmath280 , every minimal @xmath31-manifold is a minimal @xmath294-manifold , where @xmath294 is the dihedral group of order @xmath223 . if @xmath280 is prime , every minimal @xmath31-manifold is a minimal @xmath295-manifold @xcite . each manifold of genus 2 is a @xmath289-manifold of genus 2 , where @xmath289 is one of the four groups @xmath296 , @xmath297 , @xmath298 or @xmath299 , and the @xmath299-manifolds are maximally symmetric @xcite . the cyclic branched coverings of 2-bridge knots / links play an important role in this theory . [ theorem minimal ] @xcite ( i ) : : the minimal @xmath295-manifolds , of genus @xmath300 , are exactly the @xmath7-fold strictly - cyclic branched coverings of 2-bridge knots / links . the minimal @xmath31 or @xmath294-manifolds that are not minimal @xmath295-manifolds are cyclic branched coverings of 2-bridge links with two components of different branching index . ( ii ) : : the maximally symmetric @xmath299-manifolds are exactly the 3-fold cyclic branched coverings of 2-bridge knots / links . in this section we describe a symmetric surgery presentation of cyclic branched coverings of 2-bridge knots . this presentation comes from a more general construction of _ generalized takahashi manifolds _ introduced in @xcite . for any pair of positive integers @xmath2 and @xmath7 , we consider the link @xmath301 with @xmath302 components drawn in figure 10 . all its components @xmath303 , @xmath304 , @xmath305 , are unknotted circles and they form @xmath223 subfamilies of @xmath2 unlinked circles @xmath303 , @xmath305 , with a common center . the link @xmath306 has a cyclic symmetry of order @xmath7 which permutes these @xmath223 subfamilies of circles . consider the manifold obtained by dehn surgery on @xmath0 , along the link @xmath306 , such that the surgery coefficients @xmath307 correspond to the components @xmath308 , and @xmath309 correspond to the components @xmath310 , where @xmath311 and @xmath305 . without loss of generality , we can always suppose that @xmath312 , @xmath313 and @xmath314 . we will denote the resulting 3-manifold by @xmath315 . this manifold will be called a _ generalized takahashi manifold _ since for @xmath316 we get the manifolds introduced by m. takahashi in @xcite . when the surgery coefficients are @xmath7-periodic , i.e. @xmath317 , @xmath318 , @xmath319 , and @xmath320 , the resulting manifold @xmath321 is said to be a _ generalized periodic ( @xmath7-periodic ) takahashi manifold_. the following theorem generalizes the result obtained in @xcite for periodic takahashi manifolds , and gives the relation between the generalized periodic takahashi manifolds and the cyclic branched coverings of 2-bridge knots . [ twobridge ] @xcite the manifold @xmath322 is homeomorphic to @xmath323 , where @xmath324 is the rational number defined by the continued fraction @xmath325 $ ] . because every 2-bridge knot admits a conway representation with an even number of even parameters ( see exercise 2.1.14 of @xcite ) , we have the following property : [ all twobridge ] @xcite the family of generalized periodic takahashi manifolds contains all the cyclic branched coverings of 2-bridge knots . as a result , a surgery presentation with cyclic symmetry is obtained for all cyclic branched coverings of 2-bridge knots ( see figure 10 ) . , where @xmath326)$ ] . ] as an example , the surgery description of the hantzsche wendt manifold ( that is @xmath178 in the above notation ) is illustrated in figure 11 . we recall some facts about the coloured - graph presentation of 3-manifolds . a @xmath327-_coloured graph _ is a pair @xmath328 , where @xmath329 is a finite connected @xmath327-regular graph without loops , and @xmath330 is a proper edge - coloration ( i.e. , adjacent edges have different colours ) . every @xmath327-coloured graph represents a pseudosimplicial complex @xcite @xmath331 defined in the following way : ( i ) take a @xmath116-simplex @xmath332 for each vertex @xmath333 and label its vertices by the elements of @xmath334 ; ( ii ) identify , for every pair @xmath335 of @xmath336-adjacent vertices , the @xmath190-faces of @xmath332 and @xmath337 opposite the vertices labelled by @xmath336 . the underlying space @xmath338 is a connected @xmath116-dimensional pseudomanifold , which is orientable if and only if @xmath329 is bipartite @xcite . note that only isolated singular points may appear , and such spaces are said to be _ singular manifolds _ in @xcite . a @xmath116-_gem _ is a @xmath327-coloured graph representing a @xmath116-manifold ; every manifold @xmath11 is representable by gems @xcite . a gem is called a _ crystallization _ if , for each colour @xmath339 , the subgraph of @xmath329 obtained by removing all @xmath336-coloured edges is connected . there is a strict connection between crystallizations and heegaard diagrams of 3-manifolds @xcite . let @xmath340 and @xmath341 be two colours from @xmath334 , and @xmath342 , @xmath343 . a heegaard diagram of the represented manifold can be obtained just by removing one @xmath344-cycle ( i.e. , formed by edges coloured by the colours of @xmath344 ) and one @xmath196-cycle . the remaining @xmath344-cycles and @xmath196-cycles give , respectively , the two systems of curves of the heegaard diagram . in other words , the @xmath344-cycles and the @xmath196-cycles of a crystallization represent an extended heegaard diagram of the manifold ( see @xcite ) . also manifolds of arbitrary dimension can be represented by edge - coloured graphs , which give a combinatorial way of representing manifolds ; for general references see @xcite and @xcite . the family of lins - mandel 4-coloured graphs @xmath345 , with @xmath346 , @xmath347 , @xmath348 and @xmath349 , has been defined in @xcite . the set of vertices of @xmath345 is @xmath350 and the coloured edges are obtained by the following four fixed - point - free involutions on @xmath351 : @xmath352 where @xmath353 is the map defined by @xmath354 for each @xmath355 , we join the vertices @xmath356 by a @xmath357-coloured edge if and only if @xmath358 roughly speaking , the graph @xmath345 is obtained by taking @xmath7 copies @xmath240 , @xmath359 , of the @xmath360-cycle of length @xmath216 ( so that @xmath361 ) joined with @xmath362 and @xmath363 by @xmath214 edges of colour 1 , and with @xmath364 and @xmath365 by @xmath214 edges of colour 0 . each graph @xmath345 is connected and bipartite ; hence , it represents a connected , orientable 3-dimensional pseudomanifold @xmath366 . this class of graphs and spaces have been intensively studied @xcite . remark that lins - mandel spaces have been introduced as a combinatorial generalization of the lens spaces , since @xmath367 is the standard graph representing the lens space @xmath368 ( see @xcite ) . . ] the following characterization of the lins - mandel gems exists : @xcite a lins - mandel graph @xmath345 represents a 3-manifold if and only if either @xmath214 is even or @xmath369 . moreover , a lins - mandel gem @xmath345 is a crystallization if and only if @xmath370 @xcite . [ lins - mandel as cyclic ] @xcite lins - mandel spaces have the following topological structure : ( i ) : : if @xmath366 is a manifold with @xmath371 and @xmath372 , then it is homeomorphic to @xmath373 , otherwise it is homeomorphic to @xmath0 . ( ii ) : : if @xmath366 is not a manifold , then it is an @xmath7-fold cyclic covering of @xmath0 branched over a @xmath374-graph , which is embedded as the 2-bridge knot @xmath232 with an unknotting tunnel . thus , the manifold @xmath366 is a singly - cyclic branched covering of @xmath209 . in particular the covering is strictly - cyclic when @xmath375 , is almost - strictly - cyclic when @xmath376 , and is meridian - cyclic when @xmath377 . remark that by removing the cycle @xmath378 from the graph @xmath345 , with @xmath377 , we will get exactly the graph pictured in figure 7 , once the relative identifications of @xmath240 with @xmath241 , for @xmath242 , have been performed , according to the numeration of the vertices . necessary and sufficient conditions for the isomorphism between lins - mandel gems are obtained in @xcite , when @xmath379 . as a consequence , sufficient conditions for the homeomorphism between lins - mandel manifolds not homeomorphic to the sphere or the lens spaces directly follow . @xcite [ theorem 1 ] for @xmath380 the following isomorphisms of graphs hold : ( i ) : : if @xmath214 is even and @xmath381 , then @xmath382 if and only if @xmath383 ( i ) : : if @xmath214 is even and @xmath377 , then @xmath382 if and only if @xmath384 ( ii ) : : if @xmath214 is odd , then @xmath385 if and only if @xmath386 hence , if one of the above conditions holds , then the manifolds @xmath366 and @xmath387 are homeomorphic . observe that the isomorphism conditions of part ( ii ) of the previous theorem are the same as the homeomorphism conditions for lens spaces . this is not true for part ( i ) , since , in this case , the situation is complicated by the presence of the additional parameter @xmath336 . cases where @xmath214 is even are particularly interesting because the graph always represents a manifold without any restriction on @xmath336 . from theorem [ theorem 1 ] we get : @xcite [ corollary 1 ] let @xmath388 be fixed , with @xmath389 and @xmath214 even . then @xmath390 if and only if ( i ) : : @xmath391 , @xmath392 when @xmath381 and @xmath393 ; ( ii ) : : @xmath394 , @xmath392 when @xmath381 and @xmath395 ; ( iii ) : : @xmath396 , @xmath392 when @xmath377 and @xmath397 ; ( iv ) : : @xmath398 , @xmath392 when @xmath377 and @xmath395 . due to theorem [ theorem 2 ] and corollary [ corollary 1 ] , in many cases graphs distinguish manifolds : [ theorem 3 ] assume @xmath399 and @xmath400 . for @xmath377 , the manifolds @xmath401 and @xmath366 are homeomorphic if and only if the graphs @xmath402 and @xmath345 are isomorphic . the lins - mandel family contains only singly - cyclic coverings . for this reason it has been extended in @xcite , in order to obtain the whole class of cyclic branched coverings of 2-bridge knots / links . this new class of 4-coloured graphs @xmath403 depends on five integer parameters , with @xmath346 , @xmath404 , @xmath405 , @xmath349 and @xmath406 . each @xmath403 is defined by the following four fixed - point - free involutions on the set @xmath407 : @xmath408 the graph @xmath403 represents a connected , orientable pseudomanifold @xmath409 . @xcite the following properties hold : ( i ) : : @xmath403 represents a 3-manifold if and only if either @xmath214 is even or at least one of the conditions ( 1 ) @xmath410 , ( 2 ) @xmath411 , ( 3 ) @xmath412 is satisfied . ( ii ) : : @xmath413 . ( iii ) : : @xmath414 . ( iv ) : : @xmath415 . we give the connection between generalized lins - mandel manifolds and cyclic branched coverings of 2-bridge knots / links . @xcite the @xmath116-manifold @xmath409 , with @xmath416 , is homeomorphic to @xmath417 . therefore , the class of generalized lins - mandel manifolds @xmath418 , with @xmath419 and @xmath372 , is precisely the class of all cyclic coverings of @xmath0 branched over the 2-bridge knots / links ( with the exception of the trivial link with two components ) . another family of 4-coloured graphs representing cyclic branched coverings of 2-bridge knots is described in @xcite . a presentation of the fundamental group of @xmath420 can be obtained from the coloured graph construction . define the following words in @xmath421 with subscripts mod @xmath7 : @xmath422 with @xmath423 , @xmath424 and @xmath425 where @xmath426 is the map defined in section 6 . @xcite the fundamental group of @xmath420 has the following presentation @xmath427 where @xmath428 and @xmath429 are as above . since the cyclic branched coverings of 2-bridge knots have a cyclic homeomorphism , it is natural to wonder about some cyclic presentations of their fundamental groups . we recall that a finite balanced presentation of a group @xmath430 is said to be a _ cyclic presentation _ if there exists a word @xmath431 in the free group @xmath432 generated by @xmath433 such that the relators of the presentation are @xmath434 , @xmath435 , where @xmath436 denotes the automorphism defined by @xmath437 ( subscripts mod @xmath7 ) , @xmath438 . let us denote this cyclic presentation ( and the related group ) by the symbol @xmath439 , so that : @xmath440 a group is said to be cyclically presented if it admits a cyclic presentation . the polynomial associated with the cyclic presentation @xmath441 is given by @xmath442 where @xmath443 is the exponent sum of @xmath444 in @xmath431 . for the theory of cyclically presented groups we refer to @xcite . two different cyclic presentations for the fundamental groups of the manifolds @xmath323 are obtained in @xcite and @xcite . remark that explicit cyclic presentations different from the above are listed in the appendix of @xcite , for 2-bridge knots up to nine crossings . from the polyhedral construction of @xmath204 the following presentation of the fundamental group holds . denote by @xmath445 with @xmath446 } , \ ] ] where @xmath447 is the inverse of the element @xmath83 in @xmath448 and @xmath449 $ ] denotes the integral part of @xmath450 . [ minkuspresentation ] ( * ? ? ? * theorem 10 ) the fundamental group of @xmath451 admits the presentation : @xmath452 if @xmath81 is odd , @xmath453 if @xmath81 is even . remark that ( 1 ) is a cyclic presentation and ( 2 ) is a non - cyclic one . there is a nice relation between the alexander polynomial of a 2-bridge knot and the polynomial associated with the above cyclic presentation . [ proposition d2 ] ( * ? ? ? * theorem 11 ) the alexander polynomial of the knot @xmath140 is equal to the polynomial associated with the cyclic presentation ( 1 ) , up to units of @xmath454 $ ] . according to ( * ? ? ? * remark 4 ) , this property holds for a wider class of knots and cyclic presentations of their cyclic branched coverings . the following cyclic presentation for @xmath455 , when @xmath81 is odd , arises from the surgery description of the manifold as takahashi manifold ( see section 5 ) . [ cyclicpresentation ] @xcite let @xmath109 be the @xmath7-fold cyclic branched covering of the 2-bridge knot @xmath456 , with @xmath457 $ ] . then its fundamental group has the following cyclic presentation : @xmath458 where @xmath459 for @xmath438 ( subscripts mod @xmath7 ) . the right part of these formulas are defined by the recurrent rule @xmath460 and @xmath461 where @xmath462 , for @xmath463 . we will illustrate the result obtained for @xmath316 and @xmath464 . if @xmath316 , then @xmath465 . observe that this case corresponds to takahashi manifolds and was discussed in @xcite . we get @xmath466 for example , if @xmath467 and @xmath468 then @xmath469 , and @xmath470 is the figure - eight knot @xmath471 @xcite . so , its @xmath7-fold cyclic branched covering has the fundamental group with the following cyclic presentation @xmath472 ( compare with @xcite ) . if @xmath464 then @xmath473 has conway parameters @xmath474 $ ] and @xmath475 has the following presentation : @xmath476 where @xmath477 @xmath478^{q_2 } x_k^{q_1 } x_{k+1}^{-q_1 } \left [ ( x_{k+1}^{q_1 } x_{k+2}^{-q_1})^{-s_1 } x_{k+1 } ( x_{k}^{q_1 } x_{k+1}^{-q_1})^{s_1 } \right]^{-q_2 } \right]^{-s_2}\ ] ] @xmath479 @xmath480^{q_2 } x_{k-1}^{q_1 } x_{k}^{-q_1 } \left [ ( x_{k}^{q_1 } x_{k+1}^{-q_1})^{-s_1 } x_{k } ( x_{k-1}^{q_1 } x_{k}^{-q_1})^{s_1 } \right]^{-q_2 } \right]^{s_2 } .\ ] ] for example , if @xmath481 and @xmath482 then @xmath483 , that corresponds to the knot @xmath484 . so , its @xmath7-fold cyclic branched covering has the fundamental group with the following cyclic presentation : @xmath485 another cyclic presentation for the fundamental group of every cyclic branched covering of a 2-bridge knots can be obtained from the heegaard diagram construction of @xcite , since the dunwoody family contains these manifolds ( see theorem [ theorem dunwoody ] ) . moreover , these presentations arise from heegaard diagrams and are therefore geometric . we point out that the problem of determining whether a balanced presentation of a group is _ geometric _ i.e. , induced by a heegaard diagram of a closed orientable 3-manifold is of considerable interest within geometric topology @xcite . [ corollary 6 ] @xcite the fundamental group of every branched cyclic covering of a 2-bridge knot admits a cyclic presentation which is geometric . in this section we present the homology groups of some classes of cyclic branched coverings of 2-bridge knots / links . recall ( see , for example @xcite ) that when @xmath486 is a finite abelian group , the order of this group is given by the absolute value @xmath487 where @xmath488 is the alexander polynomial of @xmath75 and @xmath489 is an @xmath7-th primitive root of unity . as is well known , the manifold @xmath131 is homeomorphic to the brieskorn manifold @xmath129 . thus , in this case , the homology groups can be obtained from @xcite . if @xmath81 is odd , then the @xmath7-fold cyclic branched covering is unique and the homology groups are the following . @xcite if @xmath81 is odd then : @xmath490 where @xmath491 . if @xmath81 is even the situation is rather more complicated , since @xmath357 can assume any value in @xmath492 . @xcite let @xmath81 be even . then : @xmath493 where @xmath494 , @xmath495 , @xmath496 , @xmath497 , @xmath498 and @xmath499 . as a consequence we have the homology groups in the interesting cases @xmath73 ( i.e. , meridian - cyclic coverings ) . let @xmath81 be even and @xmath73 . then : @xmath500 where @xmath495 , @xmath496 , @xmath501 and @xmath502 . the previous corollary contains , as a particular case , the homology groups of the brieskorn manifolds @xmath503 ( see also @xcite ) . if @xmath81 is even , then : @xmath504 where @xmath491 . the homology of cyclic branched coverings of 2-bridge knots of genus one has been obtained in @xcite and @xcite , using an algorithm of fox @xcite . recall that the 2-bridge knot @xmath90 has genus one if and only if @xmath505 divides @xmath506 when @xmath507 mod @xmath327 and @xmath505 divides @xmath508 when @xmath509 mod @xmath327 ( up to equivalence we can assume that @xmath83 is even for any 2-bridge knot @xmath90 ) . @xcite let @xmath90 be a 2-bridge knot of genus one . then : @xmath510 where @xmath511 are the integers defined by : @xmath512 @xmath513 and @xmath514 thus , the homology does not depend on @xmath83 . in some particular cases , explicit formulae are obtained by j. minkus . * corollary 11.2 ) ( i ) : : if @xmath8 is even then @xmath515 ( ii ) : : if @xmath516 is odd then @xmath517 is a homology sphere . cyclic branched coverings of the whitehead link @xmath119 have been intensively studied ( see @xcite ) . [ whitehead - monodromy ] @xcite let @xmath177 and let @xmath73 , then : @xmath518 for @xmath519 and @xmath520 , the homology groups of @xmath521 are listed in @xcite and @xcite . in this section we prove that each singly - cyclic branched covering of a 2-bridge link is the composition of a meridian - cyclic branched covering of a certain link @xmath522 described below and a cyclic branched covering of a trivial knot . this gives a generalization of a result obtained in @xcite for the case of the whitehead link . we can always assume that @xmath523 $ ] if @xmath140 is a knot , and @xmath524 $ ] if @xmath140 is a link . for any @xmath525 and @xmath526 , define a link @xmath527 as in figure 13 , where @xmath443 and @xmath528 denote numbers of half - twists ( in the vertical direction ) in the corresponding boxes , and the fragment `` in degree '' @xmath529 must be repeated @xmath529 times . the link @xmath522 has @xmath530 components , where @xmath531}\ ] ] is the linking number of @xmath90 ( * ? ? ? * remark 12.7 ) ( @xmath532 $ ] denotes the integral part of @xmath450 ) . for example @xmath533 and therefore @xmath534 has @xmath535 components . observe that , roughly speaking , each 2-bridge link can be obtained as a quotient of a suitable 2-bridge knot / link by an involution whose axis does not intersect the knot / link ( for example , see the 2-periodic presentations of 2-bridge knots / links in @xcite ) . hence , any 2-bridge link @xmath140 can be presented in the form @xmath536 , where @xmath537 . for example , if @xmath538 is the whitehead link , then @xmath539 is equivalent to @xmath540 , and @xmath541 $ ] with @xmath542 . hence we get the presentation of the whitehead link in the form @xmath543 ( see , for example , figure 3.4 in @xcite ) . denote by @xmath544 the orbifold with underlying space @xmath0 and singular set @xmath522 , with all branched indices equal to @xmath2 . moreover , denote by @xmath545 the orbifold having @xmath0 as underlying space and the 2-bridge link @xmath546 as singular set , with branched indices @xmath7 and @xmath547 respectively for the two components . ( 90,75)(0,10 ) ( 40,80)(0,0)[cc]@xmath549 ( 50,70)(2,-1)30 ( 40,70)(0,-1)50 ( 68,68)(0,0)[cc]@xmath550 ( 35,45)(0,0)[cc]@xmath7 ( 80,45)(0,0)[cc]@xmath551 ( 80,35)(-2,-1)30 ( 68,25)(0,0)[cc]@xmath529 ( 40,10)(0,0)[cc]@xmath552 is commutative . moreover , the @xmath550covering is a meridian - cyclic covering of the link @xmath553 and the @xmath529-covering is the @xmath529-fold cyclic branched covering of a component of @xmath554 ( which is a trivial knot ) . the fundamental group @xmath329 of the orbifold @xmath555 admits the presentation @xmath556 where @xmath557 and @xmath558 are the homotopy classes of two meridians @xmath138 and @xmath139 around the components @xmath559 and @xmath560 of the link , and @xmath561 is the relation deriving from the standard presentation of the group of @xmath140 . from the definition of a singly - cyclic covering we have @xmath562 for the epimorphism @xmath563 defined by @xmath564 and @xmath565 . consider the subgroup @xmath566 such that @xmath567 , where @xmath568 . let @xmath569 be the epimorphism defined by @xmath570 and @xmath571 . this epimorphism induces the @xmath529-fold cyclic covering @xmath572 such that the axis of the cyclic group action is the component @xmath559 of @xmath573 corresponding to the meridian @xmath557 . therefore , the underlying space of the orbifold @xmath574 is @xmath0 and the singular set is @xmath575 . obviously , @xmath576 is a trivial knot in @xmath0 with singularity index @xmath550 , and @xmath577 is a @xmath529-periodic knot / link which also has singularity index @xmath550 . since @xmath140 is equivalent to @xmath578 with @xmath579 , then we have @xmath580 . using @xmath581 and @xmath582 , the diagram of coverings is commutative . this completes our proof . conway , j.h . : an enumeration of knots and links and some of their related properties . in : computational problems in abstract algebra , oxford 1967 ( ed . j.leech ) , pergamon press ( 1969 ) , 329358 . fomenko , a.t . , matveev , s.v . : constant energy surfaces of hamiltonian systems , enumeration of three - dimensional manifolds in increasing order of complexity , and computation of volumes of closed hyperbolic manifolds . russian math . surveys ( 1 ) * 43 * ( 1988 ) , 324 . hilden , h.m . , lozano , m.t . , montesinos - amilibia , j.m . : the arithmeticity of the figure eight knot orbifolds . in : ` topology 90 ' ( b. apanasov , w. d. neumann , a. w. reid , and l. siebenmann , eds . ) ohio state univ . , research inst . publ . , walter de gruyter ed . , berlin * 1 * ( 1992 ) , 169183 . hilden , h.m . , lozano , m.t . , montesinos - amilibia , j.m . : volumes and chern simons invariants of cyclic coverings over rational knots . in : proc . of the 37-th tanigughi symposium on topology and teichmller spaces held in finland , july 1995 , ed . by s. kojima et al .. world scientific publishing co. ( 1996 ) , 3155 . kim , a.c . , kim , y. , vesnin , a. : on a class of cyclically presented groups . in : proc . groups - korea 98 ( y.g . baik , d.l . johnson , a.c . kim eds . ) , walter de gruyter , berlin - new york ( 2000 ) , 211220 . ruini , b. , spaggiari , f. : on the structure of takahashi manifolds . tsukuba j. math . * 22 * ( 1998 ) , 723739 . corrigendum to `` on the structure of takahashi manifolds '' . tsukuba j. math . * 24 * ( 2000 ) , 433434 . song , h.j . , kim , s.h . : dunwoody 3-manifolds and @xmath244-decomposible knots . in : proceedings of workshop in pure mathematics , ( jongsu kim and sungbok hom ed . ) , `` geometry and topology '' , vol . 19 ( 2000 ) , 193211 .

we consider a complex covariant form of the macroscopic maxwell equations , in a moving medium or at rest , following the original ideas of minkowski . a compact , lorentz invariant , derivation of the energy - momentum tensor and the corresponding differential balance equations are given . conservation laws and quantization of the electromagnetic field will be discussed in this covariant approach elsewhere .

1604.01876

although a systematic study of electromagnetic phenomena in media is not possible without methods of quantum mechanics , statistical physics and kinetics , in practice a standard mathematical model based on phenomenological maxwell s equations provides a good approximation to many important problems . as is well - known , one should be able to obtain the electromagnetic laws for continuous media from those for the interaction of fields and point particles @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite . as a result of the hard work of several generations of researchers and engineers , the classical electrodynamics , especially in its current complex covariant form , undoubtedly satisfies dirac s criteria of mathematical beauty , being a state of the art mathematical description of nature . in macroscopic electrodynamics , the volume ( mechanical or ponderomotive ) forces , acting on a medium , and the corresponding energy density and energy flux are introduced with the help of the energy - momentum tensors and differential balance relations @xcite , @xcite , @xcite , @xcite , @xcite , @xcite . these forces occur in the equations of motion for a medium or individual charges and , in principle , they can be experimentally tested @xcite , @xcite , @xcite , @xcite ( see also the references therein ) . but interpretation of the results should depend on the accepted model of the interaction between the matter and radiation . in this methodological note , we discuss a complex version of minkowski s phenomenological electrodynamics ( at rest or in a moving medium ) without assuming any particular form of material equations as far as possible . lorentz invariance of the corresponding differential balance equations is emphasized in view of long - standing uncertainties about the electromagnetic stresses and momentum density , the so - called abraham - minkowski controversy ( see , for example , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite and the references therein ) . the paper is organized as follows . in sections 2 to 4 , we describe the @xmath0-complex version of maxwell s equations and derive the corresponding differential balance density laws for the electromagnetic fields . their covariant versions are given in sections 5 to 9 . the case of a uniformly moving medium is discussed in section 10 and complex lagrangians are introduced in section 11 . some useful tools are collected in appendices a to c for the reader s benefit . traditionally , the macroscopic maxwell equations in a fixed frame of reference are given by@xmath1@xmath2 here , @xmath3 is the electric field , @xmath4 is the displacement field ; @xmath5 is the magnetic field , @xmath6 is the induction field . these equations , which are obtained by averaging of microscopic maxwell s equations in the vacuum , provide a good mathematical description of electromagnetic phenomena in various media , when complemented by the corresponding material equations . in the simplest case of an isotropic medium at rest , one usually has@xmath7 where @xmath8 is the dielectric constant , @xmath9 is the magnetic permeability , and @xmath10 describes the conductivity of the medium ( see , for example , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite for fundamentals of classical electrodynamics ) . introduction of two complex fields@xmath11 allows one to rewrite the phenomenological maxwell equations in the following compact form@xmath12@xmath13 where the asterisk stands for complex conjugation ( see also @xcite , @xcite and @xcite ) . as we shall demonstrate , different complex forms of maxwell s equations are particularly convenient for study of the corresponding energy - momentum balance equations for the electromagnetic fields in the presence of the free charges and currents in a medium . we begin from a complex @xmath0-interpretation of the traditional symmetric energy - momentum tensor @xcite . by definition,@xmath14 = t_{qp}\qquad\left ( p , q=1,2,3\right ) \nonumber\end{aligned}\ ] ] and the corresponding momentum balance equation,@xmath15 _ { p}\label{momentumbalance}\\ & \ = \frac{\partial t_{pq}}{\partial x_{q}}+\frac{1}{16\pi}\left [ \operatorname{curl}\left ( \mathbf{f\times g}^{\ast}+\mathbf{f}^{\ast}\times\mathbf{g}\right ) \right ] _ { p}\nonumber\\ & \quad+\frac{1}{16\pi}\left ( f_{q}\frac{\partial g_{q}^{\ast}}{\partial x_{p}}-g_{q}\frac{\partial f_{q}^{\ast}}{\partial x_{p}}+f_{q}^{\ast}\frac{\partial g_{q}}{\partial x_{p}}-g_{q}^{\ast}\frac{\partial f_{q}}{\partial x_{p}}\right ) , \nonumber\end{aligned}\ ] ] can be obtained from maxwell s equations ( [ maxwellcomplexhom])([maxwellcomplexnothom ] ) as a result of elementary but rather tedious vector calculus calculations usually omitted in textbooks . ( we use einstein summation convention over any two repeated indices unless otherwise stated . in this paper , greek indices run from @xmath16 to @xmath17 while latin indices may have values from @xmath18 to @xmath19 inclusive . ) * proof . * indeed , in a @xmath0-complex form,@xmath20 due to an identity @xcite:@xmath21 taking into account the complex conjugate , we derive@xmath22 \label{facti}\\ & \quad=\frac{1}{2}\left ( \mathbf{f}\operatorname{div}\mathbf{g}^{\ast } -\mathbf{g}^{\ast}\times\operatorname{curl}\mathbf{f}+\mathbf{f}^{\ast } \operatorname{div}\mathbf{g}-\mathbf{g}\times\operatorname{curl}\mathbf{f}^{\ast}\right ) _ { p}\nonumber\\ & \qquad+\frac{1}{2}\left ( \mathbf{g}\operatorname{div}\mathbf{f}^{\ast } -\mathbf{f}^{\ast}\times\operatorname{curl}\mathbf{g}+\mathbf{g}^{\ast } \operatorname{div}\mathbf{f}-\mathbf{f}\times\operatorname{curl}\mathbf{g}^{\ast}\right ) _ { p}\nonumber\end{aligned}\ ] ] as our first important fact . on the other hand , in view of maxwell s equations ( [ maxwellcomplexhom])([maxwellcomplexnothom ] ) , one gets@xmath23 and , with the help of its complex conjugate,@xmath24 or @xmath25 providing the second important fact . ( up to the constant , the first term in the right - hand side represents the density of lorentz s force acting on the free charges and currents in the medium under consideration @xcite , @xcite . ) in view of ( [ factii ] ) and ( [ facti ] ) , we can write@xmath26 \nonumber\\ & \quad\quad-\frac{1}{2}\left ( \mathbf{g}\operatorname{div}\mathbf{f}^{\ast } -\mathbf{f}^{\ast}\times\operatorname{curl}\mathbf{g}+\mathbf{g}^{\ast } \operatorname{div}\mathbf{f}-\mathbf{f}\times\operatorname{curl}\mathbf{g}^{\ast}\right ) _ { p}\nonumber\\ & \quad=\frac{1}{4}\frac{\partial}{\partial x_{q}}\left [ f_{p}g_{q}^{\ast } + f_{p}^{\ast}g_{q}+f_{q}g_{p}^{\ast}+f_{q}^{\ast}g_{p}-\delta_{pq}\left ( \mathbf{f}\cdot\mathbf{g}^{\ast}+\mathbf{f}^{\ast}\cdot\mathbf{g}\right ) \right ] \nonumber\\ & \quad\quad-\frac{1}{4}\left ( \mathbf{g}\operatorname{div}\mathbf{f}^{\ast } -\mathbf{f}^{\ast}\times\operatorname{curl}\mathbf{g}+\mathbf{g}^{\ast } \operatorname{div}\mathbf{f}-\mathbf{f}\times\operatorname{curl}\mathbf{g}^{\ast}\right ) _ { p}\nonumber\\ & \quad\quad+\frac{1}{4}\left ( \mathbf{f}\operatorname{div}\mathbf{g}^{\ast } -\mathbf{g}^{\ast}\times\operatorname{curl}\mathbf{f}+\mathbf{f}^{\ast } \operatorname{div}\mathbf{g}-\mathbf{g}\times\operatorname{curl}\mathbf{f}^{\ast}\right ) _ { p}\nonumber\\ & \quad=4\pi\frac{\partial t_{pq}}{\partial x_{q}}+\frac{1}{4}\left ( \mathbf{f}\operatorname{div}\mathbf{g}^{\ast}-\mathbf{g}^{\ast}\operatorname{div}\mathbf{f}+\mathbf{f}^{\ast}\operatorname{div}\mathbf{g}-\mathbf{g}\operatorname{div}\mathbf{f}^{\ast}\right ) _ { p}\nonumber\\ & \quad\quad+\frac{1}{4}\left ( \mathbf{f}\times\operatorname{curl}\mathbf{g}^{\ast}-\mathbf{g}^{\ast}\times\operatorname{curl}\mathbf{f}+\mathbf{f}^{\ast}\times\operatorname{curl}\mathbf{g}-\mathbf{g}\times\operatorname{curl}\mathbf{f}^{\ast}\right ) _ { p}.\nonumber\end{aligned}\ ] ] finally , in the last two lines , one can utilize the following differential vector calculus identity,@xmath27 _ { p}\label{vectoridentity}\\ & \quad = a_{q}\frac{\partial b_{q}}{\partial x_{p}}-b_{q}\frac{\partial a_{q}}{\partial x_{p}},\nonumber\end{aligned}\ ] ] see ( [ a5 ] ) , with @xmath28 @xmath29 and its complex conjugates , in order to obtain ( [ momentumbalance ] ) and/or ( [ momentumbalancedreal ] ) , which completes the proof . ( an independent proof will be given in section 7 . ) derivation of the corresponding differential energy balance equation is much simpler . by ( [ maxwellcomplexhom]),@xmath30 due to a familiar vector calculus identity ( [ a1 ] ) : @xmath31 in a traditional form,@xmath32 ( see , for example , @xcite , @xcite ) , where one can substitute@xmath33 as a result , @xmath0-differential energy - momentum balance equations are given by@xmath34 and@xmath35 _ { p}+\frac{\partial t_{pq}}{\partial x_{q}}-\left ( \rho\mathbf{e}+\frac{1}{c}\mathbf{j}\times\mathbf{b}\right ) _ { p}\label{momentumbalancedreal}\\ & \ + \frac{1}{8\pi}\left [ \operatorname{curl}\left ( \mathbf{e\times d}+\mathbf{h}\times\mathbf{b}\right ) \right ] _ { p}\nonumber\\ & \quad+\frac{1}{8\pi}\left ( \mathbf{e}\cdot\frac{\partial\mathbf{d}}{\partial x_{p}}-\mathbf{d}\cdot\frac{\partial\mathbf{e}}{\partial x_{p}}+\mathbf{h}\cdot\frac{\partial\mathbf{b}}{\partial x_{p}}-\mathbf{b}\cdot\frac{\partial\mathbf{h}}{\partial x_{p}}\right ) = 0,\nonumber\end{aligned}\ ] ] respectively ( see also @xcite , @xcite ) . the real form of the symmetric stress tensor ( [ stresssymcomplex ] ) , namely,@xmath36 \qquad\left ( p , q=1,2,3\right ) , \nonumber\end{aligned}\ ] ] is due to hertz @xcite . equations ( [ energybalance])([momentumbalancedreal ] ) are related to a fundamental concept of conservation of mechanical and electromagnetic energy and momentum . here , these balance conditions are presented in differential forms in terms of the corresponding local field densities . they can be integrated over a given volume in @xmath37 in order to obtain , in a traditional way , the corresponding conservation laws of the electromagnetic fields ( see , for example , @xcite , @xcite , @xcite , @xcite , @xcite ) . these laws made it necessary to ascribe a definite linear momentum and energy to the field of an electromagnetic wave , which can be observed , for example , as light pressure . * at this point , the lorentz invariance of these differential balance equations is not obvious in our @xmath0-analysis . but one can introduce the four - vector @xmath38 and try to match ( [ energybalance])([momentumbalancedreal ] ) with the expression,@xmath39 as an initial step , in order to guess the corresponding four - tensor form . an independent covariant derivation will be given in section 7 . * in an isotropic nonhomogeneous variable medium ( without dispersion and/or compression ) , when @xmath40 and @xmath41 the ponderomotive forces in ( [ energybalance ] ) and ( [ momentumbalancedreal ] ) take the form @xcite:@xmath42{c}\dfrac{1}{c}\left ( \dfrac{\partial\varepsilon}{\partial t}\mathbf{e}^{2}+\dfrac{\partial\mu}{\partial t}\mathbf{h}^{2}\right ) \bigskip\\ \mathbf{e}^{2}\nabla\varepsilon+\mathbf{h}^{2}\nabla\mu \end{array } \right ) , \nonumber\end{aligned}\ ] ] which may be interpreted as a four - vector energy - force acting from an inhomogeneous and time - variable medium . its covariance is analyzed in section 7 . the @xmath0-linear momentum differential balance equation ( [ momentumbalancedreal ] ) , can be rewritten in a more compact form,@xmath43 with the help of the hertz symmetric stress tensor @xmath44 defined by ( [ stresssymreal ] ) . a net force is given by@xmath45 _ { p}\label{netforce}\\ & -\frac{1}{8\pi}\left ( \mathbf{e}\cdot\frac{\partial\mathbf{d}}{\partial x_{p}}-\mathbf{d}\cdot\frac{\partial\mathbf{e}}{\partial x_{p}}+\mathbf{h}\cdot\frac{\partial\mathbf{b}}{\partial x_{p}}-\mathbf{b}\cdot\frac { \partial\mathbf{h}}{\partial x_{p}}\right ) .\nonumber\end{aligned}\ ] ] in this notation , we state the @xmath0-angular momentum differential balance equation as follows@xmath46 where the field angular momentum density is defined by@xmath47 and the flux of angular momentum is described by the following tensor @xcite:@xmath48 ( here , @xmath49 is the totally anti - symmetric levi - civita symbol with @xmath50 ) . an elementary example of conservation of the total angular momentum is discussed in @xcite . * indeed , in view of ( [ compactmomentumbalance ] ) , one can write@xmath51 which completes the proof . * note . * once again , in @xmath0-form , the lorentz invariance of this differential balance equation for the local densities is not obvious . an independent covariant derivation will be given in section 8 . with the help of complex fields @xmath52 and @xmath53 we introduce the following anti - symmetric four - tensor,@xmath54{cccc}0 & -g_{1 } & -g_{2 } & -g_{3}\\ g_{1 } & 0 & if_{3 } & -if_{2}\\ g_{2 } & -if_{3 } & 0 & if_{1}\\ g_{3 } & if_{2 } & -if_{1 } & 0 \end{array } \right ) \label{complexqtensor}\ ] ] and use the standard four - vectors , @xmath38 and @xmath55 for contravariant coordinates and current , respectively . maxwell s equations then take the covariant form @xcite , @xcite : @xmath56 with summation over two repeated indices . indeed , in block form , we have@xmath57{cc}0 & -g_{q}\medskip\\ g_{p } & ie_{pqr}f_{r}\end{array } \right ) = \left ( \begin{array } [ c]{c}-\operatorname{div}\mathbf{g}=-4\pi\rho\medskip\\ \dfrac{1}{c}\dfrac{\partial\mathbf{g}}{\partial t}+i\operatorname{curl}\mathbf{f}=-\dfrac{4\pi}{c}\mathbf{j}\end{array } \right ) , \label{cmproof}\ ] ] which verifies this fact . the continuity equation,@xmath58 or in the @xmath0-form,@xmath59 describes conservation of the electrical charge . the latter equation can also be derived in the complex @xmath0-form from ( [ maxwellcomplexhom])([maxwellcomplexnothom ] ) . * note . * in vacuum , when @xmath60 and @xmath61 @xmath62 one can write due to ( [ b1])([b2]):@xmath63 as a result , the following self - duality property holds@xmath64 ( see , for example , @xcite , @xcite and appendix b ) . two covariant forms of maxwell s equations are given by@xmath65 where @xmath66 @xmath67 and @xmath68diag@xmath69 the last equation can be derived from a more general equation , involving a rank three tensor,@xmath70 ( @xmath71 are fixed ; no summation is assumed over these two indices ) , which is related to the pauli - lubaski vector from the representation theory of the poincar group @xcite . different spinor forms of maxwell s equations are analyzed in @xcite ( see also the references therein ) . two dual anti - symmetric field tensors of complex fields , @xmath52 and @xmath53 are given by @xmath72{cccc}0 & -g_{1 } & -g_{2 } & -g_{3}\\ g_{1 } & 0 & if_{3 } & -if_{2}\\ g_{2 } & -if_{3 } & 0 & if_{1}\\ g_{3 } & if_{2 } & -if_{1 } & 0 \end{array } \right ) = r^{\mu\nu}+is^{\mu\nu}\label{complexq}\\ & = \left ( \begin{array } [ c]{cccc}0 & -d_{1 } & -d_{2 } & -d_{3}\\ d_{1 } & 0 & -h_{3 } & h_{2}\\ d_{2 } & h_{3 } & 0 & -h_{1}\\ d_{3 } & -h_{2 } & h_{1 } & 0 \end{array } \right ) + i\left ( \begin{array } [ c]{cccc}0 & -b_{1 } & -b_{2 } & -b_{3}\\ b_{1 } & 0 & e_{3 } & -e_{2}\\ b_{2 } & -e_{3 } & 0 & e_{1}\\ b_{3 } & e_{2 } & -e_{1 } & 0 \end{array } \right ) \nonumber\end{aligned}\ ] ] and@xmath73{cccc}0 & f_{1 } & f_{2 } & f_{3}\\ -f_{1 } & 0 & ig_{3 } & -ig_{2}\\ -f_{2 } & -ig_{3 } & 0 & ig_{1}\\ -f_{3 } & ig_{2 } & -ig_{1 } & 0 \end{array } \right ) = f_{\mu\nu}+ig_{\mu\nu}\label{complexp}\\ & = \left ( \begin{array } [ c]{cccc}0 & e_{1 } & e_{2 } & e_{3}\\ -e_{1 } & 0 & -b_{3 } & b_{2}\\ -e_{2 } & b_{3 } & 0 & -b_{1}\\ -e_{3 } & -b_{2 } & b_{1 } & 0 \end{array } \right ) + i\left ( \begin{array } [ c]{cccc}0 & h_{1 } & h_{2 } & h_{3}\\ -h_{1 } & 0 & d_{3 } & -d_{2}\\ -h_{2 } & -d_{3 } & 0 & d_{1}\\ -h_{3 } & d_{2 } & -d_{1 } & 0 \end{array } \right ) .\nonumber\end{aligned}\ ] ] the real part of the latter represents the standard electromagnetic field tensor in a medium @xcite , @xcite , @xcite . as for the imaginary part of ( [ complexq ] ) , which , ironically , pauli called an artificiality in view of its non - standard behavior under spatial inversion @xcite , the use of complex conjugation restores this important symmetry for our complex field tensors . the dual tensor identities are given by@xmath74 here @xmath75 and @xmath76 is the levi - civita four - symbol @xcite . then@xmath77 and both pairs of maxwell s equations can also be presented in the form @xcite@xmath78 in addition to the one given above@xmath79 the real part of the first equation traditionally represents the first ( homogeneous ) pair of maxwell s equation and the real part of the second one gives the remaining pair . in our approach both pairs of maxwell s equations appear together ( see also @xcite , @xcite , @xcite , @xcite , and @xcite for the case in vacuum ) . moreover , a generalization to complex - valued four - current may naturally represent magnetic charge and magnetic current not yet observed in nature @xcite . an important cofactor matrix identity,@xmath80 was originally established , in a general form , by minkowski @xcite . once again , the dual tensors are given by@xmath81{cc}0 & f_{q}\medskip\\ -f_{p } & ie_{pqr}g_{r}\end{array } \right ) , \quad q^{\mu\nu}=\left ( \begin{array } [ c]{cc}0 & -g_{q}\medskip\\ g_{p } & ie_{pqr}f_{r}\end{array } \right ) , \label{pqblockform}\ ] ] in block form . a complete list of relevant tensor and matrix identities is given in appendix b. as has been announced in @xcite ( see also @xcite ) , the covariant form of the differential balance equations can be presented as follows@xmath82 \label{covariantenergymomentumbalance}\\ & \ + \frac{1}{32\pi}\left ( p_{\sigma\tau}^{\ast}\frac{\partial q^{\tau \sigma}}{\partial x^{\mu}}+p_{\sigma\tau}\frac{\partial\overset{\ast}{\left . q^{\tau\sigma}\right . } } { \partial x^{\mu}}\right ) \nonumber\\ & \ = -\frac{1}{c}f_{\mu\lambda}j^{\lambda}=\left ( \begin{array } [ c]{c}-\mathbf{j}\cdot\mathbf{e}/c\smallskip\\ \rho\mathbf{e}+\mathbf{j}\times\mathbf{b}/c \end{array } \right ) .\nonumber\end{aligned}\ ] ] in our complex form , when @xmath52 and @xmath53 the energy - momentum tensor is given by@xmath83{cc}\mathbf{f}\cdot\mathbf{g}^{\ast}+\mathbf{f}^{\ast}\cdot\mathbf{g } & 2i\left ( \mathbf{f\times f}^{\ast}\right ) _ { q}\bigskip\\ -2i\left ( \mathbf{g\times g}^{\ast}\right ) _ { p } & \quad2\left ( f_{p}g_{q}^{\ast}+f_{p}^{\ast}g_{q}\right ) -\delta_{pq}\left ( \mathbf{f}\cdot\mathbf{g}^{\ast}+\mathbf{f}^{\ast}\cdot\mathbf{g}\right ) \end{array } \right ) .\nonumber\end{aligned}\ ] ] here , we point out for the reader s convenience that @xmath84 and , in real form,@xmath85{cc}\left ( \mathbf{e}\cdot\mathbf{d}+\mathbf{h}\cdot\mathbf{b}\right ) /2 & \left ( \mathbf{e\times h}\right ) _ { q}\bigskip\\ -\left ( \mathbf{d\times b}\right ) _ { p } & \quad e_{p}d_{q}+h_{p}b_{q}-\delta_{pq}\left ( \mathbf{e}\cdot\mathbf{d}+\mathbf{h}\cdot\mathbf{b}\right ) /2 \end{array } \right ) . \label{4energymomentumtensorreal}\ ] ] the covariant form of the differential balance equation allows one to clarify the meanings of different energy - momentum tensors . for instance , it is worth noting that the non - symmetric maxwell and heaviside form of the @xmath0-stress tensor @xcite , @xmath86 appears here in the corresponding momentum balance equation @xcite:@xmath35 _ { p}+\frac{\partial\widetilde{t}_{pq}}{\partial x_{q}}-\left ( \rho\mathbf{e}+\frac{1}{c}\mathbf{j}\times \mathbf{b}\right ) _ { p}\label{momentumbalancedrealnonsymmetric}\\ & \quad+\frac{1}{8\pi}\left ( \mathbf{e}\cdot\frac{\partial\mathbf{d}}{\partial x_{p}}-\mathbf{d}\cdot\frac{\partial\mathbf{e}}{\partial x_{p}}+\mathbf{h}\cdot\frac{\partial\mathbf{b}}{\partial x_{p}}-\mathbf{b}\cdot\frac{\partial\mathbf{h}}{\partial x_{p}}\right ) = 0.\nonumber\end{aligned}\ ] ] at the same time , in view of ( [ momentumbalancedreal ] ) , use of the form ( [ 3denergymomentummaxwellhtensor ] ) differs from hertz s symmetric tensors in ( [ stresssymcomplex ] ) and ( [ stresssymreal ] ) only in the case of anisotropic media ( crystals ) @xcite , @xcite . indeed,@xmath87 _ { p}. \label{divergencetwotensors}\ ] ] moreover , with the help of elementary identities,@xmath88 _ { p}=\frac{\partial}{\partial x_{q}}\left ( a_{p}b_{q}-a_{q}b_{p}\right ) \label{curlidentity}\ ] ] and@xmath89 _ { p } , \label{curlidentitytwo}\ ] ] one can transform the latter balance equation into its symmetric form , which provides an independent proof of ( [ momentumbalancedreal ] ) . the fact that maxwell s equations can be united with the help of a complex second rank ( anti - symmetric ) tensor allows us to utilize the standard sturm - liouville type argument in order to establish the energy - momentum differential balance equations in covariant form . indeed , by adding matrix equation@xmath90 and its complex conjugate@xmath91 one gets@xmath92 a simple decomposition,@xmath93 with @xmath94 and @xmath95 ( and their complex conjugates ) , results in @xmath82 \label{pqforce}\\ & \ + \frac{1}{16\pi}\left [ \left ( p_{\mu\lambda}^{\ast}\frac{\partial q^{\lambda\nu}}{\partial x^{\nu}}-\frac{\partial p_{\mu\lambda}}{\partial x^{\nu}}\overset{\ast}{\left . q^{\lambda\nu}\right . } \right ) + \left ( \text{c.c.}\right ) \right ] = -\frac{1}{c}f_{\mu\lambda}j^{\lambda}.\nonumber\end{aligned}\ ] ] by a direct substitution , one can verify that @xmath96 ( an independent covariant proof of these identities is given in appendix c. ) finally , introducing@xmath97 we obtain ( [ covariantenergymomentumbalance ] ) with the explicitly covariant expression for the ponderomotive force ( [ pforce ] ) , which completes the proof . as a result , the covariant energy - momentum balance equation is given by@xmath98 in a compact form . if these differential balance equations are written for a stationary medium , then the corresponding equations for moving bodies are uniquely determined , since the components of a tensor in any inertial coordinate system can be derived by a proper lorentz transformation @xcite . by definition , @xmath99 and @xmath100 where @xmath101diag@xmath102 in view of ( [ covariantembalancecompact ] ) , we derive@xmath103 as a required differential balance equation . with the help of familiar dual relations ( [ b04 ] ) , one can get another covariant form of the angular momentum balance equation:@xmath104 in @xmath0-form , the latter relation can be reduced to ( [ vectormomentumbalance])([tensormomentum ] ) . indeed , when @xmath105 and @xmath106 one gets@xmath107 + \frac{\partial}{\partial x_{s}}\left ( e_{pqr}x_{q}\widetilde{t}_{rs}\right ) \label{4dto3dangmom}\\ & \qquad+e_{pqr}\widetilde{t}_{qr}+e_{pqr}x_{q}\left ( x_{r}+y_{r}\right ) = 0,\nonumber\end{aligned}\ ] ] where @xmath108 is the familiar lorentz force . substitution , @xmath109 results in ( [ vectormomentumbalance ] ) in view of identity ( [ divergencetwotensors ] ) . the remaining cases , when @xmath110 can be analyzed in a similar fashion . in @xmath0-form , the corresponding equations can be reduced to ( [ energybalance ] ) and ( [ momentumbalancedrealnonsymmetric ] ) . details are left to the reader . thus the angular momentum law has the form of a local balance equation , not a conservation law , since in general the energy - momentum tensor will not be symmetric @xcite . a torque , for instance , may occur , which can not be compensated for by a change in the electromagnetic angular momentum , though not in contradiction with experiment @xcite . let @xmath111 be a constant real - valued velocity vector representing uniform motion of one frame of reference with respect to another one . let us consider the following orthogonal decompositions,@xmath112 such that our complex vectors @xmath113 are collinear with the velocity vector @xmath111 and @xmath114 are perpendicular to it ( figure 1 ) . the lorentz transformation of electric and magnetic fields @xmath115 take the following complex form@xmath116 and@xmath117 although this transformation was found by lorentz , it was minkowski who realized that this is the law of transformation of the second rank anti - symmetric four - tensors @xcite , @xcite ; a brief historical overview is given in @xcite . ) this complex @xmath0-form of the lorentz transformation of electric and magnetic fields was known to minkowski ( 1908 ) , but apparently only in vacuum , when @xmath60 ( see also @xcite ) . moreover,@xmath118 in the same notation @xcite . the latter equations can be rewritten as follows@xmath119 \mathbf{v},\qquad t^{\prime}=\gamma\left ( t-\frac{\mathbf{v}\cdot\mathbf{r}}{c^{2}}\right ) , \label{lorentzcoordinatetransformshort}\ ] ] where @xmath120 in a similar fashion , one gets@xmath121@xmath122 as a compact @xmath0-version of the lorentz transformation for the complex electromagnetic fields . in complex four - tensor form,@xmath123 although minkowski considered the transformation of electric and magnetic fields in a complex @xmath0-vector form , see eqs . ( 8)(9 ) and ( 15 ) in @xcite ( or eqs . ( 25.5)(25.6 ) in @xcite ) , he seems never to have combined the corresponding four - tensors into the complex forms ( [ complexq])([complexp ] ) . in the second article @xcite , max born , who used minkowski s notes , did nt mention the complex fields . as a result , the complex field tensor seems only to have appeared , for the first time , in @xcite ( see also @xcite ) . the complex identity , @xmath124invariant@xmath125under a similarity transformation , follows from minkowski s determinant relations ( [ b18])([b20 ] ) . * example*. let @xmath126 be an orthonormal basis in @xmath127 we choose @xmath128 and write @xmath129 with@xmath130{cccc}\gamma & -\beta\gamma & 0 & 0\\ -\beta\gamma & \gamma & 0 & 0\\ 0 & 0 & 1 & 0\\ 0 & 0 & 0 & 1 \end{array } \right ) , \qquad\quad\beta=\frac{v}{c},\quad\gamma=\frac{1}{\sqrt{1-\beta^{2 } } } \label{lorentzboost}\ ] ] for the corresponding lorentz boost ( figure 2 ) . in view of ( [ complexfieldtransform ] ) , by matrix multiplication one gets@xmath131{cccc}\gamma & -\beta\gamma & 0 & 0\\ -\beta\gamma & \gamma & 0 & 0\\ 0 & 0 & 1 & 0\\ 0 & 0 & 0 & 1 \end{array } \right ) \left ( \begin{array } [ c]{cccc}0 & -g_{1 } & -g_{2 } & -g_{3}\\ g_{1 } & 0 & if_{3 } & -if_{2}\\ g_{2 } & -if_{3 } & 0 & if_{1}\\ g_{3 } & if_{2 } & -if_{1 } & 0 \end{array } \right ) \left ( \begin{array } [ c]{cccc}\gamma & -\beta\gamma & 0 & 0\\ -\beta\gamma & \gamma & 0 & 0\\ 0 & 0 & 1 & 0\\ 0 & 0 & 0 & 1 \end{array } \right ) \nonumber\\ & \qquad=\left ( \begin{array } [ c]{cccc}0 & -g_{1 } & -\gamma g_{2}-i\beta\gamma f_{3 } & -\gamma g_{3}+i\beta\gamma f_{2}\\ g_{1 } & 0 & \beta\gamma g_{2}+i\gamma f_{3 } & \beta\gamma g_{3}-i\gamma f_{2}\\ \gamma g_{2}+i\beta\gamma f_{3 } & -\beta\gamma g_{2}-i\gamma f_{3 } & 0 & if_{1}\\ \gamma g_{3}-i\beta\gamma f_{2 } & -\beta\gamma g_{3}+i\gamma f_{2 } & -if_{1 } & 0 \end{array } \right ) . \label{martixmuttiplication}\ ] ] thus @xmath132 and@xmath133 in a similar fashion , @xmath134 and@xmath135 the latter relations are in agreement with the field transformations ( [ paralleltransform])([perpendiculartransform ] ) . in block form , one gets@xmath136{c}f_{1}^{\prime}\\ f_{2}^{\prime}\\ g_{3}^{\prime}\\ g_{2}^{\prime}\\ f_{3}^{\prime}\\ g_{1}^{\prime}\end{array } \right ) = \left ( \begin{array } [ c]{cccccc}1 & 0 & 0 & 0 & 0 & 0\\ 0 & \cos\left ( i\theta\right ) & \sin\left ( i\theta\right ) & 0 & 0 & 0\\ 0 & -\sin\left ( i\theta\right ) & \cos\left ( i\theta\right ) & 0 & 0 & 0\\ 0 & 0 & 0 & \cos\left ( i\theta\right ) & \sin\left ( i\theta\right ) & 0\\ 0 & 0 & 0 & -\sin\left ( i\theta\right ) & \cos\left ( i\theta\right ) & 0\\ 0 & 0 & 0 & 0 & 0 & 1 \end{array } \right ) \left ( \begin{array } [ c]{c}f_{1}\\ f_{2}\\ g_{3}\\ g_{2}\\ f_{3}\\ g_{1}\end{array } \right ) , \label{emfieldstransformmatrix}\ ] ] where , by definition,@xmath137 as a result , the transformation law of the complex electromagnetic fields @xmath138 under the lorentz boost can be thought of as a complex rotation in @xmath139 corresponding to a reducible representation of the one - parameter subgroup of @xmath140 ( cyclic permutation of the spatial indices cover the two remaining cases ; see also @xcite . ) electromagnetic phenomena in moving media are important in relativistic astrophysics , the study of accelerated plasma clusters and high - energy electron beams @xcite , @xcite , @xcite , @xcite . minkowski s field- and connecting - equations @xcite , @xcite were derived from the corresponding laws for the bodies at rest by means of a lorentz transformation ( see @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite ) . explicitly covariant forms , which are applicable both in the rest frame and for moving media , are analyzed in @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite ( see also the references therein ) . in standard notation,@xmath141 one can write @xcite , @xcite , @xcite , @xcite : @xmath142 , \label{3dmaterialeqs}\\ \mathbf{h } & = \frac{1}{\mu}\mathbf{b}+\frac{\kappa\gamma^{2}}{\mu}\left [ -\beta^{2}\mathbf{b}+\dfrac{\mathbf{v}}{c^{2}}\left ( \mathbf{v\cdot b}\right ) + \dfrac{1}{c}\left ( \mathbf{v\times e}\right ) \right ] .\nonumber\end{aligned}\ ] ] in covariant form , these relations are given by@xmath143 ( see @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite and the references therein ) . here,@xmath144 is the four - tensor of electric and magnetic permeabilities and@xmath145 is the four - velocity of the medium ( @xcite , @xcite , a computer algebra verification of these relations is given in @xcite ) . in a complex covariant form,@xmath146 in view of ( [ covariantmaterialeqs ] ) and ( [ b1])([b2 ] ) , we get@xmath147 in terms of the real - valued electromagnetic field tensor . in practice , one can choose@xmath148 for the real - valued four - vector potential @xmath149 then@xmath150 by ( [ fourpermeabilitytensor ] ) . substitution into maxwell s equations ( [ maxwellii ] ) or ( [ maxwelli ] ) results in @xmath151 a_{\sigma}\right . \label{covariantau}\\ & \qquad\left . + \partial_{\sigma}\left ( \partial^{\tau}a_{\tau}+\kappa u^{\nu}u^{\tau}\partial_{\nu}a_{\tau}\right ) ^{\ } \!\right\ } = -\frac{4\pi \mu}{c}j^{\lambda},\nonumber\end{aligned}\ ] ] where @xmath152 is the dalembert operator . in view of an inverse matrix identity,@xmath153 the latter equations take the form ) and ( [ covariantauj ] ) , together with the gauge condition ( [ covariantaugauge ] ) , may be considered as the fundamentals of the theory @xcite . our complex fields are given by ( [ qpmaterialf]).]@xmath154 a_{\sigma}-\partial_{\sigma}\left ( \partial^{\tau}a_{\tau}+\kappa u^{\nu}u^{\tau}\partial_{\nu}a_{\tau}\right ) \label{covariantauj}\\ & \qquad=\frac{4\pi\mu}{c}\left ( g_{\sigma\lambda}-\frac{\kappa}{1+\kappa } u_{\sigma}u_{\lambda}\right ) j^{\lambda}.\nonumber\end{aligned}\ ] ] subject to the subsidiary condition,@xmath155 these equations were studied in detail for the sake of development of the phenomenological classical and quantum electrodynamics in a moving medium ( see @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite and the references therein ) . in particular , green s function of the photon in a moving medium was studied in @xcite , @xcite , @xcite ( with applications to quantum electrodynamics ) . we follow @xcite , @xcite , @xcite with somewhat different details . the substitution,@xmath156 ( a generalization of hertz s potentials for a moving medium @xcite , @xcite ) , into the gauge condition ( [ covariantaugauge ] ) results in @xmath157 in view of@xmath158 then , equations ( [ covariantauj ] ) take the form@xmath159 \partial_{\sigma}z^{\lambda\sigma}=-\frac{4\pi\mu}{c}j^{\lambda}. \label{dzjhertzequation}\ ] ] indeed , the left - hand side of ( [ covariantauj ] ) is given by@xmath154 a_{\sigma}=\left [ \partial^{\tau}\partial_{\tau } + \kappa\left ( u^{\tau}\partial_{\tau}\right ) ^{2}\right ] g_{\sigma\mu } a^{\mu}\\ & \quad=\left [ \partial^{\tau}\partial_{\tau}+\kappa\left ( u^{\tau}\partial_{\tau}\right ) ^{2}\right ] g_{\sigma\mu}\left ( \frac{\kappa } { 1+\kappa}u^{\mu}u_{\lambda}-\delta_{\lambda}^{\mu}\right ) \partial_{\rho } z^{\lambda\rho}\\ & \quad=\left [ \partial^{\tau}\partial_{\tau}+\kappa\left ( u^{\tau}\partial_{\tau}\right ) ^{2}\right ] \left ( \frac{\kappa}{1+\kappa}u_{\sigma } u_{\lambda}-g_{\sigma\lambda}\right ) \partial_{\rho}z^{\lambda\rho}\\ & \qquad=\frac{4\pi\mu}{c}\left ( g_{\sigma\lambda}-\frac{\kappa}{1+\kappa } u_{\sigma}u_{\lambda}\right ) j^{\lambda},\end{aligned}\ ] ] from which the result follows due to ( [ inversea ] ) . finally , with the help of the standard substitution,@xmath160 ( in view of @xmath161 we arrive at@xmath162 z^{\lambda\sigma}+4\pi\mu p^{\lambda\sigma}\right\ } = 0 . \label{dzphertz}\ ] ] therefore , one can choose@xmath159 z^{\lambda\nu}=-4\pi p^{\lambda\nu}. \label{zphertz}\ ] ] here , by definition,@xmath163{cccc}0 & -p_{1 } & -p_{2 } & -p_{3}\\ p_{1 } & 0 & m_{3 } & -m_{2}\\ p_{2 } & -m_{3 } & 0 & m_{1}\\ p_{3 } & m_{2 } & -m_{1 } & 0 \end{array } \right ) \label{ptensor}\ ] ] is an anti - symmetric four - tensor @xcite , @xcite , @xcite . the electric and magnetic hertz vectors , @xmath164 and @xmath165 are also introduced in terms of a single four - tensor,@xmath166{cccc}0 & z_{1}^{\left ( e\right ) } & z_{2}^{\left ( e\right ) } & z_{3}^{\left ( e\right ) } \\ -z_{1}^{\left ( e\right ) } & 0 & -z_{3}^{\left ( m\right ) } & z_{2}^{\left ( m\right ) } \\ -z_{2}^{\left ( e\right ) } & z_{3}^{\left ( m\right ) } & 0 & -z_{1}^{\left ( m\right ) } \\ -z_{3}^{\left ( e\right ) } & -z_{2}^{\left ( m\right ) } & z_{1}^{\left ( m\right ) } & 0 \end{array } \right ) . \label{ztensor}\ ] ] in view of ( [ azhertz ] ) , for the four - vector potential , @xmath167 we obtain@xmath168 and@xmath169 .\nonumber\end{aligned}\ ] ] then , equations ( [ zphertz ] ) take the form@xmath159 \mathbf{z}^{\left ( e\right ) } = 4\pi\mu\mathbf{p},\qquad\left [ \partial^{\tau}\partial_{\tau}+\kappa\left ( u^{\tau}\partial_{\tau}\right ) ^{2}\right ] \mathbf{z}^{\left ( m\right ) } = 4\pi \mu\mathbf{m } \label{zempm}\ ] ] and , for the four - current , @xmath170 one gets@xmath171 ( see @xcite , @xcite , @xcite for more details ) . the hertz vector and tensor potentials , for a moving medium and at rest , were utilized in @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite ( see also the references therein ) . many classical problems of radiation and propagation can be consistently solved by using these potentials . in the case of the covariant version of the energy - momentum tensor given by ( [ fourenergymomentumtensor ] ) , the differential balance equations under consideration are independent of the particular choice of the frame of reference . therefore , our relations ( [ qpmaterialf ] ) are useful for derivation of the expressions for the energy - momentum tensor and the ponderomotive force for moving bodies from those for bodies at rest which were extensively studied in the literature . for example , one gets@xmath172 with the help of ( [ covariantmaterialeqs])([fourpermeabilitytensor ] ) and ( [ b10 ] ) ( see also @xcite ) . in modern presentations of the classical and quantum field theories , the lagrangian approach is usually utilized . we introduce two quadratic lagrangian densities@xmath173 and@xmath174 then , by formal differentiation,@xmath175 and@xmath176 in view of ( [ b3 ] ) . the complex covariant maxwell equations ( [ covariantmaxwell ] ) take the forms@xmath177 and the covariant energy - momentum balance relations ( [ covariantenergymomentumbalance ] ) are given by@xmath178 \label{energymomentumlagrangian0}\\ & \quad+\frac{1}{32\pi}\left [ p_{\sigma\tau}^{\ast}\frac{\partial}{\partial x^{\mu}}\left ( \frac{\partial\mathcal{l}_{0}}{\partial p_{\sigma\tau}}\right ) + p_{\sigma\tau}\frac{\partial}{\partial x^{\mu}}\left ( \frac{\partial\mathcal{l}_{0}}{\partial p_{\sigma\tau}^{\ast}}\right ) \right ] = -\frac{1}{c}f_{\mu\lambda}j^{\lambda}\nonumber\end{aligned}\ ] ] and@xmath179 \label{energymomentumlagrangian1}\\ & \quad+\frac{1}{32\pi}\left [ p_{\sigma\tau}\frac{\partial}{\partial x^{\mu } } \left ( \frac{\partial\mathcal{l}_{1}}{\partial p_{\sigma\tau}}\right ) + p_{\sigma\tau}^{\ast}\frac{\partial}{\partial x^{\mu}}\left ( \frac { \partial\mathcal{l}_{1}^{\ast}}{\partial p_{\sigma\tau}^{\ast}}\right ) \right ] = \frac{1}{c}f_{\mu\lambda}j^{\lambda}\nonumber\end{aligned}\ ] ] in terms of the complex lagrangians under consideration , respectively . finally , with the help of the following densities,@xmath180 one can derive analogs of the euler - lagrange equations for electromagnetic fields in media:@xmath181 in the case of a moving isotropic medium , a relation between @xmath182 and @xmath183 is given by our equations ( [ qpmaterialf])([4vectorpotential ] ) . taking the real and imaginary parts , maxwell s equations ( [ maxwellii ] ) can be written as follows@xmath184 here,@xmath185 with the help of ( [ qpidentity ] ) and ( [ 4vectorpotential ] ) . thus the second set of equations is automatically satisfied when we introduce the four - vector potential . for the inhomogeneous pair of maxwell s equations , the lagrangian density is given by@xmath186 in view of ( [ covariantmaterialeqs ] ) . then , for conjugate momenta to the four - potential field @xmath187 one gets@xmath188 and the corresponding euler - lagrange equations take a familiar form@xmath189 the latter equation can also be derived with the help of the least action principle @xcite , @xcite , @xcite . the corresponding hamiltonian and quantization are discussed in @xcite , @xcite , @xcite among other classical accounts . in conclusion , it is worth noting the role of complex fields in quantum electrodynamics , quadratic invariants and quantization ( see , for instance , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite ) . the classical and quantum theory of cherenkov radiation is reviewed in @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite . for paraxial approximation in optics , see @xcite , @xcite , @xcite , @xcite , @xcite , and the references therein . maxwell s equations in the gravitational field are discussed in @xcite @xcite . one may hope that our detailed mathematical consideration of several aspects of macroscopic electrodynamics will be useful for future investigations and pedagogy . * acknowledgements*. we dedicate this article to the memory of professor alladi ramakrishnan who made significant contributions to probability and statistics , elementary particle physics , cosmic rays and astrophysics , matrix theory , and the special theory of relativity @xcite . we are grateful to krishna alladi , albert boggess , mark p. faifman , john r. klauder , vladimir i. manko , and igor n. toptygin for valuable comments and help . among useful differential relations are@xmath190@xmath191@xmath192@xmath193@xmath194 ( see also @xcite , @xcite and @xcite . ) here , @xmath195 and @xmath196 in this article , @xmath75 and @xmath76 is the levi - civita four - symbol @xcite with familiar contractions:@xmath197{ccc}\delta_{\kappa}^{\nu } & \delta_{\lambda}^{\nu } & \delta_{\rho}^{\nu}\\ \delta_{\kappa}^{\sigma } & \delta_{\lambda}^{\sigma } & \delta_{\rho}^{\sigma } \\ \delta_{\kappa}^{\tau } & \delta_{\lambda}^{\tau } & \delta_{\rho}^{\tau}\end{array } \right\vert , \label{b01}\]]@xmath198{cc}\delta_{\lambda}^{\sigma } & \delta_{\rho}^{\sigma}\\ \delta_{\lambda}^{\tau } & \delta_{\rho}^{\tau}\end{array } \right\vert = -2\left ( \delta_{\lambda}^{\sigma}\delta_{\rho}^{\tau}-\delta_{\rho}^{\sigma}\delta_{\lambda}^{\tau}\right ) , \label{b02}\]]@xmath199 dual second rank four - tensor identities are given by @xcite:@xmath200 in particular,@xmath201@xmath202@xmath203@xmath204@xmath205 by direct calculation,@xmath206@xmath207 as a result,@xmath208 an important decomposition,@xmath209 is complemented by an identity,@xmath210 in matrix form,@xmath211 here,@xmath212@xmath213@xmath214@xmath215 where @xmath216diag@xmath217 is the identity matrix . also,@xmath218 and@xmath219 other useful dual four - tensor identities are given by @xcite : @xmath220 in particular,@xmath221 and@xmath222 ( see also @xcite ) . in view of ( [ pqdual ] ) , or ( [ b3 ] ) , and ( [ b22 ] ) , we can write@xmath223 or@xmath224 by ( [ b3 ] ) . therefore,@xmath225 in addition , with the help of ( [ b3 ] ) one gets@xmath226 which completes the proof . b. acosta - humnez , s. i. kryuchkov , e. suazo , and s. k. suslov , _ degenerate parametric amplification of squeezed photons : explicit solutions , statistics , means and variances _ , j. nonlinear opt . phys . mater . * 24 * ( 2015 ) # 2 , 1550021 ( 27pp ) . biaynicki - birula i and biaynicki - birula z _ the role of riemann - silberstein vector in classical and quantum theories of electromagnetism _ , j. phys . _ _ * 46 * ( 2013 ) , 053001 ( 32 pp ) . b. m. bolotovskii and s. n. stolyarov , _ current status of the electrodynamics of moving media ( infinite media ) _ , sov . phys.usp . * 17 * ( 1975 ) # 6 , 875895 ; see also a revised version with the annotated bibliography in : _ einstein s collection 1974 _ , pp . 179275 , nauka , moscow , 1976 [ in russian ] . v. v. dodonov and v. i. manko , _ invariants and correlated states of nonstationary quantum systems _ , in : _ invariants and the evolution of nonstationary quantum systems _ , proceedings of lebedev physics institute , vol . 71181 , nauka , moscow , 1987 [ in russian ] ; english translation published by nova science , commack , new york , 1989 , pp . 103261 . a. einstein and j. laub , _ ber die elektromagnetischen grundgleichungen fr bewegte krper _ , ann . phys . * 26 * ( 1908 ) , 532540 ; _ bemerkungen zu unserer arbeit @xmath227@xmath227ber die elektromagnetischen grundgleichungen fr bewegte krper@xmath228@xmath228 _ , ann * 28 * ( 1908 ) , 445447 . v. l. ginzburg , _ the laws of conservation of energy and momentum in emission of electromagnetic waves ( photons ) in a medium and the energy - momentum tensor in macroscopic electrodynamics _ , sov . phys.usp . * 16 * ( 1973 ) # 3 , 434439 . w. israel and j. m. steward , _ progress in relativistic thermodynamics and electrodynamics of continous media _ , in : _ general relativity and gravitation ( one hundred years after the birth of albert einstein ) _ ( a. held , ed . ) , vol . 2 , pp . 491525 , plenum press , new york , 1980 . s. i. kryuchkov , s. k. suslov , and j. m. vega - guzmn , _ the minimum - uncertainty squeezed states for atoms and photons in a cavity _ , j. phys . b : at . mol . * 46 * ( 2013 ) , 104007 ( 15pp ) ; iop select and highlight of 2013 . h. a. lorentz , a. einstein , and h. minkowski , _ das relativittsprinzip , eine sammlung von abhandlungen _ , teubner , leipzig , berlin , 1913 ; english translation in : a. einstein et al , _ _ the principle of relativity , a collection of original papers on the special and general theory of relativity _ _ , 4th edn , dover , new york , 1932 . h. minkowski , _ die grundlagen fr die elektromagnetischen vorgnge in bewegten krpern _ . knig . ges . gttingen , math .- phys . ( 1908 ) , 53111 ; _ the fundamental equations for electromagnetic processes in moving bodies _ , english translation in : _ the principle of relativity _ ( m. saha , ed . ) , university press , calcutta , 169 , 1920 .

we demonstrate that the processes underlying on - line auction price bids and many other longitudinal data can be represented by an empirical first order stochastic ordinary differential equation with time - varying coefficients and a smooth drift process . this equation may be empirically obtained from longitudinal observations for a sample of subjects and does not presuppose specific knowledge of the underlying processes . for the nonparametric estimation of the components of the differential equation , it suffices to have available sparsely observed longitudinal measurements which may be noisy and are generated by underlying smooth random trajectories for each subject or experimental unit in the sample . the drift process that drives the equation determines how closely individual process trajectories follow a deterministic approximation of the differential equation . we provide estimates for trajectories and especially the variance function of the drift process . at each fixed time point , the proposed empirical dynamic model implies a decomposition of the derivative of the process underlying the longitudinal data into a component explained by a linear component determined by a varying coefficient function dynamic equation and an orthogonal complement that corresponds to the drift process . an enhanced perturbation result enables us to obtain improved asymptotic convergence rates for eigenfunction derivative estimation and consistency for the varying coefficient function and the components of the drift process . we illustrate the differential equation with an application to the dynamics of on - line auction data.=-1 and . .

1211.2630

recently , there has been increasing interest in analyzing on - line auction data and in inferring the underlying dynamics that drive the bidding process . each series of price bids for a given auction corresponds to pairs of random bidding times and corresponding bid prices generated whenever a bidder places a bid [ jank and shmueli ( @xcite , @xcite ) , @xcite , @xcite ] . related longitudinal data where similar sparsely and irregularly sampled noisy measurements are obtained are abundant in the social and life sciences ; for example , they arise in longitudinal growth studies . while more traditional approaches of functional data analysis require fully or at least densely observed trajectories [ @xcite , @xcite , @xcite ] , more recent extensions cover the case of sparsely observed and noise - contaminated longitudinal data [ @xcite , @xcite ] . a common assumption of approaches for longitudinal data grounded in functional data analysis is that such data are generated by an underlying smooth and square integrable stochastic process [ @xcite , @xcite , @xcite , @xcite , @xcite ] . the derivatives of the trajectories of such processes are central for assessing the dynamics of the underlying processes [ @xcite , @xcite ] . although this is difficult for sparsely recorded data , various approaches for estimating derivatives of individual trajectories nonparametrically by pooling data from samples of curves and using these derivatives for quantifying the underlying dynamics have been developed [ @xcite , @xcite , @xcite , @xcite ] . related work on nonparametric methods for derivative estimation can be found in @xcite , @xcite and on the role of derivatives for the functional linear model in @xcite . we expand here on some of these approaches and investigate an empirical dynamic equation . this equation is distinguished from previous models that involve differential equations in that it is empirically determined from a sample of trajectories , and does not presuppose knowledge of a specific parametric form of a differential equation which generates the data , except that we choose it to be a first order equation . this stands in contrast to current approaches of modeling dynamic systems , which are `` parametric '' in the sense that a prespecified differential equation is assumed . a typical example for such an approach has been developed by @xcite , where a prior specification of a differential equation is used to guide the modeling of the data , which is done primarily for just one observed trajectory . a problem with parametric approaches is that diagnostic tools to determine whether these equations fit the data either do not exist , or where they do , are not widely used , especially as nonparametric alternatives to derive differential equations have not been available . this applies especially to the case where one has data on many time courses available , providing strong motivation to explore nonparametric approaches to quantify dynamics . our starting point is a nonparametric approach to derivative estimation by local polynomial fitting of the derivative of the mean function and of partial derivatives of the covariance function of the process by pooling data across all subjects [ @xcite ] . we show that each trajectory satisfies a first order stochastic differential equation where the random part of the equation resides in an additive smooth drift process which drives the equation ; the size of the variance of this process determines to what extent the time evolution of a specific trajectory is determined by the nonrandom part of the equation over various time subdomains , and therefore is of tantamount interest . we quantify the size of the drift process by its variance as a function of time . whenever the variance of the drift process @xmath0 is small relative to the variance of the process @xmath1 , a deterministic version of the differential equation is particularly useful as it then explains a large fraction of the variance of the process . the empirical stochastic differential equation can be easily obtained for various types of longitudinal data . this approach thus provides a novel perspective to assess the dynamics of longitudinal data and permits insights about the underlying forces that shape the processes generating the observations , which would be hard to obtain with other methods . we illustrate these empirical dynamics by constructing the stochastic differential equations that govern online auctions with sporadic bidding patterns . we now describe a data model for longitudinally collected observations , which reflects that the data consist of sparse , irregular and noise - corrupted measurements of an underlying smooth random trajectory for each subject or experimental unit [ @xcite ] , the dynamics of which is of interest . given @xmath2 realizations @xmath3 of the underlying process @xmath1 on a domain @xmath4 and @xmath5 of an integer - valued bounded random variable @xmath6 , we assume that @xmath5 measurements @xmath7 , @xmath8 , are obtained at random times @xmath9 , according to @xmath10 where @xmath11 are zero mean i.i.d . measurement errors , with @xmath12 , independent of all other random components . the paper is organized as follows . in section [ sec2 ] , we review expansions in eigenfunctions and functional principal components , which we use directly as the basic tool for dimension reduction alternative implementations with b - splines or p - splines could also be considered [ @xcite , @xcite , @xcite ] . we also introduce the empirical stochastic differential equation and discuss the decomposition of variance it entails . asymptotic properties of estimates for the components of the differential equation , including variance function of the drift process , coefficient of determination associated with the dynamic system and auxiliary results on improved rates of convergence for eigenfunction derivatives are the theme of section [ sec3 ] . background on related perturbation results can be found in @xcite , @xcite , @xcite , @xcite . section [ sec4 ] contains the illustration of the differential equation with auction data , followed by a brief discussion of some salient features of the proposed approach in section [ sec5 ] . additional discussion of some preliminary formulas is provided in appendix [ seca1 ] , estimation procedures are described in appendix [ seca2 ] , assumptions and auxiliary results are in appendix [ seca3 ] and proofs in appendix [ seca4 ] . @xmath13a key methodology for dimension reduction and modeling of the underlying stochastic processes @xmath1 that generate the longitudinal data , which usually are sparse , irregular and noisy as in ( [ kl2 ] ) , is functional principal component analysis ( fpca ) . processes are assumed to be square integrable with mean function @xmath14 and auto - covariance function @xmath15 , @xmath16 , which is smooth , symmetric and nonnegative definite . using @xmath17 as kernel in a linear operator leads to the hilbert schmidt operator @xmath18 . we denote the ordered eigenvalues ( in declining order ) of this operator by @xmath19 and the corresponding orthonormal eigenfunctions by @xmath20 . we assume that all eigenvalues are of multiplicity @xmath21 in the sequel . it is well known that the kernel @xmath17 has the representation @xmath22 and the trajectories generated by the process satisfy the karhunen love representation [ @xcite ] @xmath23 . here the @xmath24 , @xmath25 @xmath26 , are the functional principal components ( fpcs ) of the random trajectories @xmath3 . the @xmath27 are uncorrelated random variables with @xmath28 and @xmath29 , with @xmath30 . upon differentiating both sides , one obtains @xmath31 where @xmath32 and @xmath33 are the derivatives of mean and eigenfunctions . the eigenfunctions @xmath34 are the solutions of the eigen - equations @xmath35 , under the constraint of orthonormality . under suitable regularity conditions , one observes @xmath36\\[-8pt ] \phi_k^{(1)}(t ) & = & \frac{1}{\lambda_k}\int_{{\mathcal{t } } } \frac{\partial}{\partial t}\,g(t , s)\phi_k(s)\,ds,\nonumber\end{aligned}\ ] ] which motivates corresponding eigenfunction derivative estimates . a useful representation is @xmath37\\[-8pt ] \eqntext{\nu_1 , \nu_2 \in\{0,1\ } , s , t \in{\mathcal{t}},}\end{aligned}\ ] ] which is an immediate consequence of the basic properties of the functional principal components @xmath27 . for more details and discussion , we refer to appendix [ seca1 ] . it is worthwhile to note that the representation ( [ kl3 ] ) does not correspond to the karhunen love representation of the derivatives , which would be based on orthonormal eigenfunctions of a linear hilbert schmidt operator defined by the covariance kernel @xmath38 . a method to obtain this representation might proceed by first estimating @xmath39 using ( [ rep ] ) for @xmath40 and suitable estimates @xmath41 for eigenfunction derivatives , then directly decomposing @xmath39 into eigenfunctions and eigenvalues . this leads to @xmath42 and the karhunen love representation @xmath43 , with orthonormal eigenfunctions @xmath44 [ @xcite ] . in the following we consider differentiable gaussian processes , for which the differential equation introduced below automatically applies . in the absence of the gaussian assumption , one may invoke an alternative least squares - type interpretation . gaussianity of the processes implies the joint normality of centered processes @xmath45 at all points @xmath46 , so that @xmath47 this joint normality immediately implies a `` population '' differential equation of the form @xmath48 , as has been observed in @xcite ; for additional details see appendix [ seca1 ] . however , it is considerably more interesting to find a dynamic equation which applies to the individual trajectories of processes @xmath1 . this goal necessitates inclusion of a stochastic term which leads to an empirical stochastic differential equation that governs the dynamics of individual trajectories @xmath3 . [ thm1 ] for a differentiable gaussian process , it holds that @xmath49 where @xmath50\\[-8pt ] & = & \frac{1}{2}\frac{d}{dt}\log[{\operatorname{var}}\{x(t)\}],\qquad t\in{\mathcal{t}},\nonumber\end{aligned}\ ] ] and @xmath0 is a gaussian process such that @xmath51 are independent at each @xmath52 and where @xmath0 is characterized by @xmath53 and @xmath54 , with @xmath55\\[-8pt ] & & { } - \beta(s)\sum_{k=1}^\infty\lambda_k\phi^{(1)}_k(t ) \phi_k(s ) + \beta(t)\beta(s)\sum_{k=1}^\infty\lambda_k\phi_k(t ) \phi_k(s).\nonumber\end{aligned}\ ] ] equation ( [ de ] ) provides a first order linear differential equation which includes a time - varying linear coefficient function @xmath56 and a random drift process @xmath57 . the process @xmath0 `` drives '' the equation at each time @xmath58 . it is square integrable and possesses a smooth covariance function and smooth trajectories . it also provides an alternative characterization of the individual trajectories of the process . the size of its variance function @xmath59 determines the importance of the role of the stochastic drift component . we note that the assumption of differentiability of the process @xmath1 in theorem [ thm1 ] can be relaxed . it is sufficient to require weak differentiability , assuming that @xmath60 , where @xmath61 denotes the sobolev space of square integrable functions with square integrable weak derivative [ @xcite ] . along these lines , equation ( [ de ] ) may be interpreted as a stochastic sobolev embedding . observe also that the drift term @xmath0 can be represented as an integrated diffusion process . upon combining ( [ kl3 ] ) and ( [ de ] ) , and observing that functional principal components can be represented as @xmath62 , where @xmath63 is the @xmath64th eigenfunction of the wiener process @xmath65 on domain @xmath66 $ ] and @xmath67 the associated eigenvalue , such a representation is given by @xmath68 another observation is that the joint normality in ( [ jt ] ) can be extended to joint normality for any finite number of derivatives , assuming these are well defined . therefore , higher order stochastic differential equations can be derived analogously to ( [ de ] ) . however , these higher - order analogues are likely to be much less relevant practically , as higher - order derivatives of mean and eigenfunctions can not be well estimated for the case of sparse noisy data or even denser noisy data . finally , it is easy to see that the differential equation ( [ de ] ) is equivalent to the following stochastic integral equation : @xmath69 in the sense that @xmath1 is the solution of both equations . for a domain with left endpoint at time @xmath70 , setting @xmath71 in ( [ int ] ) then defines a classical initial value problem . given a trajectory of the drift process @xmath0 and a varying coefficient function @xmath72 , one may obtain a solution for @xmath1 numerically by euler or runga kutta integration or directly by applying the known solution formula for the initial value problem of an inhomogeneous linear differential equation . we note that equations ( [ de ] ) and ( [ int ] ) are of particular interest on domains @xmath4 or subdomains defined by those times @xmath58 for which the variance function @xmath73 is `` small . '' from ( [ beta ] ) and ( [ gz ] ) one finds @xmath74 ^ 2}\bigr)/{\operatorname{var}}\{x(t)\}\nonumber\\[-8pt]\\[-8pt ] & = & \biggl(\sum_{k=1}^\infty\lambda_k\bigl(\phi_k^{(1)}(t)\bigr)^2 \sum_{k=1}^\infty\lambda_k \phi_k^2(t ) - \biggl\{\sum_{k=1}^\infty\lambda_k\phi_k^{(1)}(t ) \phi_k(t ) \biggr\}^2 \biggr)\nonumber\\ & & { } \bigg/\sum_{k=1}^\infty \lambda_k\phi_k^2(t).\nonumber\end{aligned}\ ] ] on subdomains with small variance function , the solutions of ( [ de ] ) will not deviate too much from the solutions of the approximating equation @xmath75 in this situation , the future changes in value of individual trajectories are highly predictable and the interpretations of the dynamic behavior of processes @xmath1 obtained from the shape of the varying coefficient function @xmath56 apply at the individual level . if @xmath76 , the dynamic behavior can be characterized as `` dynamic regression to the mean '' ; a trajectory which is away from ( above or below ) the mean function @xmath77 at time @xmath58 is bound to move closer toward the mean function @xmath77 as time progresses beyond @xmath58 . similarly , if @xmath78 , trajectories will exhibit `` explosive '' behavior , since deviations from the mean ( above or below ) at time @xmath58 will be reinforced further as time progresses , so that trajectories are bound to move further and further away from the population mean trajectory . intermediate cases arise when the function @xmath72 changes sign , in which case the behavior will switch between explosive and regression to the mean , depending on the time subdomain . another situation occurs on subdomains where both @xmath72 and @xmath59 are very small , in which case the deviation of the derivative of an individual trajectory from the population mean derivative will also be small which means that trajectory derivatives will closely track the population mean derivative on such subdomains . the independence of @xmath57 and @xmath79 means that the right - hand side of ( [ de ] ) provides an orthogonal decomposition of @xmath80 into the two components @xmath81 and @xmath57 such that @xmath82 it is therefore of interest to determine the fraction of the variance of @xmath80 that is explained by the differential equation itself , that is , the `` coefficient of determination '' @xmath83 which is seen to be equivalent to the squared correlation between @xmath84 , @xmath85 ^ 2}{{\operatorname{var}}\{x(t)\}{\operatorname{var}}\ { x^{(1)}(t)\}}= \frac{\{\sum_{k=1}^\infty\lambda_k\phi_k^{(1)}(t ) \phi_k(t)\}^2}{\sum_{k=1}^\infty \lambda_k\phi_k(t)^2 \sum_{k=1}^\infty \lambda_k\phi_k^{(1)}(t)^2}.\ ] ] we are then particularly interested in subdomains of @xmath4 where @xmath86 is large , say , exceeds a prespecified threshold of 0.8 or 0.9 . on such subdomains the drift process @xmath0 is relatively small compared to @xmath80 so that the approximating deterministic first order linear differential equation ( [ approx ] ) can substitute for the stochastic dynamic equation ( [ de ] ) . in this case , short - term prediction of @xmath87 may be possible for small @xmath88 , by directly perusing the approximating differential equation ( [ approx ] ) . it is instructive to visualize an example of the function @xmath86 for the case of fully specified eigenfunctions and eigenvalues . assuming that the eigenfunctions correspond to the trigonometric orthonormal system @xmath89 on @xmath90 $ ] , we find from ( [ r2 ] ) @xmath91 ^ 2\\ & & { } \big/ \bigl[\sum\lambda_k ( \cos ( 2k\pi t))^2 \sum\lambda_k k ( \sin(2k\pi t))^2 \bigr],\qquad t \in [ 0,1].\end{aligned}\ ] ] choosing @xmath92 and numerically approximating these sums , one obtains the functions @xmath86 as depicted in figure [ rfig ] . this illustration shows ( [ r1 ] ) , ( [ r2 ] ) , quantifying the fraction of variance explained by the deterministic part of the dynamic equation ( [ de ] ) , illustrated for the trigonometric basis @xmath93 on @xmath90 $ ] and eigenvalue sequences @xmath94 ( solid ) and @xmath95 ( dashed ) . ] that the behavior of this function often will fluctuate between small and large values and also depends critically on both the eigenvalues and the shape of the eigenfunctions . we obtain asymptotic consistency results for estimators of the varying coefficient functions @xmath72 , for the variance function @xmath73 of the drift process and for the variance explained at time @xmath58 by the deterministic part ( [ approx ] ) of the stochastic equation ( [ de ] ) , quantified by @xmath86 . corresponding estimators result from plugging in estimators for the eigenvalues @xmath96 , eigenfunctions @xmath34 and eigenfunction derivatives @xmath41 into the representations ( [ beta ] ) for the function @xmath56 , ( [ v ] ) for the variance function of @xmath0 and ( [ r2 ] ) for @xmath86 . here one needs to truncate the expansions at a finite number @xmath97 of included eigen - components . details about the estimation procedures , which are based on local linear smoothing of one- and two - dimensional functions , are deferred to appendix [ seca2 ] . our asymptotic consistency results focus on @xmath98 convergence rates . they peruse auxiliary results on the convergence of estimates of eigenvalues , eigenfunctions and eigenfunction derivatives , complementing and improving upon related results of @xcite , which were derived for convergence in the sup norm . improved rates of convergence in the @xmath98 distance are the consequence of a special decomposition that we employ in the proofs to overcome the difficulty caused by the dependence of the repeated measurements . required regularity conditions include assumptions for the distribution of the design points , behavior of eigenfunctions @xmath34 and eigenvalues @xmath96 as their order @xmath64 increases and the large sample behavior of the bandwidths @xmath99 for the estimation of the mean function @xmath77 and its first derivative @xmath32 , and @xmath100 for the estimation of the covariance surface and its partial derivative . we note that extremely sparse designs are covered , with only two measurements per trajectory ; besides being bounded , the number of measurements @xmath5 for the @xmath101th trajectory is required to satisfy @xmath102 . specifically , for the observations @xmath103 , @xmath104 , @xmath105 , made for the @xmath101th trajectory , we require that : 1 . @xmath5 are random variables with @xmath106 , where @xmath6 is a bounded positive discrete random variable with and @xmath107 , and @xmath108 are independent of @xmath5 , for @xmath109 . writing @xmath110 and @xmath111 , the triples @xmath112 are assumed to be i.i.d . for the bandwidths used in the smoothing steps for @xmath113 and @xmath32 in ( [ smooth1 ] ) , @xmath114 and @xmath115 in ( [ smooth2 ] ) , we require that , as @xmath116 , 1 . @xmath117 , @xmath118 , @xmath119 , @xmath120 , @xmath121 . to characterize the behavior of estimated eigenfunction derivatives @xmath122 , define @xmath123 for the kernels used in the local linear smoothing steps and underlying density and moment functions , we require assumptions ( b1 ) and ( b2 ) in the . denote the @xmath98 norm by @xmath124 , the hilbert schmidt norm by @xmath125 and also define @xmath126 . the following result provides asymptotic rates of convergence in the @xmath98 norm for the auxiliary estimates of mean functions and their derivatives as well as covariance functions and their partial derivatives , which are briefly discussed in appendix [ seca2 ] . a consequence is a convergence result for the eigenfunction derivative estimates @xmath127 , with constants and rates that hold uniformly in the order @xmath128 . [ thm2 ] under and and , for @xmath129 , @xmath130\\[-8pt ] \bigl\|\hat{g}^{(\nu,0)}-g^{(\nu,0)}\bigr\|_s & = & o_p \biggl(\frac{1}{\sqrt { n}h_{g,\nu}^{\nu+1}}+h_{g,\nu}^2 \biggr).\nonumber\end{aligned}\ ] ] for @xmath131 corresponding to @xmath96 of multiplicity @xmath21 , @xmath132\\[-8pt ] & & \qquad = o_p \biggl(\frac { 1}{\lambda_k } \biggl\{\frac{1}{\sqrt{n}h_{g , 1}^{2}}+h_{g,1}^2+\frac{1}{\delta_k } \biggl(\frac{1}{\sqrt{n}h_{g , 0}}+h_{g,0}^2 \biggr ) \biggr\ } \biggr),\nonumber\end{aligned}\ ] ] where the @xmath133 term in ( [ thm2-eq2 ] ) is uniform in @xmath128 . an additional requirement is that variances of processes @xmath1 and @xmath134 are bounded above and below , since these appear in the denominators of various representations , for example , in ( [ v ] ) and ( [ r2 ] ) , 1 . @xmath135 and @xmath136 for @xmath137 , implying that @xmath138 by the cauchy schwarz inequality . define remainder terms @xmath139 by the cauchy schwarz inequality , @xmath140 . in order to obtain consistent estimates of various quantities , a necessary requirement is that the first @xmath141 eigen - terms approximate the infinite - dimensional process sufficiently well . the increase in the sequence @xmath97 as @xmath142 therefore needs to be tied to the spacing and decay of eigenvalues , 1 . @xmath143 k&=&o\bigl(\min\bigl\{\sqrt{n}h_{g,1}^{2 } , h_{g,1}^{-2}\bigr\}\bigr),\\ \sum_{k=1}^k \delta_k^{-1}&=&o\bigl(\min\bigl(\sqrt{n}h_{g,0 } , h_{g,0}^{-2}\bigr\}\bigr),\\ \max_{\nu=0 , 1}\|r_{k , \nu}\|&\rightarrow&0\qquad\mbox{as $ n\rightarrow \infty$.}\end{aligned}\ ] ] if the eigenvalues decrease rapidly and merely a few leading terms are needed , condition ( a4 ) is easily satisfied . we use `` @xmath144 '' to connect two terms which are asymptotically of the same order in probability , that is , the terms are @xmath145 of each other . define the sequence @xmath146 note that @xmath147 with corresponding plug - in estimate @xmath148 , where @xmath97 is the included number of eigenfunctions . the plug - in estimate for @xmath56 is based on ( [ beta ] ) and given by @xmath149 and analogously the plug - in estimate @xmath150 of @xmath151 is based on representation ( [ gz ] ) , using the estimate @xmath152 . in a completely analogous fashion one obtains the estimates @xmath153 of @xmath86 from ( [ r2 ] ) and @xmath154 of the variance function @xmath155 of the drift process from ( [ v ] ) . the @xmath98 convergence rates of these estimators of various components of the dynamic model ( [ de ] ) are given in the following result . [ thm3 ] under and , @xmath156\\[-8pt ] \bigl\|\hat{g}_k^{(1,1)}-g^{(1,1)}\bigr\|_u&=&o_p(\alpha_n+\|r_{k,1}\| ) , \nonumber\\ \|\hat{\beta}_k-\beta\| & \stackrel{p}{\asymp } & \|\hat{g}_{z , k}- g_z\|_s \stackrel{p}{\asymp } \|\hat{g}_{z , k}-g_z\|_u\nonumber\\ \label{thm3-eq2 } & \stackrel{p}{\asymp } & \|\hat{r}^2_k - r^2\| \stackrel{p}{\asymp } \|\hat{v}_k - v\| \\ & = & o_p(\alpha_n+\|r_{k,0}\|+\|r_{k,1}\|).\nonumber\end{aligned}\ ] ] the weak convergence and @xmath98 consistency for the estimated eigenvalues @xmath157 and eigenfunctions @xmath158 of the drift process @xmath0 is an immediate consequence of this result . to see this , one may use @xmath159 where @xmath160 is any estimate of @xmath151 [ @xcite ] . here the @xmath133 terms are uniform in @xmath64 and @xmath161 for @xmath162 . to illustrate our methods , we analyze the dynamic system corresponding to online auction data , specifically using ebay bidding data for 156 online auctions of palm personal digital assistants in 2003 ( courtesy of wolfgang jank ) . the data are `` live bids '' that are entered by bidders at irregular times and correspond to the actual price a winning bidder would pay for the item . this price is usually lower than the `` willingness - to - pay '' price , which is the value a bidder enters . further details regarding the proxy bidding mechanism for the 7-day second - price auction design that applies to these data can be found in jank and shmueli ( @xcite , @xcite ) , liu and mller ( @xcite , @xcite ) . the time unit of these 7-day auctions is hours and the domain is the interval @xmath163 $ ] . adopting the customary approach , the bid prices are log - transformed prior to the analysis . the values of the live bids @xmath7 are sampled at bid arrival times @xmath9 , where @xmath164 refers to the auction index and @xmath165 to the total number of bids submitted during the @xmath101th auction ; the number of bids per auction is found to be between 6 and 49 for these data . we adopt the point of view that the observed bid prices result from an underlying price process which is smooth , where the bids themselves are subject to small random aberrations around underlying continuous trajectories . since there is substantial variability of little interest in both bids and price curves during the first three days of an auction , when bid prices start to increase rapidly from a very low starting point to more realistic levels , we restrict our analysis to the interval [ @xmath166 ( in hours ) , thus omitting the first three days of bidding . this allows us to focus on the more interesting dynamics in the price curves taking place during the last four days of these auctions . our aim is to explore the price dynamics through the empirical stochastic differential equation ( [ de ] ) . our study emphasizes description of the dynamics over prediction of future auction prices and consists of two parts : a description of the dynamics of the price process at the `` population level '' which focuses on patterns and trends in the population average and is reflected by dynamic equations for conditional expectations . the second and major results concern the quantification of the dynamics of auctions at the individual or `` auction - specific level '' where one studies the dynamic behavior for each auction separately , but uses the information gained across the entire sample of auctions . only the latter analysis involves the stochastic drift term @xmath0 in the stochastic differential equation ( [ de ] ) . we begin by reviewing the population level analysis , which is characterized by the deterministic part of ( [ de ] ) , corresponding to the equation @xmath167 . this equation describes a relationship that holds for conditional means but not necessarily for individual trajectories . for the population level analysis , we require estimates of the mean price curve @xmath77 and its first derivative @xmath168 , and these are obtained by applying linear smoothers to ( [ smooth1 ] ) to the pooled scatterplots that are displayed in figure [ auc - mu ] ( for more details , see appendix [ seca2 ] ) . one finds that both log prices and log price derivatives are increasing throughout , so that at the log - scale the price increases are accelerating in the mean as the auctions proceed . ( left panel ) and of their derivatives ( right panel ) , @xmath169 ( solid ) , @xmath170 ( dashed ) and @xmath171 ( dash - dotted ) . ] a second ingredient for our analysis are estimates for the eigenfunctions and eigenvalues ( details in appendix [ seca2 ] ) . since the first three eigenfunctions were found to explain 84.3% , 14.6% and 1.1% of the total variance , three components were selected . the eigenfunction estimates are shown in the left panel of figure [ auc - xeig ] , along with the estimates of the corresponding eigenfunction derivatives in the right panel . for the interpretation of the eigenfunctions it is helpful to note that the sign of the eigenfunctions is arbitrary . we also note that variation in the direction of the first eigenfunction @xmath172 corresponds to the major part of the variance . the variances @xmath173 that are attributable to this eigenfunction are seen to steadily decrease as @xmath58 is increasing , so that this eigenfunction represents a strong trend of higher earlier and smaller later variance in the log price trajectories . the contrast between large variance of the trajectories at earlier times and smaller variances later reflects the fact that auction price trajectories are less determined early on when both relatively high as well as low prices are observed , while at later stages prices differ less as the end of the auction is approached and prices are constrained into a narrower range . correspondingly , the first eigenfunction derivative is steadily increasing ( decreasing if the sign is switched ) , with notably larger increases ( decreases ) both at the beginning and at the end and a relatively flat positive plateau in the middle part . the second eigenfunction corresponds to a contrast between trajectory levels during the earlier and the later part of the domain , as is indicated by its steady increase and the sign change , followed by a slight decrease at the very end . this component thus reflects a negative correlation between early and late log price levels . the corresponding derivative is positive and flat , with a decline and negativity toward the right endpoint . the third eigenfunction , explaining only a small fraction of the overall variance , reflects a more complex contrast between early and late phases on one hand and a middle period on the other , with equally more complex behavior reflected in the first derivative . the eigenfunctions and their derivatives in conjunction with the eigenvalues determine the varying coefficient function @xmath72 , according to ( [ beta ] ) . the estimate of this function is obtained by plugging in the estimates for these quantities and is visualized in the left panel of figure [ auc - beta - zeig ] , demonstrating small negative values for the function @xmath72 throughout most of the domain , with a sharp dip of the function into the negative realm near the right end of the auctions . for subdomains of functional data , where the varying coefficient or `` dynamic transfer '' function @xmath72 is negative , as is the case for the auction data throughout the entire time domain , one may interpret the population equation @xmath174 as indicating `` dynamic regression to the mean . '' by this we mean the following : when a trajectory value at a current time @xmath58 falls above ( resp . , below ) the population mean trajectory value at @xmath58 , then the conditional mean derivative of the trajectory at @xmath58 falls below ( resp . , above ) the mean . the overall effect of this negative association is that the direction of the derivative is such that trajectories tend to move toward the overall population mean trajectory as time progresses . . ] thus , our findings for the auction data indicate that `` dynamic regression to the mean '' takes place to a small extent throughout the auction period and to a larger extent near the right tail , at the time when the final auction price is determined [ see also @xcite ] . one interpretation is that at the population level , prices are self - stabilizing , which tends to prevent price trajectories running away toward levels way above or below the mean trajectory . this self - stabilization feature gets stronger toward the end of the auction , where the actual `` value '' of the item that is being auctioned serves as a strong homogenizing influence . this means that in a situation where the current price level appears particularly attractive , the expectation is that the current price derivative is much higher than for an auction with an unattractive ( from the perspective of a buyer ) current price , for which then the corresponding current price derivative is likely lower . the net effect is a trend for log price trajectories to regress to the mean trajectory as time progresses . we illustrate here the proposed stochastic differential equation ( [ de ] ) . first estimating the function @xmath72 , we obtain the trajectories @xmath175 of the drift process . these trajectories are presented in figure [ auc - z ] for the entire sample of auctions . they quantify the component of the derivative process @xmath134 that is left unexplained by the varying coefficient function and linear part of the dynamic model ( [ de ] ) . the trajectories @xmath175 exhibit fluctuating variances across various subdomains . the subdomains for which these variances are small are those where the deterministic approximation ( [ approx ] ) to the stochastic differential equation works best . it is noteworthy that the variance is particularly small on the subdomain starting at around 158 hours toward the endpoint of the auction at 168 hours , since auction dynamics are of most interest during these last hours . it is well known that toward the end of the auctions , intensive bidding takes place , in some cases referred to as `` bid sniping , '' where bidders work each other into a frenzy to outbid each other in order to secure the item that is auctioned . . right : smooth estimates of the first ( solid ) , second ( dashed ) and third ( dash - dotted ) eigenfunction of @xmath0 based on ( [ gz ] ) . ] the right panel of figure [ auc - beta - zeig ] shows the first three eigenfunctions of @xmath0 , which are derived from the eigenequations derived from estimates @xmath150 of covariance kernels @xmath176 ( [ gz ] ) that are obtained as described after ( [ smooth2 ] ) . in accordance with the visual impression of the trajectories of @xmath0 in figure [ auc - z ] , the first eigenfunction reflects maximum variance in the middle portion of the domain and very low variance at both ends . interestingly , the second eigenfunction reflects high variance at the left end of the domain where prices are still moving upward quite rapidly , and very low variance near the end of the auction . this confirms that overall variation is large in the middle portion of the auctions , so that the drift process in ( [ de ] ) plays an important role in that period . further explorations of the modes of variation of the drift process @xmath0 can be based on the functional principal component scores of @xmath175 . following @xcite , we identify the three auctions with the largest absolute values of the scores . a scatterplot of second and first principal component scores with these auctions highlighted can be seen in the left upper panel of figure [ auc - zscore - ext ] . the corresponding individual ( centered ) trajectories of the drift process @xmath0 are in the right upper panel , and the corresponding trajectories of centered processes @xmath1 and @xmath134 in the left and right lower panels . the highlighted trajectories of @xmath0 are indeed similar to the corresponding eigenfunctions ( up to sign changes ) , and we find that they all exhibit the typical features of small variance near the end of the auction for @xmath0 and @xmath1 and of large variance for @xmath134 . , where the point marked by a circle corresponds to the auction with the largest ( in absolute value ) first score , the point marked with a square to the auction with the largest second score and the point marked with a `` triangle '' to the auction with the largest third score , respectively . top right : the trajectories @xmath175 of the drift process for these three auctions , where the solid curve corresponds to the trajectory of the `` circle '' auction , the dashed curve to the `` square '' auction and the dash - dotted curve to the `` triangle '' auction . bottom left : corresponding centered trajectories @xmath3 . bottom right : corresponding centered trajectory derivatives @xmath177 . ] for the two trajectories corresponding to maximal scores for first and second eigenfunction of @xmath0 we find that near the end of the auctions their centered derivatives turn negative . this is in line with dynamic regression to the mean , or equivalently , negative varying coefficient function @xmath72 , as described in section [ sec41 ] . here the trajectories for @xmath1 at a current time @xmath58 are above the mean trajectory , which means the item is pricier than the average price at @xmath58 . as predicted by dynamic regression to the mean , log price derivative trajectories at @xmath58 are indeed seen to be below the mean derivative trajectories at @xmath58 . the trajectory corresponding to maximal score for the third eigenfunction also follows dynamic regression to the mean : here the trajectory for @xmath1 is below the overall mean trajectory , so that the negative varying coefficient function @xmath72 predicts that the derivative trajectory @xmath134 should be above the mean , which indeed is the case . ( dashed ) and @xmath57 ( solid ) . right : smooth estimate of @xmath86 ( [ r1 ] ) , the variance explained by the deterministic part of the dynamic equation at time @xmath58 . ] that the variance of the drift process @xmath0 is small near the endpoint of the auction is also evident from the estimated variance function @xmath155 in the left panel of figure [ auc - var - r2 ] , overlaid with the estimated variance function @xmath178 of @xmath134 . the latter is rapidly increasing toward the end of the auction , indicating that the variance of the derivative process is very large near the auction s end . this means that price increases vary substantially near the end across auctions . the large variances of derivatives coupled with the fact that @xmath59 is small near the end of the auction implies that the deterministic part ( [ approx ] ) of the empirical differential equation ( [ de ] ) explains a very high fraction of the variance in the data . this corresponds to a very high , indeed close to the upper bound 1 , value of the coefficient of determination @xmath86 ( [ r1 ] ) , ( [ r2 ] ) in an interval of about 10 hours before the endpoint of an auction , as seen in the right panel of figure [ auc - var - r2 ] . we therefore find that the dynamics during the endrun of an auction can be adequately modeled by the simple deterministic approximation ( [ approx ] ) to the stochastic dynamic equation ( [ de ] ) , which always applies . this finding is corroborated by visualizing the regressions of @xmath80 versus @xmath79 at various fixed times @xmath58 . these regressions are linear in the gaussian case and may be approximated by a linear regression in the least squares sense in the non - gaussian case . the scatterplots of @xmath179 versus @xmath180 for times @xmath181 hours and @xmath182 hours ( where the time domain of the auctions is between 0 and 168 hours ) are displayed in figure [ auc - reg ] . this reveals the relationships to be indeed very close to linear . these are regressions through the origin . the regression slope parameters are not estimated from these scatterplot data which are contaminated by noise , but rather are obtained directly from ( [ de ] ) , as they correspond to @xmath56 . thus one simply may use the already available slope estimates , @xmath183 and @xmath184 . the associated coefficients of determination , also directly estimated via ( [ r2 ] ) and the corresponding estimation procedure , are found to be @xmath185 and @xmath186 . on @xmath187 ( both centered ) at @xmath181 hours ( left panel ) and @xmath182 hours ( right panel ) , respectively , with regression slopes @xmath188 and coefficient of determination @xmath189 , respectively , @xmath190 and @xmath191 , demonstrating that the deterministic part ( [ approx ] ) of the empirical differential equation ( [ de ] ) explains almost the entire variance of @xmath134 at @xmath182 hours but only a fraction of variance at @xmath181 hours . ] as the regression line fitted near the end of the auction at @xmath182 hours explains almost all the variance , the approximating deterministic differential equation ( [ approx ] ) can be assumed to hold at that time ( and at later times as well , all the way to the end of the auction ) . at @xmath181 the regression line explains only a fraction of the variance , while a sizable portion of variance resides in the drift process @xmath0 , so that the stochastic part in the dynamic system ( [ de ] ) can not be comfortably ignored in this time range . these relationships can be used to predict derivatives of trajectories and thus price changes at time @xmath58 for individual auctions , given their log price trajectory values at @xmath58 . we note that such predictions apply to fitted trajectories , not for the actually observed prices which contain an additional random component that is unpredictable , according to model ( [ kl2 ] ) . we find that at time @xmath182 , regression to the mean is observed at the level of individual auctions : an above ( below ) average log price level is closely associated with a below ( above ) average log price derivative . this implies that a seemingly very good ( or bad ) deal tends to be not quite so good ( or bad ) when the auction ends . the main motivation of using the dynamic system approach based on ( [ de ] ) is that it provides a better description of the mechanisms that drive longitudinal data but are not directly observable . the empirical dynamic equation may also suggest constraints on the form of parametric differential equations that are compatible with the data . in the auction example , the dynamic equation quantifies both the nature and extent of how expected price increases depend on auction stage and current price level . this approach is primarily phenomenological and does not directly lend itself to the task of predicting future values of individual trajectories . that expected conditional trajectory derivatives satisfy a first - order differential equation model ( which we refer to as the `` population level '' since this statement is about conditional expectations ) simply follows from gaussianity and in particular does not require additional assumptions . this suffices to infer the stochastic differential equation described in ( [ jt ] ) which we term `` empirical differential equation '' as it is determined by the data . then the function @xmath192 , quantifying the relative contribution of the drift process @xmath0 to the variance of @xmath134 , determines how closely individual trajectories follow the deterministic part of the equation . we could equally consider stochastic differential equations of other orders , but practical considerations favor the modeling with first - order equations . we find in the application example that online auctions follow a dynamic regression to the mean regime for the entire time domain , which becomes more acute near the end of the auction . this allows us to construct predictions of log price trajectory derivatives from trajectory levels at the same @xmath58 . these predictions get better toward the right endpoint of the auctions . this provides a cautionary message to bidders , since an auction that looks particularly promising since it has a current low log price trajectory is likely not to stay that way and larger than average price increases are expected down the line . conversely , an auction with a seemingly above average log price trajectory is likely found to have smaller than average price increases down the line . this suggests that bidders take a somewhat detached stance , watching auctions patiently as they evolve . in particular , discarding auctions that appear overpriced is likely not a good strategy as further price increases are going to be smaller than the average for such auctions . it also implies that bid snipers are ill advised : a seemingly good deal is not likely to stay that way , suggesting a more relaxed stance . conversely , a seller who anxiously follows the price development of an item , need not despair if the price seems too low at a time before closing , as it is likely to increase rapidly toward the end of the auction . for prediction purposes , drift processes @xmath175 for individual auctions are of great interest . in time domains where their variance is large , any log price development is possible . interestingly , the variance of drift processes is very small toward the right tail of the auctions , which means that the deterministic part of the differential equation ( [ de ] ) is relatively more important , and log price derivatives during the final period of an auction become nearly deterministic and thus predictable . other current approaches of statistical modeling of differential equations for time course data [ e.g. , @xcite ] share the idea of modeling with a first order equation . in all other regards these approaches are quite different , as they are based on the prior notion that a differential equation of a particular and known form pertains to the observed time courses and moreover usually have been developed for the modeling of single time courses . this established methodology does not take into account the covariance structure of the underlying stochastic process . in contrast , this covariance structure is a central object in our approach and is estimated nonparametrically from the entire ensemble of available data , across all subjects or experiments . formula ( [ rep ] ) is an extension of the covariance kernel representation in terms of eigenfunctions , given by @xmath193 [ @xcite ] , which itself is a stochastic process version of the classical multivariate representation of a covariance matrix @xmath194 in terms of its eigenvectors @xmath195 and eigenvalues @xmath96 , @xmath196 . specifically , using representation ( [ kl3 ] ) , one finds @xmath197 , and ( [ rep ] ) follows upon observing that @xmath198 for @xmath199 and @xmath200 for @xmath201 . regarding the `` population differential equation '' @xmath202 , observe that for any jointly normal random vectors @xmath203 with mean @xmath70 and covariance matrix @xmath194 with elements @xmath204 , it holds that @xmath205 . applying this to the jointly normal random vectors in ( [ jt ] ) then implies this population equation . the specific form for the function @xmath72 in ( [ beta ] ) is obtained by plugging in the specific terms of the covariance matrix given on the right - hand side of ( [ jt ] ) . applying ( [ rep ] ) , observing @xmath206 , and then taking the log - derivative leads to @xmath207/[\sum_k \lambda_k \phi_k^2(t)]$ ] , establishing the last equality in representation ( [ beta ] ) . turning to estimation , in a first step we aggregate measurements across all subjects into one `` big '' scatterplot and apply a smoothing method that allows us to obtain the @xmath208th derivative of a regression function from scatterplot data . for example , in the case of local polynomial fitting , given a univariate density function @xmath209 and bandwidth @xmath210 , one would minimize @xmath211 for each @xmath58 with respect to @xmath212 for @xmath213 , from which one obtains @xmath214 [ @xcite ] . according to ( [ kl4 ] ) , we will also need estimates of @xmath215 . there are various techniques available for this task . following @xcite , to which we refer for further details , using again local polynomial fitting , we minimize the pooled scatterplot of pairwise raw covariances @xmath216\\[-8pt ] & & \qquad\hspace*{30.8pt}{}\times \biggl\{g_i(t_{ij},t_{il})-\biggl(\sum_{m = 0}^{\nu+1}\alpha_{1 m } ( t_{ij}-t)^m+\alpha_{21}(t_{il}-s)\biggr ) \biggr\}^2\nonumber\end{aligned}\ ] ] for fixed @xmath217 with respect to @xmath218 and @xmath219 for @xmath220 , where @xmath221 , @xmath222 , @xmath223 is a kernel chosen as a bivariate density function , and @xmath224 is a bandwidth . this leads to @xmath225 . the pooling that takes place in the scatterplots for estimating the derivatives of @xmath77 and of @xmath17 is the means to accomplish the borrowing of information across the sample , which is essential to overcome the limitations of the sparse sampling designs . we note that the case of no derivative @xmath226 is always included , and solving the eigenequations on the left - hand side of ( [ kl4 ] ) numerically for that case leads to the required estimates @xmath227 of the eigenvalues and @xmath228 of the eigenfunctions . the estimates @xmath229 of the eigenfunction derivatives are then obtained from the right - hand side of ( [ kl4 ] ) , plugging in the estimates for eigenfunctions and eigenvalues , followed by a numerical integration step . the plug - in estimates , @xmath230 , are then obtained from the corresponding representations , ( [ beta ] ) , ( [ gz ] ) , ( [ v ] ) , ( [ r2 ] ) , by including @xmath141 leading components in the respective sums . while for theoretical analysis and asymptotic consistency one requires @xmath231 , the number of included eigen - terms @xmath141 in practical data analysis can be chosen by various criteria , for example , aic / bic based on marginal / conditional pseudo - likelihood or thresholding of the total variation explained by the included components [ @xcite ] . one key feature of the covariance surface smoothing step in ( [ smooth2 ] ) is the exclusion of the diagonal elements ( for which @xmath232 ) ; the expected value for these elements includes the measurement error variance @xmath233 in addition to the variance of the process . the difference between a smoother that uses the diagonal elements only and the resulting diagonal from the smoothing step ( [ smooth2 ] ) when no derivatives are involved can then be used to find consistent estimates for the error variance @xmath233 [ @xcite ] . to obtain estimates for the derivatives of the trajectories @xmath3 , a realistic target is the conditional expectation @xmath234 . it turns out that this conditional expectation can be consistently estimated in the case of gaussian processes by applying principal analysis by conditional expectation ( pace ) [ @xcite ] . for @xmath235 , @xmath236 , @xmath237 , @xmath238 , if @xmath239 and @xmath11 in ( [ kl2 ] ) are jointly gaussian , then by standard properties of the gaussian distribution , @xmath240 where @xmath241 this implies @xmath242 . the unknown quantities can be estimated by simply plugging in the variance , eigenvalue , eigenfunction and eigenfunction derivative estimates discussed above , again coupled with truncating the number of included components at @xmath141 . denote the densities of @xmath243 and @xmath244 by @xmath245 , @xmath246 , and define an interior domain by @xmath247 $ ] with @xmath248 $ ] for some @xmath249 . regularity conditions for the densities and the targeted moment functions as well as their derivatives are as follows , where @xmath250 are nonnegative integers : 1 . @xmath251 exists and is continuous on @xmath252 with @xmath253 , @xmath254 exists and is continuous on @xmath255 for @xmath256 ; 2 . @xmath257 exists and is continuous on @xmath252 , @xmath258 exists and is continuous on @xmath255 for @xmath256 . we say that a bivariate kernel function @xmath223 is of order @xmath259 , where @xmath260 is a multi - index @xmath261 , if @xmath262\\[-8pt ] & & \qquad= \cases{0 , & \quad $ 0 \leq\ell_1+\ell_2 < \ell , \ell_1\neq\nu _ 1 , \ell_2 \neq\nu_2 $ , \cr ( -1)^{|{\bf\nu}| } \nu_1 ! \nu_2 ! , & \quad $ \ell_1=\nu_1 , \ell_2=\nu _ 2$,\cr \neq0 , & \quad $ \ell_1+\ell_2=\ell$,}\nonumber \ ] ] where @xmath263 . the univariate kernel @xmath209 is said to be of order @xmath264 for a univariate @xmath265 , if ( [ order ] ) holds with @xmath266 on the right - hand side , integrating only over the argument @xmath267 on the left - hand side . for the kernel functions @xmath209 , @xmath223 used in the smoothing steps to obtain estimates for @xmath113 and @xmath32 in ( [ smooth1 ] ) and for @xmath114 and @xmath115 in ( [ smooth2 ] ) we assume 1 . kernel functions @xmath209 and @xmath223 are nonnegative with compact supports , bounded and of order @xmath268 and @xmath269 , respectively . the following lemma provides the weak @xmath98 convergence rate for univariate and bivariate weighted averages defined below . for arbitrary real functions @xmath270 and @xmath271 , define @xmath272 and @xmath273 , let @xmath274 for a single index @xmath208 and @xmath275 for a multi - index @xmath276 , and define the weighted kernel averages , employing bandwidths @xmath277 , @xmath278 \label{kwa1 } \hat{\theta}^\ast(t , s)&=&\frac{1}{e\{n(n-1)\ } n h_2^{|\bolds\nu|+2}}\sum_{i=1}^n \sum_{1\leq j\neq l \leq n_i } \theta^\ast(t_{ij } , t_{il } , y_{ij } , y_{il } ) \kappa_2\nonumber\\[-9pt]\\[-9pt ] & & \hspace*{152.4pt}{}\times\biggl(\frac{t - t_{ij}}{h_2 } , \frac{s - t_{il}}{h_2 } \biggr).\nonumber\end{aligned}\ ] ] for establishing convergence results for the general weighted averages ( [ kwa ] ) , assume that : 1 . derivatives @xmath279 exist and are continuous on @xmath252 , @xmath280 exists and is continuous on @xmath255 for @xmath281 . the univariate kernel @xmath209 is of order @xmath264 and the bivariate kernel @xmath223 is of order @xmath259 . [ lem1 ] under , , , and if @xmath282 , @xmath283 and @xmath284 , as @xmath285 , @xmath286\\[-9pt ] \|\hat{\theta}^\ast-\theta_{\bolds\nu}^\ast\|_s&=&o_p\biggl(\frac { 1}{\sqrt{n}h_2^{|\bolds\nu|+1}}+h_2^{\ell-|\bolds\nu|}\biggr).\nonumber\end{aligned}\ ] ] proof of theorem [ thm1 ] since @xmath287 are jointly gaussian processes , it is clear that @xmath0 is gaussian . formula ( [ beta ] ) for @xmath56 is obtained by observing that for joint gaussian r.v.s , @xmath288\{x(t)-\mu(t)\ } $ ] . then the properties of the functional principal component scores lead directly to @xmath289 whence @xmath290 . this implies @xmath53 . according to ( [ de ] ) , @xmath291 , for all @xmath46 , using ( [ cov ] ) and ( [ beta ] ) . this implies the independence of @xmath51 , due to the gaussianity . next observe @xmath292 , from which one obtains the result by straightforward calculation . proof of lemma [ lem1 ] since @xmath293 and @xmath6 is a bounded and integer - valued random variable . denote the upper bound by @xmath294 . to handle the one - dimensional case in ( [ kwa ] ) , we observe @xmath295 where @xmath296 is the indication function . note that for each @xmath297 , @xmath298 is obtained from an i.i.d . sample . slightly modifying the proof of theorem 2 in @xcite for a kernel of order @xmath264 provides the weak convergence rate @xmath299 . it is easy to check that @xmath300 , as @xmath6 is a positive integer - valued random variable . therefore , @xmath301 analogously , for the two - dimensional case in ( [ kwa1 ] ) , let @xmath302 and then @xmath303^{-1 } \hat{\theta}_{\mathbf{j}}^\ast$ ] . similarly to the above , one has @xmath304 . again it is easy to verify that @xmath305 . the triangle inequality for the @xmath98 distance entails @xmath306 . proof of theorem [ thm2 ] note that the estimators @xmath307 , @xmath308 , @xmath309 and @xmath310 all can be written as functions of the general averages defined in ( [ kwa ] ) , ( [ kwa1 ] ) . slightly modifying the proof of theorem 1 in @xcite , with @xmath311 rates replaced by the @xmath98 rates given in lemma [ lem1 ] , then leads to the optimal @xmath98 weak convergence rates for @xmath312 and @xmath313 in ( [ thm2-eq1 ] ) . for the convergence rate of @xmath314 , lemma 4.3 in @xcite implies that @xmath315 where @xmath316 is defined in ( [ spacing ] ) and @xmath309 is an arbitrary estimate ( or perturbation ) of @xmath17 . denote the linear operators generated from the kernels @xmath317 and @xmath318 by @xmath319 , respectively , @xmath320 . noting that @xmath321 , one finds @xmath322 which implies ( [ thm2-eq2 ] ) . proof of theorem [ thm3 ] from ( [ bosq ] ) it is easy to see that @xmath323 , and from both ( [ bosq ] ) and ( [ pf1 ] ) that @xmath324 uniformly in @xmath64 . one then finds that @xmath325 is bounded in probability by @xmath326 which implies that @xmath327 , where @xmath328 is defined in ( [ rate ] ) and the remainder terms in ( [ rem ] ) . similar arguments lead to @xmath329 , noting @xmath330 due to the cauchy schwarz inequality . regarding @xmath331 , one has @xmath332 applying the cauchy schwarz inequality to @xmath333 . observing @xmath334 yields @xmath335 . to study @xmath336 , we investigate the @xmath98 convergence rates of @xmath337 where @xmath72 ( resp . , @xmath338 ) and @xmath34 ( resp . , @xmath339 ) share the same argument , and we define @xmath340 . in analogy to the above arguments , @xmath341 , @xmath342 . this leads to @xmath343 . the same argument also applies to @xmath344 . next we study @xmath345 and find that @xmath346 . analogous arguments apply to @xmath347 , completing the proof . we are grateful to two referees for helpful comments that led to an improved version of the paper . dauxois , j. , pousse , a. and romain , y. ( 1982 ) . asymptotic theory for the principal component analysis of a vector random function : some applications to statistical inference . _ j. multivariate anal . _ * 12 * 136154 . mas , a. and menneteau , l. ( 2003 ) . perturbation approach applied to the asymptotic study of random operators . in _ high dimensional probability , iii ( sandjberg , 2002)_. _ progress in probability _ * 55 * 127134 . birkhuser , basel . ramsay , j. o. , hooker , g. , campbell , d. and cao , j. ( 2007 ) . parameter estimation for differential equations : a generalized smoothing approach ( with discussion ) . _ j. r. stat . methodol . _ * 69 * 741796 . reithinger , f. , jank , w. , tutz , g. and shmueli , g. ( 2008 ) . modelling price paths in on - line auctions : smoothing sparse and unevenly sampled curves by using semiparametric mixed models . c _ * 57 * 127148 . shi , m. , weiss , r. e. and taylor , j. m. g. ( 1996 ) . an analysis of paediatric cd4 counts for acquired immune deficiency syndrome using flexible random curves . _ j. roy . c _ * 45 * 151163 .

the in - medium pion properties , _ i.e. _ the temporal pion decay constant , the pion mass and the wave function renormalization , in symmetric nuclear matter are calculated in an in - medium chiral perturbation theory up to the next - to - leading order of the density expansion . the chiral lagrangian for the pion - nucleon interaction is determined in vacuum , and the low energy constants are fixed by the experimental observables . we carefully define the in - medium state of pion and find that the pion wave function renormalization plays an essential role for the in - medium pion properties . we show that the linear density correction is dominant and the next - to - leading corrections are not so large at the saturation density , while their contributions can be significant in higher densities . the main contribution of the next - to - leading order comes from the double scattering term . we also discuss whether the low energy theorems , the gell - mann oakes renner relation and the glashow weinberg relation , are satisfied in nuclear medium beyond the linear density approximation . we also find that the wave function renormalization is enhanced as largely as 50% at the saturation density including the next - to - leading contribution and the wave function renormalization could be measured in the in - medium decay .

1312.0832

spontaneous breakdown of chiral symmetry ( @xmath4ssb ) @xmath5 characterizes the vacuum and low - energy dynamics of quantum chromodynamics ( qcd ) @xcite . the non - vanishing chiral condensate @xmath6 is considered as one of the order parameters of @xmath7ssb and gives a characteristic scale for hadron physics . @xmath7ssb is considered to be responsible for the origin of constituent quark mass after the current quark mass is given by the higgs condensate slightly . according to the spontaneous breakdown , the pseudoscalar mesons such as @xmath8 appear as the nambu - goldstone ( ng ) bosons . recently , in order to investigate the mechanism of the dynamical mass generation , partial restoration of the chiral symmetry in the nuclear medium has gained considerable attention . this phenomenon is incomplete restoration of chiral symmetry with sufficient reduction of the absolute value of the chiral condensate in the medium and will lead to various changes of hadron properties . once we understand the partial restoration of chiral symmetry , we can predict other in - medium hadronic quantities through low energy theorems and vice versa . from this point of view , vast theoretical and experimental efforts are devoted for this topic . the density dependence of the chiral condensate is evaluated in various approaches , for example , the well - known linear density approximation providing the model - independent low density theorem @xcite , the relativistic bruckner - hartree - fock theory approach@xcite and systematic calculations by in - medium chiral perturbation theory beyond the linear density calculation @xcite . according to the model - independent linear density approximation , the leading density correction to the chiral condensate is determined by the @xmath9n sigma term @xmath10 and with the empirical value of the sigma term it has been found that the leading correction gives enough large contribution at the normal nuclear density : @xmath11 where @xmath12 , @xmath13 is the in - medium and in - vacuum condensate and @xmath14 is the normal nuclear density . recently , an in - medium sum rule satisfied in any density region has been derived model - independently by current algebra method and the low energy theorems , such as the gell mann oakes renner relation , the glashow weinberg relation and the weinberg tomozawa relation , are discussed within the linear density order @xcite . to investigate the pion properties in nuclei , deeply bound pionic atoms have received much attention @xcite . theoretically , the binding energies and decay widths of the 1s and 2p deeply - bound pionic atom states are estimated @xcite and in refs . @xcite hadronic quantities , such as pion optical potential , have been calculated beyond the linear density . experimentally the reduction of the chiral condensate is estimated quantitatively through reduction of the @xmath15-wave isovector parameter @xmath16 in the @xmath17-nucleus optical potential @xcite . the @xmath18 parameter is regarded as an in - medium isovector @xmath17n scattering length . these results show that the reduction of @xmath16 means the repulsive enhancement of the @xmath15-wave @xmath17-nucleon interaction in nucleus . another examples are low energy @xmath17-nucleus scattering and @xmath19 interaction in nuclei in the scalar - isoscalar channel . the low energy @xmath17-nucleus scattering also show that s - wave @xmath17-nucleus interaction are enhanced repulsively @xcite . according to the theoretical discussion given in refs . @xcite , the in - medium @xmath19 interaction in the scalar - isoscalar part will also has attractive enhancement thanks to the partial restoration of chiral symmetry in nuclear medium and the experimental observation of the invariant mass spectrum of the @xmath20 production off nuclear targets performed in refs . @xcite could have a hint of such a enhancement . in particular the pion decay constant is a fundamental quantity of chiral symmetry breaking . the in - medium decay constant also has been investigated in the linear density approximation @xcite and recently the chiral condensate and the decay constant have been evaluated in the next - to - leading order based on chiral order counting @xcite . in this paper , we discuss a general in - medium pion state and evaluate in - medium pionic quantities such as the decay constant , mass and the pseudo scalar coupling beyond linear density approximation . this paper is organized as follows . in sec.[in - medium chpt ] , we explain the general formulation of the in - medium chiral perturbation theory and discuss an expansion by fermi momentum counting . in sec.[decayconst ] , we discuss in - medium pion state and define the in - medium pionic quantities . here we will find that the pion wave function renormalization plays an important role for the in - medium pionic quantities . in sec.[results ] , we evaluate the in - medium pion self energy , wave function renormalization , pion decay constant and pseudo - scalar coupling and show the numerical results of the density dependence of them up to @xmath21 in fermi momentum expansion in symmetric nuclear matter and in isospin limit . we also discuss whether the in - medium low energy theorems , the gell - mann oakes renner relation and the glashow weinberg relation , are satisfied or not . finally we discuss the in - medium @xmath22 process caused by chiral anomaly . in sec.[summary ] , we summarize our paper . chiral perturbation theory ( chpt ) is a powerful tool describing low energy dynamics of the nambu - goldstone ( ng ) bosons and nucleons as an effective field theory of qcd @xcite . chpt is constructed based on chiral symmetry and its spontaneous breaking and consists in systematic expansion of ng boson momentum and the quark mass . the chiral order counting scheme makes it possible to categorize lagrangian and feynman diagrams in terms of powers of momentum and the quark mass and to estimate magnitude of possible corrections for the amplitude , such as the current - current correlation functions . thus , chpt describes quantitatively the s - matrix elements of the qcd currents . in this decade , chiral effective theory for nuclear matter has been developed @xcite and is applied to study nuclear matter properties , such as the nuclear matter energy density@xcite and also used to study partial restoration of chiral symmetry and the in - medium changes of the pion properties , such as in - medium pion mass , decay constant @xcite and 1s and 2p energy levels of deeply - bound pionic atoms @xcite . for the calculations of the in - medium pion quantities , we evaluate the green s functions in the ground state of nuclear matter , and , in particular , our interests are hadron properties in nuclear medium as a bound state of the nucleon many body system . in quantum field theory , the transition amplitude between the ground state in the presence of the external fields is the fundamental quantity , and the generating functional is defined by the transition amplitude , which can be calculated by the path integral formalism of the quantum field theory . in the in - medium chiral perturbation theory , one prepares the ground state of nucleon fermi gas at asymptotic time as a reference state @xcite . the generating functional @xmath23 for the connected green functions in nuclear matter is given as follows : @xmath24 & = & \exp{i w [ j ] } = \langle \omega_{\rm out } | \omega_{\rm in } \rangle_j \\ & = & \int du dn dn^{\dagger } \langle \omega_{\rm out } | n(+\infty ) \rangle e^{i \int dx ( { { \mathcal}l}_{\pi}+ { { \mathcal}l}_{\pi n}+ { { \mathcal}l}_{n } ) } \langle n(-\infty ) | \omega_{\rm in } \rangle , \end{aligned}\ ] ] where @xmath25 is the chiral field parametrized by the ng boson field in the nonlinear realization of chiral symmetry , @xmath26 is the nucleon field , and @xmath27 represents the scalar , pseudo scalar , vector and axial vector external fields @xmath28 . the lagrangian is described by the free pion and nucleon fields and should include , in principle , all the interaction among pions and nucleons within the lagrangian . since we use the lagrangian described by the free nucleon field and the lagrangian prescribes the nucleon interaction , the reference state can be the ground state of the free fermi gas of nucleons defined by @xmath29 where @xmath30 is the nucleon creation operator with momentum @xmath31 and the index @xmath32 represents the spin and isospin and @xmath26 is the number of the momentum states below the nucleon fermi momentum @xmath33 , which is obtained by the nucleon density @xmath34 as @xmath35 , and @xmath36 is the 0-particle state . the connected @xmath32-point green functions can be calculated by taking functional derivatives of @xmath37 $ ] with respect to the external sources @xmath38 : @xmath39 , \label{green}\ ] ] where @xmath40 is the corresponding current operator to the external source @xmath41 . considering the symmetry property of the current operator under the chiral rotation , one can identify the quark contents of the current operators , such as the pseudo - scalar current @xmath42 and the axial - vector current @xmath43 . if we evaluate the generating functional non - perturbatively with appropriate nucleon - nucleon interactions , we can describe nuclear matter in principle and deduce the qcd current green functions in nuclear matter . note that this prescription resembles the description of deuteron in terms of the free nucleons , in which one starts with the free nucleon field and with appropriate nucleon - nucleon interaction and one can find deuteron as a bound state of two nucleons by solving the bethe - salpeter equation non - perturbatively . reference @xcite propose a method to derive the in - medium chiral lagrangian in a systematic expansion in terms of the chiral counting and the fermi sea insertion . in this method , the fermi momentum is regarded as a small parameter as the chiral order . performing the integral in terms of the nucleon field by using the gauss integral formula for the bilinear form of the nucleon interaction , one obtains the generating functional characterized by double expansion of fermi sea insertions and chiral orders : @xmath44= \int du \exp \big{\ { } i\int dx \big{[}\mathcal{l}_{\pi \pi } - \int \frac{d \bf{p}}{(2\pi)^3 2e_{p } } { \mathcal f\ , } { \rm tr } \big { ( } i\gamma(x , y ) ( \slash{p } + m_n ) n(p ) \big { ) } } & & \nonumber \\ & & \hspace{-0.8 cm } - \frac{i}{2 } \int \frac{d \bf{p}}{(2\pi)^3 2e_{p } } \frac{d \bf{q}}{(2\pi)^3 2e_{q } } { \mathcal f}\ , { \rm tr } \big { ( } i\gamma(x , x ' ) ( \slash{q } + m_n ) n(q ) i\gamma(y',y ) ( \slash{p } + m_n ) n(p ) \big { ) } + \cdots , \big { ] } \big{\ } } \label{eq : zexp}\end{aligned}\ ] ] where @xmath45 denotes fourier transformation of the spacial variables except for @xmath46 , @xmath47 is the nucleon energy @xmath48 for the momentum @xmath49 and @xmath50 is the nonlocal vertex defined by the in - vacuum quantities as @xmath51^{-1}$ ] , where the interactions @xmath52 and the free nucleon propagator @xmath53 are given by the in - vacuum chiral perturbation theory . the isodoublet matrix @xmath54 restricts the momentum integral for the nucleon momenta up to the fermi momentum : @xmath55 the nonlocal vertex @xmath50 is expanded in terms of the bilinear local vertex @xmath52 as @xmath56 and the vertex @xmath52 is also expanded in terms of the chiral order . the in - medium pion lagrangian @xmath57 can be obtained by the generating functional as @xmath58= \exp [ { i \int d^{4}x \tilde{\mathcal l}_{\pi \pi}}]$ ] . for the calculation of the amplitude , one uses the free nucleon propagators for the chiral expansion of the nonlocal vertex function while the fermi sea nucleon term @xmath59 in the fermi sea insertion among the nonlocal vertices @xmath50 . it has been shown explicitly in ref . @xcite that this expansion scheme is consistent with the conventional relativistic many body theory in the sense that one can use directly the in - medium nucleon propagator , that is the fermi gas propagator , @xmath60 in the calculation . here , the free propagator and the medium part of the fermi gas propagator are denoted by @xmath61 , @xmath62 with the isospin index @xmath63 . the medium part @xmath62 represents fermi sea effect . we can calculate the connected functions using usual perturbative expansions with the in - medium nucleon propagator . in this way , one can deal with density contributions from nuclear medium by perturbative expansion of nuclear fermi momentum . in the in - medium chpt , we can classify the current green functions in terms of the order of the small parameters in the expansion of the pion momentum , the quark mass and the fermi momentum in the similar way to the in - vacuum chpt . the chiral order of a specific diagram is counted as @xcite @xmath64 where @xmath65 is the number of pion loops , @xmath66 is the number of the pion propagators , @xmath67 is the chiral dimension from the pion chiral lagrangian , @xmath68 is the chiral dimension of the nonlocal in - medium vertex with @xmath32 fermi sea insertions and @xmath69 is the chiral dimension of the @xmath70 vertex . in this chiral counting , we count fermi gas propagator as @xmath71 like free nucleon propagator , so that the counting rule is the same as in - vacuum chiral perturbation theory . in this work , we focus on the fermi momentum dependence of the green functions and count only by fermi momentum orders . we assume that the in - vacuum loop effects are renormalized into counter terms in the chiral lagrangian and we use the in - vacuum physical values to fix the low energy constants ( lecs ) . for example , we translate lec @xmath72 appearing in @xmath73 into the physical @xmath74 sigma term @xmath75 : @xmath76 since the @xmath77 term gives the leading order of the @xmath78 sigma term in the chiral perturbation theory and once one calculates the higher orders for the @xmath78 sigma term , they should also contribute to the in - medium quantities in the same manner as @xmath77 . an empirical value of the sigma term is @xmath79 @xcite . and a recent analysis based on relativistic formulation of @xmath78 chiral perturbation theory suggests @xmath80 @xcite . another example is the isoscalar @xmath78 scattering length @xmath81 . in tree order , @xmath81 is given with lecs @xmath82 from @xmath83 as @xmath84 the scattering length is used to determine the combination of the lecs @xmath82 . recent calculations based on chpt give @xmath85 at better than the 95@xmath86 confidence level @xcite . in this way , we take the in - vacuum physical values to determine lecs and perform systematic calculations for density effects of the pionic observables based on the counting of fermi momentum orders . as we have discussed in ref . @xcite , beyond the order of @xmath87 in the density expansion , the @xmath78 dynamics alone can not predict the in - medium quantities , because we encounter divergence in loop calculations . these divergence can be removed , once we introduce counter terms expressed by @xmath88 contact interactions . the @xmath88 interactions are to be determined by the in - vacuum @xmath88 dynamics . we use the following chiral lagrangian in this work . the chiral lagrangian for the meson sector is given with the chiral field @xmath25 parametrized by the pion field as @xmath89 with the covariant derivative for the chiral field @xmath25 @xmath90 given by the vector and axial vector external fields @xmath91 and @xmath92 counted as @xmath93 , and @xmath7 field @xmath94 given by the scalar and pseudoscalar fields @xmath15 and @xmath95 counted as @xmath96 . the scalar field @xmath15 is replaced by the quark mass matrix in calculation . this lagrangian is counted as @xmath96 . in this paper , in order to make the perturbative calculation simple , as we have done in the previous paper @xcite , we use the following parametrization of the chiral field @xmath25 proposed by @xcite : @xmath97\ ] ] where @xmath98 satisfies @xmath99 the chiral lagrangian for the nucleon sector is as follows : @xmath100 where @xmath52 is the chiral interaction for the nucleon in the bilinear form . the chiral interaction can be expanded in terms of the chiral order as @xmath101 and @xmath102 is counted as @xmath103 . the explicit form of the leading term @xmath104 reads @xmath105 with the vector current @xmath106 - \frac{i}{2 } u^\dag ( v_{\mu } + a_{\mu } ) u - \frac{i}{2 } u ( v_{\mu } - a_{\mu } ) u^\dag , \ ] ] and the axial current @xmath107},\ ] ] where the field @xmath108 is defined by a square root of the chiral field @xmath25 : @xmath109 . the expression of the next leading term @xmath110 is given as @xmath111 with @xmath112 , @xmath113 and the covariant derivative for the nucleon field @xmath114 . here we have omitted irrelevant terms in the present work for the in - medium pion properties in symmetry nuclear matter . the pion propagation in medium can be calculated by the two - point function of the pseudoscalar density : @xmath115 around the in - medium pion pole the two - point function can be written in terms of the in - medium quantities , such as the in - medium pion mass @xmath116 , velocity @xmath117 and pseudoscalar coupling @xmath118 as @xmath119 where @xmath120 does not include any singularity at the pion pole . in this way , with @xmath121 , the in - medium pion mass is defined by the pole position of the two point function and the coupling of pion to the pseudoscalar density @xmath122 in medium is defined by the square root of the residue of the two - point function at the pion pole . the two - point function can be calculated using the in - medium chiral perturbation theory given in the previous section . the calculation will be done in terms of the in - vacuum quantities like @xmath123 where @xmath124 is the in - vacuum pion mass , @xmath125 is the pion self - energy in medium and @xmath126 is the vertex correction of the pion coupling to the pseudoscalar density . the vertex correction @xmath126 does not contain the pion pole and it is calculated by considering one - particle irreducible diagrams . expanding the self - energy @xmath125 in eq . around @xmath127 and @xmath128 , @xmath129 we write the two - point function in the following way @xmath130 with @xmath131 comparing eqs . and , we obtain , at the pion pole , @xmath132 using the in - medium chiral perturbation theory we can calculate the self - energy @xmath125 and @xmath126 , we obtain the in - medium pion properties with the above equations . the in - medium pion propagator can be also written by the pion field operator @xmath17 as @xmath133 comparing eqs . and , for the calculation of the pion pole we regard @xmath134 the pion operator @xmath17 creates one pion in medium with the mass @xmath135 and the wave function normalization @xmath136 , satisfying @xmath137 where we have introduced the one - pion state with a momentum @xmath95 in the nuclear medium by denoting @xmath138 . the in - medium coupling constant @xmath120 defined in eq . as the residue of the pion propagator induced by the pseudoscalar density may be also written as the following matrix element : @xmath139 relation can be understood by eq . with the reduction formula : @xmath140 where we have understood @xmath141 , and we have used eq . in the second equality and eq . in the third equality . we define the in - medium decay constant in analogy of the in - vacuum decay constant as a matrix element of the axial vector current @xmath142 : @xmath143 } \delta^{ab } \label{eq : pidecaycon}\ ] ] where we have introduced a vector characterizing the medium rest frame @xmath144 with @xmath145 and there are two form factors , @xmath146 and @xmath147 , in the presence of nuclear matter . these form factors , actually , should be functions of @xmath148 and @xmath149 according to lorentz covariance . the pion decay constants are obtained at the mass shell point with @xmath150 and @xmath151 . we define the temporal and spatial components of the decay constant @xmath152 by taking @xmath153 as @xmath154 the decay constants are obtained by @xmath155 these decay constants can be calculated in the in - medium chiral perturbation theory . making good use of the reduction formula in momentum space again , we write down the matrix element in terms of the one - particle irreducible vertex correction and the wave function renormalization : @xmath156 where we mean @xmath157 and @xmath158 with @xmath159 is the vertex correction of the decay constant , which can be calculated by one - particle irreducible diagrams in the chiral perturbation theory . in this section , we investigate explicitly the in - medium pion properties with the in - medium chiral perturbation theory up to the order of @xmath160 in the density expansion . in last section , we find that the in - medium pion decay constant can be calculated by the pion wave function renormalization @xmath136 and the one - particle irreducible vertex correction @xmath161 . the wave function renormalization is evaluated by taking derivative of the pion self - energy @xmath162 with respect to the energy squared . we calculate the pion self energy first . with the self energy , we evaluate next the in - medium pion mass and the wave function renormalization , and show their density dependence . calculating the one - particle irreducible correction for the pion decay constant , we evaluate the density dependence of the decay constant with the wave function renormalization . we also check whether the low energy relations are satisfied also in the nuclear medium . finally we discuss the @xmath163 decay rate in nuclear medium . for simplicity , we concentrate on the calculation under the unpolarized symmetric nuclear matter , where the matter has spin 0 and isospin 0 with the equal proton and neutron densities , @xmath164 . in the following , we write down the feynman diagrams for each quantity and classify them according to the density dependence . as we have already discussed in the previous section , we presume that the in - vacuum lecs are fixed by the experimental observables . the vertex and mass corrections from the quantum loops of pion and nucleon are renormalized into the physical quantities . on this understanding , we replace lecs to the observed quantities and use the experimental values . in the present work the relevant replacements are the followings : @xmath165 where the left hand sides are given by lecs , while the right hand sides are written in terms of the observed quantities in vacuum . in these expression , @xmath166 with the isoscalar @xmath78 scattering length @xmath167 . in the replacement higher order contributions in the chiral counting are already involved . in this sense , we loose strict counting of the chiral order . taking this scheme , we do not have to calculate the loop integrals which contain only the in - vacuum propagators , because they are supposed to be already counted inside the experimental value . this fact reduces the number of the relevant diagrams which we should calculate . the in - medium quantities which we are going to calculate should be evaluated at the pion pole . the in - vacuum chiral order can be counted by the pion mass . therefore , in the perturbative expansion , the term of which the leading chiral counting in vacuum is @xmath168 has @xmath169 dependence . the factor @xmath170 comes from the fermi motion correction of the nucleon when one calculates the fermi sea loop integral . the order of the density expansion is counted as @xmath171 , while the order of the small parameter is given by @xmath172 . if one takes the density contribution up to @xmath173 , it is enough to consider the diagrams with @xmath174 . the details of the loop calculation is summarized in appendix [ app : calc ] . we show the feynman diagrams contributing the in - medium self energy with the density corrections up to @xmath175 in fig . [ self ] . in these diagrams , the dashed doubled and thick lines are pion propagator , the fermi gas propagator @xmath176 and the nucleon propagation in the fermi gas @xmath177 , and the filled and unfilled circles are the leading and next - to - leading order vertices from the chiral lagrangian , respectively . in the following calculations , we fix the external momentum as @xmath178 . the diagrams for the leading order of the density expansion @xmath179 are given in fig . [ self](a ) . the self energy coming from the left diagram @xmath180 in fig.[self ] ( a ) can be calculated as @xmath181 } \nonumber \\ & = & - \rho \frac{i g_a^2}{4 f^2 m_n } q_0 ^ 2,\end{aligned}\ ] ] where @xmath182 is the fermi gas propagator given in eq . , the amplitude @xmath183 can be obtained by the interaction lagrangian given in eq . of appendix [ app : lagrangian ] . next , we calculate the right diagram @xmath184 in fig.[self](a ) : @xmath185 } \nonumber \\ & = & -\frac{2 i\rho}{f^2 } \big { ( } 4 c_1 b_{0 } m_{q } - ( c_2 + c_3 ) q_0 ^ 2 \big { ) } .\end{aligned}\ ] ] the next - to - leading order contribution coming from the left diagram @xmath186 in fig . [ self](b ) is given as @xmath187 with the pion propagator @xmath188 , the symmetric factor @xmath189 and @xmath190 defined as @xmath191 } .\ ] ] finally we obtain @xmath192 with function @xmath193 defined by @xmath194 the self energies coming from the double scattering corrections @xcite given by the middle and right diagrams of fig . [ self](b ) , @xmath195 and @xmath196 , have the two @xmath197 vertices from the leading weinberg - tomozawa interaction and the next - to - leading interaction with lecs @xmath198 , respectively . these are evaluated as @xmath199 } \nonumber \\ & = & -\frac{2 i q_0 ^ 2 } { f^4 } \frac{k_f^4}{6 \pi^4 } g(a^{2 } ) , \\ -i\sigma_5 ( q_0 ) & = & - \int \frac{d^4 p}{(2 \pi ) ^4 } \frac{d^4 k}{(2 \pi ) ^4 } { \rm tr } \big { [ } ( -i a^{(2)}_{\pi \pi } ) i g_m ( k- \frac{p}{2 } ) ( -i a^{(2)}_{\pi \pi } ) ig_m ( k+ \frac{p}{2 } ) id_\pi ( q - p ) \big { ] } \nonumber \\ & = & -i \big { ( } 8c_1 b_0 m_q - 2c_2 q_0 ^ 2 - 2c_3 q_0 ^ 2 \big{)}^2 \frac{2 k_f^4}{3 \pi^4 f^4 } g(a^{2 } ) \end{aligned}\ ] ] with @xmath200 and function @xmath201 defined by @xmath202 summing up all the contributions , we obtain the the self energy up to @xmath203 as @xmath204 \frac{8k_f^4}{3 f^4 \pi^4 } g(a^{2 } ) \end{aligned}\ ] ] with @xmath200 . the in - medium pion mass is obtained by the summation of the in - vacuum mass and the self energy evaluated at the pion on the mass shell @xmath205 . this brings us a self - consistent equation : @xmath206 nevertheless , because the in - medium correction starts with the linear density @xmath207 and we are evaluating the pion mass up to @xmath208 , the density correction on the mass in the argument of the self energy gives higher orders in the density expansion of the in - medium pion mass . thus , we are allowed to evaluate the self - energy at the in - vacuum on - shell @xmath209 for the present purpose . we evaluate the in - medium pion mass up to @xmath208 as @xmath210 \frac{4 k_f^4}{3 \pi^4 f^4 } g(0 ) \right\ } \nonumber \\ & = & m_\pi \left\ { 1 + \rho \frac{b^{+}}{2m_\pi^2 } - \frac { g_a^2 k_f^4 } { 24 \pi^4 f_\pi^4 } f(\frac{m_\pi}{2k_f } ) + \left [ \frac{1}{8 } + m_\pi^2 \big { ( } \frac{b^{+}}{2m_\pi^2}-\frac{g_a^2}{8m_n } \big{)}^2 \right]\frac { 2k_f^4}{\pi^4 f_\pi^4 } \right\ } .\end{aligned}\ ] ] the lecs , @xmath211 , @xmath77 , @xmath212 , @xmath213 , have been determined by the in - vacuum physical quantities , the pion mass @xmath124 , the @xmath78 sigma term @xmath214 and factored scattering length @xmath215 . the linear density correction of the pion mass stems from the scattering length . since the isosinglet @xmath78 scattering length is known to be a small number compared to the inverse pion mass , @xmath216 , the leading correction is as small as 5% at the saturation density @xmath217 . in fig . [ mass_dep ] , we show the density dependence of the in - medium pion mass as a function of the density normalized by the normal nuclear density @xmath14 . in the figure , the dotted line shows the result up to the leading linear density and the solid line is for the result containing the next - to - leading order . we take the following values of the in - vacuum quantities ; @xmath218 , @xmath219 , @xmath220 , @xmath221 and @xmath222 @xcite . one can see from fig . [ mass_dep ] that the nlo correction is not small . the main contribution comes from the double scattering terms . the density correction of the pion mass at twice or three times the normal nuclear density becomes about 15 to 20% . since the in - medium chpt which is a low energy effective theory it would be not applicable in higher density region , nevertheless we expect that this theory would be applicable up to 3@xmath14 where fermi momentum corresponds to about 400 mev . next , the wave function renormalization is obtained by evaluating the derivative of the self energy @xmath223 with respect to @xmath224 at @xmath225 . again since the difference between the in - medium and in - vacuum pion masses in the self energy is counted as the higher order of the density expansion , we evaluate the derivative of the self energy at @xmath226 for the present purpose : @xmath227}^{-1 } \nonumber \\ & = & \left [ 1 + \frac{2\rho}{f^2 } ( c_2 + c_3 - \frac { g_a^2}{8 m_n } ) - \left(\frac{1}{8m_\pi^2 } - 2 ( 2c_1 - c_2 - c_3 ) ( c_2 + c_3)\right ) \frac{8 m_\pi^2 k_f^4}{3 \pi^4 f^4}g(0 ) \right]^{-1 } \nonumber \\ & = & 1 + \frac{2\rho}{f_\pi^2 } \big { ( } \frac{\sigma_{\pi n } } { 2 m_\pi^2 } + \frac { f_\pi^2 b^{+}}{2 m_\pi^2 } \big { ) } + \frac{m_\pi^2 k_f^4}{\pi^4 f_\pi^4 } \left [ \frac{1}{8m_\pi^2 } + 2\big { ( } \frac { f_\pi^2b^{+}}{2m_\pi^2 } - \frac{g_a^2}{8m_n } \big { ) } \big { ( } \frac{\sigma_{\pi n}}{2 m_\pi^2 } + \frac{f_\pi^2 b^{+}}{2m_\pi^2 } - \frac{g_a^2}{8m_n } \big { ) } \right ] . \nonumber \\\end{aligned}\ ] ] in the calculation of the wave function renormalization , we encounter a singularity in the derivative of function @xmath228 at @xmath229 . as discussed in detail in appendix [ infra ] , this singularity may be considered as infrared singularity in quantum field theory and brings an problematic density dependence starting from @xmath230 , which is inconsistent with the low density expansion starting from @xmath231 . thus , we should expect certain cancellation of the singularity . the infrared singularity should cancel when one calculates scattering rates , not in the amplitude itself . therefore we have dropped the singular term containing the derivative of function @xmath228 . at the linear density , we find that @xmath232 which implies that the wave function is enhanced by 40% at the saturation density . the density dependence of wave function renormalization is shown in fig.[wave ] . as one can see in the figure , the density correction of the wave function is not so small , being an order of 50% at the saturation density . it would be very interesting , if one could observe the strong enhancement of the wave function renormalization phenomenologically , such as formation cross sections of deeply bound pionic atoms . later , we will discuss again possible observation of the wave function renormalization itself in the in - medium @xmath233 decay into @xmath234 . we evaluate the temporal pion decay constant in medium . according to the discussion in the previous section , the in - medium decay constant is given by the wave function renormalization @xmath136 and the one - particle irreducible vertex correction @xmath235 as @xmath236 . the wave function renormalization has been already evaluated in the previous subsection , here let us calculate the vertex correction @xmath237 in density expansion . in fig . [ decaygraph ] , we show the relevant feynman diagrams for the density corrections up to @xmath175 . the wavy lines in the fig.[decaygraph ] denote axial vector currents . the leading graphs @xmath179 are shown in fig . [ decaygraph](a ) , while the next - to - leading order contributions @xmath203 are given in fig . [ decaygraph](b ) . since the temporal decay constant is given by the matrix element of the time - component of the axial vector current @xmath238 , we evaluate the one - particle irreducible ( 1pi ) matrix element @xmath239 for the rest pion @xmath240 . the loop calculation is essentially same as the calculation done in the self - energy . here we will show only the definition and result of each contribution . the contribution from the left diagram @xmath241 in fig . [ decaygraph](a ) reads @xmath242 } = - q_0\frac { g_a^2}{4 f m_n } \rho .\ ] ] the vertex correction coming from the right diagram in fig . [ decaygraph](a ) is @xmath243 } = q_0 \frac{2 \rho}{f } ( c_2 + c_3 ) .\ ] ] for the left diagram in fig . [ decaygraph](b ) , we have @xmath244 with symmetric factor @xmath189 : @xmath245 function @xmath246 is defined in eq . . the vertex corrections represented as the middle and right diagrams in fig . [ decaygraph](b ) are calculated as @xmath247}\nonumber \\ & = & - \frac{q_0 k_f^4}{3 \pi^4 f^4 } g(a^{2 } ) \\ \hat{f}_5 q_0 & = & - \int \frac{d^4 p}{(2 \pi ) ^4 } \frac{d^4 k}{(2 \pi ) ^4 } { \rm tr } \big { [ } ( -i a^{(2)}_{\pi a } ) i g_m ( k- \frac{p}{2 } ) ( -i a^{(2)}_{\pi \pi } ) ig_m ( k+ \frac{p}{2 } ) id_\pi ( q - p ) \big { ] } \nonumber \\ & = & ( c_2 + c_3 ) \big { ( } 2c_1m_\pi^2 - ( c_2 + c_3)q_0 ^ 2 \big { ) } \frac{8 q_0 k_f^4}{3 \pi^4 f^3 } g(a^{2})\end{aligned}\ ] ] where @xmath248 and function @xmath228 is defined in eq . . the vertex correction @xmath235 should be evaluated at @xmath249 in principle , because the decay constant is the matrix element of the axial current of the in - medium pion state having the on - shell condition @xmath250 . but , in practice , we are allowed to evaluate the vertex correction at @xmath251 , because the difference is in the higher density orders . summing up all the contributions , we obtain the density dependence of @xmath252 up to @xmath253 : @xmath254\frac{8 m_\pi^2 k_f^4}{3 \pi^4 f^3 } g(0 ) \nonumber \\ & = & f_\pi \big { [ } 1 - \frac{2 \rho}{f_\pi^2 } \big { ( } \frac{\sigma_{\pi n } } { 2 m_\pi^2 } + \frac { f_\pi^2 b^{+}}{2 m_\pi^2 } \big { ) } + \frac{g_a^2 k_f^4}{6 \pi^4 f_\pi^4 } f(\frac{m_\pi}{2k_f } ) \nonumber \\ & & - \frac{m_\pi^2 k_f^4}{\pi^4 f_\pi^4 } \left[\frac{1}{8 } + \big { ( } \frac { f_\pi^2b^{+}}{2m_\pi^2 } - \frac{g_a^2}{8m_n } \big { ) } \big { ( } \frac{\sigma_{\pi n}}{2 m_\pi^2 } + \frac{f_\pi^2 b^{+}}{2m_\pi^2 } - \frac{g_a^2}{8m_n } \big { ) } \right ] \end{aligned}\ ] ] where we have used @xmath255 . with @xmath235 , we can calculate in - medium pion decay constant : @xmath256}.\end{aligned}\ ] ] the lecs are determined by the @xmath78 sigma term @xmath214 and the isoscalar @xmath78 scattering length @xmath257 . we show in fig.[decay_dep ] the density dependence of the temporal decay constant . in this figure , the dotted line is for the result of the leading linear density and the solid line is for the result including the next - to - leading order @xmath2 . we find that nlo correction gives a few percents change in low density region but in high density region about three times the normal nuclear density nlo correction contributes 10 percents and not small . let us make some comments on the perturbative expansion in terms of the fermi momentum . in the in - medium chpt , the expansion parameter is to be @xmath258 and it amounts to around @xmath259 to @xmath260 even at higher density region than the saturation density . with this small value , we could say that the perturbative expansion in terms of the fermi momentum may have good convergence . nevertheless , if one wants to judge whether the fermi momentum expansion is good or not in nuclear matter , one should describe the nuclear matter and reproduce the saturation properties first , and see the higher density effects . the nuclear matter could not be described essentially in perturbation theory of the in - vacuum fields . one could need nonperturbative treatment to obtain the nuclear matter . in such a case , naive perturbative expansion in terms of the fermi momentum would break down before the nuclear density . the pseudo - scalar coupling @xmath261 is also calculated as @xmath262 in fig . [ fig : gpi ] , we list up the feynman graphs which contribute to the one - particle irreducible vertex correction @xmath126 . performing the same calculation as the pion decay constant @xmath235 , we obtain the density dependence of the in - medium pseudo - scalar coupling : @xmath263 it is known that the gell - mann oakes renner ( gor ) relation is satisfied in medium up to the linear density ( see , for example ref . @xcite ) . having obtained the density dependences of the pion decay constant and the pion mass in medium up to @xmath203 , we are going to examine the gor relation satisfied at @xmath203 . the density dependence of the quark condensate is obtained in ref . @xcite up to @xmath203 . using the expressions of the in - medium quantities , we obtain @xmath264 } \\ g_1 ( x^2 ) = g ( -x^2 ) & = & \frac{3}{8 } - \frac{x^2}{4 } -x \arctan \frac{1}{x } + \frac{x^2}{4 } ( x^2 + 3 ) \ln |\frac{1+x^2}{x^2}| . \label{gor}\end{aligned}\ ] ] from this equation , we find that the gor relation is not satisfied at this order and the double scattering terms break the relation . the reason of the breaking is the following : the gor relation connects the chiral condensate and the pion quantities . the chiral condensate is calculated in the soft limit by taking @xmath265 in the correlation function , while the pionic quantities are evaluated for pion on the mass shell . off the chiral limit , where pion is massive , the two energy points are different . this difference makes the gor relation broken . in the chiral limit in vacuum , one has the glashow weinberg ( gw ) relation @xmath266 . this relation is known to be satisfied in medium in the chiral limit and at the linear density @xcite . here we also directly check whether the relation is satisfied in the order @xmath203 off the chiral limit : @xmath267}. \label{gwrel}\ ] ] as seen in this equation , the gw relation is broken by the double scattering term and the breaking terms correspond to the right - hand side in eq.([gor ] ) . it is also interesting to discuss the in - medium sum rule for the in - medium quark condensate derived in ref . @xcite model - independently in the chiral limit : @xmath268 where @xmath269 is the label of the zero - modes in nuclear matter , such as the pions and particle - hole excitations , and summation is taken over all of the zero modes . it has been shown that , in the linear density , only the pion zero mode can contribute to the sum rule and one obtains the in - medium glashow weinberg relation @xmath270 . the other zero modes , such as particle - hole excitations , contributes only as higher density correction . as discussed in ref . @xcite , the @xmath88 correlation effects come into the chiral effective theory from the order of @xmath271 . thus , we expect that the pion mode dominates the sum rule beyond the linear density but below the square density order in the chiral limit . we can check it by taking the chiral limit in eq . . this means that the gw relation is satisfied up to @xmath203 in the chiral limit : @xmath272 with this relation , realizing that both @xmath273 and @xmath120 decrease as the density increases , we find that the chiral condensate decreases more rapidly than the pion decay constant . although one might have some corrections from @xmath271 contributions , this observation could be a hint that at a certain higher density the chiral condensate would become zero before the pion decay constant would be zero . in this situation , even though the quark condensate is zero the chiral symmetry is spontaneously broken . the neutral @xmath233 meson decays dominantly into @xmath275 in vacuum through the quantum effect known as the axial anomaly . here let us estimate the medium modification of the decay rate of @xmath233 to @xmath234 by following the formulation presented in sec.[decayconst ] . the decay amplitude may be written , as requested by lorentz invariance , in the form @xmath276 where @xmath277 stands for the decay amplitude in medium , @xmath278 is the totally antisymmetric tensor , @xmath279 and @xmath280 are the polarization vectors of the emitted photons , and @xmath281 and @xmath282 are the momenta of the photon . according to the argument given in sec.[decayconst ] , the in - medium amplitude @xmath283 can be written as @xmath284 with the wave function renormalization @xmath136 and the medium correction of the one - particle irreducible @xmath285 vertex @xmath286 . in the nuclear medium , the @xmath233 meson may decay into @xmath234 through non - anomalous diagrams because of the presence of nucleons , which can bring out intrinsic parity violated in the @xmath163 process . nevertheless , in ref . @xcite , it has been found that there are no medium corrections to the @xmath163 amplitude up to @xmath287 , that is , the linear density correction in the density expansion . the argument is based on the spin and flavor structure of the vertices . the nucleon one loop with the @xmath288 contact vertex should vanish because the contact vertex contains @xmath289 and there are no other vertices can bring @xmath289 in the nucleon loop , which makes the trace of the dirac matrices vanish . the diagrams in which the photons and the neutral pion emit sequentially from a nucleon also turn out to vanish according to careful calculation of the flavor matrix . hence , the in - medium correction up to the linear density vanishes and the in - medium vertex correction @xmath290 amplitude is equivalent to the in - vacuum amplitude @xmath291 . this leads to the conclusion that the in - medium correction of the decay amplitude @xmath292 can be calculated solely by the wave function renormalization as @xmath293 ignoring the small density correction of the pion mass in the phase space of the decay process , we find that the modification of the @xmath233 decay rate to @xmath234 in nuclear matter is given by the wave function renormalization as @xmath294 this implies that the @xmath233 decay into @xmath234 in nuclear matter measures directly the in - medium renormalization of the @xmath233 wave function in lower density . using the result discussed in the previous subsections , we obtain in the linear density approximation as @xmath295 at normal nuclear density the decay rate is 1.4 times enhanced and the partial width becomes about 10 ev . we have discussed the in - medium pion properties , such as the pion decay constant , the pion mass and the wave function renormalization based on the in - medium chiral perturbation theory . first , we have provided a general formalism of the in - medium chiral perturbation theory and have discussed an expansion in terms of fermi momentum . assuming that the renormalization for the in - vacuum quantities is performed in an appropriate way , we use the observed values to determine the low energy constants ( lecs ) in the chiral lagrangian . since we have used the physical values , the higher order corrections for the momentum expansion are implicitly included in the calculation . thus , we focus on the expansion of the fermi momentum of the physical quantities , which are calculated by the qcd current green functions . to calculate the in - medium quantities , we carefully define the in - medium pion state , and we have found that the in - medium wave function renormalization plays an essential role to define the in - medium coupling constants , such as the decay constant and pseudo - scalar coupling constant . we have evaluated the density dependence of the decay constant , the pion mass , the pion wave function renormalization and the pseudo - scalar coupling including the next - to - leading order of the density expansion beyond the well - known linear density approximation based on the in - medium chiral perturbation theory . we have found that the @xmath2 corrections give a few percents changes in the low density region for the decay constant and the pion mass , while in higher density such as three times saturation density , the corrections can be as the order of 10 to 20 percents and can give significant contribution . we have also found that the wave function renormalization is enhanced as largely as 50 percents at the saturation density . the main contribution among the corrections comes from the double scattering term . in addition , we have checked whether the low energy theorems , the gell - mann oakes renner relation and the glashow weinberg relation are satisfied in medium beyond the linear density approximation . we have found that these relations are not satisfied at @xmath2 off the chiral limit . the origin of the breaking is that we use the different energy values to evaluate the pion quantities and the chiral condensate ; we take the soft limit to obtain the chiral condensate , while we take the pion on shell point , that is @xmath296 , to evaluate the pion quantities . finally , we have discussed the density dependence of the @xmath297 decay width . considering the spinor and flavor vertex structure of the chiral interactions , we have found that the density dependence of the @xmath274 width comes from the wave function renormalization alone at linear density order . with this observation , the wave function renormalization @xmath136 would be measured directly in the decay process . this work was partially supported by the grants - in - aid for scientific research ( no . 25400254 , no . 24105706 and no . 24540274 ) . in order to make the perturbative calculation simple , we use the following parametrization of pion field @xmath25 given by ref . @xcite , as we have done in the previous paper @xcite : @xmath97\ ] ] where @xmath98 satisfies @xmath99 the expansion of the chiral field @xmath25 in terms of the pion field is given in ref . @xcite as @xmath298 and for the chiral field @xmath108 @xmath299 in this parametrization , the soft pion theorems , such as adler s condition , remain satisfied in simple perturbation theory . in addition , we show chiral interactions which we use in the following calculation . with this parametrization , one finds the explicit expression of each interaction term in the chiral lagrangian . the chiral interactions for pions are the followings : @xmath300 the interactions containing nucleon in the bilinear form read @xmath301 = - \frac{1}{2f } \gamma^\mu \phi^i a_\mu^j \epsilon^{ijk } \tau^k \\ a_{\pi a}^{(2 ) } & = & -\frac{2c_2}{f m_n^2 } \partial_{\mu } \phi^i a_{\nu}^i \partial^{\mu } \partial^{\nu } + \frac{2c_3}{f } \partial_{\mu } \phi^i a^{\mu i } .\end{aligned}\ ] ] in this appendix , we show the details of the calculations of the self energy shown in sec.[results ] . we consider the isospin symmetric limit and the symmetric nuclear matter . the external momentum is fixed as @xmath302 for simplicity . we write the nucleon propagator in the fermi sea as @xmath303 for the pauli - blocked nucleon propagator in the symmetric nuclear matter @xmath182 , we use the following expression , which is equivalent to eq . : @xmath304 we also define the propagator @xmath305 in which the spinor structure and the nucleon energy are factored out from @xmath182 . in the following calculation , the trace symbol @xmath306 implies to take trace in both spinor and isospin spaces , while @xmath307 means the trace for only the spinor space . we first calculate the pion self energy given by the diagram in the left of fig.[self](a ) : @xmath308 } \nonumber \\ & = & - \int \frac{d^4 p}{(2 \pi ) ^4 } { \rm tr } \big { [ } ( -i ) \frac{g_a}{2f } \gamma_5 i \slash{q } \tau^i i \frac{\slash{p}+\slash{q } + m_n } { 2e({\bf p}+{\bf q } ) } g_r ( p+q ) \nonumber \\ & & \phantom{- \int \frac{d^4 p}{(2 \pi ) ^4 } { \rm tr } \big { [ } } \times ( -i ) \frac{g_a}{2f } ( - \gamma_5 ) i \slash{q } \tau^j i \frac{\slash{p } + m_n } { 2e({\bf p } ) } g_r ( p ) \big { ] } \nonumber \\ & = & - \delta^{ij } \frac{g_a^2}{2f^2 } \int \frac{d^4 p}{(2 \pi ) ^4 } { \rm tr}_s \big { [ } \slash{q } ( \slash{p}+\slash{q } -m_n ) \slash{q } ( \slash{p}+m_n ) \big { ] } \frac{g_r ( p+q ) g_r ( p)}{2e({\bf p } ) 2e({\bf p+q } ) } .\ \ \ \ \end{aligned}\ ] ] the spinor trace can be calculated as @xmath309 } = 4(p\cdot q ) \ { ( p+q)^2 -p^2 \ } -4q^2 ( p^2 + m_n^2 ) .\end{aligned}\ ] ] for the calculation of the integral , we perform the @xmath310 integral first using the cauchy theorem along a contour of upper semicircle in the complex @xmath310-plane : @xmath311 } \nonumber \\ & = & 8i q_{0}^{2 } m_{n } \theta(k_{f}-|{\bf p}| ) . \nonumber \end{aligned}\ ] ] here we have picked up only the terms including the step function , because the vacuum part should be subtracted for the calculation of the in - medium quantities , and in the final expression we have taken the leading term in the @xmath312 expansion . finally we obtain @xmath313 where we have used @xmath314 . next we consider the self energy given by the right diagram in fig . [ self](a ) : @xmath185 } \nonumber \\ & = & \int \frac{d^3 p}{(2 \pi ) ^3 } ( -4i ) \big { ( } \frac{8b_0 c_1 m_q}{f^2 } -\frac{2c_2}{f^2 m_n^2 } ( q\cdot p)^2 - \frac{2c_3}{f^2 } q^2 \big { ) } \theta ( k_f - |{\bf p}| ) \nonumber \\ & = & - \frac{2 i \rho}{f^2 } ( 4 c_1b_0 m_q - ( c_2 + c_3 ) q_0 ^ 2 ) .\end{aligned}\ ] ] here we have taken the leading term of the @xmath315 expansion . next we calculate the next - to - leading contributions in the density expansion . here we denote the pion propagator as @xmath316 . first we consider the left diagram in fig . [ self](b ) @xmath317 : with symmetric factor @xmath189 , the contribution is written as @xmath318 where , @xmath190 is defined as @xmath191 } .\ ] ] using the on - shell conditions @xmath319 , @xmath320 for the nucleon propagators in the fermi sea , the spinor trace is calculated as follows @xmath321 } = -8 m_n^2 k^2 .\ ] ] hence , we obtain @xmath223 : @xmath322 where we have used the following integral formula : @xmath323 with function @xmath324 : @xmath325 we consider the contribution coming from the middle diagram in fig . [ self](b ) , @xmath326 : @xmath327 } \\ & = & - \int \frac{d^4 p}{(2 \pi ) ^4 } \frac{d^4 k}{(2 \pi ) ^4 } { \rm tr } \big { [ } \frac{\epsilon^{ikm } \tau^m}{4f^2 } ( 2\slash{q } - \slash{p } ) i g_m ( k- \frac{p}{2 } ) \frac{\epsilon^{kjm } \tau^m}{4f^2 } ( 2\slash{q } - \slash{p } ) ig_m ( k+ \frac{p}{2 } ) \big { ] } \\\end{aligned}\ ] ] performing integration in terms of @xmath328 and @xmath329 , we obtain @xmath330 which provide also @xmath331 , @xmath332 . with these expression , the spinor trace is calculated as @xmath333 } = 32 ( q\cdot k)^2 - 8 ( q \cdot p)^2 + 8 q^2 p^2 \end{aligned}\ ] ] in the leading term of the @xmath312 expansion , we have the relation @xmath334 , @xmath335 and we obtain @xmath336 using the following integral formula @xcite @xmath337 with @xmath338 , we get @xmath339 where we have used @xmath340 finally we consider @xmath341 given by the right diagram in fig . [ self](b ) . this graph corresponds to the double scattering correction called as ericson - ericson term @xcite and is calculated as : @xmath342 } \\ & = & \int \frac{d^3 p}{(2 \pi ) ^3 } \frac{d^3 k}{(2 \pi ) ^3 } { \rm tr}_s \big { [ } ( \slash{k } - \frac{\slash{p}}{2 } + m_n ) ( \slash{k } + \frac{\slash{p}}{2 } + m_n ) \big { ] } \frac { \delta^{ij}}{2m_n^2 } id_\pi ( p+q ) \\ & & \times \big { ( } \frac{8c_1 b_0 m_q } { f^2 } - \frac{2c_2}{f^2 m_n^2 } \ { q\cdot ( k- \frac{p}{2 } ) \ } \ { ( p+q)\cdot ( k-\frac{p}{2 } ) - \frac{2c_3}{f^2 } q \cdot ( p+q ) \ } \big { ) } \\ & & \times \big { ( } \frac{8c_1 b_0 m_q } { f^2 } - \frac{2c_2}{f^2 m_n^2 } \ { q\cdot ( k+ \frac{p}{2 } ) \ } \ { ( p+q)\cdot ( k+\frac{p}{2 } ) - \frac{2c_3}{f^2 } q \cdot ( p+q ) \ } \big { ) } \\ & & \times \theta ( k_f - |{\bf k } - \frac{{\bf p}}{2}| ) \theta ( k_f - |{\bf k } + \frac{{\bf p}}{2 } |).\end{aligned}\ ] ] the spinor trace is reduced to @xmath343 } & = & 4 k^2 - p^2 + m_n^2 = 8m_n^2 - 2 p^2\end{aligned}\ ] ] finally @xmath344 is obtained as @xmath345 in the last equality , we have performed the integral in the same way as @xmath195 . the loop integral of the double scattering term reads @xmath346 where @xmath347 is the energy of the external line and we have taken @xmath348 for the external momentum . performing the following integral @xmath349 where @xmath350 , we write the loop integral @xmath351 as @xmath352 where we have define @xmath353 . the integral with respect to @xmath46 can be done straightforwardly and we obtain @xmath354\end{aligned}\ ] ] this function is finite in the limit of @xmath355 , _ i.e. _ @xmath356 . but the derivative of @xmath357 with respect to @xmath224 has a singularity at @xmath356 . the derivative is obtained as @xmath358\end{aligned}\ ] ] this function is logarithmically divergent at @xmath356 . in addition this loop integral starts from @xmath230 in the expansion of small @xmath359 . this behavior is contradict with the low density expansion where the leading order should be @xmath207 . consequently , the wave function renormalization could have such a strange density dependence . thus this contribution is pathologic . we do not take the exact limit of @xmath360 in the evaluation . nevertheless , we have to deal with the singularity , because the wave function renormalization would have inconsistent density dependence with the low density expansion and we may make evaluation of the physical quantities very close to the singular point , where the results are numerically unreliable . the origin of the singularity seen in the derivative of the loop function at @xmath361 can be identified when one performs derivative of eq . in terms of @xmath347 and takes @xmath362 : @xmath363 the integral gets divergent when the integrand takes @xmath364 at the end point of the integral . this singularity is very similar with the infrared divergence in quantum field theory . such infrared divergence should be cancelled with other diagrams emitting soft pions , when one calculates scattering rates , not in the amplitude itself . thus , the singularity found in the derivative of the loop function should be cancelled with other terms , when one calculates cross sections . relying on the above argument , we simply drop the term which includes the infrared singularity in this work . j. m. alarcon , j. martin camalich and j. a. oller , phys . d * 85 * , 051503 ( 2012 ) . j. m. alarcon , l. s. geng , j. martin camalich and j. a. oller , arxiv:1209.2870 [ hep - ph ] . et al _ , phys . b*694 * , 473 ( 2011 ) . charap , phys . d * 3 * , 1998 ( 1971 ) . i.s . gerstein , r. jackiw , s. weinberg , and b.w . lee , phys.rev . d * 3 * , 2486 ( 1971 ) . a.l . fetter and j.d . walecka , _ quntaum theory of many - particle systems _ ( dover , new york , 2003 ) .

we present a measurement of the systemic proper motion of the large magellanic cloud ( lmc ) from astrometry with the high resolution camera ( hrc ) of the advanced camera for surveys ( acs ) on the hubble space telescope ( ) . we observed lmc fields centered on 21 background qsos that were discovered from their optical variability in the macho database . the qsos are distributed homogeneously behind the central few degrees of the lmc . with 2 epochs of hrc data and a year baseline we determine the proper motion of the lmc to better than% accuracy : . this is the most accurate proper motion measurement for any milky way satellite thus far . when combined with hi data from the magellanic stream this should provide new constraints on both the mass distribution of the galactic halo and models of the stream .

astro-ph0508457

the magellanic clouds ( mcs ) are the closest , easily observable galaxies from our vantage point in the milky way . they are of fundamental importance for studies of stellar populations , the interstellar medium , and the cosmological distance scale . moreover , due to their locations at 50 kpc ( lmc ) and 61 kpc ( smc ) from the sun , and @xmath4 kpc from the galactic plane , they provide one of our best probes of the composition and properties of the galactic dark halo . precise measurements for the proper motions ( pms ) of the magellanic clouds combined with observational knowledge of the distribution and velocities of hi gas in the magellanic stream can constrain models of the shape and radial density distribution of the galactic dark halo as well as theoretical models for the formation of the magellanic stream . few developments in modern astronomy are as important as the discovery of cosmic dark matter . most of the matter in the universe is invisible to astronomers . in the milky way , the dark component is about twenty times more massive than the visible disk of stars and gas . the dark matter forms a vast , diffuse halo that occupies more than a thousand times the volume of the stellar disk . the composition of this dark halo is unknown , but it may comprise a mixture of ancient degenerate dwarf stars and exotic , hypothetical elementary particles ( for a review see alcock 2000 ) . the investigation of the galactic halo is complicated by our immersion deep inside it . interior to the solar circle the rotation curve may be measured with confidence ( see e.g. fich & tremaine 1991 ) . while the estimates out to 20 kpc are not as robust , these observations indicate that the rotation curve is flat ( merrifield 1992 ; yoshiaki & rubin 2001 ) . much has been learned about the halo in the direction out of the plane from the study of the motions of gas clouds and stars . these imply that the halo is more spherical than disk - like , but no precise measurement for its axial ratio exists ( van der marel 2001 ) . analyses of the kinematics of some globular clusters and small satellite galaxies suggest that the halo extends out to @xmath5 kpc ( e.g. , wilkinson & evans 1999 ) . for the large majority of these objects there is only radial velocity data , and where available , proper motions are much less precisely determined than the radial velocities . so while these studies constrain the total halo mass , they say little about its shape or distribution . spherical symmetry and a particular density distribution are generally _ assumed_. the magellanic stream is an hi emission feature that spans more than @xmath6 across the sky ( br " uns et al . it consists of gas that trails the magellanic clouds as they orbit the milky way and provides an opportunity for detailed study of tidal disruption of galaxies as well as the milky way dark halo . many detailed theoretical models have been constructed for the magellanic stream ( gardiner & noguchi 1996 ; lin , jones & klemola 1995 ; heller & rohlfs 1994 ) . these models describe how the magellanic clouds orbit the milky way , and how they lose material through tidal effects and other physical mechanisms . the model parameters are adjusted to best reproduce the observed density , morphology and velocities of the hi gas seen along the magellanic stream . the most sophisticated calculations in the class of models that invoke tidal stripping are those by gardiner & noguchi ( 1996 ) . in their models the lmc and smc form a gravitationally bound system that orbits the milky way . the magellanic stream represents material that was stripped from the smc @xmath7 gyr ago . this was the time of the previous perigalactic passage , which coincided with a close encounter between the clouds . in magellanic stream models , the orbit of the clouds is one of the prime variables that is adjusted to fit the data . as a result of newton s law , the orbit is fully determined by : ( a ) the position of the clouds on the sky ; ( b ) the present distance of the clouds from the sun ; ( c ) the radial velocities of the clouds ; ( d ) the proper motions of the clouds ; and ( e ) the gravitational potential of the galactic dark halo . the first three of these are known relatively well ; the latter two are not . even small differences in the proper motions can give vastly different orbits for the clouds , especially when considered over the long period of time over which the magellanic stream developed . the usual approach has therefore been to estimate the proper motions of the clouds from the properties of the magellanic stream , assuming a fixed galactic halo gravitational potential ( usually a simple spherical isothermal halo ) . the results imply that the magellanic clouds are just past their pericenter ( itself at @xmath8 kpc from the galactic center ) , that the apocenter to pericenter ratio is @xmath9 , and the orbital period is @xmath7 gyr . the inferred tangential velocity in the galactocentric rest - frame has differed substantially from model to model , ranging from , e.g. , @xmath10 km / s ( heller & rohlfs 1994 ) to @xmath11 km / s ( gardiner & noguchi 1996 ) . it was realized by heller & rohlfs ( 1994 ) and lin , jones & klemola ( 1995 ) that the arguments used to model the magellanic stream can be turned around . if the proper motions of the magellanic clouds are known , then the gravitational potential of the galactic dark halo can be determined from models of the magellanic stream . they explored this using the available proper motion data , but found that its accuracy was insufficient to obtain strong constraints on the dark halo properties . kroupa et al . ( 1994 ) found that a proper motion better than @xmath12 is required for strong results . so far the proper motion of the lmc has not been known accurately enough to strongly constrain properties of the milky way dark halo . measurements , however , are steadily improving and determinations are available from the following sources : kroupa et al . ( 1994 ) , using stars from the ppm catalogue ; jones et al . ( 1994 ) , using photographic plates with a 14 year epoch span ; kroupa & bastian ( 1997 ) , using hipparcos data ; drake et al . ( 2002 ) , using data from the macho project ; anguita , loyola & pedreros ( 2000 ) and pedreros et al . ( 2002 ) , using ccd frames with an 11 year epoch span ; and momany & zaggia ( 2005 ) , using the usno ccd astrograph all - sky catalog ( ucac2 ) . the measurements are all consistent with each other to within the error bars with the exception of two outliers : anguita et al . ( 2000 ) and momany & zaggia ( 2005 ) present values of @xmath13 . when these two results are ignored , the weighted average of the remaining measurements yields proper motions towards the west and north of @xmath14 ( van der marel et al . 2002 ) i.e. to @xmath15% accuracy . this implies @xmath16 km / s , which is consistent with most of the published magellanic stream models , but is not accurate enough to discriminate between them . the proper motion of the smc is much less well known than that of the lmc , and only one reasonably accurate measurement exists . kroupa & bastian ( 1997 ) obtain a value of @xmath17 for the smc with an error of ( 0.84 , 0.75)@xmath18 . a sound measurement of the proper motion of the lmc requires all of the following key factors : ( 1 ) an instrument which can perform the astrometry with adequate precision ; ( 2 ) a reference frame consisting of point sources , such as quasars , distributed widely behind the clouds ; ( 3 ) secure determination of the membership of stars in the lmc ; and ( 4 ) a reliable kinematic model of the internal rotation of the clouds . the previous estimates of the motions ( referred to above ) do not satisfy all of these requirements . however , a much improved measurement is now possible . some teams have identified qsos behind our neighboring galaxies in order to provide good inertial reference frames against which the motions of these galaxies can be measured ( geha et al . 2003 ; dobrzycki et al . 2002 ; 2003 ) ; the @xmath0 has been shown to be very stable for astrometry ( anderson & king 2003 ) and in particular the advanced camera for surveys ( acs ) has higher resolution , and is better calibrated and more stable than even wfpc2 ( anderson & king 2004 ; hereafter ak04 ) . we were awarded two epochs on the high resolution camera ( hrc ) on the acs in cycle 11 and cycle 13 for a study of the proper motions of the mcs . we report our proper motion result for the lmc in this paper . smc results as well as implications for the mc system as a whole will be presented in subsequent papers . the paper is outlined as follows : 2 deals with the sample and the observations ; 3 describes our analysis ; 4 presents the center of mass proper motion of the lmc ; 5 is a discussion and 6 is a summary . a total of 54 qsos were identified behind the magellanic clouds from their optical variability in the macho database ( geha et al . 2003 ) . out of this sample , 44 qsos were behind the lmc . in our first epoch program ( go-9462 ; cycle 11 ) , we proposed to image fields around these qsos with acs / hrc in snapshot mode . in a snapshot proposal the targets are randomly selected from the proposed list , but with manually assigned priorities . we successfully observed 34 of the 44 lmc qsos in the first epoch . in the second epoch program go-10130 ( cycle 13 ) we re - imaged the targets that were observed in epoch 1 and for maximally efficient use of @xmath0 resources we again chose to observe in snapshot mode . we achieved a completion rate of @xmath19% for epoch 2 and @xmath20% overall , so that our final sample consists of 21 targets . table 1 describes this sample . it lists the i d number for each qso ( each qso field will be referred to by its i d number subsequently ) , as well as its macho i d ( for referencing purposes with geha et al . 2003 ) , ra , dec ( j2000 ) , @xmath21 magnitude ( an average of our measurements from both epochs ) and redshift . table 1 also lists the execution date , visit number , exposure time and orientation for each epoch s observations , as well as the time baseline that was ultimately achieved for each target . these are discussed in 2.2 . figure [ figure1 ] is a plot of the qsos behind the lmc . qsos for which we obtained two epochs of data are marked with white circles . their distribution behind the lmc is reasonably uniform . qsos which did not make it through one or both of our snapshot programs are marked with white squares . in a snapshot program the targets are executed in the cycle so as to fill gaps in the @xmath0 schedule . the majority of our observations were scheduled in the first part of the cycle . it is interesting to study some of the specifics of the scheduling times and the quantities that they affect . figure [ figure2](a ) is a histogram of the time baseline for each qso field ( in years ) . the median baseline that we achieved is @xmath22 years but there is a spread from @xmath23 to 2.7 years . for observations with @xmath0 , the date of observation determines the orientation of the detector with respect to the sky ( see the @xmath0 primer , karakla 2005 ) . again , because of the snapshot nature of our project , the orientations were not fixed _ a priori _ and hence were not identical from one epoch to the next . figure [ figure2](b ) is a histogram of the orientations on the sky ( orientat ) of each of our qso fields for both epochs ( orientat is defined as the angle on the sky , east of north , of the detector @xmath24-axis ) . figure [ figure2](c ) is a histogram of the relative telescope rotation between the epochs for each qso field , plotted between -180and 180 . it shows that the observations were not taken at completely random orientations , for there is a broad peak near @xmath25 . nonetheless , there is considerable randomness in the orientation of the telescope with respect to the sky , both in an absolute and in a relative sense . this is good because it implies that any possible systematic proper motion errors tied to the ccd frame will average to zero roughly as @xmath26 . the possibility of actual systematic effects are investigated in detail in 3.3 . our observing strategy involved the use of the hrc because it provides the highest resolution available ( the average pixel scale is 28.27 mas / pixel - see ak04 ) and is well calibrated for astrometry . it is also well sampled even in the bluest filters and thus a star s integrated flux does not depend strongly on where it lands in a pixel . it is optimized towards the visual and red part of the spectrum and so for our main astrometry goals we chose the f606w filter which is a broad @xmath21 filter with high throughput . this filter is also not too red and thus avoids possible issues with the hrc red halo at wavelengths @xmath27 ( see the acs instrument handbook , pavlovksy et al . in addition to f606w , we used f814w which is a broad @xmath28 filter for imaging in the first epoch only , not for the main astrometry , but to be able to place sources on a color - magnitude diagram ( cmd ) . we thought that this might help assist in the rejection of foreground stars . for the f606w filter , in epoch 1 , exposure times were chosen so as to achieve a s / n of at least 100 for the qsos based on the known macho magnitudes . in epoch 2 the exposure times were chosen to be somewhat larger than in epoch 1 to take into account the fact that the macho sources were considerably blended and that the actual @xmath21 magnitudes measured from first epoch @xmath0 data tended to be dimmer than the macho magnitudes . some of the isolated qsos actually turned out to be brighter because of their intrinsic variability . note that on the timescale of our proper motion measurements , the expected photometric variability is only approximately a few tenths of a magnitude in the @xmath21-band and so we quote an average from both epochs in table 1 . for the f814w filter , the exposure time was set to 1.7 minutes total which is sufficient for photometry . we chose the same dither pattern in the f606w filter for both epochs . this amounted to 8 exposures per epoch for each qso ( 16 exposures total for both epochs ) taken using two four - point half - pixel shifts which were shifted relatively by an integer 8-pixel amount . figure [ dithers ] is a schematic of this dither pattern . since we did not plan to do astrometry with the f814w images we chose a simple cr - split in order to be able to reject cosmic rays and other transients . in summary , each qso field has 18 images associated with it - 16 in @xmath21-band and 2 in @xmath28-band - and each of the @xmath21-band images were dithered to minimize any pixel - location based systematic errors . for our basic reduction purposes we used the bias - subtracted , dark - subtracted , flat - fielded images ( _ flt.fits ) provided by the stsci / acs data reduction pipeline . ak04 found that these images are well - behaved for astrometric purposes . they are not , however , corrected for geometric distortion . we did not use the geometrically corrected products created by the multidrizzle software ( _ drz.fits ) since they involve a re - sampling that might degrade the astrometry . instead of geometrically correcting the images , we geometrically corrected the star positions on the images , after doing psf - fitting , as we will discuss in 3 below . the hrc has a large amount of geometric distortion due to its off - axis location and it is necessary to understand this fully because a high - precision astrometric program such as ours requires a distortion solution that is at least as accurate as the centroiding . although the distortion is large , it is very stable , and can therefore be calibrated accurately . this calibration has been done by ak04 , and our analysis relies heavily on their work . they and krist ( 2003 ) have shown that for acs / hrc the point spread function ( psf ) does nt vary significantly over the field ( in contrast to wfpc2 ) . although the @xmath0 psf does vary with time due to `` breathing '' ( thermal effects associated with day / night transitions ) and long - term changes in focus , these changes affect mostly the relative amounts of light in the core and the wings of the psf . the actual shape of the core is not altered significantly . developed multi - filter psfs for the hrc using a large dataset of 40 observations each with 20 different pointings of the globular cluster 47 tuc ( @xmath0 calibration programs go-9028 , go-9443 & go-9019 ) . ak04 also developed a detailed solution to the geometric distortion which they modeled as the combination of a large - scale distortion characterized by a fourth - order polynomial and fine - scale variations characterized by a look - up table that can be linearly interpolated to give the adjustment required at any point in the image . they showed that the second and higher order terms are stable over short and long timescales . however , the linear terms vary with time ( even within a single @xmath0 orbit ) due to breathing , differential velocity aberration , and other effects . therefore any two exposures must be brought to a common frame with the use of a 6 parameter linear transformation ( the 6 parameters are scale , skew , translation and rotation ) . when this is done , the geometric distortion correction is accurate and stable over long timescales to @xmath290.005 pixel . for comparison , given our median time baseline ( see figure [ figure2 ] ) , the proper motion of the lmc , estimated on the basis of previous work ( van der marel et al . 2002 ) , corresponds to a position difference of 0.1 pixels between epoch 1 and 2 . thus we have a very good handle on systematics in this regard . to sum up , ak04 have provided software that finds the point sources on an image , fits the psf to determine the raw stellar position and flux , applies a distortion correction to calculate the geometrically corrected stellar position ( i.e. , on a frame that is locally cartesian on the sky ) and finally , applies a photometric correction to the sources for geometric distortion and finite aperture effects . we ran the ak04 code on all 18 images for each field . at this stage , some cuts were made based on crowding ( by specifying how far away the next brightest source can be ) and threshold ( by specifying the minimum peak brightness in excess of the sky background ) . these cuts were not stringent so as to identify the largest possible number of real sources . using the first @xmath21-band image in the first epoch as a reference , we cross - identified the detected sources in all other images with it . we then created a master - list of sources for each qso field by accepting only those sources detected in all 18 images and for which the magnitudes in all @xmath21 images were in mutual agreement to within @xmath30 . this rejected all cosmic rays , other transients and bad pixels that were detected as sources ( note that individual exposures were short so that cosmic ray rejection would be reliable ) . this also rejected all bona fide stars that had fallen off the field in just a few of the images because of the dither pattern or the ccd rotation between epochs 1 & 2 , and stars in which one or more of the measurements were sub - optimal . so in this respect this cut was conservative . however , it ensured that we used exactly the same subset of stars for every image in every epoch , which in turn minimized any potential for systematic differences and errors . table 2 lists for each field the number of real sources ( detected in at least half of the images ) and the number of sources in the master - list ( detected in every image in every epoch with magnitudes consistent to within @xmath30 ) . as can be seen from the table , this generally cuts down the number of sources by a factor of two . we obtained an average @xmath21 and @xmath28 magnitude and photometric errors for each source in the master - list by combining the measurements of the individual exposures . for each field we identified the qso in the master - list from registering with the macho discovery images ( see geha et al . we then removed the qso from the master - list to obtain a list that consisted primarily of lmc objects . we were not interested in measuring ( nor were we able to measure ) the relative motion of lmc stars within a given field with respect to each other . we could only measure the average motion of all the stars in a given field with respect to the qso . to do this we determined and applied geometric transformations to align the star positions in every image with those of the reference image . then subsequently , we used the same transformations to correct the position of the qso . the motion of the qso , over the time baseline , was then obtained simply as the difference between its average position in each epoch . since in reality the qso is so distant that it is essentially fixed on the sky , the resulting measurement is just the reflex motion due to the average motion of lmc stars . we used a six parameter linear fit to transform the star positions and align them as best as possible ( in a @xmath31 sense ) with the positions of the same stars in the reference image . we ran the linear transformations iteratively and accepted for use in the calculation of the transformation only those stars for which the proper motion ( pm ) and the error in the proper motion ( @xmath32pm ) , measured relative to the average of the other stars , were both @xmath33 pixels . this rejected foreground stars ( in general those are expected to move by @xmath34 towards the north ; see momany & zaggia 2005 ) and stars with large centroiding errors . the latter are generally faint stars that were unlikely to improve the accuracy of the result . the terms in the transformations were modified iteratively until the number of stars used in the fit was unchanging . after determination of the linear transformations for all images , we calculated the pm and @xmath32pm for all sources in the master - list , as well as for the qso . figure [ flowchart ] is a schematic illustrating the basic steps involved in getting a qso pm in pixels . one component of the final astrometric error is how well we are able to align the stars between epochs . we therefore determined the pm of each star individually as the difference between its average positions in epoch 2 and 1 ( just as we do for the qso ) . we then ran some statistics to calculate the error in the average pm of all the stars , which we called @xmath35 ( the average itself is zero by definition ) . this error quantifies how accurately we were able to align the star - fields between the two epochs and gives us an idea of our systematic errors . table 2 lists the final number of sources ( @xmath36 ) used in the linear transformations for each field . once the images are aligned the relative motion of the qso and the star - field is simply the difference in the average position of the qso from epoch 1 to epoch 2 . we called the random error in this difference @xmath37 . the final error in the relative motion between the qso and the star - field is then @xmath38 . figure [ figure3 ] shows the distribution and contributions of these two components of the final error . @xmath37 is plotted on the @xmath39-axis and @xmath35 is on the @xmath24-axis . the straight line marks @xmath40 . the figure shows that our final errors are dominated by the centroiding errors of the qso and that the error introduced by aligning the star - field is smaller . we transformed the results to a pm and @xmath32pm of lmc stars in the directions north and west in @xmath18 using the orientation of the reference image with respect to the sky ( orientat ) , the time baseline , and the fact that the pm of the lmc is the opposite of what we measure for the qso ( technically what we measure is the reflex motion of the qso ) . absolute orientations of @xmath0 observations are fairly accurate ( @xmath41 ) and any uncertainty in orientat has no effect on our final pm errors . table 2 tabulates the pm for each field and the corresponding error ( columns 5 - 8) . figure [ figure4 ] shows the ( @xmath42 , @xmath21 ) cmd for the lmc . the qsos are marked in green and the stars with pm & @xmath32pm @xmath33 pixels are marked in red . the cmd looks consistent with other cmds of the lmc ( see for e.g. alcock et al . the main sequence and the giant branch are clearly visible , and the red clump ( which is often used as a distance indicator ) is also where we would expect it to be , at @xmath43 ( see smecker - hane et al . we believed that we might make some additional cuts for a foreground population of stars based on this cmd , but such a population is not clearly delineated , and any cuts we could make ( like the very red , dim tail of objects ) seem to be obviated by the pm cuts that we employ . to test the robustness of our results we tried a number of different variations and cuts in our analysis . apart from the cuts in pm space to eliminate foreground stars and stars with large centroiding errors , these variations included : a ) constraints on the distance of a star from the center of the field ( i.e. eliminating stars in the outer edges from our analysis in case the geometric distortion is more complicated than understood in those regions ) ; b ) inclusion of higher order terms in the transformation to account for any possible time - variations in the distortion terms ; c ) making both steeper and looser cuts in the pm and @xmath32pm space of the stars ( e.g. pm and @xmath32pm @xmath44 and @xmath45 pixel ) ; d ) using only those stars whose pm/@xmath32pm @xmath46 to further ensure that the transformations are not unduly influenced by stars whose pms are not statistically consistent with zero . none of these cuts were found to significantly alter the results . in particular they did not affect the fields that could be seen as ` outliers ' at that point in the analysis . figure [ systematics](a ) & ( b ) show pm and @xmath32pm in pixels as a function of @xmath21 magnitude for all stars in all master - lists of all fields respectively . these plots look reasonable and have the expected shapes with errors @xmath47 . figure [ systematics](c ) shows @xmath32pm vs. pm for all stars in all master - lists . it illustrates that the pm residuals are not much larger than the expected random errors . figure [ pmvectors ] shows pm vectors for all the stars in all master - lists of all fields in @xmath39 and @xmath24-space in pixels , magnified by a factor of hundred . this enables us to inspect how and if the shifts vary with position on the chip and if there are any other systematic trends in the transformations . all vectors appear to be random both in magnitude and orientation across the field . figure [ pmvspos ] seeks to address the same concerns , but plots the pms for the stars with pm and @xmath48 pixels versus @xmath39 and @xmath24 position on the chip separately . the scatter in @xmath39 and @xmath24 look comparable and there is no obvious trend with position on the chip . figure [ pmbins ] is again a plot of pm vs. @xmath39 and @xmath24 for the stars , but now the pm values have been binned for every 100 pixels , and the average pm value for each bin is plotted . this allows us to really get down into the noise and investigate systematic errors at the level of a hundredth of a pixel . we conclude from this plot that there is no evidence for systematic errors in the geometric distortion correction larger than @xmath49 pixels , which is consistent with the findings of ak04 . thus , all the results look sensible , robust , and well behaved to within the expected accuracy . one way to estimate the proper motion of the lmc center of mass is just to average the results from the 21 fields listed in table 2 . however , this is not the most accurate approach . the fields are separated by several degrees and this causes real variations in the pms of lmc stars in different fields . in general , the motion observed for an individual qso field can be written as @xmath50 here @xmath51 is the proper motion of the lmc center of mass and @xmath52 is the field - dependent residual . the latter contains contributions from different effects : ( a ) : : variations as a function of position in the components of the three - dimensional velocity vector of the center of mass that are seen in the plane of the sky ( `` viewing perspective '' ) ; ( b ) : : the internal rotation of the lmc ; ( c ) : : time variations @xmath53 and @xmath54 in the lmc viewing angles . the formulae for these contributions can be found in van der marel et al . ( 2002 , hereafter vdm02 ) . with the help of a model for these contributions one can use the observed proper motion @xmath55 for any given field to obtain an estimate @xmath56 for the proper motion of the lmc center of mass using @xmath57 the estimates @xmath58 for all fields are listed in table 2 ( columns 9 & 10 ) . the field - dependent correction terms @xmath52 are small compared to the overall lmc proper motion , but are significant given the accuracy that we are trying to achieve . the average two - dimensional size of the correction terms is @xmath59 and the maximum over all fields is @xmath60 . the corrections are smaller for fields that are closer to the center of the lmc . the largest contribution to the corrections comes from the viewing perspective effect , which contributes on average @xmath61 , with a maximum over all fields of @xmath62 . by contrast , the internal rotation contributes on average @xmath63 , with a maximum over all fields of @xmath64 . we used an iterative procedure to evaluate equation ( [ correctionformula ] ) . this is necessary because the residuals @xmath52 themselves depend either directly or indirectly on the proper motion of the lmc center of mass . at each iteration step we averaged the estimates @xmath56 for the different fields to obtain an optimal estimate for @xmath51 . this value was then used in the next iteration step to evaluate the residuals @xmath52 . for the internal rotation model of the lmc we fitted the line - of - sight velocity data for 1041 carbon stars ( at each iteration step ) as in vdm02 . this fit depends on the proper motion of the lmc center of mass , because the latter introduces a spurious solid - body component in the observed line - of - sight velocities that must be subtracted while modeling the internal rotation . we did not include any time - dependence of the viewing angles ( indicative of precession and nutation of the lmc disk plane ) in the model . as we show in 4.4 below , the available observational constraints are consistent with @xmath65 . where necessary , in the model for @xmath66 we used : distance modulus @xmath67 , systemic line - of - sight velocity @xmath68 , center position @xmath69 and @xmath70 , inclination @xmath71 and line - of - nodes position angle @xmath72 as in vdm02 . the field - dependent corrections @xmath52 are more - or - less symmetrical around the lmc center . so the corrections approximately average to zero for a distribution of quasar fields that is symmetric around the lmc center . this is not exactly the case for the fields that we observed , but nonetheless , the fields are reasonably homogeneously distributed around the center ( see figure [ figure1 ] ) . therefore , the average of the corrections over all fields is smaller than the individual corrections themselves . moreover , this average is determined primarily by the geometry of the field distribution . the specific assumptions made about the model of the lmc and its rotation have almost a negligible effect . we have run tests in which we varied the values of the model parameters over ranges that are realistic estimates of the observational uncertainties : @xmath73 in @xmath74 , @xmath75 in inclination , @xmath76 for the range in line - of - nodes position angle , @xmath77 in lmc center position , and @xmath78 mag in distance modulus . none of these changes affected our final estimate of the proper motion of the lmc center of mass by more than @xmath79 . the quantities @xmath58 for each field all provide an independent estimate of the same quantity . for each field we calculated the residual vector @xmath80 between its value of @xmath58 and the final estimate of the lmc center of mass proper motion that we derived from these estimates ( to be discussed in 4.3 below ) . the scatter in these residuals provides insight into the observational uncertainties . to address the possibility of systematic trends we show in figure [ finalcuts ] the size @xmath81 of the residuals as a function of the following quantities : the @xmath21 magnitude of the qso ; @xmath42 of the qso ; the number of stars @xmath82 used to align the stars in each field between the two epochs ; the quantity @xmath83 that indicates how well the stars could be aligned between the epochs ; and the distance of the qso to its nearest neighboring star ( without application of cuts to the star - list ) . there is a clear systematic effect as a function of @xmath83 . there is one field ( i d l13 ) that has a much higher value of @xmath83 than the other fields . this field also happens to have the largest residual @xmath80 . it also has the nearest neighboring star of all fields ( only 9 pixels from the qso ) . so it is not clear whether this is the reason for the large @xmath80 or whether it is the poor accuracy ( measured by @xmath83 ) with which the star - field could be aligned . there is also a systematic effect as a function of @xmath84 . fields with a very low value of @xmath85 do tend in general to have larger residuals than fields with a high value of @xmath85 . this makes sense , since one expects to be able to align the data from the two epochs better if there are more stars in the field , and we certainly do not want to trust the fields in which very few stars were used to determine the linear transformations . thus , based on these plots we decided to retain for our final estimate of the lmc proper motion only the 13 fields for which @xmath86 and @xmath87 . these fields are shown with filled circles in figure [ finalcuts ] while the rejected fields are shown with open circles . once we have applied these cuts there are no trends in @xmath80 for the remaining fields with qso magnitude , color , or distance to the nearest star . hence , no additional cuts were applied . the pm values for all the qso fields are shown in the ( @xmath88)-plane in figure [ pms ] . they are compared to the residual proper motions of all the lmc stars that were used in the transformations . the stars are shown with open circles and the qsos with filled ones . panel ( a ) shows the observed pm values for all fields . panel ( b ) shows the estimates @xmath58 derived from the observed values for the final 13 `` high - quality '' fields only . the reflex motion of the qsos clearly separates from the star - fields in both panels . the solid lines in ( b ) mark the weighted average of the 13 fields ( see 4.3 , equation ( [ pmfinal ] ) ) . figure [ mu_resid ] shows the residual vectors @xmath80 as a function of field position along with the @xmath89 error bars for each field . rejected and non - rejected fields are shown with open and filled circles respectively . the thick solid vector anchored by a plus sign shows for comparison the size of the inferred center of mass proper motion of the lmc , at the adopted lmc center . for the high quality fields there is no obvious trend as a function of absolute telescope pointing . low quality fields that were rejected on the basis of their @xmath36 or @xmath90 do tend to have large residuals and appear to be concentrated mainly in the north - east . this reflects the low stellar density in this region ( see figure [ figure1 ] ) . on the other hand , the lmc is known to have internal motions and these fields appear to both correspond to the location of the lmc spiral arm ( nikolaev et al . 2004 ) and show a similar residual trend in direction . by not including these fields we thus risk removing a real and interesting trend from the data . however , given the very low numbers of stars in these fields we do not feel that we can detect such a trend with confidence . thus , these fields are not included in our final estimates . note that if we were to include them , they would not make a statistically significant difference to the bulk motion of the lmc ( see 4.3 ) . this is illustrated by the bold dashed line in figure [ mu_resid ] which is a straight average of all 21 fields ( columns 5 & 6 of table 2 ) . one possible source of systematic errors in our analysis is the degrading charge transfer efficiency ( cte ) of the acs / hrc ccd . when charge is read out from a ccd pixel , some charge is left behind and released at a later time . this causes sources in an image to show faint tails along the detector @xmath24-direction . the impact of imperfect cte is larger for sources that are further from the read - out amplifier the impact also increases with time , because the continuing bombardment by cosmic rays in the harsh environment of space increases the number of charge traps in the ccd . the effect of imperfect cte on photometry has been studied in detail by many authors , and has been quantified for the acs / hrc by riess ( 2003 ) . more recently , some attention has been drawn to the impact of imperfect cte on astrometry ( bristow , piatek & pryor 2005 ; piatek et al . these authors quantified this effect for the stis instrument on @xmath0 in the context of a study of the proper motion of the ursa minor dwarf spheroidal galaxy . they found that the inferred reflex pm of the qso in their stis field changes by @xmath91 due to imperfect cte . the acs / hrc ccd was a flight - spare for the stis instrument , and its detector properties are similar to those of the stis ccd . however , the effect of imperfect cte on our lmc pm should be much less than that found in the stis ursa minor study , for several reasons . first , the acs / hrc pixels span only @xmath92 mas on the sky , as compared to @xmath93 mas for the stis ccd . so the same star trail in pixel space will produce an astrometric shift in mas that is smaller by a factor of @xmath94 . second , the median observation date of our second epoch lmc data was only 2.5 years after installation of the acs on @xmath0 , as compared to 5.0 years for the last epoch in the stis ursa minor study . if the effect of imperfect cte on astrometry varies linearly with time since installation ( as it does for photometry ; mutchler & sirianni 2005 ) then our study should be affected by a factor of @xmath95 less . and third , we have observations for @xmath96 different fields ( or @xmath97 for the `` high quality '' sample ) , as compared to only 1 in the stis ursa minor study . since our fields have more or less random detector orientations on the sky , any cte induced astrometric shift will average down to zero @xmath98 . so even if we were to make no allowance for cte one would not expect our study to be affected by more than @xmath99 . the stis ursa minor study could correct explictly for the astrometric effects of cte owing to the existence of a detailed model for the underlying stis ccd detector physics and charge trap properties ( bristow & alexov 2002 ) . such a model does not ( yet ) exist for the acs / hrc ccd , and it is therefore not possible to correct our data explictly for imperfect cte . however , a simple correction can be applied by fitting a shift in the detector @xmath24-direction of sources that is linearly proportional to both the @xmath24-position on the detector and the time since installation of the instrument in space . this is akin to what piatek et al . ( 2005 ) did for the wfpc2 data that they obtained for ursa minor . they had to include this term explictly , because they otherwise allowed only for a translation , rotation and scale change between epochs ( i.e. , a four - parameter linear fit ) . instead , in our study we perform a fully general six - parameter linear fit to align the star - fields from different epochs . therefore , the type of astrometric shift expected from imperfect cte is automatically allowed for and fitted to the data . this seems to work well , given that the average pms of the lmc stars do not show a trend with @xmath24-position on the detector ( see figure [ pmbins ] ) . the only thing that we do not allow for is a variation of the astrometric shift with brightness . brighter sources are less affected by imperfect cte than fainter sources . since the qso is generally brighter than the average lmc star in the field , there might be a small residual cte effect in the pm for each field . however , we will estimate the final errors in the lmc pm from the scatter between the pm measurements for different fields . therefore , any residual systematic errors due to imperfect cte are explictly accounted for in our final error estimates . so in summary : ( 1 ) even when not modeled at all , imperfect cte is expected to have at most a small effect on our acs / hrc results for the lmc pm ; ( 2 ) our methodology for aligning star fields between epochs allows for a low - order correction of the astrometric shifts due to imperfect cte ; and ( 3 ) any residual effects of imperfect cte are explictly accounted for in our final error estimates . to obtain our final estimate of the lmc proper motion we took the weighted average of the 13 high - quality fields to obtain @xmath100 the errors in these numbers are based on the rms scatter between different fields , divided by @xmath101 . to be conservative we did not use the weighted average errors @xmath102^{-1/2}$ ] . these errors have a value of @xmath103 to the north and to the west , and are a factor of @xmath104 smaller in the northward direction and @xmath105 smaller in the westward direction than the errors in equation ( [ pmfinal ] ) . the fact that the weighted errors , which are due to pure noise propagation , are smaller than the errors in equation ( [ pmfinal ] ) indicates that there is more scatter in the results from different fields than can be accounted for by random errors alone . this is not entirely unreasonable . an error of @xmath106 over a baseline of 1.6 years ( our mean ) corresponds to 0.004 pixels , and it has not been established that there are no systematic errors in , for example , the geometric distortion , at that level ( see figure [ pmbins ] and ak04 ) . one possible cause of systematic errors might be variations in the higher order geometric distortion terms with time at the level of thousandths of a pixel or residual astrometric effects due to imperfect cte . note that if so , such systematic errors would probably have a fixed orientation with respect to the detector . since our fields were observed at somewhat random orientations because of the snapshot nature of our program ( see figure [ figure2 ] ) , such systematic errors would still approximately decrease as @xmath26 when averaging ( as we assumed in getting equation ( [ pmfinal ] ) ) . it is important to note that our result does not depend sensitively on the details of either the modeling of the field - dependence of the pms , or the criteria applied to reject lower - quality fields , or the actual statistic applied to obtain the average . a straight unweighted average of the observed proper motions for all 21 fields , as listed in table 2 ( columns 5 & 6 ) , yields @xmath107 . the median is @xmath108 . these estimates are both consistent with the result in equation ( [ pmfinal ] ) to within the errors . our data might in principle allow us to detect the rotation of the lmc in the plane of the sky . to address this we plot in figure [ rotation](a ) the residuals @xmath109 where @xmath110 is the observed proper motion for a field and @xmath51 is the center - of - mass proper motion from equation ( [ pmfinal ] ) . the quantity @xmath52 is the residual proper motion as in 4.1 , but now includes only the contribution from viewing perspective , and _ not _ the contribution from internal rotation . the residuals @xmath111 are therefore the observational estimates of the internal motions of the lmc . for comparison , we show in figure [ rotation](b ) the internal motions predicted by the model employed in 4.1 . it is clear that the observations are dominated by noise , and rotation is not immediately obvious from visual inspection . on the other hand , some signal can be measured by averaging the results from the individual fields . for each field we determined the unit vectors @xmath112 and @xmath113 along and perpendicular to the direction of rotation predicted by the model . we determined the components of the residual vectors @xmath114 along these unit vectors , which are given by the inner products @xmath115 and @xmath116 . we then determined the averages of these components for our 13 high - quality fields . this yields @xmath117 and @xmath118 . for comparison , the model predicts @xmath119 and @xmath120 . although the errors are significant , we do detect the expected sign and the expected magnitude for the rotation . this provides additional confidence in the accuracy of our results . in our models we have assumed throughout that @xmath121 . if these terms were to have non - zero values then this would have left a signature in the residuals @xmath122 . in particular , using model calculations similar to those in figure 10c , d of vdm02 , and assuming that the model for the internal rotation of the lmc is correct , we would have expected that in @xmath18 , @xmath123 and @xmath124 . in these equations both @xmath54 and @xmath53 are also in units of @xmath18 ( where @xmath125 ) . the observed values therefore provide the following constraints on the precession and nutation of the lmc : @xmath126 both quantities are consistent with zero , although the constraints are not particularly informative . the error bars are considerably larger than the precession and nutation that are expected on the basis of @xmath127-body calculations ( weinberg 2000 ) . we are now in a position to make comparisons with previous proper motion work . figure [ comppm ] shows all the pm determinations to date with their @xmath128% confidence ellipses . we do not include the very discrepant results of anguita et al . ( 2000 ) and momany & zaggia ( 2005 ) on this plot as there must be unidentified systematic errors associated with these measurements ( see discussions in pedreros et al . 2000 and momany & zaggia 2005 ) . the figure shows the region of the @xmath129-plane spanned by the rest of the results . these results are more - or - less consistent with each other . our @xmath0 determination appears to be closest to the results of the pedreros et al . ( 2002 ) study and the kroupa & bastian ( 1997 ) hipparcos study . our value for @xmath130 is on the high end of what was found in previous studies . the measurements from all these studies can be combined to get a weighted average of the proper motion of the lmc . in doing so we first increase ( as in vdm02 ) the error bars of all the studies by a fixed factor so that they become statistically consistent , as judged by @xmath131 . the resulting grand average of all the lmc proper motion measurements is @xmath132 in the discussion that follows we adopt the @xmath0-only values given in equation ( [ pmfinal ] ) . the observed proper motion of the lmc quoted in equation ( [ pmfinal ] ) yields an estimate of its transverse velocity via @xmath133 where the @xmath39- and @xmath24-directions point towards the direction of west and north respectively . @xmath134 is the lmc distance , which we assume to be @xmath135 kpc as in 4.1 ( freedman et al . this yields @xmath136 the listed errors do not include the potential contribution from the uncertainty in the distance modulus . the values of @xmath137 and @xmath138 correspond to a transverse velocity of @xmath139 in the direction of position angle @xmath140 . the systemic line - of - sight velocity is very well known from a number of different tracers including a detailed analysis of 1041 carbon stars by vdm02 who obtained @xmath141 . we are therefore in a position to ask what these measurements imply for the three - dimensional space motion of the lmc . this issue has been addressed previously by several authors but our results provide the most accurate measurement so far . as many previous authors have done we adopt a cartesian coordinate system ( @xmath142 , @xmath143 , @xmath144 ) with the origin at the galactic center , with the @xmath144-axis pointing toward the galactic north pole , the @xmath142-axis pointing in the direction from the sun to the galactic center , and the @xmath143-axis pointing in the direction of the sun s galactic rotation . it is necessary to correct for the reflex motion of the sun to determine the motion of the lmc with respect to the milky way . following the equations set up in vdm02 , 9.3 , we get @xmath145 for the 3-d velocity of the lmc and its radial and tangential components . the fact that @xmath146 is larger than the milky way circular velocity of @xmath147 and that @xmath148 is small and positive confirms that the lmc is just past perigalacticon , as found by previous authors . our value for the galactocentric tangential velocity is higher than that implied by a weighted average of earlier pm measurements , which is @xmath149 . it is also a little higher than values inferred from most magellanic stream models . there are two main approaches to modeling the stream , one that invokes ram - pressure stripping of mc gas by matter in the galactic halo , and another that invokes the tidal force of the milky way on the mc system . as stated in the introduction , gardiner & noguchi ( 1996 ) have constructed detailed simulations for the tidal distortion of the smc owing to the galaxy and the lmc , and get the following constraints for the motion of the lmc : @xmath150 and @xmath151 . our observed value for the tangential velocity is noticably higher , while the radial velocity seems to agree well . our values appear more consistent with those of heller & rohlfs ( 1994 ) who use a combination of the tidal and ram - pressure stripping models to obtain : @xmath152 and @xmath153 . the errors in the proper motion of the lmc are now technically good enough to allow distinctions between various magellanic stream models . our @xmath0 proper motion determination for the smc will soon follow in a subsequent paper . taken together these two data points should provide powerful new constraints on the mc system and the gravitational potential of the galactic halo . our results also lead to revised values for several other quantities that characterize the lmc . given a revised pm we can obtain a different rotation curve @xmath154 for the carbon stars in the lmc . this results in an amplitude of @xmath155 for the lmc rotation curve which is an increase of @xmath156% compared to the vdm02 value . this will in turn cause small changes in other estimates that are based on @xmath154 such as the total mass of the lmc and its tidal radius . updated modeling of the carbon star velocity field using our revised pm values also yields an improved constraint on @xmath53 . using the same methodology as in vdm02 we obtain @xmath157 , which is now more consistent with zero than previously believed . this is a much stronger constraint than what we were able to obtain with the analysis of the field - dependent pms alone ( equation ( 5 ) ) . we undertook a project using two epochs of @xmath0/acs data of magellanic cloud fields centered on background qsos to determine the systemic proper motion of the clouds . the lmc results are presented in this paper . we have determined the proper motion of the lmc to be @xmath158 . this is accurate to better than 5% . when combined with hi data for the magellanic stream , this should allow improved constraints on both the mass distribution in the galactic halo , and theoretical models for the origin of the magellanic stream . our data provides the most accurate proper motion measurement for any milky way satellite . in future papers , we will present results for the smc as well as implications for the lmc - smc system and the magellanic stream . improvements to our work may be possible by using @xmath0 with a longer baseline . a long baseline would facilitate measurements of the internal motions of the clouds as well as an even more accurate measurement of their systemic motions . in addition we may be able to make a distance determination using the method of rotational parallax ( olling & peterson , 2000 ) . rotational parallax is the method of determining distances to local group galaxies by measuring a rotation curve using both proper motions and radial velocities ( see brunthaler et al . 2005 for an application of this method to m33 ) . since the former are distance - dependent and the latter are not , using both methods provides a distance measurement . more generally , considerable improvement may be expected when the next generation of astrometric satellites such as @xmath159 and @xmath160 come on - line . the authors would like to thank jay anderson and ivan king for their geometric distortion calibration of the hrc , and in particular , jay anderson for making his analysis software available to the community . support for this work was provided by nasa through grant numbers go-09462 and go-10130 from the space telescope science institute ( stsci ) , which is operated by the association of universities for research in astronomy , inc . , under nasa contract nas5 - 26555 . m. g. is supported by nasa through hubble fellowship grant hf-01159.01-a awarded by stsci . khc s work was performed under the auspices of the u.s . department of energy , national nuclear security administration by the university of california , lawrence livermore national laboratory under contract no . w-7405-eng-48 . alcock , c. 2000 , science , 287 , 74 + alcock , c. et al . 2000a , apj , 542 , 281 + alcock , c. , et al . 2000b , , 119 , 2194 + anderson , j. & king , i. r. 2004 , acs instrument science report 04 - 15 ( baltimore : space telescope science institute ) ( ak04 ) + anderson , j. , & king , i. r. 2003 , , 115 , 113 + anguita , c. , loyola , p. , & pedreros , m. h. 2000 , aj , 120 , 845 + bristow , p. , & alexov , a. 2002 , st - ecf instrument science report ce - stis 2002 - 01 ( garching bei munchen : space telescope european coordinating facility ) + bristow , p. , piatek , s. , & pryor , c. 2005 , in st - ecf newsletter 38 , p. 12 ( garching bei munchen : space telescope european coordinating facility ) + br " uns , c. , et al . 2005 , , 432 , 45 + brunthaler , a. , reid , m. j. , falcke , h. , greenhill , l. j. , & henkel , c. 2005 , science , 307 , 1440 + drake , a. j. , cook , k. h. , alcock , c. , axelrod , t. s. , geha , m. , & macho collaboration 2001 , baas , 33 , 1379 + dobrzycki , a. , macri , l. m. , stanek , k. z. , & groot , p. j. 2003 , , 125 , 1330 + dobrzycki , a. , groot , p. j. , macri , l. m. , & stanek , k. z. 2002 , , 569 , l15 + fich , m. & tremaine , s. 1991 , ara&a , 29 , 409 + freedman , w. l. , et al . 2001 , , 553 , 47 + gardiner , l. t. , & noguchi , m. 1996 , , 278 , 191 + geha , m. et al . 2003 , aj , 125 , 1 + heller , p. & rohlfs , k. 1994 , a&a , 291 , 743 + jones , b. f. , klemola , a. r. , & lin , d. n. c. 1994 , aj , 107 , 1333 + karakla , d. 2005 , @xmath0 primer ( baltimore : space telescope science institute ) + krist , j. 2003 , acs instrument science report 03 - 06 ( baltimore : space telescope science institute ) + kroupa , p. , r " oser , s. , & bastian , u. 1994 , mnras , 266 , 412 + kroupa , p. & bastian , u. 1997 , new astronomy , 2 , 77 + lin , d. n. c. , jones , b. f. , & klemola , a. r. 1995 , apj , 439 , 652 + merrifield , m. 1992 , aj , 103 , 1552 + momany , y. , & zaggia , s. 2005 , , 437 , 339 + mutchler , m. , & sirianni , m. 2005 , acs instrument science report 2005 - 03 ( baltimore : space telescope science institute ) + nikolaev , s. , drake , a. j. , keller , s. c. , cook , k. h. , dalal , n. , griest , k. , welch , d. l. , & kanbur , s. m. 2004 , apj , 601 , 260 + olling , r. p. & peterson , d. m. , 2000 , astro - ph/0005484 + pavlovsky , c. 2005 , acs instrument handbook ( baltimore : space telescope science institute ) + pedreros , m. h. , anguita , c. , & maza , j. 2002 , aj , 123 , 1971 + smecker - hane , t. a. , cole , a. a. , gallagher , j. s. , & stetson , p. b. 2002 , , 566 , 239 + piatek , s. , pryor , c. , bristow , p. , olszewski , e. w. , harris , h. c. , mateo , m. , minniti , d. , & tinney , c. g. 2005 , aj , 130 , 95 + riess , a. 2003 , acs instrument science report 2003 - 09 ( baltimore : space telescope science institute ) + van der marel , r. p. 2002 , in ` the shapes of galaxies and their halos ' , natarajan p. , ed . , 202 , 2002 ( singapore : world scientific ) + van der marel , r. p. , alves , d. r. , hardy , e. , & suntzeff , n. b.2002 , aj , 124 , 2639 ( vdm02 ) + weinberg , m. d. 2000 , , 532 , 922 + wilkinson , m.i . , & evans , n.w . 1999 , mnras , 310 , 645 + sofue , y. , & rubin , v. 2001 , , 39 , 137 + llllcccccccccccc & & & & & & date & visit & @xmath161 & @xmath161 & orientat & date & visit & @xmath161 & orientat & + & & & & & & & & f606w & f814w & & & & f606w & & + & & ( h , m , s ) & ( deg , , ) & & & & & ( min ) & ( min ) & ( deg ) & & & ( min ) & ( deg ) & ( yrs ) + l1 & 68.10972.36 & 5 47 50.2 & -67 28 1.3 & 16.4 & 1.01 & 2002 - 09 - 05 & 1 & 6.7 & 1.7 & 75.3 & 2005 - 04 - 25 & 1 & 6.7 & -51.3 & 2.6 + l2 & 42.860.123 & 4 46 11.1 & -72 5 9.0 & 17.3 & 0.26 & 2002 - 09 - 12 & 2 & 6.7 & 1.7 & 96.4 & 2004 - 09 - 14 & 2 & 6.7 & 96.5 & 2.0 + l3 & 2.5873.82 & 5 16 28.9 & -68 37 1.8 & 17.4 & 0.46 & 2002 - 09 - 24 & 3 & 6.7 & 1.7 & 100.9 & 2004 - 08 - 11 & 3 & 6.7 & 59.6 & 1.9 + l4 & 58.5903.69 & 5 16 36.8 & -66 34 35.8 & 18.0 & 2.24 & 2002 - 09 - 22 & 4 & 6.7 & 1.7 & 99.6 & 2004 - 09 - 12 & 4 & 6.7 & 90.3 & 2.0 + l5 & 13.5717.178 & 5 15 36.1 & -70 54 0.8 & 18.8 & 1.66 & 2002 - 09 - 07 & 5 & 6.7 & 1.7 & 68.8 & 2004 - 08 - 10 & 5 & 10.1 & 59.0 & 1.9 + l6 & 75.13376.66 & 6 2 34.3 & -68 30 41.1 & 18.7 & 1.07 & 2002 - 10 - 11 & 6 & 6.7 & 1.7 & 106.9 & 2004 - 08 - 13 & 6 & 7.3 & 57.1 & 1.8 + l7 & 78.5855.788 & 5 16 26.3 & -69 48 19.0 & 18.6 & 0.63 & 2002 - 10 - 28 & 7 & 6.7 & 1.7 & 135.2 & 2004 - 08 - 10 & 7 & 7.3 & 58.7 & 1.8 + l8 & 53.3360.344 & 5 0 54.0 & -66 43 59.8 & 18.9 & 1.86 & 2002 - 09 - 03 & 9 & 6.9 & 1.7 & 90.4 & 2004 - 09 - 05 & 9 & 11.5 & 90.5 & 2.0 + l9 & 17.2227.488 & 4 53 56.5 & -69 40 35.4 & 19.0 & 0.28 & 2002 - 09 - 05 & 11 & 9.5 & 1.7 & 88.6 & 2004 - 09 - 09 & 11 & 10.1 & 92.3 & 2.0 + l10 & 30.11301.499 & 5 49 41.6 & -69 44 15.1 & 19.9 & 0.46 & 2002 - 09 - 03 & 12 & 12.0 & 1.7 & 73.5 & 2004 - 08 - 10 & 12 & 21.5 & 48.9 & 1.9 + l11 & 69.12549.21 & 5 57 22.4 & -67 13 21.5 & 16.8 & 0.14 & 2003 - 06 - 04 & 19 & 6.7 & 1.7 & -15.9 & 2005 - 04 - 28 & 18 & 6.7 & -51.4 & 1.9 + l12 & 25.3469.117 & 5 1 46.7 & -67 32 39.8 & 18.3 & 0.38 & 2003 - 06 - 04 & 20 & 6.7 & 1.7 & -3.4 & 2004 - 08 - 18 & 19 & 6.7 & 69.1 & 1.2 + l13 & 6.6572.268 & 5 20 57.0 & -70 24 52.6 & 18.1 & 1.81 & 2003 - 02 - 16 & 21 & 6.7 & 1.7 & -113.5 & 2004 - 08 - 10 & 20 & 6.7 & 57.8 & 1.5 + l14 & 25.3712.72 & 5 2 53.7 & -67 25 45.0 & 18.6 & 2.17 & 2003 - 06 - 07 & 24 & 6.7 & 1.7 & 12.1 & 2004 - 07 - 13 & 23 & 6.9 & 36.1 & 1.1 + l15 & 9.5484.258 & 5 14 12.1 & -70 20 25.8 & 18.5 & 2.32 & 2003 - 06 - 04 & 25 & 6.7 & 1.7 & -5.9 & 2004 - 07 - 14 & 24 & 8.0 & 33.3 & 1.1 + l16 & 53.3970.140 & 5 4 36.0 & -66 24 15.7 & 18.0 & 2.04 & 2002 - 12 - 26 & 26 & 6.7 & 1.7 & -162.3 & 2005 - 04 - 07 & 25 & 6.7 & -59.5 & 2.3 + l17 & 63.7365.151 & 5 25 14.4 & -65 54 45.7 & 18.6 & 0.65 & 2003 - 05 - 08 & 27 & 6.7 & 1.7 & -34.9 & 2004 - 07 - 05 & 26 & 8.8 & 30.3 & 1.2 + l18 & 61.8199.302 & 5 30 26.8 & -66 48 52.9 & 18.7 & 1.79 & 2002 - 12 - 25 & 29 & 6.7 & 1.7 & -169.3 & 2004 - 08 - 10 & 28 & 9.5 & 64.9 & 1.6 + l19 & 82.8403.551 & 5 31 59.7 & -69 19 51.5 & 19.8 & 0.15 & 2003 - 05 - 20 & 30 & 8.0 & 1.7 & -24.1 & 2004 - 07 - 11 & 29 & 11.6 & 26.1 & 1.1 + l20 & 48.2620.2719 & 4 56 14.3 & -67 39 9.0 & 19.3 & 0.26 & 2002 - 12 - 01 & 33 & 9.6 & 1.7 & 174.3 & 2005 - 06 - 04 & 31 & 12.3 & -1.1 & 2.5 + l21 & 5.4892.1971 & 5 10 32.5 & -69 27 15.5 & 18.9 & 1.58 & 2003 - 06 - 04 & 40 & 6.7 & 1.7 & -4.9 & 2004 - 07 - 18 & 35 & 9.9 & 38.0 & 1.1 + lcccccccccc & & & & @xmath162 & @xmath130 & @xmath163 & @xmath164 & @xmath162 & @xmath130 & + & & & & ( @xmath18 ) & ( @xmath18 ) & ( @xmath18 ) & ( @xmath18 ) & ( @xmath18 ) & ( @xmath18 ) & + l1 & 85 & 35 & 15 & 0.800 & -1.162 & 0.089 & 0.121 & 0.550 & -1.334 & 0 + l2 & 70 & 35 & 10 & -0.274 & - 2.220 & 0.067 & 0.070 & 0.238 & -2.098 & 0 + l3 & 298 & 109 & 51 & 0.413 & -1.976 & 0.061 & 0.082 & 0.553 & -1.969 & 1 + l4 & 97 & 48 & 17 & 0.521 & -2.121 & 0.118 & 0.150 & 0.692 & -2.218 & 1 + l5 & 240 & 84 & 27 & 0.102 & -2.054 & 0.097 & 0.086 & 0.234 & -2.007 & 1 + l6 & 41 & 16 & 7 & 0.983 & -1.658 & 0.298 & 0.191 & 0.543 & -1.858 & 0 + l7 & 622 & 269 & 116 & 0.426 & -2.068 & 0.064 & 0.073 & 0.543 & -2.045 & 1 + l8 & 135 & 49 & 20 & -0.050 & -1.961 & 0.058 & 0.081 & 0.329 & -2.018 & 1 + l9 & 213 & 81 & 42 & -0.047 & -2.023 & 0.084 & 0.077 & 0.406 & -1.967 & 1 + l10 & 214 & 52 & 38 & 0.600 & -1.934 & 0.087 & 0.099 & 0.337 & -2.014 & 1 + l11 & 59 & 22 & 8 & 1.186 & -1.089 & 0.187 & 0.352 & 0.827 & -1.305 & 0 + l12 & 153 & 58 & 20 & -0.024 & -2.447 & 0.183 & 0.169 & 0.338 & -2.478 & 1 + l13 & 426 & 163 & 68 & 1.168 & -2.455 & 0.078 & 0.134 & 1.230 & -2.441 & 0 + l14 & 97 & 34 & 18 & 0.136 & -1.837 & 0.225 & 0.174 & 0.484 & -1.874 & 1 + l15 & 340 & 147 & 46 & 0.573 & -2.738 & 0.217 & 0.173 & 0.719 & -2.704 & 1 + l16 & 95 & 33 & 8 & 0.277 & -1.525 & 0.117 & 0.141 & 0.609 & -1.597 & 0 + l17 & 59 & 25 & 5 & 1.004 & -2.313 & 0.185 & 0.289 & 1.067 & -2.449 & 0 + l18 & 100 & 39 & 12 & 0.914 & -1.779 & 0.128 & 0.133 & 0.897 & -1.908 & 0 + l19 & 486 & 163 & 89 & 0.844 & -1.735 & 0.208 & 0.163 & 0.808 & -1.744 & 1 + l20 & 178 & 44 & 25 & 0.005 & -1.813 & 0.116 & 0.071 & 0.441 & -1.833 & 1 + l21 & 498 & 210 & 113 & 0.054 & -2.464 & 0.122 & 0.115 & 0.254 & -2.434 & 1 +

this paper proposes and discusses a test of the chemical composition of ultra - high energy cosmic rays that relies on the anisotropy patterns measured as a function of energy . in particular , we show that if one records an anisotropy signal produced by heavy nuclei of charge above an energy , one should record an even stronger ( possibly much stronger ) anisotropy at energies due to the proton component that is expected to be associated with the sources of the heavy nuclei . this conclusion remains robust with respect to the parameters characterizing the sources and it does not depend at all on the modelling of astrophysical magnetic fields . as a concrete example , we apply this test to the most recent data of the pierre auger observatory . assuming that the anisotropy reported above 55 eev is not a statistical accident , and that no significant anisotropy has been observed at energieseev , we show that the apparent clustering toward cen a can not be attributed to heavy nuclei . similar conclusions are drawn regarding the apparent excess correlation with nearby active galactic nuclei . we then discuss a robust lower bound to the magnetic luminosity that a source must possess in order to be able to accelerate particles of charge up toeev ,erg / s . using this bound in conjunction with the above conclusions , we argue that the current pao data does not support the model of cosmic ray origin in active radio - quiet or even radio - loud galaxies . finally , we demonstrate that the apparent clustering in the direction of cen a can be explained by the contribution of the last few gamma - ray bursts or magnetars in the host galaxy thanks to the scattering of the cosmic rays on the magnetized lobes .

0907.1354

the sources of ultra - high energy cosmic rays have remained elusive in spite of the enormous progress reached on the experimental side , with present day detectors reaching apertures @xmath6km@xmath7 sr yr . the differential energy spectrum , the chemical composition and the distribution of arrival directions on the sky are as many clues to the nature of the source . there is now a consensus on the existence of the gzk cut - off ( greisen 1966 ; zatsepin & kuzmin 1966 ) , which has been observed by two different experiments ( abbasi et al . 2008 , abraham et al . the existence of this gzk cut - off puts on solid ground the models which attribute the origin of ultra - high energy cosmic rays to powerful astrophysical objects distributed on cosmological scales , such as powerful radio - galaxies ( rachen & biermann 1993 ) , gamma - ray bursts ( milgrom & usov 1995 , vietri 1995 , waxman 1995 ) , or magnetars ( arons 2003 ) . experimental results on the chemical composition and anisotropies are however far more confusing . while the most recent analysis of hires data points to a proton composition above the ankle ( abbasi et al . 2005 , belz 2009 ) , the fluorescence data of the pierre auger observatory ( pao ) rather indicates an increasingly heavier composition above this energy , with the last data points at @xmath8eev close to expectations for iron ( unger et al . 2007 , wahlberg et al . 2009 ) . regarding the distribution of arrival directions on the sky , there exist various contradictory claims , see for instance the correlation with the super - galactic plane reported by stanev et al . ( 1995 ) , contradicted by the results of the agasa experiment ( takeda et al . 1999 ) ; the claim for a correlation with bl lac objects in tinyakov & tkachev ( 2001 , 2002 ) , gorbunov et al . ( 2002 , 2004 ) , which has been debated ( evans , ferrer & sarkar 2003 , 2004 ; tinyakov & tkachev 2004 ) ; as well as the possible detections of multiplets in various datasets , e.g. takeda et al . ( 1999 ) , uchihori et al . ( 2000 ) , farrar , berlind & hogg ( 2006 ) , the significance of which is questioned in abbasi et al . ( 2004 ) and finley & westerhoff ( 2004 ) ; or , finally , the recent data of the pao , which reveal a correlation with nearby active galactic nuclei ( agn ) , the maximal signal being obtained for a search radius of @xmath9 around agn located closer than 75 mpc and for energies above @xmath10eev ( abraham et al . 2007 , 2008b ) . in general , these issues of anisotropies and chemical composition are discussed separately . however , as we argue in the present paper , one can construct a powerful test of the chemical composition by using the anisotropy signal at various energies . this becomes particularly interesting when one notes that measurements of chemical composition are relatively sensitive to the details of the shower reconstruction , in particular to the extrapolation of hadronic models at energies beyond those currently tested in accelerators ( see wibig 2008 , ulrich et al . 2009a , b and wibig 2009 for recent discussions of this issue ) . moreover , measurements of chemical composition rely on the use of fluorescence detectors whose duty cycle is rather low and can not be made event by event , so that they are limited by statistics . the essence of the test proposed in this paper is to exploit the fact that a source in the sky emitting heavy nuclei of charge @xmath0 at an energy @xmath11 is expected to produce a similar anisotropy pattern at energies @xmath12 via the proton component which is expected to be associated with the same source . our expectation for the existence of a proton component relies on the theoretical expectation , that if protons are present in the plasma in which particle acceleration takes place , then they should be accelerated along side with the heavy nuclei . as discussed in some detail at the end of [ sec : test - chem ] , a source accelerating heavy nuclei to energy @xmath11 is quite generally expected to accelerate protons to energy @xmath12 . although the isotropic background `` noise '' increases at lower energies , we show in [ sec : test - chem ] that the signal - to - noise ratio for the detection of the anisotropy also increases . this prediction does not depend at all on the modelling of astrophysical magnetic fields as it only relies on the property that protons of energy @xmath12 follow the same path in the intervening magnetic fields and produce the same angular image as heavy nuclei of charge @xmath0 and energy @xmath11 . the proposed test is discussed in more detail in section [ sec : test - chem ] . when applied to the most recent data of the pao ( section [ sec : pao ] ) , it shows that the signal that is responsible for the apparent anisotropy pattern at energies @xmath13eev must not be heavy but light nuclei , provided this anisotropy is confirmed by future data , and provided the pao does not see evidence for anisotropy at lower energies , as seems to be the case ( abraham et al . 2007 ) . we discuss the implications of these results in the last two sections . in section [ sec : agn ] , we argue that local agn ( including fr i radio - galaxies ) do not possess the power required to accelerate protons to ultra - high energies . when combined with the previous conclusions , this leads to the conclusion that the current pao data do not support agn as sources of the highest energy cosmic rays . finally , section [ sec : disc ] summarizes the findings and concludes that the current data on composition and anisotropy suggest one of the following : ( i ) the shower modelling or , what would be more interesting , the hadronic theoretical models of shower reconstruction are in error at high energy ; ( ii ) the composition switches abruptly from heavy to light above @xmath14eev ; ( iii ) the source injects primarily heavy nuclei ( which seems unlikely ) ; ( iv ) the anisotropy seen by the pao is a statistical artefact . as argued below , one can use the results of searches for anisotropy at various energies in order to constrain the chemical composition of ultra - high energy cosmic rays . the basic claim is the following : if one detects anisotropy above some energy @xmath1 , but not below , then the chemical composition is most likely light above @xmath15 , because one would have otherwise observed a similar anisotropy pattern at an energy @xmath16 , with @xmath0 the assumed average charge of the cosmic rays at energies @xmath17 . consider a region of the sky , of angular size @xmath18 and solid angle @xmath19 , showing an excess of particles above isotropic expectations above some energy threshold @xmath15 . the most natural interpretation is to infer the existence of one or more sources in this direction , the image of which is spread over @xmath20 , because of smearing by @xmath18 in the intervening magnetic fields , or because @xmath20 subtends the source distribution . for the sake of the argument , we assume that the spectrum produced by the source consists of protons and of heavy nuclei of charge @xmath0 . in all known source models , one expects that the ratio of the elemental spectra @xmath21 and more generally @xmath22 at a given energy @xmath23 , where @xmath24 denotes the maximal energy at the source for protons . for instance , the composition ratio of protons to iron peak elements ( @xmath25 ) in the galactic cosmic ray spectrum , which is roughly consistent with the solar abundance when compared at a given energy per nucleon , implies a source composition at a given energy @xmath26 ( as taken from the recent atic-2 data , panov et al . the only energy regime in which one may obtain @xmath27 is @xmath28 , with @xmath29 the maximal energy for the nuclei of charge @xmath0 ( see also further below ) . we also assume here that the maximal energy scales as @xmath0 ; as argued further below , this is a conservative choice . now , let us ask what would be seen if the composition above @xmath1 were heavy . consider the quantity @xmath30 which characterizes the signal - to - noise ratio of the anisotropy signal : @xmath31 represents the excess number of events over isotropic expectations , i.e. the difference between the total number of events observed above @xmath1 within @xmath20 and @xmath32 , the number of events expected from the isotropic background . in the above model , @xmath33 corresponds on average to the number of events produced by the sources inside @xmath20 . a detection of anisotropy thus corresponds to @xmath34 . one can now compute a similar quantity , but corresponding to the proton component at energies @xmath2 , denoted @xmath35 . these protons have a same magnetic rigidity than the heavy component above @xmath1 , therefore they follow the same path in the galactic and extra - galactic magnetic fields , and consequently they are to produce a similar anisotropy pattern in the sky , up to the composition ratio of the elemental spectra and up to the level of background noise associated with the isotropic component . assuming that the elemental spectra at injection have the same power - law form @xmath36 at @xmath37 , and that the total number of events observed scales as @xmath38 , with @xmath39 the slope of the observed all - sky differential spectrum , one finds @xmath40 the fudge factor @xmath41 incorporates the effect of energy losses ; it is discussed in more detail in the following . at energies above the ankle , one has @xmath42 ( abbasi et al . 2008 , abraham et al . therefore , if the source emits a power - law with @xmath43 , one finds that the signal - to - noise ratio for detecting the anisotropy is larger at @xmath16 than at @xmath1 by a factor @xmath44 , generally expected to be significantly larger than unity ! clearly , the softer the source spectrum , the larger the gain in signal - to - noise ratio with decreasing @xmath0 . the factor @xmath45 accounts for the difference between the propagated and injected spectra . given that the energy loss distance for protons of energy @xmath16 is much larger than the energy loss distance for heavy nuclei of energy @xmath1 , if @xmath46ev , one expects @xmath47 . indeed , regarding the ( primary ) proton component , energy losses can be safely neglected at @xmath16 energies , since the attenuation length is of order of a gpc or more at eev energies , which is much larger than the depth up to which which anisotropies can be produced , this latter being of order of the inhomogeneity scale of the large scale structure @xmath48mpc or of the distance to the closest source , whichever is larger . therefore , @xmath45 approximately corresponds to the reciprocal of the attenuation factor for heavy nuclei of charge @xmath0 and energy @xmath17 propagating over the distance scale @xmath49 to the sources inside @xmath20 . the attenuation factor is understood as the ratio of the flux above @xmath17 after propagation on a distance @xmath49 to the injected flux above this energy . for the case of iron primaries , this attenuation factor is @xmath50 above @xmath51eev at a distance of 10 mpc , becoming @xmath52 at a distance of 100 mpc ( counting in the propagated flux all secondary nuclei with @xmath53 ) . for oxygen , the attenuation factor is @xmath54 at 10 mpc , becoming @xmath55 at 100 mpc ( counting in all c , n and o secondary nuclei ) ; see bertone et al . ( 2002 ) for an illustration of the relative fragility of ultra - high energy cno nuclei vs iron peak elements . actually , this enhancement factor @xmath45 should be further increased by the number of secondary protons with energy @xmath2 produced by the photo - disintegration of the primary heavy nuclei with energy @xmath17 . this number is not easy to quantify but it can be significant . for instance , recent simulations by aloisio , berezinsky & gazizov ( 2008 ) indicate that the signal should be dominated by protons at energies @xmath56ev even if the source composition were strongly enriched in iron nuclei [ assuming of course that @xmath57ev ] . allard et al . ( 2008 ) also conclude that heavy nuclei could be found in abundance at the detector only if the composition were essentially dominated by fe group nuclei at energies beyond @xmath58ev . as discussed in this latter reference , and above , the generic case that gives rise to a heavy composition at ultra - high energies is when the proton maximal energy at the source is smaller than the gzk cut - off . in the following , we adopt a conservative point of view and neglect the effect of energy losses , i.e. we set @xmath59 . let us now discuss the scaling of @xmath60 with charge . the maximal energy is given by the comparison of the acceleration timescale ( which depends on rigidity @xmath12 ) with the minimum of the escape timescale ( which also depends on @xmath12 ) , the age of the source or dynamical timescale ( which do not depend on @xmath11 , @xmath0 ) and the energy losses timescale ( which depends on @xmath11 , @xmath0 and mass number @xmath61 ) . if the latter timescale does not provide the dominant limitation to the acceleration mechanism , then the maximal energy scales as the rigidity , as assumed above . if energy losses are dominant however , one expects @xmath62 , as explained in the following . then the above argument would remain valid , as it only requires that protons exist at energies @xmath2 , i.e. that @xmath63 . note that in the more extreme case in which protons and heavy nuclei had similar maximal energies , the composition at ultra - high energies @xmath17 would be dominated by protons , and therefore the anisotropy would be due to those protons since they have a higher rigidity than the heavy nuclei of a same energy . for photo - hadronic interactions , the energy loss timescales for protons and iron - peak nuclei are known to be comparable , while that of intermediate elements is generally smaller ; in that case , one would expect @xmath64 . regarding synchrotron losses , the energy loss timescale scales as @xmath65 , therefore @xmath66 if the acceleration timescale @xmath67 and synchrotron losses dominate at the highest energies . in that case , the scaling @xmath68 remains correct to within a factor of order unity . the above conclusions agree with the recent simulations of allard & protheroe ( 2009 ) . finally , accounting for a more complex chemical composition with more than two species would not modify significantly the results shown in eq . ( [ eq : sigma ] ) as long as the anisotropy above @xmath1 is indeed due to the species of charge @xmath0 . in practice , it will be useful to test the composition above @xmath1 by checking the anisotropies at various @xmath69 , with @xmath70 representative of the most abundant elements . in the following , we provide specific examples of the above test applied to the most recent data of the pao . we will show in particular that the above argument remains valid when one considers the all - sky anisotropy pattern due to an ensemble of close - by sources . abraham et al . ( 2007 , 2008b ) have reported a correlation of 20/27 arrival directions of the highest energy events ( @xmath71ev ) with nearby ( @xmath72mpc ) active galactic nuclei . however , the typical model of ultra - high energy cosmic ray origin in agn refers to strongly beamed fanaroff - riley ii ( fr ii ) sources with giant radio lobes ( e.g. rachen & biermann 1993 ) , while 19 out of these 20 correlating agn in the pierre auger dataset belong to the seyfert or liner class , only one being a fanaroff - riley i ( fr i ) radio - galaxy ( cen a ) . these correlating agns are thus more likely to be tracers of the local large scale structure in which the actual sources of ultra - high energy cosmic rays camouflage and indeed , the arrival directions are compatible with a distribution of sources following the large scale structure ( kashti & waxman 2008 , ghisellini et al . 2008 , zaw , farrar & greene 2009 ) . the extended dataset up to 2009 of the pao reveals a significantly weaker correlation , with only 26 out of 58 events above @xmath51eev lying within @xmath9 of an agn located within 75mpc ( hague et al . this level of correlation is compatible with that expected for sources distributed as the large scale structure ( see discussion in kotera & lemoine 2008 and aublin et al . 2009 for a detailed discussion of the correlation with cosmological catalogs ) . it is important to note that the departure from isotropy lies at the @xmath73 level . therefore one should be careful not to overinterpret this excess . in the following , we use these pao results as a concrete example for the test developed in the previous section . interestingly , the pao data up to 2007 has also shown evidence for clustering of events in the region around cen a , with as many as @xmath74 events within @xmath75 ( gorbunov et al . 2008 ) , as well as correlation of @xmath76 out of 27 events with a sample of radio - galaxies ( moskalenko et al . 2008 ; nagar & matulich 2008 ) . this has triggered a surge of interest in models of ultra - high energy cosmic ray origin in cen a as well as forecast studies of neutrino and gamma - ray expected signals from cen a ( e.g. cuoco & hannestad 2008 , gupta 2008 , halzen & omurchadha 2008 , holder et al . 2008 , kachelriess et al . however , such a correlation does not necessarily imply that cen a is the source of these cosmic rays . first of all , and as noted in abraham et al . ( 2008b ) , hague et al . ( 2009 ) , the evidence for clustering is based on an a posteriori analysis , so that it is difficult to quantify the level of significance . even if future data confirm this clustering , one must keep in mind that cen a lies in front of one of the largest concentrations of matter in the local universe , at @xmath77mpc from us , so that the correlation could be accidental . finally , as we argue in section [ sec : cena ] , the cen a agn / jet / lobe system is too weak to accelerate protons up to the observed energies @xmath78eev . even more intriguing is the fact that the chemical composition measured by the pao becomes increasingly heavier above the ankle at @xmath79eev , with the last data points at @xmath14eev lying close to expectations for iron ( unger et al . 2007 , wahlberg et al . 2009 ) . such a measurement is in obvious conflict with the most recent analysis of hires data , which indicate a pure proton composition up to @xmath80eev ( belz 2009 ) . as we argue in the following , it is also at odds with the observed anisotropy pattern above @xmath51eev , unless the composition abruptly changes from heavy to light above @xmath14eev . let us first discuss the anisotropy toward cen a , using the most recent pao dataset , in particular , @xmath81 events located within @xmath82 out of a total of @xmath83 events above @xmath51eev ( hague et al . 2009 ) , assuming that the clustering is not a statistical accident . in a first approximation , one can model the anisotropy pattern as in the previous section , with one ( or several clustered ) source(s ) contributing a fraction @xmath84 of the total flux , the remaining @xmath85 being accounted for by an isotropic background . we do not need to assume that cen a itself is a source , but we center the anisotropic signal on the location of cen a in order to match the observed pattern . given that @xmath86 events are expected on average from this region of the sky if the sources are distributed isotropically , we set @xmath87 in the following . we assume that the sources inject protons and iron nuclei with powerlaw spectra @xmath88\right\}$ ] , producing an angular image of size @xmath89 . as before , this angle could represent smearing due to angular deflection , in which case @xmath90 , or the angular size of the ensemble of sources . although we assume that all heavy nuclei are iron , the following could be easily generalized to any kind of mixed composition without changing the basic result . we also neglect the energy losses for the spectrum of the anisotropic component , as before . this is motivated by the proximity of matter in this direction of the sky and , as discussed before , incorporating energy losses would not diminish the signal we are to extract out of the data , while it would introduce other free parameters . the remaining @xmath85 fraction of the flux is modelled with an isotropic background , e.g. as produced by far away sources . let us first assume that the composition is dominated by iron at the highest energies . we set @xmath91eev , and @xmath92 above @xmath1 . we also set @xmath93eev , so that essentially no proton contribute to the signal above @xmath1 . in the top panel of fig . [ fig : ptsa ] we plot the histogram of the number of events as a function of the angular distance to the source ( solid line ) . the colored region indicates the expectations for a purely isotropic signal . the departure from isotropy is clear , although the uncertainty on the signal in each bin is substantial given the small number of events involved . here the number of events within @xmath82 of cen a is @xmath94 for the model ( slightly lower than the observed 12 events , yet within the uncertainty ) and @xmath86 for the isotropic expectations so that the inferred @xmath95 . in the middle panel of fig . [ fig : ptsa ] , we plot the same histogram , but at an energy @xmath96eev . in order to simulate the anisotropy signal , we have proceeded as follows . we have assumed that the total number of events increases ( with decreasing energy ) according to the observed data ( in particular the pao spectrum published in yamamoto et al . 2007 ) , and we have calculated the contribution to this flux above @xmath97 for all the protons injected by the source above this energy . we have assumed that the ratio of protons to iron is 1:0.06 at a given energy well below @xmath60 , corresponding to the composition ratio of proton to iron peak elements ( @xmath25 ) in the galactic cosmic ray spectrum reconstructed at the source . the iron nuclei injected by the source at energies @xmath98eev add up to the isotropic background and do not contribute to the anisotropy signal . as made obvious in fig . [ fig : ptsa ] , the gain in signal - to - noise ratio is very large : the number of events observed above @xmath99eev within @xmath82 is now 5150 , for 2330 expected , so that @xmath100 . this gain by a factor @xmath101 can be recovered from eq . ( [ eq : sigmap ] ) . finally , the lower panel of fig . [ fig : ptsa ] shows the ratio of the differential histograms @xmath102 for the source proton contribution plus isotropic background ( filled circles ) to the isotropic background @xmath103 normalized to a same total number of events , including the error bars evaluated through the monte carlo simulation . the empty circles show the same quantity , but for a model in which the source injects only protons , even up to the highest energies , assuming that @xmath89 represents magnetic deflection , i.e. @xmath90 : we have set @xmath92 above @xmath51eev in this model , so that the anisotropy signal above @xmath51eev is similar to that of the previous iron+proton model shown in the top panel of fig . [ fig : ptsa ] , but the low energy signal is essentially isotropic , see the bottom panel . the conclusions to be drawn are clear . if the clustering seen toward cen a is real , the events responsible for this anisotropy pattern toward cen a above @xmath51eev are unlikely to be heavy nuclei , otherwise pao would have observed a much stronger anisotropy at energies @xmath104eev ( abraham et al . 2007 , 2008b ) . in the present section , we investigate how the anisotropy signal with respect to nearby agn , as reported in hague et al . ( 2009 ) can be used to discriminate the chemical composition at the highest energies . we proceed in a similar way as before , but we now assume that the spectrum is composed of two components : one that follows the local large scale structure , contributing a fraction @xmath84 to the flux above @xmath51eev , and one isotropic component that makes up for the remainder , @xmath85 . as a proxy for the local large scale structure , we use the pscz catalog of galaxies ( saunders et al . 1991 ) , as in previous studies ( kashti & waxman 2008 , kotera & lemoine 2008 ) . out of simplicity , we assume that the sources follow the pscz distribution without bias . however , we allow for the possibility of a deflection angle @xmath89 in intervening magnetic fields and we explore two possible values @xmath105 and @xmath92 above @xmath51eev . this choice is motivated by the resolution of the pscz catalog , approximately @xmath106 , and by the possibility that this signal is composed of heavy nuclei . in practice , we proceed through monte carlo simulations as follows . we construct many mock catalogs of angular sky distributions , each mock catalog being a sample of @xmath83 events above @xmath51eev ( or the corresponding number at any other energy ) . the arrival direction of each event in the sample is drawn at random , according to a probability law reproducing the pscz distribution and pao coverage if it belongs to the @xmath84 anisotropic fraction , or reproducing isotropic arrival directions and pao coverage if the event belongs to the remaining @xmath85 fraction of isotropic background . in order to draw an event from the pscz catalog , we include the appropriate weight factors for the flux dilution with inverse distance squared , for the pscz exposure function and for the pao aperture . then we randomly displace this arrival direction by @xmath89 in the plane of the sky . the energy losses are accounted for by the depth of the catalog , which is limited to @xmath107mpc . we have set @xmath108 , which allows to produce an anisotropy signal close to the level observed : the pao indeed records 24 events out of 58 events above @xmath51eev located within @xmath109 of an agn closer than @xmath110mpc , while the above model produces @xmath111 events within @xmath109 if @xmath105 and @xmath112 events if @xmath92 . the model with 90% of the flux above @xmath51eev coming from sources distributed as the large scale structure with negligible deflection thus gives a number of correlating agn within @xmath109 that is comparable to that seen by pao , to within the uncertainty . a deflection angle @xmath92 provides a signal that is marginally too low , but given the simplifying assumptions made above , this is not crucial . it certainly indicates that larger deflection angles will probably result in too small anisotropies . it is interesting to note that the anisotropy signal peaks at a separation which is substantially smaller than @xmath89 , i.e. @xmath113 for @xmath92 . this of course results from the pao procedure , which selects the closest agn to the arrival direction and not to the source direction , i.e. after the direction at the source has been modified by @xmath89 . the differential histograms corresponding to these models and isotropic expectations are shown in fig . [ fig : lssa ] as a function of the angular separation to the closest agn drawn from the vron - cetty & vron ( 2006 ) catalog , with an arbitrary cut at @xmath110mpc , following the pao analysis . these histograms are shown for the two possible values of @xmath89 , as indicated . here , we find @xmath114 and @xmath115 for @xmath105 and @xmath92 respectively . the signal is thus marginal in terms of departure from isotropy . we now plot the ratios of the differential histograms expected in these two models at low energy @xmath116ev to the histogram expected for purely isotropic arrival directions , as before . we have assumed a conservative composition ratio 1:0.06 between proton and iron as in the previous section , and we have neglected energy losses on the propagated spectra . the results are shown in fig . [ fig : lssb ] , which clearly reveals a stronger anisotropy signal at these energies . in terms of signal - to - noise , we find @xmath117 and @xmath118 respectively for @xmath105 and @xmath92 ( @xmath119 and @xmath120 respectively , @xmath121 ) , which significantly exceed the signal - to - noise ratios obtained above @xmath51eev . this generalizes the argument presented in section [ sec : test - chem ] for the case of a more complex distribution of point sources on the sky , superimposed on an isotropic background . it also shows that the present angular data of the pao disfavors a heavy composition at ultra - high energies . strictly speaking , this test demonstrates that the _ anisotropic _ part of the signal can not be due to heavy nuclei _ but _ one can not exclude that the isotropic background is made up of heavy nuclei . however , our above model also suggests that a significant fraction ( 90% in fig . [ fig : lssa],[fig : lssb ] ) of the flux at @xmath13eev is produced by the anisotropic component if one takes at face value the anisotropy results of the pao . this suggests that the bulk of ultra - high energy cosmic rays above @xmath51eev are protons . on general grounds , one can construct a lower bound to the magnetic luminosity that a source must exhibit in order to be able to accelerate particles up to @xmath58ev ( norman et al . 1995 , waxman 1995 , lyutikov & ouyed 2005 , waxman 2005 ) . let us first briefly describe the general , model independent argument given in waxman ( 1995 ) . consider an astrophysical source driving a flow of magnetized plasma , with characteristic magnetic field strength @xmath122 and velocity @xmath123 . imagine now a conducting wire encircling the source at radius @xmath124 . the potential generated by the moving plasma is given by the time derivative of the magnetic flux , and is therefore given by @xmath125 where @xmath126 . a proton which is allowed to be accelerated by this potential drop would reach energy @xmath127 . the situation is somewhat more complicated in the case of a relativistic outflow , where @xmath128 . in this case , the proton is allowed to be accelerated only over a fraction of the radius @xmath124 , comparable to @xmath129 . to see this , one must realize that as the plasma expands , its magnetic field decreases , so the time available for acceleration corresponds , say , to the time of expansion from @xmath124 to @xmath130 . in the observer frame this time is @xmath131 , while in the plasma rest frame it is @xmath132 . thus , a proton moving with the magnetized plasma can be accelerated over a transverse distance @xmath133 . this sets a lower limit to the product of the magnetic field and source size , which is required to allow acceleration to @xmath134 , @xmath135 . this also sets a lower limit to the rate @xmath136 at which energy is carried by the out flowing plasma , and which must be provided by the source ( waxman 1995 , 2005 ) , @xmath137 where we have generalized the last equation to particles of charge @xmath0 . let us consider next the application of the general argument given above to the commonly assumed scenario , where particle acceleration within the outflowing plasma is achieved via diffusive acceleration . in this case , one writes the acceleration timescale @xmath138 as a multiple @xmath139 of the larmor time , and one compares it to the dynamical timescale available for acceleration . let us discuss this bound in some detail . for the sake of definiteness , we assume an outflow of lorentz factor @xmath140 and opening angle @xmath141 ( the isotropic case is recovered in the limit @xmath142 ) . the quantity @xmath139 depends on the diffusion coefficient of the particle , hence on its energy and the general properties of the magnetic field of the accelerating region . in acceleration mechanisms which involve scattering against magnetic inhomogeneities ( such as fermi i or fermi ii ) , @xmath143 , where @xmath144 is the scattering timescale and @xmath145 the diffusion coefficient . the parallel diffusion coefficient can be written as @xmath146 ( see for instance casse , lemoine & pelletier 2002 ) , with @xmath147 the fraction of turbulent magnetic energy density ( in units of total magnetic energy density ) and @xmath148 the coherence length . the exponent @xmath149 depends on the spectral index @xmath150 of the ( one dimensional ) turbulence power spectrum : @xmath151 for @xmath152 , @xmath153 for @xmath154 . from this , one easily verifies that @xmath155 is general , meaning that the so - called bohm regime @xmath156 is a limiting regime . in non - relativistic fermi shock acceleration ( @xmath157 or @xmath158 ) , @xmath159 , with @xmath160 the shock velocity in the comoving frame and @xmath145 refers to quantities measured in the upstream frame . hence @xmath161 . one recovers a similar result for non - relativistic fermi ii acceleration , with @xmath162 , @xmath163 denoting the alfvn speed . if acceleration takes place in a ( non - relativistic ) shear flow with velocity gradient @xmath164 , the timescale @xmath165 ( rieger , bosch - ramon & duffy 2007 ) . the deconfinement limit corresponds to @xmath166 , in which case the limiting acceleration timescale becomes comparable to @xmath167 , as above . note that the acceleration timescale is larger ( hence acceleration is less efficient ) when @xmath168 . in the ultra - relativistic shock limit @xmath169 , one expects @xmath170 ( up to a factor of order unity , see pelletier , lemoine & marcowith 2009 ) , with the condition @xmath171 . since @xmath139 increases with @xmath172 , this process is in general not efficient . finally , in the mildly relativistic regime @xmath173 , one expects @xmath174 , hence @xmath175 , whose minimum value is of order one provided the bohm regime diffusion applies and the magnetic field is nearly fully turbulent . to summarize , @xmath176 in general , and @xmath177 corresponds to a maximally efficient acceleration process . the constraint @xmath178 , with @xmath179 the dynamical timescale at radius @xmath124 ( in the lab frame ) , now gives : @xmath180 with @xmath181 the observed energy in units of @xmath58ev . the presence of @xmath182 does not mean that in the limit @xmath183 , this lower bound vanishes , since @xmath184 more than compensates for this term . similarly , the above bound does not vanish in the limit @xmath185 , since lateral escape losses become prominent when @xmath186 . the corresponding timescale @xmath187 , hence @xmath188 . in the case of agn jets , one finds in general @xmath189 , hence both limits are comparable . according to the above discussion , the most optimistic values for the parameters entering these equations are @xmath190 , @xmath191 and @xmath192 . therefore @xmath193erg / s can be considered as a firm lower bound on the source magnetic luminosity . since most of the agn seen to correlate within @xmath109 of the pao events are seyfert galaxies , with bolometric luminosities well below @xmath194ergs / s , the agn that they harbor can not be the source of light particles at @xmath58ev . we also note that zaw , farrar & greene ( 2009 ) have shown that one third of this sample is actually made of star forming galaxies with little or no agn activity . hence the correlation with stricto - sensu agn objects is actually weaker than reported . actually even fr i radio - galaxies , tev blazars and bl lac objects do not possess the required power to accelerate @xmath195 particles up to @xmath58ev , since their magnetic luminosities are of order @xmath196ergs / s ( celotti & ghisellini 2008 ) . such objects might possibly accelerate heavy nuclei up to energies around the gzk cut - off but , as discussed above , the anisotropy pattern at high and especially low energies would then appear inconsistent with the pao results . further anisotropy studies in both energy ranges would certainly further constrain their possible contribution . the compilation of celotti & ghisellini ( 2008 ) has been done in the frame of leptonic synchrotron and inverse compton emission . in hadronic blazar models , the magnetic field in the blazar zone is typically one order of magnitude larger than in leptonic models , so that these objects might fulfill the magnetic luminosity bound to accelerate protons up to @xmath58ev . however , in this case , acceleration would occur in the blazar zone hence the emission should be beamed forward . in order to escape further expansion losses in the magnetized jets , the accelerated protons would have to be converted into neutrons , which would decay back to protons on a distance scale @xmath197mpc , i.e. outside the jet . one should therefore observe a correlation of the arrival directions with blazars , not with radio - galaxies seen offside ( rachen 2008 ) . however the pierre auger observatory reports no correlation with blazars ( harari et al . 2007 ) , so that this possibility is equally at odds with present observational results . keeping in line with the modelling of celotti & ghisellini ( 2008 ) , only flat spectrum radio quasars ( the likely equivalent to the most powerful fr ii sources ) seem capable of producing jets with @xmath198ergs / s . but , considering the highest energy pao event ( @xmath199ev ) , one finds that the smallest angular separation of this event to the fr ii sources located within 130 mpc , as compiled by massaglia ( 2007 ) is already @xmath200 ( for ngc 4261 ) . the next objects are 3c 296 ( @xmath201 away ) and pks1343 - 60 ( @xmath202 away ) . the closest blazar ( classified as bl lac in the vron & vron - cetty catalog ) located closer than @xmath203mpc lies @xmath204 away from this highest energy event ( tex 0554 + 534 ) . recently , it has been proposed that transients in active galactic nuclei could power up the engine up to the luminosities required , thereby evading the above constraints ( see for instance farrar & gruzinov 2009 , dermer et al . however , as demonstrated by waxman & loeb ( 2009 ) , such transients should produce counterpart flares in x - rays through the concomittant acceleration of electrons , which should have been picked up by existing surveys . their non - detection strongly argues against such flaring scenarios . given its proximity and the apparent clustering of pao events in this direction of the sky , cen a has received a lot of attention . in particular , several authors have suggested that this agn could host a site of acceleration to ultra - high energies ( cavallo 1978 ; romero et al . 1996 ; farrar & piran 2000 ; gorbunov et al . 2008 ; hardcastle et al . 2009 , osullivan , reville & taylor 2009 ) . let us therefore discuss it in some more detail . centaurus a is classified as a fr i / misaligned blazar ( see israel 1998 for a review ) . its bolometric luminosity @xmath205erg / s and its jet kinetic power is inferred to be @xmath206erg / s . therefore , this source should not be able to accelerate light particles to @xmath58ev . the detailed modelling of the spectral energy distribution of its nucleus by chiaberge , capetti & celotti ( 2001 ) provides further information on the emission zone : @xmath207 cm , @xmath208 g and @xmath209 . this corresponds to a magnetic luminosity @xmath210ergs / s , a value that seems quite reasonable in view of the jet kinetic power and of the bolometric luminosity , and which agrees relatively well with estimates of the magnetic field in the inner jet at larger distances from the core , when assuming @xmath211constant . the papers which have discussed the issue of acceleration to ultra - high energy in some detail are romero et al . ( 1996 ) , which has argued that acceleration can take place in the x - ray knots of the inner jet through diffuse shock acceleration , hardcastle et al . ( 2009 ) and osullivan , reville & taylor ( 2009 ) , which have discussed the stochastic acceleration of protons in the giant lobes of cen a and gorbunov et al . ( 2008 ) relying on the model of neronov et al . ( 2007 ) of particle acceleration in a black hole magnetosphere . romero et al . ( 1996 ) uses the inferred values of @xmath212 g ( upstream field ) , @xmath213kpc , @xmath214 and @xmath215 . however , since the larmor radius @xmath216 ( with @xmath217ev , @xmath218 g ) , even the hillas criterion ( hillas 1984 ) is not satisfied . balancing the acceleration timescale with the escape timescale @xmath131 , one derives a maximal energy @xmath219ev ( @xmath220 , @xmath221 ) . as discussed above , this maximal energy lies in the correct ballpark for accelerating nuclei to ultra - high energies . however the anisotropy signal would then appear incompatible with the pao data ; in particular , a strong anisotropy signal at eev energies should have been detected . hardcastle et al . ( 2009 ) have argued that protons could be accelerated up to @xmath222ev through stochastic acceleration in the magnetized turbulence of the giant lobes of cen a. however , one can show that their estimate of the acceleration timescale is far too optimistic because , when applied to the electrons , it produces a maximal energy well in excess of the inferred @xmath223ev . in detail , using the estimate of hardcastle et al . ( 2009 ) for the acceleration timescale , @xmath224 , and their bohm scaling @xmath225 , one infers the maximal energy for electrons by comparing @xmath226 to the synchrotron and inverse compton energy loss timescales , leading to @xmath227ev . the above maximal electron energy is five orders of magnitude larger than the inferred maximal energy , hence the above acceleration timescale is far too optimistic and , consequently , the maximal proton energy must be much lower than @xmath58ev . this conclusion agrees with those of osullivan , reville & taylor ( 2009 ) ; see also casse , lemoine & pelletier ( 2002 ) for a similar discussion regarding the acceleration in the lobes of powerful radio - galaxies . finally , neronov et al . ( 2007 ) have argued that in a magnetic field of strength @xmath228 g threading a maximally rotating black hole of mass @xmath229 , heavy nuclei could be accelerated to energies of order @xmath58ev by exploiting the potential drop @xmath230v . however , this maximal energy is quite optimistic because it ignores radiative energy losses , which are highly efficient in the nuclei of powerful agn ( norman et al . 1995 , henri et al . 1999 ) . in any case , the poynting flux that is required , @xmath231erg / s far exceeds any other luminosity measured elsewhere in the galaxy , and in particular the estimate obtained from multi - band spectral modelling of the nucleus emission ( chiaberge , capetti & celotti 2001 ) . let us start by summarizing the results obtained so far . we have discussed how one can can test the chemical composition of ultra - high rays on the sky by comparing the anisotropy signals at various energies . our main result is to show that if anisotropies are observed above some energy @xmath1 and the composition is assumed to be heavy at that energy ( nuclei of charge @xmath0 ) , one should observe at energies @xmath2 a substantially stronger anisotropy signal , which is associated with the proton component emitted by the sources that are responsible for the anisotropy pattern . these conclusions do not depend at all on the modelling of the intervening magnetic fields and they are robust with respect to the parameters characterizing the sources . we have then applied this test to the most recent data of the pao . we find that , if these data are taken at face value ( and if the pao arrival directions around the ankle appear isotropic ) , the events that are responsible for the anisotropy signal reported toward cen a and toward the nearby agn should not be heavy nuclei . interestingly , this result appears to be at odds with the current pao results on the chemical composition at the highest energies . one can not exclude at the present time that the observed anisotropies are statistical accidents ; only the acquisition of a larger set of data will tell . one can not exclude either that the composition switches abruptly from heavy to light at @xmath232eev . finally , one can not exclude that a systematic bias affects the composition measurements of the pao . in any case , the present discussion indicates that , if both the anisotropy and the heavy composition are confirmed by future data , some possibly important information is to be extracted out of this apparent contradiction . the theoretical and experimental uncertainties in the extrapolation of the @xmath233 cross - section to center - of - mass energies @xmath234tev are one of the possible sources of biases in shower reconstruction . recently , it has been argued that , if this cross - section were underestimated by some @xmath235% at these energies , one might reconcile the existing @xmath236 measurements with a pure proton composition at energies above the ankle ( ulrich et al . 2009a , 2009b ; see however wibig 2009 ) . in this respect , and in the light of the above discussion , the apparent anisotropy towards cen a offers a very valuable opportunity to test this possibility . it suffices , in principle , to test the anisotropy pattern against a future ( independent ) data set at various energies by comparing the number of events detected within a predetermined solid angle in a predetermined direction to the expected number for isotropic arrival directions ; at the same time , one needs to perform a dedicated composition measurement for cosmic rays from this particular region of the sky , in order to avoid as much as possible the contamination due to the isotropic contribution to the all - sky flux . then , one should apply the test discussed above . the natural question that follows is what can be inferred about the sources of ultra - high energy cosmic rays if , following the anisotropy results of pao and the above discussion , one assumes that most ultra - high energy cosmic rays are protons . let us now explore this line of reasoning . first of all , one concludes that the existing data of the pao disfavor the agn model of ultra - high energy cosmic ray origin . indeed , we have discussed in section [ sec : agn ] a robust lower bound to the magnetic luminosity that a source must possess in order to be able to accelerate particles of charge @xmath0 up to @xmath58ev , @xmath237erg / s . this bound allows to rule out the local radio quiet agn as sources of @xmath58ev protons . we have also argued that , while fr i radio - galaxies do not seem to possess the required luminosity to accelerate protons up to @xmath58ev , fr ii radio - galaxies and blazars do not correlate with the highest energy events . although radio - loud galaxies could possibly accelerate heavy nuclei up to energies close to the gzk cut - off , as we have discussed they would produce an anisotropy pattern predominantly at @xmath238eev energies rather than at ultra - high energies , in conflict with the pao results . we have discussed the particular case of cen a in some detail , as this source has recently received a lot of attention due to the observed clustering of events around it . we have provided a critical discussion of acceleration in this object and argued that its agn / jet / lobe system can not produce protons at the energies required , in good agreement with the general expectations from the above luminosity bound . that being said , the excess clustering in this particular direction , if real , remains to be interpreted . first , if sources are distributed according to the large scale structure , one expects a certain number of events to coincide accidentally with the direction of cen a , because of the location of the centaurus and shapley superclusters in this direction . using the pscz catalog of galaxies as a proxy for the distribution of the sources , with a depth of @xmath239mpc , and accounting for the pao exposure , one expects @xmath240 events above @xmath51eev within @xmath241 of cen a nucleus ( and @xmath242 for isotropic arrival directions ) for @xmath243 observed , or @xmath244 events within @xmath82 ( and @xmath86 for isotropic arrival directions ) for @xmath81 observed . note that the large scale structure is poorly sampled in the proximity of cen a due to its low galactic latitude . nevertheless , the above numbers already indicate that the contamination from the surrounding large scale structure in the direction to cen a is substantial . in particular , at the present level of statistics , one can not claim observing a significant departure from a model in which the sources are distributed as the large scale structure . it should also be recalled here that the pao analysis of clustering around cen a is an a posteriori one , hence one may not therefore assign a reliable significance to the detection of the anisotropy . moreover , if sources are gamma - ray bursts or magnetars , one should include in the flux prediction the contribution of such sources located in the cen a host galaxy itself . it is well known that the probability of detecting ultra - high energy cosmic rays from nearby gamma - ray bursts is extremely small unless the arrival times of these ultra - high energy cosmic rays are sufficiently dispersed ( waxman 1995 , waxman & miralda - escud 1996 ) . consider for instance gamma - ray bursts with rate of occurence @xmath245 in cen a ; these can be seen if either of their jets points into a solid angle @xmath246 where @xmath89 is the typical deflection acquired by the cosmic ray en route to the detector , giving a detection probability @xmath247 , with @xmath248 ; @xmath249 represents the spread of arrival times and @xmath250 the angular deflection for particles of energy @xmath251eev traveling from cen a to the detector through a magnetic field of strength @xmath252 g and coherence length @xmath253kpc ( waxman & miralda - escud 1996 ) . the above probability suggests that one should not detect events from gamma - ray bursts emitted directly toward the observed . however , if the jet of the gamma - ray burst hits one of the lobes of cen a , the particles can be redistributed in all directions , in particular towards the detector . the magnetic field inside the lobe , @xmath254 g is indeed sufficient to impart a deflection of order unity on the lobe distance scale @xmath48kpc to protons of @xmath255eev . in this case , the expected time delay and time dispersion are of the order of the travel time to the lobes , @xmath256yr , so that the expected number of gamma - ray bursts that one can see at any time , through rescattering of the particles on the lobes of cen a is : @xmath257 for an apparent solid angle of the lobes @xmath258 as viewed from the host galaxy of cen a , which is close to @xmath259 . the last gamma - ray burst(s ) of cen a may then contribute to the flux above the gzk cut - off up to : @xmath260 with @xmath261yrs , @xmath262 for an injection spectrum with @xmath263 , and @xmath264ergs the energy injected by one gamma - ray burst in ultra - high energy cosmic rays . this flux should be spread over the angular scale of the lobes , @xmath265 . at @xmath255eev , pao records a diffuse flux @xmath266 . after correcting for the solid angle of the cen a image and the pao aperture towards cen a , one finds that the last gamma - ray bursts in cen a should contribute up to @xmath267% of the observed pao flux , well within the range of the observed anisotropy . we also note that this contribution from sources inside cen a would improve the overall correlation of all the pao data above @xmath51eev with sources distributed as the local large scale structure . finally , we remark that the above effect of rescattering on the lobes provides a clear example of the possible biases discussed in kotera & lemoine ( 2008 ) , namely that the pao could be preferentially observing the last scattering centers ( in the present case , the lobes of cen a ) rather than the source itself ( the gamma - ray bursts located in the core of cen a ) . abbasi , r. u. , et al . ( the hires collaboration ) , 2004 , , 610 , l73 abbasi , r. u. , et al . ( the hires collaboration ) , 2005 , , 623 , 164 abbasi , r. u. , et al . ( the hires collaboration ) , 2008 , physical review letters , 100 , 101101 abraham , j. et al . ( the pierre auger collaboration ) , 2007 , science , 318 , 938 abraham , j. , et al . ( the pierre auger collaboration ) , 2008a , physical review letters , 101 , 061101 abraham , j. et al . ( the pierre auger collaboration ) , 2008b , astroparticle physics , 29 , 188 allard , d. , busca , n. g. , decerprit , g. , olinto , a. v. , & parizot , e. 2008 , journal of cosmology and astro - particle physics , 10 , 33 allard , d. , & protheroe , r. j. 2009 , arxiv:0902.4538 aloisio , r. , berezinsky , v. , & gazizov , a. 2008 , arxiv:0803.2494 arons , j. 2003 , , 589 , 871 belz , j. ( the hires collaboration ) , _ proc . ( malfa , italy , sept . 2008 ) aublin , j. et al . ( the pierre auger collaboration ) , 2009 , arxiv:0906.2347 , icrc 2009 bertone , g. , isola , c. , lemoine , m. , & sigl , g. 2002 , , 66 , 103003 casse , f. , lemoine , m. , & pelletier , g. 2002 , , 65 , 023002 cavallo , g. 1978 , , 65 , 415 celotti , a. , ghisellini , g. , 2008 , mnras , 385 , 283 chiaberge , m. , capetti , a. , & celotti , a. 2001 , , 324 , l33 cuoco , a. , & hannestad , s. 2008 , , 78 , 023007 dermer , c. d. , razzaque , s. , finke , j. d. , & atoyan , a. , 2008 , arxiv:0811.1160 evans , n. w. , ferrer , f. , sarkar , s. , 2003 , prd 69 , 3501 . evans , n. w. , ferrer , f. , sarkar , s. , 2004 , prd 69 , 8302 . farrar , g. r. , & piran , t. 2000 , arxiv : astro - ph/0010370 farrar , g. r. , berlind , a. a. , & hogg , d. w. 2006 , , 642 , l89 farrar , g. r. , & gruzinov , a. , 2009 , , 693 , 329 finley , c. b. , & westerhoff , s. 2004 , astroparticle physics , 21 , 359 ghisellini , g. , ghirlanda , g. , tavecchio , f. , fraternali , f. , & pareschi , g. 2008 , arxiv:0806.2393 gorbunov , d. s. , tinyakov , p. g. , tkachev , i. i. , troitsky , s. v. , 2002 , apj 577 , l93 . gorbunov , d. s. , tinyakov , p. g. , tkachev , i. i. , troitsky , s. v. , 2004 , astro - ph/0406654 . gorbunov , d. s. , tinyakov , p. g. , tkachev , i. i. , & troitsky , s. v. 2008 , arxiv:0804.1088 greisen , k. , 1966 , prl 16 , 748 . gupta , n. 2008 , journal of cosmology and astro - particle physics , 6 , 22 hague , j. d. et al . ( the pierre auger collaboration ) , arxiv:0906.2347 , icrc 2009 halzen , f. , & omurchadha , a. 2008 , arxiv:0802.0887 harari , d. , et al . ( the pierre auger collaboration ) , 2007 , arxiv:0706.1715 hardcastle , m. j. , cheung , c. c. , feain , i. j. , & stawarz , l. 2009 , mnras , 393 , 1041 henri , g. , pelletier , g. , petrucci , p. o. , & renaud , n. 1999 , astroparticle physics , 11 , 347 hillas , a. m. 1984 , , 22 , 425 holder , j. , & for the veritas collaboration 2008 , arxiv:0810.0471 israel , f. p. 1998 , , 8 , 237 kachelriess , m. , ostapchenko , s. , & tomas , r. 2008 , arxiv:0805.2608 kashti , t. , & waxman , e. 2008 , journal of cosmology and astro - particle physics , 5 , 6 kotera , k. , & lemoine , m. 2008 , , 77 , 123003 lyutikov , m. , & ouyed , r. 2005 , arxiv : astro - ph/0507620 massaglia , s. , 2007 , nuc . 168 , 302 milgrom , m. , & usov , v. 1995 , , 449 , l37 moskalenko , i. v. , stawarz , l. , porter , t. a. , & cheung , c. c. 2008 , arxiv:0805.1260 nagar , n. m. , & matulich , j. 2008 , arxiv:0806.3220 neronov , a. , semikoz , d. , & tkachev , i. 2007 , arxiv:0712.1737 norman , c. a. , melrose , d. b. , & achterberg , a. 1995 , , 454 , 60 osullivan , s. , reville , b. , & taylor , a. m. , 2009 , arxiv:0903.1259 pelletier , g. , lemoine , m. , & marcowith , a. , 2008 , arxiv:0707.3459 panov , a. d. , et al . ( atic-2 collaboration ) 2006 , arxiv : astro - ph/0612377 rachen , j. p. , & biermann , p. l. 1993 , , 272 , 161 rachen , j. , 2008 , in proceedings of `` rencontres de blois 2008 - challenges in particle astrophysics '' , blois ( france ) , may 18 - 23 2008 , ed . j. dumarchez rieger , f. m. , bosch - ramon , v. , & duffy , p. 2007 , , 309 , 119 romero , g. e. , combi , j. a. , perez bergliaffa , s. e. , & anchordoqui , l. a. 1996 , astroparticle physics , 5 , 279 saunders , w. , frenk , c. , rowan - robinson , m. , lawrence , a. , & efstathiou , g. 1991 , , 349 , 32 stanev , t. , biermann , p. l. , lloyd - evans , j. , rachen , j. p. , watson , a. a. , 1995 , prl 75 , 3056 . takeda , m. , et al . 1999 , , 522 , 225 tinyakov , p. g. , tkachev , i. i. , 2001 , pisma zh . 74 , 3 [ jetp lett . tinyakov , p. g. , tkachev , i. i. , 2002 , astropart . tinyakov , p. g. , tkachev , i. i. , 2004 , prd 69 , 8301 . uchihori , y. , nagano , m. , takeda , m. , teshima , m. , lloyd - evans , j. , watson , a. a. , 2000 , astropart . ulrich , r. , engel , r. , mller , s. , pierog , t. , schssler , f. , & unger , m. 2009a , arxiv:0906.0418 ulrich , r. , engel , r. , mller , s. , schssler , f. , & unger , m. 2009b , arxiv:0906.3075 unger , m. et al . ( the pierre auger collaboration ) , icrc-2007 , arxiv:0706.1495 vron - cetty , m .- , & vron , p. 2006 , , 455 , 773 vietri , m. 1995 , , 453 , 883 wahlberg , h. et al . ( the pierre auger collaboration ) , 2009 , arxiv:0906.2319 , icrc 2009 waxman , e. 1995 , phys . lett . , 75 , 386 . waxman , e. , & miralda - escud , j. 1996 , , 472 , l89 waxman , e. , 2005 , phys . scripta , t121 , 147 waxman , e. , & loeb , a. , 2008 , arxiv:0809.3788 wibig , t. 2008 , arxiv:0810.5281 wibig , t. 2009 , , 79 , 094008 yamamoto , t. et al . ( the pierre auger collaboration ) , icrc 2007 zatsepin , g. t. , kuzmin , v. a. , 1966 , pisma zh . fiz . 4 , 114 [ jetp . lett . 4 , 78 ] . zaw , i. , farrar , g. r. , & greene , j. e. 2009 , , 696 , 1218

we present new millimeter and radio observations of nine z quasars discovered in deep optical and near - infrared surveys . we observed the 250 ghz continuum in eight of the nine objects and detected three of them . new 1.4 ghz radio continuum data have been obtained for four sources , and one has been detected . we searched for molecular co ( 6 - 5 ) line emission in the three 250 ghz detections and detected two of them . combining with previous millimeter and radio observations , we study the fir and radio emission and quasar - host galaxy evolution with a sample of 18 z quasars that are faint at uv and optical wavelengths ( rest - frame 1450 magnitudes of ) . the average fir - to - agn uv luminosity ratio of this faint quasar sample is about two times higher than that of the bright quasars at z ( ) . a fit to the average fir and agn bolometric luminosities of both the uv / optically faint and bright z quasars , and the average luminosities of samples of submillimeter / millimeter - observed quasars at z to 5 , yields a relationship of . five of the 18 faint z quasars have been detected at 250 ghz . these 250 ghz detections , as well as most of the millimeter - detected optically bright z quasars , follow a shallower trend of defined by the starburst - agn systems in local and high - z universe . the millimeter continuum detections in the five objects and molecular co detections in three of them reveal a few of fir - emitting warm dust and of molecular gas in the quasar host galaxies . all these results argue for massive star formation in the quasar host galaxies , with estimated star formation rates of a few hundred . additionally , the higher fir - to - agn luminosity ratio found in these 250 ghz - detected faint quasars also suggests a higher ratio between star formation rate and supermassive black hole accretion rate than the uv / optically most luminous quasars at z .

1107.5065

understanding the formation and evolution of the first galaxies and the supermassive black holes ( smbhs ) they host is a key goal of both observational and theoretical astronomy . the first sample of more than twenty bright quasars at z@xmath06 were selected from the imaging data of the sloan digital sky survey ( hereafter sdss main survey , e.g. , fan et al . 2000 , 2006 ) . these objects are very bright in quasar uv and optical emission ( sdss magnitudes of @xmath9 ) and have smbhs with masses of @xmath10 a few @xmath11 ( jiang et al . 2007 ; kurk et al . 2007 ) . the dust and molecular gas properties of this bright quasar sample at z@xmath06 have been studied at millimeter wavelengths , using the max planck millimeter bolometer array ( mambo ) , the iram plateau de bure interferometer ( pdbi ) , and the very large array ( vla , bertoldi et al . 2003a ; 2003b ; petric et al . 2003 ; walter et al . 2003 , 2004 , 2009 ; carilli et al . 2004 , 2007 ; riechers et al . 2009 ; wang et al . 2007 , 2008b , 2010 ) . about 30% of them have strong continuum emission at 250 ghz from 40-to-60 k warm dust ( bertoldi et al . 2003a ; petric et al . 2003 ; wang et al . 2007 , 2008b ) . the fir luminosities of the mambo detections are a few @xmath12 to @xmath13 , and the estimated dust masses are a few @xmath6 ( beelen et al . 2006 ; bertoldi et al . 2003a ; wang et al . most of the 250 ghz - detected sources have also been detected in molecular co line emission with pdbi , yielding molecular gas masses of @xmath14 ( walter et al . 2003 ; carilli et al . 2007 ; riechers et al . 2009 ; wang et al . 2010 , 2011 ) . the fir - co and fir - radio luminosity ratios of these objects are consistent with the trend defined by dusty starburst systems at low and high redshifts ( riechers et al . 2006 ; beelen et al . 2006 ; wang et al . the results provide direct evidence for massive star formation at a rate of @xmath15 , indicating that bulges are being formed at the same time that the smbh is rapidly accreting in at least 1/3 of the optically bright quasar systems at z@xmath06 ( wang et al . 2008b ) . molecular co ( 3 - 2 ) and ( 7 - 6 ) line emission was resolved in one of the fir and co luminous quasars , sdss j114816.64 + 525150.3 at z=6.42 , revealing a source size of @xmath16 ( walter et al . 2004 , 2009 ) . the dynamical mass estimated with the size and line width of the co emission suggest a smbh - spheroidal bulge mass ratio an order of magnitude higher than the typical present - day value ( walter et al . 2004 ) . z@xmath06 quasars that are fainter at uv and optical wavelengths have been discovered from deep optical - near infrared surveys ( the sdss southern deep imaging survey , @xmath17 , jiang et al . 2008 , 2009 , the canada - france high redshift quasar survey [ cfhqs ] , @xmath18 , willott et al . 2007 , 2009a , 2009b ) . near - infrared spectroscopic observations of some of these faint quasars reveal that they are likely to be accreting at the eddington rate , and have less massive black holes ( i.e. , 10@xmath19 to @xmath20 ) than the sdss main survey z@xmath06 quasar sample ( kurk et al . 2009 ; willott et al . the steepness of the quasar luminosity function at high redshifts is such that lower - luminosity / mass objects are more common . thus , these faint z@xmath06 quasars provide an important sample to understand smbh - host galaxy formation at the earliest epoch . the millimeter dust continuum emission from the faint z@xmath06 quasars was first studied with a small sample of four objects discovered by cfhqs ( willott et al . . one of them was detected at 250 ghz with a flux density of @xmath21 , indicating a fir continuum luminosity from dust of @xmath22 . the mean 250 ghz flux density of the four quasars when stacked is @xmath23 , which is about a factor of two smaller than the average emission of the z@xmath06 luminous quasar sample ( willott et al . 2007 ; wang et al . we subsequently detected 250 ghz dust continuum and molecular co ( 6 - 5 ) line emission in another two z@xmath06 quasars with @xmath24 ( sdss j205406.42 - 000514.8 and ndwfs j142516.30 + 325409.0 , jiang et al . 2008 ; cool et al . 2006 ; wang et al . 2008b , 2010 ) and co ( 2 - 1 ) line emission from one cfhqs quasar at z=6.2 ( cfhqs j142952 + 544717 , willott et al . 2010a ; wang et al . these observations revealed a few @xmath25 of 40-to-60 k warm dust and @xmath7 of molecular gas in the quasar host galaxies , i.e. , masses comparable to those found with the bright z@xmath06 quasars . in this paper , we report new millimeter and radio observations of nine z@xmath06 quasars . we then study the fir and radio emission with a sample of 18 z@xmath06 quasars that are faint at uv and optical wavelengths , and investigate the dust and molecular gas emission properties and star forming activity in the host galaxies of the millimeter - detected objects . we summarize the current quasar sample at z@xmath06 and describe the new millimeter and radio observations in section 2 , present the results in section 3 , analyse the continuum emission properties of the faint z@xmath06 quasar sample in section 4 , discuss the star formation and quasar - host evolution in section 5 , and conclude in section 6 . a @xmath26-cdm cosmology with @xmath27 , @xmath28 and @xmath29 is adopted throughout this paper ( spergel et al . there are thirty - three z@xmath06 quasars that have published millimeter and radio observations from the literature ( bertoldi et al . 2003a ; petric et al . 2003 ; carilli et al . 2004 ; wang et al . 2007 ; 2008b ; willott et al . 2007 ) . in this paper , we present new mambo observations of the 250 ghz dust continuum emission from eight z@xmath06 quasars ( table 1 ) , and molecular co observations of three of them . seven of the eight sources were discovered from the sdss southern survey with optical magnitudes of @xmath30 ( jiang et al . 2008 , 2009 ) , and the other source , ulas j131911.29 + 095051.4 ( hereafter j1319 + 0950 ) , was selected from the ukirt infrared deep sky survey with @xmath31 ( ukidss , mortlock et al . seven of the eight objects ( all but sdss j020332.39 + 001229.3 , hereafter j0203 + 0012 ) have never been observed at millimeter wavelengths before . we also report new radio 1.4 ghz continuum observations for three of eight new mambo - observed objects and another z@xmath06 quasar from cfhqs ( willott et al . 2007 ) . the new observations , together with the previous data , now provide us a sample of forty 250 ghz - observed z@xmath06 quasars with optical magnitudes over a wide range , i.e. , @xmath32 , or quasar rest - frame 1450 @xmath1 ab magnitudes of @xmath33 . wang et al . ( 2008b ) studied the dust emission from quasars at z@xmath06 , mainly based on the most luminous sample from the sdss main survey . in this paper , we investigate the fir and radio emission properties of the fainter quasars at z@xmath06 , i.e. , a subsample of 18 objects with @xmath2 ( table 1 and 2 ) , and compare them to the bright z@xmath06 quasars ( @xmath3 ) discussed in previous papers ( bertoldi et al . 2003a ; petric et al . 2003 ; wang et al . we will discuss the optical measurements and quasar uv and bolometric luminosities based on @xmath34 , which is derived from the sdss z - band magnitudes and deep optical spectra ( e.g. , fan et al . 2000 , 2001 , 2003 , 2004 , 2006 ; jiang et al . 2008 , 2009 ; willott et al . 2007 ) , or from infrared photometry in the cases in which significant dust reddening is presented in the optical spectra ( e.g. , mcgreer et al . 2006 ) . observations of the 250 ghz dust continuum from the eight z@xmath06 quasars were carried out using the mambo - ii 117-element array on the iram 30-m telescope ( kreysa et al . 1998 ) in the winters of 2008 - 2009 and 2009 - 2010 . we adopted the standard on - off photometry mode with a chopping rate of 2 hz by @xmath35 in azimuth ( see also wang et al . 2007 , 2008 ) . the on - source observing time was about 1 - 3 hours for each object , to achieve 1@xmath36 rms sensitivities of @xmath37 to 0.7 mjy . we used the mopsic pipeline ( zylka 1998 ) to reduce the data , and three sources , j1319 + 0950 , sdss j012958.51@xmath38003539.7 ( hereafter j0129@xmath380035 ) , and j0203 + 0012 , were detected at @xmath39 . wang et al . ( 2008b ) observed j0203 + 0012 with mambo and did not detect the source with a 1@xmath36 rms of 0.7 mjy . our new observations , together with the old data , improve the rms to 0.46 mjy for this object , and we detect the 250 ghz continuum at 4@xmath36 . we have also obtained a 2.4@xmath40 signal from another source , sdss j023930.24@xmath38004505.4 ( hereafter j0239@xmath380045 ) . we also observed four z@xmath06 quasars at 1.4 ghz in winter 2008 - 2009 , using the vla in a configuration with a synthesized beam size of @xmath41 . flux scales were measured using the standard vla calibrators of 3c286 and 3c48 , and phase calibrations were performed every 30 min on nearby phase calibrators ( 1254 - 116 , 1507 - 168 , 2136 + 006 , and 2323 - 032 ) . the typical rms noise level was @xmath42 in three hours of observing time for each source . we reduced the data with the standard vla wide field data reduction software aips , and one source , j1319 + 0950 , was detected at @xmath43 . we searched for molecular co ( 6 - 5 ) line emission from the three mambo detections using the 3 mm receiver on the pdbi . the source j1319 + 0950 was observed in summer 2009 in c and d configurations ( i.e. a spatial resolution of @xmath44 ) with the narrow - band correlator , which provides a bandwidth of 1 ghz in dual polarization . the observations of j0129@xmath380035 and j0203 + 0012 were carried out using the new wide band correlator widex with a bandwidth of 3.6 ghz in dual polarization , which covers well the @xmath45-redshift uncertainty of @xmath46 ( jiang et al . the observations were performed in summer 2010 in d configuration with a synthesized beam size of @xmath47 . phase calibration was performed every 20 min with observations of phase calibrators , 1307 + 121 and 0106 + 013 , and we observed mwc 349 as the flux calibrator . we reduced the data with the iram gildas software package ( guilloteau & lucas 2000 ) . the co ( 6 - 5 ) line emission is detected from j1319 + 0950 at 4.8@xmath40 in 10.3 hours of on - source time with an rms sensitivity of @xmath48 per @xmath49 binned channel , and from j0129@xmath380035 at 5.3@xmath40 in 8.8 hours with an rms of @xmath50 per @xmath51 channel . we also detect the continuum at @xmath52 at the co ( 6 - 5 ) line frequency in both objects . the other source , j0203 + 0012 , was undetected in both co and 3 mm continuum in 13.9 hours of on - source time , with a sensitivity of @xmath53 per @xmath51 channel . we summarize the new millimeter and radio observations of the nine z@xmath06 quasars in table 1 . detailed measurements of the three new 250 ghz dust continuum / molecular co ( 6 - 5 ) detections are listed below : * j1319 + 0950 * the source was discovered in the ukirt infrared deep sky survey ( ukidss ) with @xmath54 , which lies in the typical magnitude range of the optically bright z@xmath06 quasars selected from the sdss main survey ( mortlock et al . we detect the 250 ghz dust continuum emission from this object with a flux density of @xmath55 , which is the third - strongest 250 ghz detection among the known z@xmath06 quasars . the 1.4 ghz radio continuum is also detected from our vla observation , with @xmath56 ( left panel of figure 1 ) . the radio emission appears unresolved with a peak at @xmath57 , @xmath58 , i.e. within @xmath59 of the optical quasar position . we also see a 4@xmath40 peak ( @xmath60 ) at @xmath61 , @xmath62 on the radio map , which is @xmath63 away from the quasar . for comparison , only 0.02 detections with @xmath64 are expected from a random circular sky area with a radius of @xmath65 , according to the 1.4 ghz radio source counts from previous deep vla observations ( i.e. , @xmath66 , fomalont et al . no optical counterpart has been found at the radio source position in the nasa / ipac extragalactic database or the sdss images . we observed the co ( 6 - 5 ) line emission from this object with the pdbi . the line is detected at 4.8@xmath40 on the velocity - integrated map with a line flux of @xmath67 , @xmath68 away from ( i.e. , consistent with ) the quasar position ( the upper panel of figure 2 ) . a gaussian fit to the line spectrum yields a peak line flux density of @xmath69 and a line width of @xmath70 , centered at the redshift of @xmath71 . this is slightly higher than , but consistent with the mg ii redshift of @xmath72 , considering the measurement errors ( @xmath73 , mortlock et al . 2008 ) and the typical velocity shift and scatters @xmath74 ( @xmath75 in redshift ) between the broad mg ii line emission and the quasar systemic redshift ( richards et al . 2002 ) . we also detect the continuum under the co ( 6 - 5 ) line emission with a peak value of @xmath76 on the continuum map averaged over the line - free channels ( right panel of figure 1 ) . we adopt this as the continuum flux density as the source is unresolved with the @xmath77 beam . we checked the position of the 1.4 ghz radio source @xmath63 away from the quasar in the pdbi co/97 ghz continuum images . there is no clear ( @xmath78 ) detection at this position and the measurement in the 97 ghz continuum image is @xmath79 ( right panel of figure 1 ) . * j0129@xmath380035 * this is the faintest of the sdss z@xmath06 quasars , with @xmath80 ( jiang et al . 2009 ) . we detect 250 ghz dust continuum from the quasar host galaxy , with a flux density of @xmath81 , comparable to the flux densities of the optically bright z@xmath06 quasars with 250 ghz detections . we detected the co ( 6 - 5 ) line in this source with a flux of @xmath82 , i.e. , at 5.3@xmath40 , integrated over a velocity range of 380 @xmath83 ( the middle panel of figure 2 ) . the line width fitted with a single gaussian is @xmath84 , with a line peak of 1.2@xmath850.3 mjy . the host galaxy redshift measured with the co line is @xmath86 , which is in good agreement with the optical redshift of @xmath87 measured with @xmath45 ( jiang et al . 2009 ) . the underlying dust continuum is also detected with a flux density of @xmath88 . * j0203 + 0012 * this object is a broad absorption line ( bal ) quasar discovered by both the sdss southern survey ( jiang et al . 2008 ) and the ukidss survey ( venemans et al . 2007 ; mortlock et al . 2008 ) with @xmath89 . the redshift measured with the quasar @xmath90 line emission is z=5.72 ( mortlock et al . wang et al . ( 2008b ) detected the 1.4 ghz radio continuum with @xmath91 ( wang et al . our mambo observation yields a 250 ghz dust continuum flux density of @xmath92 . we have searched for the co ( 6 - 5 ) line emission from this object over a wide redshift range , from z=5.60 to 5.84 , but did not detect it in 14 hours of on - source time . assuming a line width of 600 @xmath83 , the 3@xmath93 upper limit of the co flux is estimated to be @xmath94 ( the lower panel of figure 2 ) upper limit for the co line flux as @xmath95 , where @xmath96 is the assumed line width , and @xmath97 and @xmath98 are the rms per channel and channel width listed in section 2.2 ( seaquist et al . 1995 ) . ] . the dust continuum at the co ( 6 - 5 ) line frequency is also undetected with a continuum sensitivity of @xmath99 and a 3@xmath40 upper limit of 0.12 mjy . we calculate the fir luminosities and upper limits for the eight z@xmath06 quasars observed by mambo . we model the fir emission with an optically thin graybody and normalize the model to the mambo-250 ghz and available pdbi-3 mm dust continuum flux densities/3@xmath40 upper limits ( see also wang et al . 2007 , 2008 ) . a dust temperature of @xmath100 and emissivity index of @xmath101 are adopted in the model ; these are typical values for the fir and co luminous quasars at lower redshifts ( beelen et al . we then integrate the model fir sed from @xmath102 to @xmath103 to calculate the total fir luminosities / upper limits . the fir luminosity of the strongest mambo detection , j1319 + 0950 , is @xmath104 , making it one of the most fir - luminous objects among the z@xmath06 quasar sample . the other two mambo - detected objects are from the sdss southern survey sample and have fir luminosities of @xmath105 . we also estimate the dust masses for the three mambo detections , using @xmath106 , where @xmath107 is the planck function and @xmath108 is the dust absorption coefficient with @xmath109 at 125@xmath110 m ( hildebrand 1983 ) . the estimated dust masses of the three sources are @xmath111 . we finally calculate the 1.4 ghz radio continuum luminosity densities and upper limits for the four vla observed objects . we assume a power - law continuum of @xmath112 ( condon 1992 ) to correct the observed frequency to the quasar rest frame . the results are summarized in table 3 ; we also include the 1.4 ghz radio luminosity of j0203 + 0012 from wang et al . ( 2008b ) . we calculate the co ( 6 - 5 ) line luminosities and upper limits for the three co - observed z@xmath06 quasars , using @xmath113 ( @xmath114 ) , where @xmath115 is the line flux or 3@xmath40 upper limit in @xmath116 presented in table 1 , @xmath117 is the observed frequency of the co ( 6 - 5 ) line in ghz , @xmath118 is the luminosity distance in mpc , and z is the quasar redshift ( solomon & vanden bout 2005 ) . we then assume a co excitation ladder similar to that of j1148 + 5251 ( riechers et al . 2009 ) , for which @xmath119 , allowing us to calculate the co ( 1 - 0 ) luminosities . the fir - to - co(1 - 0 ) luminosity ratios of j1319 + 0950 and j0129@xmath380035 are 520 and 340 @xmath120 , respectively , which are similar to the values found for other z@xmath06 co - detected quasars ( wang et al . we then calculate the molecular gas masses ( @xmath121 $ ] ) , adopting a co luminosity - to - gas mass conversion factor of @xmath122 seen in low - redshift ultra - luminous infrared galaxies ( ulirgs , solomon et al . 1997 ; downes & solomon 1998 ) . the molecular gas masses for the two new co detections are all of order @xmath7 ( table 3 ) . the new observations increase the sample of mambo - observed z@xmath06 quasars to forty objects ( bertoldi et al . 2003a ; petric et al . 2003 ; willott et al . 2007 ; wang et al . 2007 , 2008b ) . in this paper , we study their fir and radio emission properties , focusing on the 18 sources that are fainter at uv and optical wavelengths ( @xmath123 , hereafter faint z@xmath06 quasars ) and compare them with the first sample of millimeter - studied z@xmath06 quasars from the sdss main survey ( hereafter bright z@xmath06 quasars , bertoldi et al . 2003a ; petric et al . 2003 ; wang et al . 2007 , 2008b ) . the mambo observations of these faint z@xmath06 quasars have typical sensitivities ( 0.4 to 0.6 mjy ) comparable to or slightly lower than that achieved in previous mambo observations of the sdss main survey z@xmath06 quasars ( @xmath124 , petric et al . 2003 ; wang et al . 2007 , 2008 ) and the optically bright quasar samples at redshifts 2 and 4 ( omont et al . 2001,2003 ; carilli et al . 2001 ) . five of the 18 sources are detected at 250 ghz at @xmath78 ( wang et al . 2007 , 2008b ; willott et al . 2007 ) , yielding a detection rate of 28@xmath8512% , or @xmath125% if we consider the subsample of twelve objects from the sdss southern survey selected with the same color selection criteria ( three detections , fan et al . 2004 ; jiang et al . 2008 ; 2009 ; wang et al . 2008b , this work ) . these are slightly lower than the mambo detection rate of @xmath12630% found with the optically bright quasar samples from z=2 to 6 ( omont et al . 2001 , 2003 ; carilli et al . 2001 ; wang et al . 2008b ) , but are consistent within the errors . we searched for molecular co ( 6 - 5 ) line emission in four of the five mambo - detected faint z@xmath06 quasars , and detected three ( wang et al . 2010 ) , indicating that highly excited molecular gas is ubiquitous in these millimeter ( fir ) luminous quasar - galaxy systems at the earliest epoch . fourteen of the 18 sources have radio 1.4 ghz continuum observations with typical 1@xmath40 rms of @xmath127 , and only two of them , j0203 + 0012 and j1427 + 3312 have radio detections ( mcgreer et al . 2006 ; wang et al . 2007 , 2008b , this work ) . this gives a much lower radio detection rate for the faint z@xmath06 quasar sample than that of the bright objects ; nine out of 22 z@xmath06 quasars with @xmath3 were detected at 1.4 ghz with similar observational sensitivities ( @xmath127 , petric et al . 2003 ; carilli et al . 2004 ; wang et al . 2007 , 2008b ) . the 1.4 ghz continuum flux density is 1.7 mjy for j1427 + 3312 ( mcgreer et al . 2006 ) and 195 @xmath128jy for j0203 + 0012 ( wang et al . 2008b ) , indicating radio - loud nuclei in the two objects ( kellermann et al . 1989 ) where fluxes are measured in the rest frame . we adopt this definition and calculate r assuming a power - law uv - to - optical continuum of @xmath129 ( fan et al . 2000 ) and radio continuum of @xmath112 ( condon 1992 ) . ] ( momjian et al . 2008 ) . deep observations are required to better constrain the radio emission from the radio - quiet uv / optical faint z@xmath06 quasars . we calculate the average 250 ghz and 1.4 ghz continuum for the 18 faint z@xmath06 quasars . the mean flux densities are derived for i ) all sources , ii ) mambo detections , and iii ) mambo non - detections , using the mambo and vla measurements weighted by inverse variance ( see also wang et al . we also present the average values for the 22 bright quasars ( @xmath3 ) for comparison . the average emission of this bright quasar sample is consistent with the results presented in wang et al . ( 2008b ) . we list all the results in table 4 . the mean 250 ghz flux density is @xmath130 for all 18 faint quasars , and @xmath131 for the five mambo - detected sources , which are about 50% and 70% of the values found for the bright quasars , respectively . the results are about the same when the two radio - loud quasars are excluded . the stacking measurement for the 13 mambo - undetected sources in the faint sample is @xmath132 , which places an upper limit of @xmath133 ( stacking average plus the 3@xmath93 rms ) for the average millimeter dust continuum emission . we convert the average 250 ghz flux densities to @xmath134 fir luminosities , adopting the assumptions described in section 3.2 , and calculate the average fir - to-1450@xmath135 luminosity ratios ( @xmath136 ) for each sample ( see table 4 ) . the average luminosity ratios for all the sources and the mambo detections in the faint - quasar sample are two and three times higher than the corresponding values found with bright quasars . this is consistent with the non - linear fir - to - agn luminosity relationships of @xmath137 , where @xmath138 found with the agn - starburst at low and high redshifts ( beelen et al . 2004 ; hao et al . 2005 , 2008 ; wang et al . 2008b ; serjeant & hatziminaoglou 2009 ; lutz et al . 2010 ) , as we will discuss further in section 4.2 . however , the different average luminosity ratios found with the mambo detections in the bright and faint quasar samples may be partly due to the selection effect that the detection limit of the mambo observations are about the same for the two samples , while the agn luminosities are different by design . the average 1.4 ghz continuum flux density , derived from 14 vla - observed sources in the faint quasar sample , is @xmath139 , and @xmath140 when the two radio - loud sources are excluded . this suggests an upper limit of @xmath141 for the radio - quiet and uv / optically faint z@xmath06 quasars . we estimate the average rest - frame 1.4 ghz luminosities / upper limits for the two samples and list the results in table 4 . we plot @xmath142 vs. @xmath143 of all the mambo - detected z@xmath06 quasars together with the average luminosities of the mambo - detections and non - detections in figure 3 , and compare them to sample of iras - selected local star forming galaxies ( yun et al . the local star forming galaxies with fir and radio emission powered by the active star formation have an average luminosity ratio of @xmath144 . the range @xmath145 contains 98% of the iras galaxies ( yun et al . the average luminosities / upper limits of the mambo - detected z@xmath06 quasars , though lower than the mean value of @xmath146 , are all within this range defined by local star forming galaxies , this argues that , in addition to the agn power , star formation also contributes significantly to the fir - to - radio emission of the mambo - detected z@xmath06 quasars . the average ratio of the mambo - undetected bright z@xmath06 quasars is offset from this range . the correlations between fir luminosities and agn uv , optical , and bolometric luminosities have been studied widely with samples of infrared and optically selected quasars to understand the origin of quasar fir emission . study of samples of the palomar - green ( pg ) quasars and seyfert galaxies shows a relationship of @xmath147 for typical optically - selected agn / quasars in the local universe ( hao et al . 2005 ; netzer et al . 2007 ; lutz et al . 2010 ) . a shallower relationship of @xmath148 was found for the local ir - luminous quasars hosted by starburst ulirgs ( hao et al . 2005 ) and high - redshift fir - and - co luminous quasars ( beelen et al . 2004 ; hao et al . 2008 ; wang et al . 2007 ; lutz et al . 2010 ) , which suggests excess fir emission powered by extreme starburst in these systems . wang et al . ( 2008b ) reported that the millimeter - detected bright z@xmath06 quasars follow the shallow luminosity trend defined by the local ir quasars . in this work , we investigate the fir - to - agn luminosity correlation of the full sample z@xmath06 quasars , and compare with the samples of mambo-250 ghz or scuba-350 ghz observed optically bright quasars at z@xmath02 to 5 ( omont et al . 2001 , 2003 ; carilli et al . 2001 ; priddey et al . 2003a ) , and local ir quasars ( zheng et al . 2002 ; hao et al . 2005 ) . we convert the uv or optical luminosities of all objects to agn bolometric luminosities assuming a uv - to - optical power - law continuum of @xmath129 , and an optical - to - agn bolometric luminosity conversion factor of @xmath149 ( richards et al . 2006 ) . in figure 4 , we plot all the z@xmath06 quasars , as well as the local ir quasar sample and the average luminosities of the submillimeter / millimeter [ ( sub)mm]-observed z@xmath02 to 5 quasar samples ( omont et al . 2003 ; hao et al . 2008 ) . the average fir luminosities of the z@xmath02 to 5 quasars are derived with the mean flux densities at 250 ghz or 350 ghz ( omont et al . 2001 , 2003 ; carilli et al . 2001 ; priddey et al . 2003a ) , adopting the same assumptions of dust temperature and emissivity index used for the z@xmath06 quasars ( see section 3.2 ) . we fit a line to the average luminosities of all these ( sub)mm - observed quasar samples at z@xmath02 to 6 using the ordinary least square method ( isobe et al . this yields an relationship of : + @xmath150 between the average fir and agn emission in these high - redshift optically - selected quasars , and the non - linear slope of 0.62 might suggest contributions from both host galaxy star formation and agn power to the fir emission ( beelen et al . 2004 ; netzer et al . 2007 ; hao et al . 2005 , 2008 ; wang et al . 2008b ; serjeant & hatziminaoglou 2009 ; lutz et al . 2010 ) the mambo-250 ghz observations at sub - mjy sensitivity have detected the most fir luminous objects among the z@xmath06 quasars ( bertoldi et al . 2003a , petric et al . 2003 ; wang et al . 2007 , 2008 , this work ) . most of these objects follow the shallower luminosity correlation trend of the local ir quasars and the ( sub)mm - detected quasars at z@xmath02 to 5 . in particular , the five mambo - detected z@xmath06 quasars with @xmath2 ( willott et al . 2007 ; wang et al . 2008b ; this work ) overlap with the high luminosity end of the ir quasars . this may suggest a starburst - dominant fir emission in these most fir luminous quasars at z@xmath06 , similar to that found in the local ir quasars . a fit to all the ( sub)mm detections in the high - z quasar sample and the local ir quasars gives a relationship of @xmath151 strong fir dust continuum emission has been detected in the host galaxies of @xmath1521/4 of the faint z@xmath06 quasars ( @xmath2 , willott et al . 2007 ; wang et al . 2008b , 2010 , this work ) . the dust continuum and molecular co detections reveal a few @xmath6 of fir - emitting warm dust and @xmath7 of molecular gas in the host galaxies of these faint quasars , which are comparable to the typical values found with the millimeter - detected optically bright sdss z@xmath06 quasars . the dust formation process associated with the evolution of massive stars on time scales @xmath153 has been discussed widely since the discovery of @xmath154 of dust in one of the most luminous / massive z@xmath06 quasars , j1148 + 5251 ( beelen et al . 2006 ; morgan & edmunds 2003 ; maiolino et al . 2004 ; dwek et al . the detections of comparble amounts of dust in these faint z@xmath06 quasars may suggest similar fast dust formation mechanism in their host galaxies . their fir luminosities estimated from the millimeter measurements are 3 to @xmath155 , falling on the fir - to - agn luminosity trend defined by the local ir quasars and the fir luminous quasars at z@xmath02 to 5 ( section 4.2 ) . the lower limit of their average fir - to - radio emission ratio of @xmath156 derived with the 250 ghz and 1.4 ghz measurements is within the typical range of @xmath145 in star forming galaxies ( section 4.1 ) . all these results suggest significant star formation in the millimeter - detected optically faint quasars at the highest redshift . conservatively assuming that 50% of the fir emission is powered by host galaxy star formation , the average star formation rate ( sfr ) of the five millimeter - detected @xmath2 quasars is about 560 @xmath8 ( kennicutt 1998 ) . three of them have been detected in molecular co ( 6 - 5 ) line emission ( wang et al . 2010 ; this work ) and the ratios between fir and co luminosities are 220 to 370 @xmath120 , which is within the typical range found with the submillimeter galaxies and co - detected quasars at lower redshifts ( solomon & vanden bout 2005 ; greve et al . 2005 ) . this result suggests a high star formation efficiency in these co - detected faint z@xmath06 quasars , similar to that of the extreme starburst systems ( solomon & vanden bout 2005 ) . additionally , if the mambo - detected faint z@xmath06 quasars are indeed less massive smbh systems accreting at their eddington limit , the higher @xmath136 found with these object ( see table 4 ) may suggest a higher ratio between sfr and smbh accretion rate in these systems compared to the mambo detections among the optically bright z@xmath06 quasars this may indicate that the stellar bulge grows faster than the central smbh in these faint / less massive quasar - host systems at z@xmath06 . we present observations of millimeter and radio continuum and co(6 - 5 ) line emission from the host galaxies of quasars at z@xmath06 . the new observations complete our mambo dust continuum survey of all the uv / optically faint z@xmath06 quasars ( @xmath2 ) discovered from sdss . combining with previous data , we calculate the average fir and radio emission for a sample of 18 z@xmath06 quasars with @xmath2 . the mean fir - to - agn uv luminosity ratio of this faint quasar sample is about two times higher than that of the bright z@xmath06 quasars ( @xmath3 ) . a fit to the average fir and agn bolometric luminosities of both the uv / optically faint and bright z@xmath06 quasars , and the average luminosities of samples of submillimeter / millimeter - observed quasars at z@xmath02 to 5 , yields a relationship of @xmath4 . the millimeter observations have detected the most fir luminous objects among these faint z@xmath06 quasars . the 250 ghz dust continuum detections in five of the 18 faint quasars and co ( 6 - 5 ) line detections in three of them ( see table 1 and 2 ) reveal dust masses of a few @xmath25 and molecular gas masses of @xmath7 in the quasar host galaxies , which are comparable to the dust and gas masses found in the bright z@xmath06 quasars ( beelen et al . 2006 ; bertoldi et al . 2003b ; carilli et al . 2007 ; wang et al . 2008a , 2010 ) . their fir luminosities are estimated to be @xmath157 , and the fir and agn bolometric luminosities follow a shallower relationship of @xmath5 , in agreement with the starburst - agn systems at lower redshifts . all these results suggest active star formation in the host galaxies of the millimeter - detected and uv / optically faint quasars at z@xmath06 with sfrs of a few hundred @xmath8 . moreover , the higher average fir - to agn uv luminosity ratios ( table 4 ) found with these objects and the shallow luminosity relationship suggest higher sfr - to - agn accretion rate ratios than that of the more luminous / massive z@xmath06 quasars . further high - resolution imaging of the dust and molecular gas components ( e.g. , with alma ) in the quasar host galaxies will be critical to constrain the host galaxy dynamical mass , black hole - spheroidal host mass ratio , gas and star formation rate surface density of these objects . this work is based on observations carried out with the max planck millimeter bolometer array ( mambo ) on the iram 30 m telescope , the plateau de bure interferometer , and the very large array ( nrao ) . iram is supported by insu / cnrs ( france ) , mpg ( germany ) and ign ( spain ) . the national radio astronomy observatory ( nrao ) is a facility of the national science foundation operated under cooperative agreement by associated universities , inc . we acknowledge support from the max - planck society and the alexander von humboldt foundation through the max - planck - forschungspreis 2005 . dominik a. riechers acknowledges support from from nasa through hubble fellowship grant hst - hf-51235.01 awarded by the space telescope science institute , which is operated by the association of universities for research in astronomy , inc . , for nasa , under contract nas 5 - 26555 . m. a. strauss thanks the support of nsf grant ast-0707266 . beelen , a. 2004 , phd thesis , observatoire de paris beelen , a. , cox , p. , benford , d. j. , dowell , c. d. , kov@xmath158cs , a. , bertoldi , f. , omont , a. , & carilli , c. l. , 2006 , apj , 642 , 694 bertoldi , f. , carilli , c. l. , cox , p. , fan , x. , strauss , m. a. , beelen , a. , omont , a. , & zylka , r. , 2003 , a&a , 406 , l55 bertoldi , f. et al . 2003b , a&a , 409 , l47 carilli , c. l. et al . 2001 , apj , 555 , 625 carilli , c. l. et al . 2004 , aj , 128 , 997 carilli , c. l. et al . 2007 , apj , 666 , l9 condon , j. j. 1992 , ara & a , 30 , 575 cool , r. j. et al . , 2006 , aj , 132 , 823 dwek , e. , galliano , f. , & jones , a. p. 2007 , apj , 662 , 927 downes , d. , & solomon , p. m. 1998 , apj , 507 , 615 fan , x. et al . 2000 , aj , 120 , 1167 fan , x. et al . 2001 , aj , 122 , 2833 fan , x. et al . 2003 , aj , 125 , 1649 fan , x. et al . 2004 , aj , 128 , 515 fan , x. et al . 2006 , aj , 131 , 1203 fomalont , e. b. , kellermann , k. i. , cowie , l. l. , capak , p. , barger , a. j. , partridge , r. b , windhorst , r. a. , & richards , e. a. 2006 , apjs , 167 , 103 guilloteau , s. , & lucas , r. 2000 , aspc , 217 , 299 greve , t. r. et al . 2005 , mnras , 359 , 1165 hao , c. n. , xia , x. y. , mao , s. et al . 2005 , apj , 625 , 78 hao , c. n. , xia , x. y. , mao , s. , deng , z. g. , & wu , h. 2008 , chjaa , 8 , 12 hildebrand , r. h. 1983 , qjras , 24 , 267 isobe , t , feigelson , e. d. , akritas , m. g. & babu , g. j. 1990 , apj , 364 , 104 jiang , l. et al . 2007 , aj , 134 , 1150 jiang , l. et al . 2008 , aj , 135 , 1057 jiang , l. et al . 2009 , aj , 138 , 305 kellermann k. i. , sramek , r. , schmidt , m. et al . 1989 , aj , 98,1195 kennicutt , r. c. 1998 , ara&a , 36 , 189 kreysa , e. et al . 1998 , spie , 3357 , 319 kurk , j. d. et al . 2007 , 669 , 32 kurk , j. d. , walter , f. , fan , x. , jiang , l. , jester , s. , rix , h .- w . , & riechers , d. a. , 2009 , apj , 702 , 833 lutz , d. et al . 2010 , apj , 712 , 1287 maiolino , r. , schneider , r. , oliva , e. , bianchi , s. , ferrara , a. , mannucci , f. , pedani , m. , roca sogorb , m. 2004 , nature , 431 , 533 mcgreer , i. d. , becker , r. h. , helfand , d. j. , & white , r.l . 2006 , apj , 652 , 157 momjian , e. , carilli , c. l. , & mcgreer , i. d. 2008 , aj , 136 , 344 morgan , h. l. , & edmunds , m. g. , 2003 , mnras , 343 , 427 mortlock , d. j. et al . 2009 , a&a , 509 , 97 netzer , h. et al . 2007 , apj , 666 , 806 omont , a. , cox , p. , bertoldi , f. , cox , p. , carilli , c. l. , priddey , r. s. , mcmahon , r. c. , & isaak , k. g. 2001 , a & a , 374 , 371 omont , a. , beelen , a. , bertoldi , f. , mcmahon , r. g. , carilli , c. l. , & isaak , k. g. 2003,a & a , 398 , 857 petric , a. o. , carilli , c. l. , bertoldi , f. , fan , x. , cox , p. , strauss , m. a. , omont , a. , & schneider , d. p. 2003 , aj , 126 , 15 priddey , r. s. , isaak , k. g. , mcmahon , r. g. , & omont , a. 2003a , mnras , 339 , 1183 priddey , r. s. , isaak , k. g. , mcmahon , r. g. , roboson , e. i. , & pearson , c. p. 2003b , mnras , 344 , l74 richards , g. t. , vanden berk , d. e. , reichard , t. a. , hall , p. b. , schneider , d. p. , subbarao , m. , thakar , a. r. , & york , d. g. 2002 , aj , 124 , 1 richards , g. t. et al . 2006 , apjs , 166 , 470 riechers , d. a. et al . 2006 , apj , 650 , 604 riechers , d. a. , et al . 2009 , apj , 703 , 1338 seaquist , e. r. , ivison , r. j. , & hall , p. j. 1995 , mnras , 276 , 867 serjeant , s. & hatziminaoglou , e. 2009 , mnras , 397 , 265 solomon , p. m. , downes , d. , radford , s. j. e. , & barrett , j. w. 1997 , apj , 478 , 144 solomon , p. m. , & vanden bout , p. a. 2005 , ara&a , 43 , 677 , * sv05 * spergel , d. n. et al . 2007 , apjs , 170 , 377 venemans , b. p. , mcmahon , r. g. , warren , s. j. , gonzalez - solares , e. a. , hewett , p. c. , mortlock , d. j. , dye , s. , & sharp , r. g. 2007 , mnras , 376 , l76 walter , f. et al . 2003 , nature , 424 , 406 walter , f. , carilli , c. l. , bertoldi , f. , menten , k. , cox , p. , lo , k. y. , fan , x. , & strauss , m. a. 2004 , apj , 615 , l17 walter , f. , riechers , d. , cox , p. , neri , r. , carilli , c. , bertoldi , f. , wei@xmath159 , a. , & maiolino , r. 2009 , nature , 457 , 699 wang , r. et al . 2007 , aj , 134 , 617 wang , r. et al . 2008a , aj , 135 , 1201 wang , r. et al . 2008b , apj , 687 , 848 wang , r. et al . 2010 , apj , 714 , 699 wang , r. et al . 2011 , apjl , in press ( astro - ph/1105.4199 ) willott , c. j. et al . 2007 , aj , 134 , 2435 willott , c. j. et al . 2009 , aj , 137 , 3541 willott , c. j. et al . 2010a , aj , 139 , 906 willott , c. j. et al . 2010b , aj , 140 , 546 yun , m. s. , reddy , n. a. , & condon , j. j. 2001 , apj , 554 , 803 zheng , x. z. xia , x. y. mao , s. et al . 2002 , aj , 124 , 18 zylka , r. 1998 , mopsi users manual , ( iram : grenoble ) lccccccc name & redshift & @xmath34 & @xmath160 & @xmath161 & @xmath115 & fwhm & @xmath162 + & & & mjy & @xmath163 & @xmath116 & @xmath83 & mjy + ( 1 ) & ( 2 ) & ( 3 ) & ( 4 ) & ( 5 ) & ( 6 ) & ( 7 ) & ( 8) + sdss j012958.51@xmath38003539.7 & 5.7794@xmath850.0008 & 22.28 & * 2.37@xmath850.49 * & & 0.37@xmath850.07 & 283@xmath8587 & 0.14@xmath850.04 + sdss j020332.39 + 001229.3 & 5.72 & 20.97 & * 1.85@xmath850.46 * & * 195@xmath8522@xmath164 * & @xmath1650.32 & & @xmath1650.12 + sdss j023930.24@xmath38004505.4 & 5.82 & 22.28 & 1.29@xmath850.54 & & & & + ulas j131911.29 + 095051.4 & 6.1321@xmath850.0012 & 19.65 & * 4.20@xmath850.65 * & * 64@xmath8517 * & 0.43@xmath850.09 & 537@xmath85123 & 0.31@xmath850.08 + cfhqs j150941.78@xmath38174926.8 & 6.12 & 19.82 & 1.00@xmath850.46@xmath166 & 23@xmath8518 & & & + sdss j205321.77 + 004706.8 & 5.92 & 21.20 & 0.09@xmath850.63 & & & & + sdss j214755.40 + 010755.0 & 5.81 & 21.65 & @xmath380.36@xmath850.61 & @xmath3828@xmath8518 & & & + sdss j230735.40 + 003149.0 & 5.87 & 21.73 & @xmath380.56@xmath850.53 & @xmath3821@xmath8517 & & & + sdss j235651.58 + 002333.3 & 6.00 & 21.77 & 0.21@xmath850.49 & & & & + note column ( 1 ) , name ; column ( 2 ) , redshift . co redshifts are presented for the two sources detected by the pdbi , and we list the redshifts measured with the uv quasar emission lines from the discovery papers for the other six sources ( jiang et al . 2008 , 2009 ) ; column ( 3 ) , rest - frame 1450 @xmath1 magnitudes from the discovery papers ( mortlock et al . 2008 , jiang et al . 2008 , 2009 ) . column ( 4 ) , mambo measurements of the 250 ghz dust continuum ; column ( 5 ) , vla measurements of the 1.4 ghz radio continuum ; column ( 6 ) and ( 7 ) , line flux and fwhm of the co ( 6 - 5 ) line emission ; column ( 8) , measurement of the dust continuum at the co ( 6 - 5 ) line frequency . the 250 ghz and 1.4 ghz detections are marked as boldface . + @xmath164wang et al . ( 2008b ) ; @xmath166willott et al . 2007 . lcccccccc name&redshift&@xmath34&@xmath160&@xmath161&@xmath167&@xmath168&@xmath169&reference + & & & mjy & @xmath170jy & @xmath171 & @xmath25 & @xmath7 & + ( 1 ) & ( 2 ) & ( 3 ) & ( 4 ) & ( 5 ) & ( 6 ) & ( 7 ) & ( 8) & ( 9 ) + sdss j000552.34@xmath38000655.8&5.85 & 20.83&0.36@xmath850.48&40@xmath85130&@xmath1653.4&&&1,2 + cfhqs j003311.40@xmath38012524.9&6.13&21.78@xmath164&*1.13@xmath850.36*&@xmath3827@xmath8519&2.6@xmath850.8&1.5&&2,3 + sdss j030331.40@xmath38001912.9&6.08&21.28&0.23@xmath850.51&@xmath3885@xmath8562&@xmath1653.5&&&2 + sdss j035349.76 + 010405.4&6.05&20.22&1.20@xmath850.46&17@xmath8519&@xmath1653.2&&&2 + ndwfs j142516.30 + 325409.0&5.89&20.62@xmath164&*2.27@xmath850.51*&20@xmath8520&5.4@xmath851.2&3.0&2.0&2,4 + first j142738.59 + 331242.0&6.12&20.33@xmath164&0.39@xmath850.66&*1730@xmath85131*&@xmath1654.6&&&2,5 + sdss j163033.90 + 401209.6&6.05&20.64&0.80@xmath850.60&14@xmath8515&@xmath1654.2&&&2,6 + cfhqs j164121.64 + 375520.5&6.05&21.30@xmath164&0.08@xmath850.46&@xmath3830@xmath8532&@xmath1653.2&&&2,3 + sdss j205406.42@xmath38000514.8&6.04&20.60&*2.38@xmath850.53*&17@xmath8523&5.5@xmath851.2&3.1&1.2&2,4 + sdss j231546.36@xmath38002357.5&6.12&21.34&0.28@xmath850.60&31@xmath8516&@xmath1654.1&&&2 + cfhqs j232908.28@xmath38030158.8&6.42&21.65@xmath164&0.01@xmath850.50&14@xmath8522&@xmath1653.4&&&2 + + note reference:(1 ) wang et al . 2007 ; ( 2 ) wang et al . 2008b ; ( 3 ) willott et al . 2007 ; ( 4 ) wang et al . 2010 ; ( 5 ) mcgreer et al . 2006 ; ( 6 ) bertoldi et al . + @xmath164the values are calculated from the absolute 1450 @xmath135 magnitudes in the discovery papers ( cool et al . 2006 ; mcgreer et al . 2006 ; willott et al . 2007 ) . the 250 ghz and 1.4 ghz detections are marked as boldface . @xmath166the upper limits of @xmath142 are derived with the 3@xmath40 upper limits of the 250 ghz flux densities . lcccccc name & @xmath142 & @xmath168 & @xmath172 & @xmath173 & @xmath174 & @xmath169 + & @xmath171 & @xmath25 & @xmath175 & @xmath176 & @xmath176 & @xmath7 + ( 1 ) & ( 2 ) & ( 3 ) & ( 4 ) & ( 5 ) & ( 6 ) & ( 7 ) + j0129 - 0035 & 5.2@xmath850.9 & 2.9 & & 1.2@xmath850.2 & 1.5@xmath850.3 & 1.2 + j0203 + 0012 & 4.4@xmath851.1 & 2.5 & 0.118@xmath850.013 & @xmath1651.0 & @xmath1651.3 & @xmath1651.0 + j0239 - 0045 & @xmath1653.9 & & & & & + j1319 + 0950 & 10.0@xmath851.3 & 5.7 & 0.045@xmath850.012 & 1.5@xmath850.3 & 1.9@xmath850.4 & 1.5 + j1509@xmath381749 & @xmath1653.2 & & @xmath1650.038 & & & + j2053 + 0047 & @xmath1654.4 & & & & & + j2147 + 0107 & @xmath1654.4 & & @xmath1650.034 & & & + j2307 + 0031 & @xmath1653.8 & & @xmath1650.032 & & & + j2356 + 0023 & @xmath1652.8 & & & & & + lccccccccc group & number@xmath164 & @xmath177 & @xmath178 & @xmath179 & @xmath180 & number@xmath181 & @xmath182 & @xmath183 & q + & & @xmath171 & mjy & @xmath171 & & & @xmath163 & @xmath175 & + ( 1 ) & ( 2 ) & ( 3 ) & ( 4 ) & ( 5 ) & ( 6 ) & ( 7 ) & ( 8) & ( 9 ) & ( 10 ) + + all objects & 18 & 4.3 & 0.85@xmath850.12 & 2.0@xmath850.3 & 0.46 & 14 & 19@xmath856 & 0.012@xmath850.003 & 1.63@xmath850.15 + ( radio quiet ) & 16 & 4.0 & 0.79@xmath850.13 & 1.9@xmath850.3 & 0.46 & 12 & 2@xmath856 & @xmath1650.013 & @xmath1841.57 + 250 ghz detections & 5 & 4.3 & 1.85@xmath850.20 & 4.3@xmath850.5 & 1.01 & 4 & 45@xmath8510 & 0.030@xmath850.07 & 1.59@xmath850.11 + ( radio quiet ) & 4 & 4.2 & 1.86@xmath850.23 & 4.3@xmath850.5 & 1.03 & 3 & 1@xmath8512 & @xmath1650.024 & @xmath1841.68 + 250 ghz non - detections & 13 & 4.3 & 0.33@xmath850.15 & @xmath1651.8 & @xmath1650.42 & 10 & 7@xmath857 & @xmath1650.019 & + ( radio quiet ) & 12 & 3.9 & 0.32@xmath850.15 & @xmath1651.8 & @xmath1650.46 & 9 & 2@xmath857 & @xmath1650.015 & + + all objects & 22 & 17.6 & 1.64@xmath850.13 & 3.8@xmath850.3 & 0.22 & 22 & 50@xmath853 & 0.033@xmath850.002 & 1.49@xmath850.04 + ( radio quiet ) & 21 & 16.9 & 1.68@xmath850.13 & 3.9@xmath850.3 & 0.23 & 21 & 38@xmath853 & 0.026@xmath850.002 & 1.61@xmath850.05 + 250 ghz detections & 8 & 19.1 & 2.71@xmath850.18 & 6.3@xmath850.4 & 0.33 & 8 & 43@xmath854 & 0.029@xmath850.003 & 1.77@xmath850.05 + 250 ghz non - detections & 14 & 16.8 & 0.55@xmath850.18 & 1.3@xmath850.4 & 0.08 & 14 & 59@xmath855 & 0.040@xmath850.003 & 0.94@xmath850.15 + ( radio quiet ) & 13&15.5 & 0.58@xmath850.18 & 1.4@xmath850.4 & 0.09 & 13 & 32@xmath855 & 0.021@xmath850.003 & 1.23@xmath850.15 + note we calculate the average flux densities using @xmath185 , @xmath186 , and @xmath187 ( omont et al . 2003 ) , where @xmath188 and @xmath189 are the measurement and 1@xmath40 rms for each object at 250 ghz or 1.4 ghz . for results with @xmath190 less than 3@xmath191 , we take the stacking average value plus 3@xmath192 as the upper limits of the average flux densities and calculate the corresponding upper limits for the luminosities . + @xmath164 number of sources observed at 250 ghz . @xmath166 average luminosity at rest - frame 1450 @xmath1 , derived from @xmath34 . @xmath181 number of sources observed at 1.4 ghz . @xmath193 a radio spectral index of -0.75 ( condon 1992 ) is adopted here to calculate the rest frame 1.4 ghz lunimosity . , which is plotted on the bottom left , and the contour levels are ( -3 , -2 , 2 , 3 , 4)@xmath19415 @xmath195 pdbi 97 ghz continuum image of the quasar averaged over the line - free channels . the synthesized beam size ( fwhm ) is 3.5@xmath196 , and the contour levels are ( -2 , -1 , 1 , 2 , 3)@xmath1940.08 @xmath197 . the cross in both plots denotes the optical position of the quasar ( mortlock et al . 2008).,title="fig:",height=249 ] -2.8 in 3.5 in 6 quasars observed with the pdbi . the continuum emission underlying the co spectra is removed for the two detections , j1319 + 0950 and j0129@xmath380035 . the synthesized beam size ( fwhm ) of the co velocity - integrated image of j1319 + 0950 is @xmath03.5@xmath196 , and @xmath198 for j0129@xmath380035 and j0203 + 0012 . the top abscissae of the co spectra give the redshift range of the spectral windows , and zero velocity in each panel corresponds to the optical / near - infrared redshifts of z=6.127 , 5.780 , 5.720 from the discovery papers ( mortlock et al . 2008 ; jiang et al . 2009 ) . the solid lines in the spectra show gaussian fits to the line emission . the co peak positions are consistent with the optical quasars for the two detections . , title="fig:",width=576,height=192 ] 6 quasars observed with the pdbi . the continuum emission underlying the co spectra is removed for the two detections , j1319 + 0950 and j0129@xmath380035 . the synthesized beam size ( fwhm ) of the co velocity - integrated image of j1319 + 0950 is @xmath03.5@xmath196 , and @xmath198 for j0129@xmath380035 and j0203 + 0012 . the top abscissae of the co spectra give the redshift range of the spectral windows , and zero velocity in each panel corresponds to the optical / near - infrared redshifts of z=6.127 , 5.780 , 5.720 from the discovery papers ( mortlock et al . 2008 ; jiang et al . the solid lines in the spectra show gaussian fits to the line emission . the co peak positions are consistent with the optical quasars for the two detections . , title="fig:",width=576,height=192 ]

we study the stellar mass tully - fisher relation ( tfr ; stellar mass versus rotation velocity ) for a morphologically blind selection of emission line galaxies in the field at redshifts 0.1 z 0.375 . kinematics ( , v ) are measured from emission lines in keck / deimos spectra and quantitative morphology is measured from v- and i - band hubble images . we find a transition stellar mass in the tfr , m = 9.5 m . above this mass , nearly all galaxies are rotation - dominated , on average more morphologically disk - like according to quantitative morphology , and lie on a relatively tight tfr . below this mass , the tfr has significant scatter to low rotation velocity and galaxies can either be rotation - dominated disks on the tfr or asymmetric or compact galaxies which scatter off . we refer to this transition mass as the mass of disk formation " , m because above it all star - forming galaxies form disks ( except for a small number of major mergers and highly star - forming systems ) , whereas below it a galaxy may or may not form a disk . [ firstpage ] galaxies : evolution - galaxies : formation -galaxies : fundamental parameters - galaxies : kinematics and dynamics

1506.04144

the tully - fisher relation ( tfr ; ) is an empirical scaling between the luminosity of a disk galaxy and its rotation velocity . later work replaced luminosity with the more physical quantity stellar mass , m@xmath4 . in the local universe , the tfr is remarkably tight @xcite and its parameterization serves as an important constraint for models of disk galaxy formation ( e.g. @xcite ) . although the tfr is well behaved for ordered disk galaxies , morphologically disturbed or compact galaxies tend to fall below the local relation , exhibiting low rotation velocity for a given stellar mass @xcite . ordered velocity fields may be disturbed in major merging events or tidal disruptions , through the accretion of non - ordered external angular momentum from cold flows @xcite and/or disruptive feedback @xcite . @xcite and @xcite demonstrated that accounting for both the disordered motions ( @xmath1 ) and ordered rotation velocity ( v@xmath2 ) in a new kinematic quantity , @xmath7 , re - establishes a tighter @xmath8-m@xmath4 scaling relation . this relation is independent of galaxy morphology and coincident with the faber - jackson relation for early type galaxies . numerical simulations have shown that s@xmath9 traces the overall potential well of galaxy - dark halo systems , even for galaxies undergoing drastic kinematic events , such as a major merger @xcite . the measurement of a disordered velocity component has been incorporated more in recent years ( e.g. ) . with further investigation , @xcite demonstrated that since z @xmath10 1 star - forming galaxies have been kinematically settling from systems with a dominant dispersion component to rotationally supported disks . such galaxies follow kinematic downsizing : more massive galaxies exhibit the most ordered kinematics ( high v@xmath2/@xmath1 ) at all epochs . while high and intermediate mass galaxies are likely settling to disks , it is unclear whether low mass galaxies follow the same evolutionary path . it is well known locally that morphological type is a strong function of stellar mass , with dwarf systems exhibiting more irregular morphology , distinct from their large disk counterparts ( e.g. ) . moreover , there is mounting evidence that disturbed galaxies are increasingly more common at low masses in the early universe @xcite . however , kinematic surveys often select against galaxies with disturbed morphologies with regards to other interests , e.g. performing distance measurements or studying dark matter ( e.g. @xcite ) . in the local universe , dwarf galaxies can show rotational signatures in both their hi and stellar components ( e.g. ) . when the morphological selection is opened to include irregular galaxies , compact galaxies , and close pairs , the presence of features such as peculiar velocity fields and thick disks are found . indeed , complex kinematics and low rotational support is frequently found for both low luminosity ( e.g. @xcite ) and highly star - forming dwarfs ( e.g. ) . however , locally , limited studies have been performed placing large samples of these disordered systems on the tfr , or relating the relative contributions of v and @xmath11 to the morphology of the galaxy . in this paper we study the resolved kinematics for a morphologically unbiased sample of star - forming galaxies at z @xmath10 0.2 . compared to nearby samples , a survey at z @xmath10 0.2 benefits from the smaller angular sizes of galaxies and the higher target density . large homogenous samples can be efficiently obtained with a multi - object spectrograph . this sample is unique from other local samples in that we include galaxies with both disturbed and disk - like morphologies . furthermore , we incorporate measurements of the non - negligible contributions from random motions as well as the rotation velocity . in section 2 , we discuss the sample selection , the method used for measuring kinematics from emission lines and the measurements of quantitative morphologies . in section 3 , we present the main result of this paper : a transition stellar mass in the tfr . in section 4 , we illustrate the correlation between gas phase kinematics and galaxy morphology . in section 5 , we compare our results with measurements of kinematics for low mass galaxies in the literature both locally and at intermediate redshift . our conclusions are presented in section 6 . in the appendix , we test our ability to recover kinematics for the galaxies with the smallest angular extent in our sample . we adopt a @xmath12cdm cosmology defined with ( h , @xmath13 , @xmath14 ) = ( 0.7 , 0.3 , 0.7 ) . in this section we detail the sample selection and the measurements of the kinematics and morphological indices . we reference the appendix for further analysis and discussion of the effects of seeing on our kinematic measurements . we focus on the lowest redshift bin in the galaxy sample used by @xcite and @xcite ( hereafter referred to as k07 and k12 , respectively ) , namely 0.1 @xmath0 z @xmath0 0.375 . at these redshifts , the k12 sample is sensitive to low mass dwarf galaxies . we briefly review how this sample was selected and defer to k12 for further details . the k12 sample is drawn from field 1 of the deep2 redshift survey @xcite . deep2 employed the deimos multi - object spectrograph @xcite on the keck - ii telescope . the 1200 line / mm grating was used and the slits were fixed at 1@xmath15 , leading to a spectral resolution of r @xmath10 5000 . the k12 sample was cut on nebular line strength ( @xmath16 erg s@xmath17 cm @xmath18 ) , available hubble space telescope ( hst ) imaging and slit alignment from the photometric major axis ( @xmath19 40@xmath20 ) . we use hst imaging in two passbands with the advanced camera for surveys ( acs ) , v ( f606w ) and i ( f814w ) , from the aegis survey @xcite . each of the images has a pixel scale of 0.03@xmath15 and a typical fwhm psf of 0.1@xmath15 ( 0.33 kpc at z = 0.2 ) . inclinations are measured with the sextractor software using the v - band hst image . the v - band traces the young stars and thus should not be very different from the nebular emission line morphology . a further cut is applied to the sample , limiting inclinations to 30@xmath21 i @xmath22 , to avoid uncertain inclination corrections for face - on galaxies and dust effects in edge on systems . however , a handful ( 4 ) of severely disturbed galaxies for which inclinations and pas are uncertain are included . the exclusion of these galaxies does not alter our conclusions . stellar masses were derived using the rest frame b - v color and absolute b - band magnitude @xcite with refined empirical corrections from sed fitting @xcite , as described in @xcite . the adopted imf is from @xcite . errors on stellar masses are approximately 0.2 dex . spectral slits must be aligned to within 40@xmath23 of a galaxy s major axis to reliably recover rotation ( @xcite ) . k07 removed galaxies with slits misaligned by more than 40@xmath23 from the hst derived photometric major axis , except for the 4 galaxies with severely disturbed morphologies . beyond this cut there are no residual correlations between measured kinematics ( rotation or dispersion ) and slit alignment in our sample . the scale of the atmospheric turbulence sets a limiting angular size for measuring rotation velocity in a galaxy . the seeing tends to smooth and eliminate small scale rotation gradients . we test the limiting resolving power imposed by beam smearing in the appendix . for the observational conditions of deep2 and for the lowest s / n in our sample , kinematics can be measured for galaxies to diameters encompassing 95@xmath24 of the light ( d@xmath25 ) of ( 0.87@xmath260.06 ) @xmath27 seeing . we make a further cut on the k07 sample and remove 12 galaxies which are not extended enough to confidently measure kinematics . our final sample contains 119 galaxies and uniformly covers the blue cloud " ( figure 1 in k12 ) . the qualifying feature of our selection is that , aside from the inclination cut , there was no explicit selection on morphology . contrary to previous tfr studies , this sample selection includes disturbed and compact galaxies . this allows us to more fully sample the population of emission line galaxies . we adopt kinematics measured from bright nebular emission lines ( h@xmath28 @xmath296563 or [ oiii ] @xmath295007 ) in k07 . the rotation velocity , uncorrected for inclination , and spatially resolved gas dispersion , @xmath1 , were measured with the program rotcurve @xcite , taking into account the effects of seeing . the seeing was measured through alignment stars on each slit mask ( see @xcite ) . for our sample , the seeing varied between 0.55@xmath15 and 1.2@xmath15 fwhm , with a median value of 0.75@xmath15 ( 1.36 kpc at z = 0.1 , 3.84 kpc at z = 0.375 ) . the kinematic measurements have been described in previous papers , so we will briefly review them here and refer to @xcite for the details . in short , rotcurve builds model arctan rotation curves where v is the velocity on the flat part of the rotation curve ( v@xmath2 @xmath27 @xmath30 ) , with an additional dispersion term ( @xmath1 ) , which is constant with radius . due to the seeing , the rise of the rotation curve is not well resolved and the turnover radius for the rotation curve ( r@xmath31 ) is not well constrained . the turnover radius is kept fixed to a value of 0.2@xmath15 . the recovered rotation velocity and turnover radius are only slightly covariant , with a marginal @xmath26 0.1 dex change in v for a @xmath260.1@xmath15 change in r@xmath31 . models with varying v and @xmath1 are blurred with the seeing and fit to the data . v and @xmath1 are explored with a grid spacing of 5 km s@xmath17 and the best fit value is determined through a @xmath32 minimization on this grid . the spectral resolution of the 1200 line / mm grating used by the deep2 survey allows measurements for v@xmath2 @xmath27 @xmath30 down to @xmath10 5 kms@xmath17 and @xmath1 to @xmath10 15 kms@xmath17 . measurements with best fits below these limitations are set to these values as upper limits . this is the case for 7 galaxies for v@xmath2 and 22 galaxies for @xmath1 . example kinematic fits and the corresponding @xmath32 spaces are shown later in this paper in figure [ fig : indmorph ] . hst imaging is used to measure inclinations in order to correct the measured rotation velocity . as mentioned previously , corrections were not applied to four galaxies with severely disturbed morphologies where the inclination was highly uncertain . they are marked with carets in figure [ fig : kinematics ] . we note that the quantity @xmath1 is _ not _ like the typical pressure supported velocity dispersion measured from stellar absorption lines in early - type galaxies . since @xmath1 is tracing hot gas which can radiate , a high dispersion system can not remain in equilibrium after a crossing time . therefore @xmath1 is effectively measuring velocity gradients below the seeing limit , as illustrated by the simulations of @xcite . the typical thermal broadening for t @xmath10 10@xmath33 k hydrogen gas is 10 kms@xmath17 , so a measure of @xmath1 @xmath34 25 kms@xmath17 is tracing the relative motions of hii regions and/or disordered motions associated with unresolved velocity gradients ( @xcite , k12 ) . high resolution hst - acs images allow us to quantify morphologies for each galaxy . in particular , we examine three indices : the gini coefficient , asymmetry , and concentration . we find little difference between the indices for the v- and i - band and so we simply adopt the i - band measurements ( approximately rest v - band at z=0.2 ) for our analysis . we describe these parameters now and examine how they correlate with kinematics in section 4 . the gini coefficient was adopted from economic theory . as applied to the study of galaxy morphology , the gini coefficient provides a non - parametric measurement of the distribution of light among the pixels associated with a galaxy @xcite . the gini coefficient approaches unity for systems with a large fraction of intensity confined to only a few pixels and approaches zero for a uniform intensity distribution . it compliments the more standard concentration index @xcite , as both trace intensity clustering , but the gini coefficient is insensitive to the location of the clustering . the typical uncertainty on the gini coefficient is 0.02 , as determined from empirical comparisons of deep ( udf ) and shallow fields ( goods ) @xcite . the concentration index ( defined in e.g. @xcite and @xcite ) measures the ratio of flux contained in two isophotes . the outer isophote is defined by the 2@xmath11 background contour with a normalized radius of 1 and the inner isophote is defined by a normalized radii of @xmath28 . concentration , asymmetry , and size were measured with the gim2d software @xcite , which simultaneously fit to both the v- and i - band images . the concentration index is measured for @xmath35 . we find no significant differences in trends between alpha levels and adopt @xmath36 for our analysis . the asymmetry index in gim2d is as defined in @xcite . asymmetry is measured by first assigning contiguous pixels to the galaxy , rotating the image by 180@xmath20 and then self - subtracting the mirrored image from the original image . the error bars we use in the figures below for asymmetry and concentration represent the typical 1@xmath11 scatter between the four gim2d measurements for each individual v - band image . these do not represent the additional uncertainty from the image s / n ( limiting magnitude @xmath10 28.7 ab ) , for which comparisons with deeper fields are needed . shallower studies than ours ( limiting magnitude @xmath10 26.5 ab ; e.g. @xcite ) find relative uncertainties in concentration and asymmetry of @xmath10 0.05 with respect to deeper data ( limiting magnitude @xmath10 27.4 ab ) . z @xmath0 0.375 is shown . below a stellar mass of @xmath3 m@xmath4/m@xmath37 9.5 , there is significant scatter to low v@xmath2 ( red points ) . the local tfr @xcite ( black line ) , adjusted for a @xcite imf , and its extrapolation to lower stellar masses ( dashed line ) is shown . b ) for the same galaxies the integrated gas velocity dispersion , a measure of disordered motions , is shown versus stellar mass . c ) accounting for both ordered and disordered motions in the s@xmath9-m@xmath4 relation results in a relatively tight scaling , coincident with the local faber - jackson relation from @xcite . galaxies are color / shape coded according to whether they are on the tf ridgeline and whether they have high or low stellar mass . this coding is repeated in other plots in the paper . the rotation velocities for the four galaxies with carats are not inclination corrected due to their highly uncertain inclination . ] in the top panel of figure [ fig : kinematics ] we show the stellar mass tfr for our sample . there is a noticeable transition at a stellar mass of @xmath3 m@xmath4/m@xmath37 9.5 . above this mass , all galaxies in our sample are on the local tfr . below this mass , galaxies may or may not lie on the local tfr , with the galaxies falling off tending to display systematically high measurements of @xmath1 . we posit that @xmath3 m@xmath4/m@xmath37 9.5 marks an important stellar mass associated with the stabilization and formation of a disk . we will refer to this mass as the mass of disk formation " , m@xmath6 . in figure [ fig : kinematics ] we separate our sample into 3 bins , motivated by their location on the tfr . we define these populations as high mass ( @xmath3 m@xmath4/m@xmath38 ) disks ( on the tfr ; black triangles ) , low mass ( @xmath3 m@xmath39 ) ordered ( @xmath40 ) disks ( on the tfr ; blue circles ) , and low mass ( @xmath3 m@xmath39 ) disordered ( @xmath41 ) systems ( scattering to low v@xmath2 from the tfr ; red diamonds ) . this notation will be continued for successive plots . we compare the slope in the tfr , @xmath3 v@xmath2 @xmath42 @xmath28 @xmath3 m@xmath4 , with a local sample ( z @xmath0 0.1 ) of 189 disk galaxies from sdss @xcite , who find a value of @xmath28 = 0.278 @xmath26 0.13 over a mass range 9.0 @xmath0 @xmath3 m@xmath4/m@xmath5 @xmath0 11.0 . we refer to this relation as the tf `` ridgeline '' and mark it and its extrapolation to lower stellar masses in our tf plots . the high and low mass disks in our sample are well fit by the local ridgeline . in the middle panel of figure [ fig : kinematics ] we plot m@xmath4 versus @xmath1 . as mentioned previously , @xmath1 integrates unresolved velocity gradients in our systems , probing disordered components to the velocity field . the subset of galaxies that fall short of the tully - fisher ridgeline ( red points ) exhibit higher integrated velocity dispersions than the galaxies on the ridgeline at similar stellar mass . as shown in k07 and k12 , combining both the velocity dispersion and rotation velocity into a new kinematic parameter , @xmath43 , establishes a relatively tight relation . in the bottom panel of figure [ fig : kinematics ] we plot s@xmath44 versus m@xmath4 for k=0.5 . the choice of k=0.5 is motivated by virial arguments for a spherically symmetric tracer distribution ( see @xcite ) . combining both the velocity dispersion and rotation velocity into s@xmath9 re - establishes a tight analogue to the tfr , independent of morphology and coincident with the faber - jackson relation ( k07 , k12 ) . /v@xmath2 ) . as expected from k07 and k12 , these quantities trace each other well . we use @xmath1/v@xmath2 to quantify the location of a galaxy on the tfr in the remainder of this paper.,width=332 ] as expected , for galaxies in our sample , the residual from the local tf ridgeline is a strong function of @xmath1/v@xmath2 . in figure [ fig : scatter ] we demonstrate this correlation , demarcating the region where ordered disk galaxies fall ( i.e. those on the tfr ) would fall . we will use @xmath1/v@xmath2 as a proxy for the location of a galaxy on the tfr plot ( figure [ fig : kinematics ] , top ) . in figure [ fig : tfmorph ] we reproduce the tfr in figure [ fig : kinematics ] , but replace the data points with hst v + i - band color images . upon visual inspection we note distinct morphological characteristics among the regions of this plot identified in figure [ fig : kinematics ] . the galaxies with stellar masses higher than @xmath3 m@xmath45m@xmath46 exhibit extended disk - like morphologies . galaxies with stellar masses @xmath3 m@xmath45m@xmath47 that are rotation dominated ( @xmath1/v@xmath2 @xmath0 1 ) also show disk morphology , although they are less extended than the higher mass systems . galaxies that fall below the tf ridgeline have dispersion dominated kinematics ( @xmath1/v@xmath2 @xmath34 1 ) and show irregular or compact morphologies . in figure [ fig : indmorph ] , we show two representative galaxies from each of the three defined regions in figure [ fig : kinematics ] ( from left to right ) : ( 1 ) low mass dispersion dominated galaxies which have compact morphology and irregular asymmetric features , ( 2 ) low mass rotationally supported galaxies which have characteristic morphologies of disks : they are more or less symmetric and elongated , and ( 3 ) high mass ordered galaxies which have similar disk - like morphologies . these trends were noted by visual inspection in previous work ( k07 ) . we will now demonstrate that gas phase kinematics are related to quantitative galaxy morphology by comparing @xmath1/v@xmath2 ( ratio of disordered to ordered motions ) to quantitative morphological indices ( gini coefficient , concentration , asymmetry ) . as described in the rest of this section , we find that the galaxies which fall on the tully - fisher ridgeline tend to exhibit low values of asymmetry , concentration , and gini . in addition , they tend to have larger sizes consistent with the disk size - mass relation . kinematically ordered galaxies ( @xmath1/v@xmath2 @xmath0 1 ) at both low and high mass are consistent with being drawn from identical parent distributions for these indices . galaxies which are kinematically disordered ( @xmath1/ v@xmath2 @xmath34 1 ) and fall from the tfr ridgeline tend to exhibit quantitative morphologies characteristic of disturbed or compact systems ( high asymmetry , concentration and gini ) and are statistically distinct from the galaxies on the tf ridgeline . below we detail these findings . ) are shown as a function of kinematics for the galaxies in our sample . the bottom panel shows the log residuals from the sdss g+r derived disk mass - size relation ( rm ; @xcite ) . the quantity @xmath1/v@xmath2 measures the relative contributions from disordered motions and ordered rotation . galaxies which scatter to low v@xmath2 off of the tf ridgeline ( red diamonds ) are on average more asymmetric , have higher gini and concentration values , and are smaller than galaxies on the tf ridgeline ( blue circles for low mass and black triangles for high mass ) . ] the top panel in figure [ fig : morph ] shows the asymmetry index versus @xmath1/v@xmath2 for our galaxy sample . low mass disordered galaxies ( red diamonds ) on average have a higher distribution in asymmetry than the galaxies with ordered kinematics which lie on the tf ridgeline ( blue circles and black triangles ) . the asymmetry index is distributed with an average and standard deviation 0.18 @xmath26 0.06 for low mass disordered systems and 0.11 @xmath26 0.06 and 0.11 @xmath26 0.04 for the low and high mass ordered systems , respectively . we run a two - sample kolmogorov - smirnov ( k - s ) test to determine the probability that the three distributions in asymmetry are drawn from identical parent distributions . the p - value returned from the k - s test represents the probability that data drawn from identical parent distributions would be as disparate as observed . the k - s test between the low and high mass ordered populations yields a p value of 0.77 , indicating a high probability of being drawn from identical parent distributions . however , the k - s tests between the low mass disordered galaxies and the high / low mass ordered distributions yield near zero probabilities of being identical ( p = 10@xmath48 and 10@xmath49 , for comparisons with high and low mass ordered samples , respectively . ) the second panel in figure [ fig : morph ] shows the gini coefficient versus @xmath1/v@xmath2 . the disordered low mass population of galaxies exhibits a trend towards higher values of gini . average values of gini increase from g = 0.48 @xmath26 0.04 and 0.48 @xmath26 0.09 for the low and high mass ordered systems , respectively , to g = 0.52 @xmath26 0.04 for the disordered low mass systems . the k - s test between the ordered populations yields a p value of 0.48 , while the dispersion dominated systems are unique compared with either distribution of ordered galaxies ( p = @xmath50 ) . while statistically significant , the relative change in gini is small , reflecting the fact that even disk galaxies exhibit a level of non - uniformity ( i.e. spiral features and clumpy star - formation ) . in large heterogeneous samples of galaxies the range of the gini coefficient is relatively small ( e.g. @xcite ) , with a floor at a value of around 0.40 for the most uniform galaxies and a peak value of only 0.65 . in relation to this small range , a change of 0.04 reflects a non - negligible difference in galaxy type . the third panel of figure [ fig : morph ] shows the concentration index versus @xmath1/v@xmath2 . similar to the gini coefficient , kinematically disordered systems typically display higher values of concentration . the concentration index increases from 0.51 @xmath26 0.09 and 0.53 @xmath26 0.07 for the low and high mass ordered systems , respectively , to 0.65 @xmath26 0.12 for the disordered low mass systems . the k - s test between the ordered populations yields a p value of only 0.02 , due to the large scatter in the low mass ordered distribution . the p value between the low mass disordered distribution and the ordered distributions is once again near zero ( p = 10@xmath48 ) . we now examine the correlation between size with disordered motions in our sample . we use the hst v - band half - light radius r@xmath51 as measured using gim2d . the low mass dispersion dominated galaxies have a mean value of r@xmath51/kpc = 1.8 @xmath26 1.2 rms . the low and high mass ordered systems are close to a factor of 2 larger , 3.1 @xmath26 1.5 and 4.5 @xmath26 2.0 kpc , respectively . to account for the well known mass dependance on size , we examine residuals from the sdss dr7 median fit to the disk mass - size relation ( equation 3 in @xcite ) . in the bottom panel of figure [ fig : morph ] we plot the log residual from the size mass relation ( @xmath52r@xmath53r@xmath54 ) versus @xmath1/v@xmath2 . the ordered galaxies in our sample are distributed around the relation with an average scatter of 0.22 dex , consistent with the intrinsic width of the relation . the residuals for the dispersion dominated systems are centered 0.15 dex below the relation , although there are a few outliers that lie on or above the relation . at a given mass , the dispersion dominated systems tend to be smaller than expected from the radius - mass relation . in the concentration - asymmetry plane ( figure [ fig : conc - asymm ] ) these differences become more apparent , with both the low and high mass ordered galaxies occupying relatively narrow regions of low asymmetry and concentration . the dispersion dominated systems , however , tend to exhibit markedly high scatter in this plane , tending towards higher values of both concentration and asymmetry . _ in summary : the ordered systems in our sample display quantitative indices characteristic of disks : low asymmetry and low concentration . kinematically disordered galaxies on the other hand exhibit concentrated emission with asymmetric features , statistically distinct from their ordered counterparts . _ large local studies of the tfr have primarily focused on massive disk galaxies with @xmath3 m@xmath4/m@xmath5 @xmath55 9.5 ( e.g. @xcite ) . we find that the majority of star - forming galaxies above this mass are morphologically disk - like and the tfr is well behaved . in fact , the only populations of dispersion dominated systems at high stellar masses in the local universe are the relatively rare and interaction induced ( u)lirgs systems ( e.g. ) and merging galaxies ( e.g. the hi rogues gallery of @xcite ) . the tfr for @xmath3 m@xmath4/m@xmath5 @xmath0 9.5 is less well - studied , especially for a morphologically unbiased sample . furthermore , velocity dispersion is not frequently measured . integrated line widths are often used as a proxy for the rotation velocity , with corrections assumed for anomalous contributions from random motions . most tfr studies of dwarf galaxies have focused on gas dominated systems ( m@xmath56 @xmath57 m@xmath4 ) . these galaxies typically show large residuals from the tfr to lower masses at a given rotation velocity , in the opposite direction to our sample . a tight baryonic tully - fisher relation is established once gas masses are included @xcite and it has been well calibrated for a local sample of gas - dominated galaxies ( e.g. @xcite ) . for local dwarf galaxies with both @xmath1 and v@xmath2 measured , evidence for a luminosity dependence on v/@xmath1 has been long established ( e.g. ) . this trend is to be expected even for galaxies on the tfr , if we assume that @xmath1 is independent of mass and v@xmath2 declines with mass . when available , though , resolved kinematics of irregular or compact dwarf systems typically show complex velocity fields which depart from a regularly rotating thin disk ( e.g. @xcite ) . for instance , the well studied smc ( m@xmath4/m@xmath58 10@xmath59 ) and lmc ( m@xmath4/m@xmath58 10@xmath60 ) are both rotationally supported in their stellar and neutral gaseous components ( v/@xmath61 3 ) , while the kinematics of the smallest local di systems , traced through neutral hi gas , typically showing no signs of rotation ( e.g. @xcite ) . recent surveys of hi in the local volume have helped further characterize the dynamics of dwarfs , e.g. things @xcite , figgs @xcite , vla - angst @xcite and little things @xcite . these surveys find mixed evidence among dwarf galaxies for both the existence of ( i ) rotationally supported disks ( e.g. @xcite ) and ( ii ) systems with thick disks or high / dominant contributions of dispersion ( e.g. @xcite ) . the current data for resolved ionized gas kinematics in low mass galaxies is more limited . interest in actively star - forming blue compact dwarfs ( bcd ) has led to measurements of velocity fields for a relatively large sample ( @xmath62 ) of these systems , usually in hi ( e.g. @xcite ) . star - bursting systems , of which bcds are a subset , make up close to 5@xmath24 of the population of local dwarf galaxies @xcite . the origin of the compactness and late stage star - formation in the low mass ( @xmath3 m@xmath4/m@xmath5 @xmath10 7.0 - 9.6 ; @xcite ) bcds is up for debate . recent hi investigations suggest high fractions of kinematically disturbed disks ( @xmath10 50% ) , although only a handful of systems fall short of the baryonic tully - fisher relation . the ionized velocity fields are often found to exhibit high or dominant contributions from dispersion as well ( e.g. @xcite ) . one of the largest surveys of resolved ionized gas kinematics in local galaxies is the ghasp survey ( gassendi h@xmath28 survey of spirals ; @xcite ) . the ghasp survey selection was focused on galaxies with both spiral and irregular morphologies , although most of the more luminous galaxies display characteristic disk morphologies . the ghasp survey samples a wide range in stellar mass 9.0 @xmath0 @xmath3 m@xmath4/m@xmath5 @xmath0 11.7 @xcite . the tfr for a subset of galaxies is found to be relatively tight down to m@xmath44 = -18 or @xmath3 m@xmath4= 8.0 m@xmath5 @xcite . while galaxies with large non - circular motions and corresponding rotation curve asymmetries show the largest scatter from the tf relation , to both low v@xmath2 and low luminosity , the scatter is small compared to the dispersion dominated systems in our sample . aside from a few outliers , no large break to low rotation velocity was found in the tfr for the ghasp sample . this may be due to a combination of both small sample sizes at masses @xmath3 m@xmath4/m@xmath5 @xmath0 9.5 ( 9 galaxies ) and/or selection functions for rotating galaxies playing a role in constructing the stellar mass tfr . the @xmath11/v for these 9 low mass galaxies ranges between 0.1 - 0.7 ( i.e. all would be kinematically ordered in this paper ) . on visual inspection of their optical morphologies in ned , we note that seven of these galaxies appear to have disk morphologies , with the presence of spiral arms in most cases . the low @xmath11/v in these galaxies is consistent with the picture presented in this paper , that disk morphologies indicate ordered kinematics . two of the galaxies ( ugc 5721 , ugc 10757 ) do appear to have disturbed morphologies , so it is interesting to note that their @xmath11/v is still less than 1 ( 0.16 and 0.35 , respectively ) . the mean velocity dispersion for the ghasp sample is 25 @xmath26 5 km s@xmath17 , typical of local disks in general . this level of dispersion is expected for a combination of the thermal broadening from h atoms ( t@xmath1010@xmath33 k ; @xmath10 10 kms@xmath17 ) and internal turbulence of hii regions ( @xmath10 20 kms@xmath17 ; ) . these measurements are comparable to the mean velocity dispersions for our ordered disks ( 21.0 @xmath26 8.1 kms@xmath17 ) and slightly lower than our systems with more disturbed kinematics ( 34 @xmath26 7.3 kms@xmath17 ) . for the few compact rotators in ghasp we ran mock tests ( see appendix a for an example with galaxy ugc 528 ) to ensure that we could accurately recover their rotation if they were in our sample at z @xmath10 0.2 . the results of these tests indicate that if these galaxies were included in our sample we would recover their rotating disks . our understanding of the ionized kinematics of dwarf galaxies at low redshifts will substantially increase in the near future with ongoing and upcoming integral field spectroscopy ( ifs ) surveys . the califa ( 0.005 @xmath0 z @xmath0 0.03 , r @xmath10 850 - 1650 ; n @xmath10 600 ) , sami ( z @xmath0 0.12 , r @xmath10 1700 - 4500 ; n @xmath10 3400 ) and manga ( z @xmath10 0.03 ; r @xmath10 2000 ; n @xmath10 10,000 ) surveys will provide velocity fields for large ensembles of local low mass galaxies . first results from califa have reported that about half of the galaxies in their sample ( 8.5 @xmath0 @xmath3 m@xmath4/m@xmath5 @xmath0 11.5 ) have structure in their velocity fields which departs from a rotating disk . however , resolved velocity dispersions were not measured in this study and the relatively coarse spectral resolution ( r @xmath10 1200 ) may not allow for measurements of the dispersion at the differences outlined in this paper . first results from sami show a large fraction of both low and high mass local galaxies falling to low v@xmath2 from the tfr @xcite . the sami survey improves upon the ghasp sample with a large coverage of low mass systems , pushing down to stellar masses @xmath3 m@xmath4/m@xmath5 @xmath10 8 . below a stellar mass @xmath3 m@xmath4/m@xmath37 9 , they find a precipitous trailing of galaxies from the well - defined 1 r@xmath63 tf relation @xcite . high spectral resolution in the red band ( 6300 - 7400 @xmath64 , r @xmath10 4500 ) allow measurements of velocity dispersions . a tight scaling relation ( m@xmath4-s@xmath65 ) is established once they account for the average measured velocity dispersion , similar to previous findings at high redshift . the sdss - iii manga survey will provide extensive additional coverage of this low mass regime , with a flat selection distribution down to a stellar mass of 10@xmath66 m@xmath5 and sampling out to at least 1.5 times the effective radius ( r@xmath63 ) @xcite . in summary , when studied in large numbers , local studies of the tfr have been biased towards massive disk galaxies and moreover will often not have the spectral resolution to resolve velocity dispersion . when resolved gas phase kinematics are available in low mass galaxies , complex velocity fields are often found in both neutral and ionized gas . the local tfr for dwarf galaxies will be expanded upon in the upcoming years , with surveys probing kinematics in ionized gas to low stellar masses in unprecedented numbers . early results indicate an increase in complexity in the resolved 2d velocity fields of these low mass galaxies . due primarily to sensitivity limits , there have been few studies of the kinematics in more distant ( z @xmath34 0.1 ) dwarf galaxies . furthermore , most studies do not measure resolved velocity dispersions and do not include irregular or compact galaxies . using deep 8 hour deimos exposures , @xcite constructed the stellar mass tfr for half of their sample ( n = 41 ) of dwarf galaxies ( 7 @xmath67 m@xmath45m@xmath68 9 ) out to z = 1 , demonstrating that at least a fraction of dwarf galaxies are settled into rotationally supported systems at intermediate redshift . the remaining galaxies were either missing emission or were too small to reliably resolve rotation . recently , interest in dispersion dominated galaxies at high redshift ( e.g. @xcite ) has motivated searches for high dispersion low redshift analogs . these studies typically select highly star - forming systems and find complex kinematics with low v/@xmath1 compared to local disk galaxies ( e.g. @xcite ) . compact lyman break analogues ( lba ) are an example of such a selection . lbas are highly star - forming galaxies at z @xmath10 0.2 , selected from their high galex nuv luminosity @xcite . although these systems have slightly higher stellar masses 9.5 @xmath19 m@xmath4/m@xmath5 @xmath19 10.7 than the break in our sample , they exhibit morphologies and kinematics which are similar to some of our more compact dwarf galaxies . @xcite report on ao - assisted ifs observations of 19 compact lbas at z @xmath10 0.2 with stellar masses near our transition mass , 9.1 @xmath0 @xmath3 m@xmath4/m@xmath5 @xmath0 10.7 . they find a very high contribution of disordered motion in these rare systems ( @xmath1 @xmath10 70 km s@xmath17 ) and evidence that disk - like structure is more common in high mass lbas . similarly , @xcite measured h@xmath28 kinematics for 67 extremely star - forming systems down to a stellar mass of @xmath69 m@xmath5 and found that one fifth of their sample shows no signs of rotation . the rotating disk galaxies in their sample fall on the tfr while still exhibiting values of v/@xmath1 lower than local disks , comparable to dynamically hot high redshift galaxies . the dispersion dominated galaxies v/@xmath1 @xmath19 1 in both the intermediate redshift ifs samples have typical stellar masses above our observed mass of disk formation ( @xmath3 m@xmath6/m@xmath5 @xmath10 9.5 ) . although the unique selections ( e.g. high star - formation , high surface mass density ) for these samples are not necessarily representative , they do demonstrate the important contribution of @xmath1 to the kinematics in active systems . at high redshifts ( 1.5 @xmath0 z @xmath0 3 ) , complex velocity fields and thick disks with high velocity dispersions are ubiquitous in the samples currently available in the literature ( e.g. @xcite ) . a stellar mass dependence on the relative contributions of @xmath1 and v@xmath2 appears to exist , although beam smearing ( see appendix ) may provide the enhanced dispersion in some of the the smallest systems @xcite . kinematic data for low mass galaxies at these redshifts is currently limited . the advent of high sensitivity near - ir instruments ( e.g. mosfire , kmos ) will soon provide large samples of kinematics for galaxies with low stellar masses . we study the stellar mass tully - fisher relation ( tfr ; stellar mass versus rotation velocity ) for a morphologically blind sample of emission line galaxies in the field . the galaxies are at a redshift of @xmath70 , or a lookback time of about 2 gyrs , and therefore we expect little evolution from the local tfr . we report on a transition mass in the tfr which we call the mass of disk formation , " m@xmath6 . this mass separates galaxies which always form disks ( masses greater than m@xmath6 ) from those which may or may not form disks ( masses less than m@xmath6 ) . for m@xmath71 , all galaxies in our sample are settled onto the local tfr . however , for m@xmath72 , galaxies can either lie on the tfr or scatter off of it to low rotation velocity . the galaxies which scatter off have higher disordered motions , as measured through integrated gas velocity dispersions ( @xmath1 ) . moreover , we find that galaxies on the tfr are morphologically distinct from those which scatter off . the quantitative morphologies of galaxies on the relation are on average less asymmetric and concentrated than those galaxies which scatter off . we perform mock observations of the 3d kinematics of local galaxies to investigate how well we are able to recover the rotation velocity of the smallest galaxies in our sample . these simulations show that we are able to recover rotation , if present , in the smallest galaxies in our sample . the authors would like to thank the anonymous referee who provided useful suggestions that improved this manuscript . rcs gratefully acknowledges support through a grant from the stsci jdf . this research has made use of the fabry perot database , operated at cesam / lam , marseille , france . the authors also acknowledge nsf grants ast 95- 29098 and 00- 71198 to uc santa cruz . rcs would like to thank a. dutton for helpful comments . we wish to extend thanks to those of hawaiian ancestry on whose sacred mountain we are privileged guests . abraham , r. g. , valdes , f. , yee , h. k. c. , & van den bergh , s. 1994 , apj , 432 , 75 abraham , r. g. , tanvir , n. r. , santiago , b. x. , et al . 1996 , mnras , 279 , l47 abraham , r. g. , van den bergh , s. , & nair , p. 2003 , apj , 588 , 218 amorn , r. , vlchez , j. m. , hgele , g. f. , et al . 2012 , apj , 754 , l22 arribas , s. , colina , l. , bellocchi , e. , maiolino , r. , & villar - martn , m. 2014 , a&a , 568 , a14 barton , e. j. , geller , m. j. , bromley , b. c. , van zee , l. , & kenyon , s. j. 2001 , aj , 121 , 625 begeman , k. g. 1987 , ph.d . thesis , begum , a. , chengalur , j. n. , karachentsev , i. d. , & sharina , m. e. 2008 , mnras , 386 , 138 begum , a. , chengalur , j. n. , karachentsev , i. d. , sharina , m. e. , & kaisin , s. s. 2008 , mnras , 386 , 1667 bell , e. f. , & de jong , r. s. 2001 , apj , 550 , 212 bell , e. f. , papovich , c. , wolf , c. , et al . 2005 , apj , 625 , 23 bershady , m. a. , verheijen , m. a. w. , swaters , r. a. , et al . 2010 , apj , 716 , 198 bertin , e. , & arnouts , s. 1996 , a&as , 117 , 393 bothwell , m. s. , kennicutt , r. c. , & lee , j. c. 2009 , mnras , 400 , 154 brooks , a. m. , governato , f. , quinn , t. , brook , c. b. , & wadsley , j. 2009 , apj , 694 , 396 bundy , k. , ellis , r. s. , conselice , c. j. , et al . 2006 , apj , 651 , 120 bundy , k. , bershady , m. a. , law , d. r. , et al . 2015 , apj , 798 , 7 cannon , j. m. , mcclure - griffiths , n. m. , skillman , e. d. , & ct , s. 2004 , apj , 607 , 274 catinella , b. , kauffmann , g. , schiminovich , d. , et al . 2012 , mnras , 420 , 1959 chabrier , g. 2003 , the publications of the astronomical society of the pacific , 115 , 763 cortese , l. , fogarty , l. m. r. , ho , i .- t . , et al . 2014 , apj , 795 , l37 courteau , s. , dutton , a. a. , van den bosch , f. c. , et al . 2007 , apj , 671 , 203 covington , m. d. , kassin , s. a. , dutton , a. a. , et al . 2010 , apj , 710 , 279 davis , m. , guhathakurta , p. , konidaris , n. p. , et al . 2007 , apj , 660 , l1 davies , r. , frster schreiber , n. m. , cresci , g. , et al . 2011 , apj , 741 , 69 de rossi , m. e. , tissera , p. b. , & pedrosa , s. e. 2012 , a&a , 546 , a52 dutton , a. a. , van den bosch , f. c. , dekel , a. , & courteau , s. 2007 , apj , 654 , 27 dutton , a. a. , van den bosch , f. c. , faber , s. m. , et al . 2011 , mnras , 410 , 1660 elmegreen , b. g. , & burkert , a. 2010 , apj , 712 , 294 epinat , b. , amram , p. , & marcelin , m. 2008 , mnras , 390 , 466 epinat , b. , amram , p. , balkowski , c. , & marcelin , m. 2010 , mnras , 401 , 2113 faber , s. m. , phillips , a. c. , kibrick , r. i. , et al . 2003 , proc . spie , 4841 , 1657 frster schreiber , n. m. , genzel , r. , bouch , n. , et al . 2009 , apj , 706 , 1364 gallazzi , a. , charlot , s. , brinchmann , j. , & white , s. d. m. 2006 , mnras , 370 , 1106 garrido , o. , marcelin , m. , amram , p. , & boissin , o. 2003 , a&a , 399 , 51 garca - lorenzo , b. , mrquez , i. , barrera - ballesteros , j. k. , et al . 2015 , a&a , 573 , a59 gonalves , t. s. , basu - zych , a. , overzier , r. , et al . 2010 , apj , 724 , 1373 green , a. w. , glazebrook , k. , mcgregor , p. j. , et al . 2014 , mnras , 437 , 1070 heckman , t. m. , hoopes , c. g. , seibert , m. , et al . 2005 , apj , 619 , l35 hibbard , j. e. , van gorkom , j. h. , rupen , m. p. , & schiminovich , d. 2001 , gas and galaxy evolution , 240 , 657 hunter , d. a. , ficut - vicas , d. , ashley , t. , et al . 2012 , aj , 144 , 134 kannappan , s. j. , fabricant , d. g. , & franx , m. 2002 , aj , 123 , 2358 kassin , s. a. , de jong , r. s. , & weiner , b. j. 2006 , apj , 643 , 804 kassin , s. a. , weiner , b. j. , faber , s. m. , et al . 2007 , apj , 660 , l35 kassin , s. a. , weiner , b. j. , faber , s. m. , et al . 2012 , apj , 758 , 106 kirby , e. n. , bullock , j. s. , boylan - kolchin , m. , kaplinghat , m. , & cohen , j. g. 2014 , mnras , 439 , 1015 kronberger , t. , kapferer , w. , schindler , s. , & ziegler , b. l. 2007 , a&a , 473 , 761 law , d. r. , steidel , c. c. , erb , d. k. , et al . 2009 , apj , 697 , 2057 lee , j. c. , kennicutt , r. c. , jr . , funes , s. j. j. g. , sakai , s. , & akiyama , s. 2009 , apj , 692 , 1305 lehnert , m. d. , nesvadba , n. p. h. , le tiran , l. , et al . 2009 , apj , 699 , 1660 lelli , f. , verheijen , m. , & fraternali , f. 2014 , a&a , 566 , aa71 lemoine - busserolle , m. , & lamareille , f. 2010 , mnras , 402 , 2291 lin , l. , koo , d. c. , weiner , b. j. , et al . 2007 , apj , 660 , l51 lotz , j. m. , primack , j. , & madau , p. 2004 , aj , 128 , 163 lotz , j. m. , madau , p. , giavalisco , m. , primack , j. , & ferguson , h. c. 2006 , apj , 636 , 592 lotz , j. m. , davis , m. , faber , s. m. , et al . 2008 , apj , 672 , 177 mac low , m .- m . , & ferrara , a. 1999 , apj , 513 , 142 masters , k. l. , springob , c. m. , & huchra , j. p. 2008 , aj , 135 , 1738 mateo , m. l. 1998 , ara&a , 36 , 435 mahajan , s. , drinkwater , m. j. , driver , s. , et al . 2015 , mnras , 446 , 2967 mcconnachie , a. w. 2012 , aj , 144 , 4 mcgaugh , s. s. 2005 , apj , 632 , 859 miller , s. h. , ellis , r. s. , newman , a. b. , & benson , a. 2014 , apj , 782 , 115 mo , h. j. , mao , s. , & white , s. d. m. 1998 , mnras , 295 , 319 mortlock , a. , conselice , c. j. , hartley , w. g. , et al . 2013 , mnras , 433 , 1185 navarro , j. f. , & steinmetz , m. 2000 , apj , 538 , 477 newman , j. a. , cooper , m. c. , davis , m. , et al . 2013 , apjs , 208 , 5 newman , s. f. , genzel , r. , frster schreiber , n. m. , et al . 2013 , apj , 767 , 104 ott , j. , stilp , a. m. , warren , s. r. , et al . 2012 , aj , 144 , 123 patil , a. , d. huard and c.j . fonnesbeck . pymc : bayesian stochastic modelling in python . journal of statistical software , 35(4 ) , pp . 1 - 81 prez - gallego , j. , guzmn , r. , castillo - morales , a. , et al . 2011 , mnras , 418 , 2350 peth , m. a. , lotz , j. m. , freeman , p. e. , et al . 2015 , arxiv:1504.01751 puech , m. , hammer , f. , flores , h. , et al . 2010 , a&a , 510 , a68 pizagno , j. , prada , f. , weinberg , d. h. , et al . 2005 , apj , 633 , 844 pizagno , j. , prada , f. , weinberg , d. h. , et al . 2007 , aj , 134 , 945 rampazzo , r. , plana , h. , amram , p. , et al . 2005 , mnras , 356 , 1177 reyes , r. , mandelbaum , r. , gunn , j. e. , pizagno , j. , & lackner , c. n. 2011 , mnras , 417 , 2347 roberts , m. s. , & haynes , m. p. 1994 , ara&a , 32 , 115 roychowdhury , s. , chengalur , j. n. , begum , a. , & karachentsev , i. d. 2010 , mnras , 404 , l60 shields , g. a. 1990 , ara&a , 28 , 525 simard , l. , willmer , c. n. a. , vogt , n. p. , et al . 2002 , apjs , 142 , 1 somerville , r. s. , & primack , j. r. 1999 , mnras , 310 , 1087 stark , d. v. , mcgaugh , s. s. , & swaters , r. a. 2009 , aj , 138 , 392 swaters , r. a. , van albada , t. s. , van der hulst , j. m. , & sancisi , r. 2002 , a&a , 390 , 829 swaters , r. a. , sancisi , r. , van albada , t. s. , & van der hulst , j. m. 2009 , a&a , 493 , 871 thuan , t. x. , hibbard , j. e. , & lvrier , f. 2004 , aj , 128 , 617 torres - flores , s. , epinat , b. , amram , p. , plana , h. , & mendes de oliveira , c. 2011 , mnras , 416 , 1936 tully , r. b. , & fisher , j. r. 1977 , a&a , 54 , 661 vaduvescu , o. , mccall , m. l. , richer , m. g. , & fingerhut , r. l. 2005 , aj , 130 , 1593 van zee , l. , skillman , e. d. , & salzer , j. j. 1998 , aj , 116 , 1186 van zee , l. , salzer , j. j. , & skillman , e. d. 2001 , aj , 122 , 121 vergani , d. , epinat , b. , contini , t. , et al . 2012 , a&a , 546 , a118 verheijen , m. a. w. 2001 , apj , 563 , 694 walter , f. , brinks , e. , de blok , w. j. g. , et al . 2008 , aj , 136 , 2563 weiner , b. j. , willmer , c. n. a. , faber , s. m. , et al . 2006a , apj , 653 , 1027 weiner , b. j. , willmer , c. n. a. , faber , s. m. , et al . 2006b , apj , 653 , 1049 wisnioski , e. , frster schreiber , n. m. , wuyts , s. , et al . 2015 , apj , 799 , 209 yegorova , i. a. , & salucci , p. 2007 , mnras , 377 , 507 young , l. m. , van zee , l. , lo , k. y. , dohm - palmer , r. c. , & beierle , m. e. 2003 , apj , 592 , 111 zhao , y. , gao , y. , & gu , q. 2013 , apj , 764 , 44 reliably measuring resolved kinematics for a galaxy depends critically on its apparent size relative to the seeing , i.e. beam smearing ( @xcite ) . atmospheric turbulence blurs together intrinsically spatially separate velocity gradients , leading to a smoothing of the rotation field and a boosted central velocity dispersion ( e.g. @xcite ) . rotcurve models the seeing and recovers rotation curves and average integrated gas velocity dispersions for galaxies with apparent sizes larger than the seeing ( see examples in @xcite ) . overcoming beam smearing with galaxies whose intrinsic size is comparable to the seeing is challenging and has been addressed by several papers with respect to ifs measurements at high redshift ( e.g. @xcite ) . to understand the effects of beam smearing on our measurements we perform mock observations of observed velocity and dispersion maps for a nearby galaxy from the literature . in particular , we use fabry - perot observations of ugc 528 from the ghasp survey @xcite . we determine the apparent galaxy size ( defined as the extent of the line emission ) at which the intrinsic rotation velocity is no longer recovered under the observing conditions and instrumental set - up used by the deep2 survey . ugc 528 is a small rotating local spiral galaxy of hubble type sab(rs)b . it has a rotation velocity v@xmath73(i ) @xmath10 60 kms@xmath17 and a relatively high average dispersion across its face of @xmath1 = 26.0 @xmath26 9.6 kms@xmath17 . we build mock spectral cubes ( 3d : spatial @xmath27 spatial @xmath27 @xmath29 ) from the available velocity , dispersion and flux maps . this allows us to perform mock observations for an arbitrary velocity sampling without needing to interpolate between slices . to do this , we first construct an empty cube with velocity sampling matching deep2 , namely @xmath74v = 12.6 kms@xmath17 at z = 0.2 next , using the flux , velocity and dispersion maps we fill the cube with model h@xmath28 emission . for each spaxel in the map with valid velocity information we create a gaussian line profile with a mean line of sight velocity v@xmath75 , dispersion @xmath1 and amplitude scaled to match the integrated flux . this cube is then projected from its original redshift to z = 0.2 by resizing the spaxel scale . each slice in the cube is then convolved with a 2d spatial gaussian kernel ( fwhm = 0.75@xmath15 ) to simulate seeing . a mock 1@xmath15 wide slit is placed along the kinematic pa of the galaxy . from this we produce a mock spectrum using the deep2 pixel scales ( 0.118 @xmath15/pixel , 0.33 @xmath64/pixel ) and spectral resolution ( r @xmath10 5000 ) . noise is added to the 2d spectrum to match the typical pixel to pixel s / n profiles of our faintest sources . we create separate cubes in this manner , each time varying the spatial size to simulate different apparent sizes for ugc 528 . the final product is a set of 100 mock observations of ugc 528 at z = 0.2 ranging in size 0.4 -2.2@xmath15 with a 0.018@xmath15 step size . each mock observation results in a 2d spectrum of the emission line for a given apparent size . next we run each of the mock spectra through rotcurve to measure kinematics . we fix the scale radius of ugc 528 in physical units to its true value of 0.6 kpc @xcite . for reasonable values , we find that the chosen scale radius has little impact on the kinematic measurements , as was also found by @xcite . we measure v@xmath2 @xmath27 @xmath76 and @xmath1 for each of the 100 mock spectra . the results of our test are demonstrated in figure [ fig : ugc528 ] . as expected , at large intrinsic sizes relative to the seeing ( d@xmath25 @xmath34 1.5 @xmath27 seeing ) we are able to recover the intrinsic rotation and integrated dispersion well . as the mock observations reach closer to and beyond the scale of the seeing , we note that the measured rotation velocity decreases and the measured dispersion increases as expected . to quantify the scale at which this turnover occurs , we fit a simple change point model using pymc @xcite . the turnover to artificially low v@xmath2 occurs at a scale of d@xmath25/seeing = 0.87 @xmath26 0.06 . this error incorporates both the measurement error imposed by the slit ( red shading in figure [ fig : ugc528 ] , left ) and the error in the fit to the model . we now compare the intrinsic sizes of the galaxies in our sample to the size of the seeing . in the right panel of figure [ fig : ugc528 ] we plot the size of the galaxy in emission , tracing areas of star - formation , against the size in continuum , tracing the young stellar population . the hubble sizes are measured from the hst v - band images and the emission sizes are measured directly from the deimos spectra . the observed emission lines ( @xmath77 ) are broadened by the combined psf of the instrument and the seeing ( @xmath78 ) . we can recover the true extent of the emission line ( @xmath79 ) : for an approximate 1-d gaussian profile , the diameter containing 95@xmath24 of the flux will be 4 times this value . in reality , intrinsic emission profiles of galaxies will be broader than a gaussian distribution and the value of @xmath79 is primarily determined by the profile of the core . a galaxy profile with broader wings than a gaussian will have a d@xmath25 that is larger than that given by 4 @xmath27 @xmath79 . the hst v - band ( rest b at z @xmath10 0.2 ) extent correlates well with the seeing - corrected emission line sizes ( figure [ fig : ugc528 ] , right ) . we demarcate the typical seeing size of 0.75@xmath15 . the smallest galaxies in our sample reach down to seeing - corrected emission extents of @xmath10 1.2@xmath15 ( or 1.6 times the typical seeing fwhm ) , well into the predicted regime for accurately recovering the kinematics ( figure [ fig : ugc528 ] ; left panel ) . in conclusion , it is unlikely that our low mass dispersion dominated galaxies are dispersion - dominated because of beam - smearing .

we present high spatial resolution observations of the blue compact dwarf galaxy izw18 performed in the h line with a scanning fabry - perot interferometer at the cfh telescope . morphological structure of the galaxy in h and in the red continuum is investigated . we also analyse the velocity field of the ionized gas . besides the two compact hii components of the main body we find a population of small hii regions in its surroundings whose diameter distribution and h luminosity function are consistent with those observed in dwarf irregular galaxies . in the main body of the galaxy besides of the nw and se red continuum peaks which are displaced with respect to the h maxima , three new red condensations have been discovered . they have no clear h counterparts . the velocity field in izw18 shows peculiar motions superimposed on a quite regular background implying solid - body rotation with a gradient of about 70 km kpc . the h line profiles exhibit an asymmetric structure , except for the nw main compact component . at least part of this asymmetry could result from accreted and/or expelled surrounding gas from the main star - forming core(s ) of the galaxy . contrary to previous suggestions that the south - west and north - east extensions of this galaxy are diffused emission produced by bipolar emitting gas we provide evidence that they are hii regions powered by star formation sites . the redshift of the zwicky s `` flare '' has been measured for the first time and corresponds to the same velocity as izw18 . in such a context the optical ridge that appears to be an isolated morphological structure has a shape that may result from the gravitational interaction with the zwicky s `` flare '' if this latter is a neighbour extreme dwarf object .

astro-ph9612079

extensive studies of blue compact dwarf galaxies ( bcdgs ) have revealed that a large fraction of them are chemically unevolved small systems undergoing violent star formation activity ( thuan 1987 ) . the triggering mechanism of starbursts in bcdgs is still completely unknown . one of the approaches for understanding this mechanism is to investigate the morphological , kinematical and dynamical relationship between star forming regions and their surrounding gas and also to search for global correlations between star forming regions and the gas in various places of a galaxy . since bcdgs are among the best candidates for genuine zero - redshift young galaxies their kinematical and dynamical state may give also some clues on the nature of dark matter ( tremaine & gunn 1979 ) . the neutral gas component of some bcdgs has been kinematically studied with radio interferometers ( viallefond & thuan 1983 ; viallefond et al . 1987 ; taylor et al . in the few objects studied so far , this gas shows core - halo distribution and some chaotic velocity superimposed on a regular velocity field indicative of a solid body rotation . on the other hand the ionized gas component in bcdgs has a clumpy distribution and is mainly concentrated in at least one compact star forming hii region immersed in an amorphous extended envelope of low luminosity ionized material ( kunth et al . the inner dynamics of hii regions in bcdgs has been investigated by gallagher & hunter ( 1983 ) and lequeux et al . velocity fields in bcdgs have been mapped using multi - slit observations ( tomita et al . 1993 ) and h@xmath0 fabry - perot interferometric observations ( thuan et al . 1987 ) . the combination of a fabry - perot interferometer with a large telescope provides high spatial and velocity resolution and allows to probe small scales in dwarf compact galaxies , yielding information on components not resolved on hi maps ( thuan et al . further , a bcdg not only has a small angular size and a compact distribution but at the same time exhibits very strong and narrow emission lines . for this reason bcdgs are ideal targets for fabry - perot interferometer observations . in this paper we present results relevant to the extreme metal poor bcdg izw18 ( mrk116 ) . izw18 first was described by zwicky ( 1966 ) as a `` double system of compact galaxies '' . later izw18 became an outstanding extragalactic object when sargent & searle ( 1970 ) characterized it as `` isolated extragalactic hii region '' and found that it is extremely metal poor as compared to the sun ( searle & sargent 1972 ) . later on after 25 years more than 100 scientific papers have been addressed to this galaxy . but among them , articles which contain data on izw18 as a extended two component system ( petrosian et al . 1978 ; mazzarella & boroson 1993 ) are rare . optical imagery of izw18 has been published by hua et al . ( 1987 ) , davidson et al . ( 1989 , hereafter dkf89 ) , dufour & hester ( 1990 , hereafter dh90 ) and martin ( 1996 , hereafter m96 ) . broadband wfpc2 on hst observations were carried out by hunter & thronson ( 1995 , hereafter ht95 ) . ghrs hst spectroscopic observations revealed very low oxygen abundance in the hi envelope of this galaxy ( kunth et al . 1994 , but see pettini & lipman 1995 for a critical comment of this result ) . hi line interferometric observations by lequeux & viallefond ( 1980 ) , viallefond et al . ( 1987 ) and radio continuum observations by klein et al . ( 1983 ) were also reported . probably due to its very low metal - content izw18 has been detected neither by iras ( kunth & sevre 1986 ; mazzarella et al . 1991 ) nor in co millimeter line ( arnault et al . 1988 ) . as part of our high spatial and spectral resolution study of genuine bcdgs , in this paper we present the h@xmath0 scanning fabry - perot interferometric observations of izw18 . we describe the observations and data reduction in section 2 . results are presented in section 3 and are discussed in section 4 . we summarize our conclusions in section 5 . we shall adopt 10 mpc as the distance to izw18 , based on a recessional velocity of 763 km @xmath1 and a hubble constant h@xmath3 = 75 km @xmath1 mpc@xmath2 . the linear scale is therefore 48 pc arcsec@xmath2 . observations have been performed at the cassegrain focus of the 3.6 m canada - france - hawaii telescope on november 22 1986 , with the cigale instrument ( boulesteix et al . it consists of a focal reducer ( f/8 - f/2 ) , and a piezo - electrically scanned fabry - perot interferometer from queensgate instruments interfaced with a two dimensional photon counting camera . the optical setup of the f/2 focal reducer allows a field of view of 6.2 x 6.2 . the original photon addresses are computed on a 512 x 512 pixel grid on the detector , giving a scale of 0.725 per pixel . the fabry - perot etalon operates with an interference order of 798 at the rest wavelength of h@xmath0 providing a free velocity range of 375 km @xmath1 . the corresponding free spectral range was scanned through 21 scanning steps ( each step corresponding to a given spacing of the plates of the interferometer ) , the 21th step being equivalent to the 1st one . thus the velocity sampling step is 18.8 km @xmath1 . each scanning step has been exposed 18 x 10s . the total exposure time was 60 minutes ( 18x10x20s ) . a neon lamp emitting the 6598.9 line was used for wavelength calibration . the final product of the 20 elementary interferograms processing is a 3 - dimensional data cube with 2 spatial and one velocity coordinates . for each pixel with reasonably high s / n ratio the radial velocity , the h@xmath0 line profiles and the h@xmath0 and continuum fluxes are derived from this cube . the night sky oh emission line at 6577.2 has a wavelength value close to that of the redshifted izw18 h@xmath0 emission . an interactive routine was used to subtract the night sky line . the radial velocity at each pixel was then extracted by a barycentric technique ( laval et al . the barycenter of the h@xmath0 line is found with an accuracy much better than the interferometer scanning step ( 18.8 km @xmath1 ) . from the data cube , h@xmath0 line profiles are built for each pixel . gaussian curves were fitted to the observed profiles to investigate their structure and derive velocity dispersions due to macroscopic gas motions . the velocity dispersion of the gaussian profile which best fits the observed profile is defined by @xmath4 where @xmath5 is the thermal broadening function and @xmath6 the instrumental profile . according to the value t@xmath7 = 18000 k of the nebular gas electron temperature of izw18 ( dufour et al . 1988 ) the thermal broadening @xmath5 is estimated as 12.2 km @xmath1 ( dopita 1972 ) . the instrumental profile determined by scanning the neon 6598.95 line , was found to be well fitted by a gaussian with velocity width @xmath6 = 14.4 km @xmath1 . the observed profile is described with @xmath8 = 0.425 fwhm . the total intensity profile for each pixel when adding up the 21 channels , contains information about the monochromatic emission as well as the continuum emission of the object . cigale observations enable us to separate monochromatic emission from continuum emission ( see details in laval et al . it is thus possible to restore pure monochromatic h@xmath0 images of izw18 , completely free from the underlying stellar continuum . the continuum subtracted h@xmath0 line images of izw18 were calibrated using the absolute flux published by davidson & kinman ( 1985 ) and dh90 . in the same way a continuum map free from h@xmath0 line contribution was built . the fwhm of the interference filter was 10 which determined the bandwidth of the image continuum of izw18 . continuum flux calibrations were based on the assumption that the continuum slope of izw18 is given by @xmath9 ( davidson & kinman 1985 ) and on absolute flux values at 6600 given by dkf89 . in making a continuum model we have neglected the underlying stellar abosrption since according to skillman and kennicutt ( 1993 ) the absorption component in h@xmath0 is of the order of 2 while the h@xmath0 emission has an equivalent width more than two order of magnitude larger ( 465 ) . for the confirmation of faint detected emission line features such as small narrow - line hii regions , a careful check was made on the series of channel maps : a feature has been considered as certain only if seen on at least two consecutive channel maps . h@xmath0 and continuum luminosities are derived from the measured fluxes and are corrected for extinction using values adopted by dufour et al . ( 1988 ) . since the image of izw18 is sufficiently small to be located in the central part of the field of view where all the causes for sensitivity variations are reduced ( optical system , filter combination , detector inhomogeneities ) the original images of the galaxy were not flat - fielded . when the interferometer is scanning the object , the monochromatic emission of this latter is modulated by the interference pattern while the continuum emission remains unaffected . this enables to distinguish line and continuum emission with a much higher contrast than with simple images obtained through interference filters . figures 1 and 2 show the logarithmic contour maps of respectively the red continuum emission and the monochromatic h@xmath0 emission in absolute flux units across the main body of izw18 . the selectivity of the fabry - prot interferometer due to its high finesse ( f=16 measured on calibration frames ) allows a very clear separation of the line and the continuum emissions . furthermore , the overall system ( fp + focal reducer ) benefits from a very low level of scattered light , much below a percent , which ensures that the continuum map is free from additional line emission . in red continuum ( fig.1 ) the morphological structure of izw18 is complex and new details appear which were not detected by dkf89 . some of these details have their counterparts on the hst wfpc2 frames ( ht95 ) . in the main body of the galaxy besides two known nw and se red continuum components ( dkf89 ) a third one ( a ) is observed half way between them close from the north of the hii region number 6 in ht95 . we identify two other condensations : ( b ) to the north and ( c ) to the north - west of the nw compact red component . among all the components nw dominates in size , surface and absolute brightness . table 1 lists @xmath10@xmath0 , @xmath10@xmath11 positions of the photometric center of the components in arcsec with respect to the nw component . positive values of @xmath10@xmath0 are to the east and positive @xmath10@xmath11 to the north , effective diameters d in parsecs , surface brightness sb in units of 10@xmath12ergs @xmath13 and red magnitudes at 6600 . table 1 also lists the same parameters for the zwicky s `` flare '' ( zwicky 1966 ) - noted as the c satellite by dkf89 and dh90 . cccccc @xmath14 & @xmath15 & @xmath16 & @xmath17 & @xmath18 & @xmath19 + & & & & + @xmath20 & 0 & 0 & 112 & 4.8 & 17.2 + @xmath21 & 2.9 & -3.6 & 79 & 3.1 & 18.7 + @xmath22 & 1.5 & -1.5 & 56 & 3.2 & 19.2 + @xmath23 & -0.7 & 2.9 & 56 & 2.8 & 19.1 + @xmath24 & -2.9 & 1.5 & 40 & 4.0 & 20.2 + @xmath25 & -18.9 & 16.7 & 177 & 0.3 & 20.6 + the comparison between the structure of the nw red component and its surroundings with that of ht95 frames shows that this component is coeval to an isolated hii region south to the main nw star - forming complex ( ht95 ) . a sharp cut - off in the nw red continuum brightness distribution is observed towards the ( a ) and se components while it fades smoothly in the opposite direction towards the ( b ) component . the ( b ) component itself is located in the northern part of nw star forming complex of ht95 . at faint levels of brightness a west side `` ledge '' is observed . the ( c ) component is embedded into this ledge and has no direct counterpart on hst frames ( ht95 ) . besides the main body of the galaxy its outer envelope is also radiating in the red continuum . this radiation is not homogenous . it shows a clumpy structure extending to the ne direction up to a distance of about 740 pc from the nw component . its patterns are located within the same region of the ne h@xmath0 shell of m96 . towards the sw the red continuum radiation extends in more homogeneous way by curving at its end to the west and its patterns closely follow the h@xmath0 arc like structure dh90 . stellar light as well as some nebular continuum contribute to the red continuum radiation of the outer envelope of the galaxy and its arc - like structure . to determine their respective contribution we have measured the l(h@xmath0)/l(6600 ) ratios across these regions and compared the obtained values with model calculations provided to us by mas - hesse ( private communication ) . models show that the underlying nebular continuum should be less than 1 percent of our measured luminosity at 6600 . our h@xmath0 monochromatic image ( fig.2 ) completely reproduces the structural features which were identified in earlier observations ( hua et al . 1987 ; dkf89 ; dh90 ; ht95 ; m96 ; ostlin et al . let us underline some details . we hereafter shall define an autonomous hii feature as the region which is inside its outermost closed contour . both nw and se condensations are isolated from the main body of the galaxy by an isophote corresponding to 4.5 10@xmath26 ergs @xmath27 arcsec@xmath28 of surface brightness . the nw component of izw18 has a size of 195 x 117 pc@xmath29 , while the se component is 157 x 117 pc@xmath29 . note that the elongation of both components coincides with the elongation of the main body of the galaxy . after correction from galactic reddening the integrated h@xmath0 flux of nw and se components come to be 1.3 10@xmath30 ergs @xmath31 and 6.3 10@xmath32 ergs @xmath27 respectively . at d = 10 mpc these values correspond to h@xmath0 luminosities of 1.6 10@xmath33ergs @xmath34 and 7.6 10@xmath35ergs @xmath1 . assuming case b conditions and t@xmath7 = 18000 k ( dufour et al . 1988 ) these luminosities require conversion of 2.2 10@xmath36 and 1.1 10@xmath36 ionizing photons @xmath1 requiring 47 and 23 o5v stars , assuming that an o5v star emits 4.7 10@xmath37 photons @xmath1 ( osterbrock 1989 ) . star formation rates ( from 0.1 to 100@xmath38 ) in the nw and se component are equal respectively to 4.7 10@xmath39@xmath38yr@xmath2 pc@xmath28 and 2.9 10@xmath39@xmath38yr@xmath2 pc@xmath28 assuming a salpeter initial mass function ( hunter & gallagher 1986 ) . masses for the ionized gas in the nw and se components have been estimated to be m(hii ) = 7.3 10@xmath40@xmath38 for the nw condensation and m(hii ) = 3.4 10@xmath40@xmath38 for the se one , but these standard estimates are only indicative since these hii regions are far from being homogeneous as seen from the hst images . the analysis of the mean spatial profiles of both components along the direction of their elongation and perpendicular to it , shows that the surface brightness peak of the nw h@xmath0 component is about 7.0 10@xmath26ergs @xmath27 arcsec@xmath28 , as compared to 6.4 10@xmath26ergs @xmath27 arcsec@xmath28 for the se component . the h@xmath0 surface brightness gradient on the nw component across two perpendicular directions is different . it is gently sloping across the major axis of the component and much steeper in the perpendicular direction where according to ht95 a gap in ionized gas distribution is observed . the smoothest gradient is observed along the tail extending to the nw and wraps around to the west of the component . this feature was noted by dkf89 , dh90 and also traced in the h@xmath0 maps of hua et al . it is very well seen in fig.1b of m96 and shows its detailed structure in fig . 3 of ht95 . as regards to the se compact component , its h@xmath0 line brightness distribution is more homogeneous . an interesting h@xmath0 feature is seen as a prominent winding ridge ( dh90 ; ht95 ; m96 ; ostlin et al . 1996 ) which expands south from the se hii compact component , bends around to the nw , then turns back to the ne , making an half wrapping around the nw hii compact component . in the nw direction from the north - western boundary of the ridge about the same distance as from the nw component lies zwicky s `` flare '' ( zwicky 1966 ) . the ridge as an isolated structure is well shaped and has a conspicuous clumpy structure ( ht95 ) . as mentioned above ( sec.3.1 ) several patterns of the ridge radiate some red continuum . the continuum patterns do not follow the h@xmath0 clumps distribution . to understand the nature of the ridge we carefully tested the possibility of its link to the se compact component . we conclude that there is not direct connection between them . from this point of view the ridge can be an isolated feature in the system . as it is noted by viallefond et al . ( 1987 ) the two major star forming areas in the main body of izw18 are offset to the east of the double - lobed hi emission peak . the ridge wraps around the hi distribution peak opposite to the side where the main body of the galaxy is located , roughly following the steep contours marking the edge of the hi concentration . figure 3 shows the h@xmath0 image of the galaxy with the 6600 red continuum contours overlaid . one sees that the centroid of the nw h@xmath0 emission is located in the plateau , between the nw and ( b ) red continuum peaks and is displaced to the north of at 1.5 ( 72pc ) from nw red continuum peak . as it was mentioned above ( sec . 3.1 ) the nw red component lies on an isolated stellar cluster - hii region complex ( ht95 ) . the ( b ) red peak is located on the nw h@xmath0 emission region and its connection with this star forming region is more evident . the centroid of the se compact h@xmath0 component is displaced by 1.0 ( 48pc ) to the west of the se red continuum centroid . this displacement between the continuum and h@xmath0 images is also evident from fig.17 of ht95 . the h@xmath0 emission extends towards the ne and sw directions from the main body of izw18 up to distance of about 1.2 kpc . the same order of extension is detected by dh90 and m96 . as we stress in sec.3.1 in the same direction but for smaller distances the clumpy red continuum emission was detected . the extended envelope detected in h@xmath0 line is not simply diffuse and shows numerous condensations ( dh90 ) . we suggest that the isolated condensations that have been identified in the extended envelope of izw18 as well as in the ridge are bonafide hii regions . this suggestion is supported by : + - the presence of clumpy red continuum radiation of stellar origin in the same fields where hii regions were identified . - the detection of individual stars in the vicinity of the hii regions fields ( table 2 of ht95 ) . - the existence of nearly 15 well isolated hii regions in the main body of the galaxy also reported by ht95 . since the h@xmath0 luminosity of any individual condensation is low ( see below ) the spectral type of the ionizing stars can not be much earlier that b0 ( osterbrock 1989 ) . at the distance of izw18 such a star has a u magnitude equal to 25.4 which is roughly the limiting magnitude for the f336w frame of ht95 . nevertheless we carefully examined the positions of each individual hii regions to find such stars using the f336w frames of the hst wfpc2 kindly given to us by deidre hunter . no such stars were found . cccccc # & @xmath10@xmath41 & @xmath10@xmath11 & @xmath17 & @xmath42 & @xmath43 + & & & & & + 1 & 10.9 & -2.2 & 56 & 0.65 & 0.78 + 2 & 10.2 & 0.0 & 97 & 1.75 & 2.11 + 3 & 10.2 & 4.4 & 97 & 6.57 & 7.89 + 4 & 8.7 & 0.7 & 40 & 0.77 & 0.92 + 5 & 6.5 & 15.2 & 56 & 1.68 & 2.01 + 6 & 4.4 & -10.9 & 56 & 0.54 & 0.64 + 7 & 3.6 & -11.6 & 40 & 0.65 & 0.78 + 8 & 3.6 & 5.1 & 56 & 6.27 & 7.52 + 9 & 3.6 & 13.1 & 97 & 2.67 & 3.21 + 10 & 2.9 & 18.9 & 69 & 2.71 & 3.25 + 11 & 2.2 & 6.5 & 40 & 3.48 & 4.17 + 12 & 2.2 & 20.3 & 56 & 2.37 & 2.84 + 13 & 1.5 & -16.7 & 97 & 4.55 & 5.46 + 14 & 1.5 & 18.9 & 40 & 1.68 & 2.01 + 15 & 0.7 & -13.8 & 40 & 1.49 & 1.79 + 16 & 0.0 & -8.7 & 40 & 6.69 & 8.03 + 17 & 0.0 & 15.2 & 56 & 2.06 & 2.47 + 18 & -0.7 & -14.5 & 79 & 3.97 & 4.77 + 19 & -0.7 & -10.9 & 69 & 7.03 & 8.44 + 20 & -1.5 & 10.9 & 112 & 7.19 & 8.62 + 21 & -1.5 & 13.1 & 69 & 1.75 & 2.11 + 22 & -2.2 & -17.4 & 56 & 2.79 & 3.35 + 23 & -2.9 & -13.8 & 97 & 9.48 & 11.30 + 24 & 88.0 & -7.3 & 56 & 6.80 & 8.17 + 25 & -5.8 & -15.2 & 40 & 0.92 & 1.10 + 26 & -6.5 & -4.4 & 56 & 7.15 & 8.58 + 27 & -7.3 & 3.6 & 40 & 3.02 & 3.62 + 28 & -8.0 & 4.4 & 56 & 3.86 & 4.63 + 29 & -9.4 & -5.1 & 97 & 17.00 & 20.40 + 30 & -10.2 & -4.4 & 112 & 19.20 & 23.00 + 31 & -10.9 & 1.5 & 56 & 1.64 & 1.97 + 32 & -12.3 & -1.5 & 40 & 1.91 & 2.29 + 33 & -13.1 & 0.0 & 79 & 7.53 & 9.04 + 34 & -14.5 & 4.4 & 97 & 5.58 & 6.70 + 35 & -14.5 & -2.2 & 56 & 6.92 & 8.30 + 36 & -16.0 & -13.1 & 97 & 5.73 & 6.88 + 37 & -18.1 & -8.7 & 56 & 1.30 & 1.56 + 38(c ) & -21.1 & 17.0 & 63 & 0.77 & 0.92 + 39 & -19.6 & -13.1 & 40 & 0.80 & 0.96 + a total of 39 hii regions were identified . they have been detected on the total h@xmath0 map and their presence on at least two consecutive channel maps has been carefully checked . a majority of them are well seen on fig.1 and 2 of ostlin et al . ( 1996 ) and also on fig . 1 of ht95 . observed hii regions are shown in the identification chart ( fig . 2 ) and listed in table 2 which gives the following informations : column 1 lists the reference number given to individual hii regions as shown in the chart . columns 2 and 3 give the x , y positions of the photometric center of the regions , in units of arcsec . the coordinate center is the h@xmath0 brightness peak in the nw component of the galaxy , positive values of @xmath10@xmath0 and @xmath10@xmath11 are to the east and north respectively . the faintest h@xmath0 fluxes correspond to a s / n threshold of 5 . columns 4 and 5 give the effective diameter @xmath17 of the regions in arcsec and parsecs . an equivalent diameter was defined in terms of the limiting isophotal area @xmath44 , of each hii region , according to the definition @xmath45 the limits of each region are defined by the contour where the intensity of the h@xmath0 emission falls down to the average local intensity of the diffuse background . columns 6 and 7 list the total dereddened h@xmath0 fluxes of the regions in units of 10@xmath46ergs @xmath27 and their h@xmath0 luminosities in units of 10@xmath47ergs @xmath1 . the faintest features listed in table 2 have an h@xmath0 flux corresponding to a minimum number of photons that gives a signal - to - noise of 5 . on fig . 2 is marked and listed in table 2 the zwicky s `` flare '' ( zwicky 1966 ) , component c by dh90 . about 1.5 arcsec displacement exists between its red ( table 1 ) and h@xmath0 ( table 2 ) peaks . this displacement also appears on fig . 2 of ostlin et al . ( 1996 ) . in fig . 4 we show the cumulative diameter distribution built according to table 2 . it is consistent with an exponential law @xmath48 = @xmath49exp(@xmath17/@xmath50 ) where @xmath50 = 24 pc . this value is comparable with @xmath50 values found for the irregular galaxies ( hodge 1983 ; hodge et al . 1989 ; hodge & lee 1990 ) . for izw18 the small number of hii regions prevents a derivation of an accurate luminosity function , but nevertheless the data are sufficient to get a first approximation . in fig . 5 the distribution of h@xmath0 luminosities binned in intervals of 1.0 10@xmath47ergs @xmath1 is presented . the observed luminosity function of izw18 shows a turnover with a maximum at @xmath51 = 3.0 10@xmath47ergs @xmath52 corresponding to the luminosity of the orion nebula ( gebel 1968 ) . this turnover is due to incompleteness towards fainter luminosities . at luminosities brighter than this turnoff the observed function is reasonably well fitted by a power law @xmath53=@xmath44@xmath54@xmath55 with a slope @xmath0 = -1.6@xmath560.3 . this is consistent with the mean value derived for dwarf irregular galaxies ( strobel et al . 1991 ) @xmath0 = - 1.5@xmath560.3 . it may be compared with the steeper slope @xmath0 = -1.7 found for larger irregulars by kennicutt et al . ( 1989 ) . figure 6 shows iso - velocity lines superimposed on the h@xmath0 gray scale image of izw18 . figure 7 is a position - velocity cut across the optical major axis of the galaxy across the nw and se compact components ( pa = 143@xmath57 ) . the shape of this position - velocity cut is in good agreement with slit 2 ( pa = 156.1@xmath57 ) position - velocity diagram of m96 . figures 6 and 7 show a clear overall regular radial velocity gradient across the main body of izw18 . this finding obtained from the ionized gas confirms an earlier result by viallefond et al . 1987 using hi aperture synthesis techniques . 7 further shows that the radial velocity gradient is globally consistent with a solid - body - like rotation of the ionized gas . the peaks of h@xmath0 emission in the nw and se components are respectively centered at 739 km @xmath1 and 779 km @xmath1 and their average velocities are equal to 742@xmath567 km @xmath1 and 783@xmath565 km @xmath1 . these values are in agreement with those obtained using high - resolution spectroscopy ( dkf89 ) . they are also very similar to the davidson & kinman ( 1985 ) and hi velocity results ( lequeux & viallefond 1980 ; viallefond et al . 1987 ) . the best symmetrical pattern of the velocity distribution is obtained along the major axis , when assuming a systemic recession velocity of 763 km @xmath1 . this value is in good agreement with the velocity of the number 5 hi cloud ( 765 km @xmath1 ) in lequeux & viallefond ( 1980 ) and relates to a region which is very close to the red continuum component ( a ) . assuming that the radial velocity gradient , ( which amounts to 73 km @xmath58 and is indicating a high mass concentration ) , is due to rotation ( m96 ) we derive a keplerian spherical mass of @xmath59 within a radius about 10 arcsec ( 480pc ) . within this radius , we derive @xmath60 for the mass of ionized gas . it corresponds to 1.6% of the total mass and about 40% of the neutral hi mass of the clump corresponding to the main body of izw18 ( lequeux & viallefond 1980 ; viallefond et al . 1987 ) . within this limited radius , such a simple approximation is sufficient to derive a realistic estimate of the total mass . any other more elaborate model will provide an estinate within a factor of 2 as compared to the keplerian spherical mass . over the regular velocity field of izw18 some complex structure is superimposed that locally modifies the isovelocity line and creates `` bumps '' at the position - velocity cut . these irregularities are particularly conspicuous in the nw region . across a 70x70 pc@xmath29 area located 150pc north from the peak intensity of the nw component we observe a local velocity excess of 15 - 17 km @xmath1 . the position of this excess coincides with the secondary peak in the spectrum of dfk89 . it is also coincident with the location of the red continuum ( b ) component ( fig . 1 , insert ) . this velocity structure has been previously discussed by m96 and by skillman & kennicutt ( 1993 ) . the latter authors suggest a merging of two ( or more ) clouds rather than a pure solid body rotation . we dispute this interpretation using our previous statements on the regularity of the general velocity pattern and further note that the violent star formation that occurs in the nw compact component is likely to induce local peculiar gas motions . other interesting velocity features relate to the ridge . in general the ridge follows the velocity distribution of the hi peak and izw18 velocity field . along the southern and part of the western length of the ridge , differences in radial velocities of the order of 10 - 15 km @xmath1 are found between its inner and outer regions . this difference is also evident across the slit 3 ( pa=78.8@xmath57 ) position - velocity diagram of m96 . no peculiar velocities were detected on the hii regions area of the ridge . coincident with zwicky s `` flare '' a small h@xmath0 knot has a radial velocity of about 730 km @xmath1 . therefore izw18 and the zwicky s `` flare '' belong to the same system . those hii regions detected in the main body of izw18 and its envelope ( sec.3.2 ) with sufficient s / n ratio exhibit radial velocities similar to that of the central velocity field . figure 8 and 9 display each individual h@xmath0 line profile superimposed to the h@xmath0 isophotal map of izw18 in order to show the line profile structure of the outer envelope and the h@xmath0 line intensity distribution across the galaxy . depending upon their location line profiles show different shapes . typical unsmoothed h@xmath0 line profiles are shown in fig . 10 together with the corresponding gaussian fits . in all cases , the observed broad wings have instrumental origin and are most probably caused by adjacent orders . table 3 reports the results of the line profile analysis conducted on areas where the signal - to noise ratio is sufficient to achieve gaussian or multi - gaussian fitting of the observed h@xmath0 line . @xmath61(main ) is the velocity dispersion of the main component of the line , corrected for thermal and instrumental broadening ( sec . 2 ) ; n(sec ) is the number of secondary gaussian profiles fitting the observed h@xmath0 line ; v(sec ) is the peak velocity difference between the main and secondary gaussian components ; i(sec)/i(main ) is the intensity ratio of the secondary and main gaussian components . the numbers in table 3 are average values taking into account several pixels for each region . cccccc & & nw region & se region & w ridge & s ridge + @xmath62 & & @xmath63 & 26.5@xmath560.4 & 16 & 16 + n(sec ) & & 0 & 1 & 2 & 2 + v(sec)(km s@xmath64 ) & @xmath65 & & -53.0@xmath562.7 & 45.3@xmath5611.9 & -40.7@xmath563.2 + & @xmath66 & & & -39.0@xmath567.0 & -76.8@xmath565.1 + i(sec)/i(main ) & @xmath65 & & 0.15 & 0.30 & 0.40 + & @xmath66 & & & 0.20 & 0.20 + the nw component h@xmath0 line profiles mainly appear symmetrical and well - represented by a one - component gaussian fit ( fig . it is interesting to note that in h@xmath0 line a small red asymmetry appears towards the direction of the se component . with the assumption that the component is a gravitationally bound system ( terlevich & melnick 1981 ) the total mass of the system is estimated about 2 10@xmath67@xmath38 and the fraction of ionized gas in it as about 3% . according to skillman & kennicutt ( 1993 ) a broad emission can be seen at the base of the h@xmath0 line profile for the nw component with a peak amplitude at about 0.2% of the main h@xmath0 line and a fwhm of about 80 ( about 3600 km @xmath1 ) . the small free spectral range ( 375 km @xmath1 ) of the fabry - perot interferometer prevents to detect this broad component . all h@xmath0 line profiles observed in pixels belonging to the se component are asymmetric with blue side excesses ( fig . this blueshifted component contributes in average to 13% of the line flux and has a mean velocity of -50 km @xmath1 from the systemic velocity of the main component . again , the small free spectral range of our observations ( 375 km @xmath1 ) forbids the detection of faint large velocity components such as the ones found by m96 with velocities up to 230 km @xmath1 and a an intensity of about 3% of the main line flux . across the western part of the ridge ( fig . 10c ) profiles with blue and red velocities excesses are typical . usually redshifted components are more intense than blueshifted ones . across the southern part of the ridge profiles with strong blue side asymmetry are typical ( fig . two secondary components with 25% and 12% contributions of the line flux and -41 km @xmath1 and -77 km @xmath1 blueshifted from the systemic velocity of the main component were discovered . the bright knot well seen on the m96 echelle spectrum at h@xmath0 along the slit position 1 ( pa=7.7@xmath57 ) ( see her fig . 2 ) is caused by the southern part of the ridge . the two blueshifted components in the h@xmath0 profiles of the ridge are well separated in velocity ( see fig . this is well in agreement with the velocity range displayed in the m96 spectrum observed across the ridge position . on the so called sw doppler ellipse ( m96 ) , nearly 4south from the ridge position , another fainter knot is observed that refers to our hii region number 23 . in the outer enveloppe of the main body of the galaxy , profiles are mainly asymmetric with red wings in the north - west and west of the se component ( m96 ) , and with blue wings elsewere . in the ne doppler ellipse of m96 ( her fig . 2 ) , three other hii regions from our list have been identified as they appear as individual knots : number 9 and number 10 merged with number 12 . all three hii regions have h@xmath0 line profiles with blueshifted components . the question still remains whether izw18 is an old system in which some intense episodes of star formation already occured , or whether it forms stars for the very first time ? loose & thuan ( 1986 ) reported observations of a faint outer component with an extended stellar continuum which they attributed to an underlying old population . thuan ( 1983 ) detected infrared emission , while hua et al . ( 1987 ) , dfk89 , dh90 from red continuum imagery , confirmed the presence of this component . the implication is that izw18 may be an older and more complex system . in the following , we make the assumption that h@xmath0 peaks relate to the current star forming regions while red continuum features correspond to older sites of star formation . from the existence of the h@xmath0 loops and filaments ht95 also suggested the existence of intermitent star formation episodes in the central and northern parts of the galaxy . their positions are in good agreement with our ( a ) and ( b ) red components . the distribution of the clumps in the main body of izw18 shows a chain like structure . out of seven identified current or more evolved star - forming knots six are distributed along the line se,(a),nw,(b ) . this chain of star forming knots is positioned along the direction of elongation of the main body of izw18 and is consistent with hi outer clumps distribution ( lequeux & viallefond 1980 ) . only knot ( c ) is displaced from this linear distribution . in the chain both nw and se currently active star forming regions are located between more evolved star forming sites . is this geometry a random configuration ? does it examplify the sequential nature of star formation events in the galaxy ( gerola et al . 1980 ; kunth et al . clearly the chain although well ordered spatially does not define any prefered age sequence . stochastic self - propagating star formation simulations of star formation sites in dwarf galaxies show that the properties of dwarf systems depend on the size ratio of the galaxy to the star formation cells . small galaxies similar to izw18 evolve mainly via a series of disconnected star forming bursts ( e.g. hunter & gallagher 1985 ) . therefore , a stochastic origin of the observed `` chain '' is more plausible ( see also ht95 ) . the ridge discovered by dh90 is thought to represent a radiation - bounded ionization front driven into the main hi cloud . our red continuum data dispute this claim . indeed the discovery of some stars in the ridge and surrounding regions ( ht95 ) shows that the ridge contains a stellar population that can ionize the gas `` in situ '' . therefore the ridge is an isolated structure in the w - sw part of the double - peaked hi gas condensation and has no direct morphological link with the main body of izw18 . extended h@xmath0 emission have been detected up to distances of 1.5 kpc from the main body of izw18 . dh90 , who noted the clumpy nature of this extended h@xmath0 emission , suggested that this emission could be powered by the central nw and se star forming complexes . m96 using imagery and high dispersion spectroscopy advocate that both north - east and south - west extensions of the outer envelope of the galaxy could be superbubbles powered by star forming complexes . however the existence of clumpy red continuum emission of stellar origin coeval with the extended h@xmath0 emission , as well as the number of stars which were discovered in the same regions ( ht95 ) consisting into compact and well individualized clumpy structure allow us to suggest that a stellar population is embedded into the ionized gas . therefore part of the young stars from this population can ionize and produce the observed extended hii emission . as noted in sec.3.3 , the diameters distribution and a tentative h@xmath0 luminosity function of these hii regions are consistent with those of normal hii regions in dwarf irregular galaxies . the existence of the stellar and hii regions populations in ne and sw extensions of the outer envelope of izw18 does not preclude the contribution of some superbubbles in the same regions . as previously suggested , star formation in the main body of izw18 possibly occurs via a series of disconnected bursts as a result of density fluctuations in the interstellar medium near the hi clouds . what is the mechanism which triggers the subsequent star formation ? the velocity field in an hii galaxy as well as in its surrounding hi cloud may indeed play an important role ( saito et al . 1992 ; tomita et al . 1993 ) . the velocity gradient observed along the main star forming regions , both in h@xmath0 ( our results in sec.3.4 ) and in hi ( from the maps of viallefond et al . 1987 ) is equal to about 70 km @xmath68 . the same gradient across the offseted hi peak distribution is only 50 km @xmath68 ( viallefond et al . it means that local hii as well as hi gas in the region of the optical galaxy rotates faster than the hi gas at its peak distribution . the ordered component of the izw18 velocity field shows a peak orbital velocity of about 50 km @xmath1 . most irregular galaxies ( giants as well as dwarfs ) , exhibit a solid body rotation with about the same peak orbital velocities of 5070 km @xmath1 but with shallow velocity gradients of 520 km @xmath68 over the optically visible region ( hopp & schulte - ladbeck 1991 ; shostak & skillman 1989 ; gallagher & hunter 1984 and literature therein ) . the much higher velocity gradient in izw18 arises from its much smaller linear scale , when compared to normal irregulars . the disordered motions superimposed over the regular velocity field of izw18 appear as sharp velocity jumps of 10 to 20 km @xmath1 occuring on scales of about 500 pc@xmath69 . these appear related to the main body of the galaxy , and suggest that the gas is not in equilibrum locally . corresponding dynamical timescales could be of the order of few 10@xmath70years ( ht95 ) . the gravitational effects of the ( b ) and ( c ) star forming centers as well as explosive star forming events may explain these local disturbances . as regards to h@xmath0 line profiles , the discrepancy between a single gaussian component with velocity dispersion of 23 km @xmath1 ( e - folding width : 33 km @xmath1 ) observed in a 40 arcsec and 21 arcsec apertures by arsenault & roy ( 1986 ) and the complex line structure seen at our resolution of 0.725 per pixel is not a surprise . gaussian - like smooth profiles obtained by arsenault & roy ( 1986 ) result from integration over many distinct components with different space velocities and emissivities . as examples we can recall high resolution observations of ngc604 by rosa & solf ( 1984 ) , hippelein & fried ( 1984 ) in comparison with results for ngc604 presented by arsenault & roy ( 1986 ) and also the very interesting case of the giant hii region ngc2363 ( arsenault & roy 1986 ; roy et al . 1991 ; roy et al . 1992 ) . the single gaussian component observed in the nw compact hii region may suggest a self- gravitating system ( terlevich & melnick 1981 ) . in this case the expected velocity dispersion ( melnick et al . 1988 ) is about twice smaller than the observed one . this difference can be caused either by the gravity of the nearby located recent star forming nw and ( b ) regions or / and the effects of turbulence . the chaotic morphology of the ionized gas in the component ( ht95 ) as well as the existence of high velocity gas expanding shell ( m96 ) may favor the latter effects . application of gravitation - turbulence model in this case would be more reliable ( arsenault & roy 1988 ) . in the se compact hii region the velocity dispersion for the main gaussian component ( about 19 km @xmath1 ) is also higher than expected from a self - gravitational model ( terlevich & melnick 1981 ) . such an excess velocity dispersion can also be caused by the peculiar motions of several stellar clusters observed in the core of this component ( ht95 ) . the weaker second gaussian component may be produced by another isolated hii region in the se area , undistinguishable because of projection effects . the ridge which has no direct morphological connection with the main body of izw18 broadly follows the velocity field structure of the hi peak . the ridge is therefore co - rotating with the hi cloud and the main body of izw18 . the ridge as an isolated configuration is probably not in equilibrium . it expands in an opposite direction with respect to the hi gas showing local well determined infall and/or outfall of ionized gas . the geometry as well as the velocity field of the ridge are consistent with the `` blister '' interpretation . the complex structure of the h@xmath0 line profiles observed in the envelope showing a variable position - dependent asymmetry ( blue and/or red wings ) suggests that we are observing gas infall onto the main body of the galaxy , whatever the direction of observation and/or local ionization fronts . the blue asymmetric components are much more prominent on the eastern side , opposite to the major hi concentrations . it may be evidence that infall of gas clouds fuels the overall star formation in izw18 . but we do not exclude the possibility that the arc - like structure of the ridge can be affected by a tidal interaction between izw18 and the zwicky s `` flare '' . from our data and data from the literature we examine the nature of izw18 following the hypothesis first introduced by searle & sargent ( 1972 ) and further discussed by kunth et al . ( 1986 ) , vigroux et al . ( 1986 ) , thuan ( 1987 ) that in general bcdgs may be considered as the low luminosity end of irregular galaxies . in table 4 global characteristics and parameters related to star formation properties in izw18 are compared with those of dwarf irregular galaxies . ccccc & izw18 & ref & dwarf irr & ref + m(b ) & -13.9 & huchra ( 1977 ) & -14.5 & hunter & gallagher ( 1985 , 1986 ) + @xmath71 & 0.9 & huchra ( 1977 ) & 5.9 & hunter & gallagher ( 1985 , 1986 ) + @xmath72 ) & 21.3 & huchra ( 1977 ) & 23.8 & hunter & gallagher ( 1985 , 1986 ) + @xmath73 & 7.8 & lequeux & viallefond ( 1980 ) & 8.3 & hunter & gallagher ( 1985 , 1986 ) + @xmath74 ) & 50 & this paper & 60 & gallagher & hunter ( 1984 ) + @xmath75 ) & 73 & this paper & 10 & gallagher & hunter ( 1984 ) + u - b & -0.76 & huchra ( 1977 ) & -0.44 & hunter & gallagher ( 1985 , 1986 ) + b - v & 0.17 & huchra ( 1977 ) & 0.36 & hunter & gallagher ( 1985 , 1986 ) + largest hii reg.(pc ) & 170 & this paper & 300 & hunter & gallagher ( 1985 ) + @xmath76 & 24 & this paper & 25 & hodge ( 1983 ) + @xmath0 & -1.6 & this paper & -1.5 & strobel et al . ( 1991 ) + @xmath77 & 0.17 & dufour et al . ( 1988 ) & 4.43 & hunter & gallagher ( 1985 , 1986 ) + @xmath78 & @xmath79 & this paper & @xmath80 & hunter & gallagher ( 1985 ) + @xmath81 & 9 & this paper & 1.5 & hunter & gallagher ( 1985 ) + @xmath82 & @xmath83 & this paper & @xmath84 & hunter & gallagher ( 1985 ) + the results of comparison may be summarized as follows . izw18 has an absolute luminosity consistent with that of an `` average '' dwarf irregular . the small linear size of izw18 accounts for the high surface brightness of the optically visible component . hi masses of izw18 and dwarf irregular galaxies of comparable blue luminosity and the clumpy structure of the hi distribution are similar . as for izw18 , star formation in dwarf irregulars is not simply related to the projected hi gas density ( e.g. allsopp 1978 ) . \3 . a rigid body rotation which is typical for the majority of dwarf irregular galaxies with similar peak orbital velocity is observed in izw18 . the size of izw18 makes the velocity gradient across it at least 5 times higher than in irregular galaxies . \4 . except for two giant star forming sites , izw18 contains a population of faint hii regions . this population by its luminosity and diameter distribution does not differ from the hii region population in dwarf irregular galaxies . \5 . in izw18 besides an extended underlying older stellar population ( loose & thuan 1986 ) we report the existence of recent star forming clumps . these regions also are important components in dwarf irregular galaxies ( hunter & gallagher 1986 ) . \6 . by their very blue color , high h@xmath0 luminosity and surface brightness and their high star formation rate per unit area , the current star forming regions in izw18 significantly differ from those in dwarf irregular galaxies . we may conclude that izw18 share many characteristics in common with dwarf irregulars except for its much higher level of star forming activity per unit area . the large velocity gradient together with the locally complex velocity field ( including peculiar gas motions responsible for the line profile asymetries ) are most probably responsible for the star formation itself . from the existence of recent star forming regions seen in continuum , with no exact spatial coincidence with hii regions , we can possibly conclude that we are observing a major second burst of star formation in this object ( e.g. kunth et al . the first burst was completed about few tenth 10@xmath85 years ago ( e.g. ht95 ; m96 ) and the present one is extremely recent ( kunth & sargent 1986 ) . the age of the present burst may be estimated as no more than a few 10@xmath85 years ( lequeux et al . 1981 ; copetti et al . 1985 ) . the main results of h@xmath0 scanning fabry - perot investigations of izw18 may be summarized as follows : + - in the main body of this galaxy , the _ current _ sites of active star formation , identified from the presence of prominent emission lines have ages of few 10@xmath70 years and are spatially separated from red continuum clumps which are sites of recent star formation episodes with ages of about few tenth of 10@xmath70 years . all star formation sites in the galaxy are separated from the compact hi cloud double - peaked core . - the h@xmath0 line profiles have a complex structure , and except in the nw hii compact region , can not be fitted by single gaussian components . the presence of blue and/or red wings in many points of the envelope suggests a generalized pattern of infalling ( expelling ) gas clouds onto ( from ) the central star forming regions . - the observed `` ridge '' is an isolated morphological structure in the system . it is a star forming site separated with respect to the izw18 main body by the hi cloud . the shape and orientation of this ridge can be due to tidal interation between izw18 and the zwicky s `` flare '' . - the sw and ne extensions of the outer envelope of the galaxy besides the ionized gas component connected to a supergiant shell - expanding superbubble ( m96 ) also contain stellar as well as hii regions components . - the near environment of the galaxy is likely to be polluted not only with gas processed by the central massive stars and driven out by superwinds but also by products of a more local stellar population . when compared to the average characteristics of dwarf irregular galaxies , the basic properties of izw18 ( luminosity , hi mass , solid body rotation , normal hii region population , existence of current and recent star formation sites ) are consistent with the hypothesis of an overall similarity between this blue compact dwarf galaxy and a typical dwarf irregular . the rigid - body rotational velocity gradient and the blue surface brightness , which depend on the linear scale are much larger in izw18 . the essential difference is that a high star formation activity ( an order of two magnitudes in the sf rate per unit area which respect to that observed in a typical dwarf irregular ) is concentrated in a small compact region . @xmath86 we would like to thank dr . deidre hunter for providing her hst - wfpc2 data for izw18 and dr . miguel mas - hesse for his model calculations of stellar contribution at 6600 with respect to the nebular continuum . part of this research was supported by pics ( programme international de cooperation scientifique ) # 247 of cnrs between france and armenia . the financial support of international association for the promotion of cooperation with scientists from the independant states ( intas ) is gratefully acknowledged . petrosian would like to thank the warm hospitality of institut dastrophysique de paris and observatoire de marseille during his stay . we also thank jean mouette for his help during the preparation of the figures . allsopp , n.j . 1978 , mn , 184 , 397 arnault , p. , casoli , f. , combes , f. , kunth , d. 1988 , a&a , 205 , 41 arsenault , r. , roy , j .- 1986 , aj , 92 , 567 arsenault , r. , roy , j .- 1988 , a&a , 201 , 199 bertola , f. , casini , c. , bettoni , d. , galletta , g. , noreau , l. , kronberg , p.p . 1984 , aj , 89 , 350 boulesteix , j. , georgelin , y.p . , marcelin , m. , monnet , g. 1983 , spie conf . astron . , 5 , 445 , 37 copetti , m.v.f . , pastoriza , m.g . , dottori , h.a . 1985 , a&a , 152 , 427 davidson , k. , kinman , t.d . 1985 , apjs , 58 , 321 davidson , k. , kinman , t.d . , friedman , s.d . 1989 , aj , 97 , 1591 ( dfk89 ) dopita , m.a . 1972 , a&a , 17 , 165 dufour , r.j . 1986 , pasp , 98 , 1025 dufour , r.j . hester j.j . , 1990 , apj , 350 , 149 ( dh90 ) dufour , r.j . garnett d.r . , shields g.a . , 1988 , apj , 332 , 752 gallagher , iii j.s . , hunter , d.a . 1983 , apj , 274 , 141 gallagher , iii j.s . , hunter , d.a . 1984 , ara&a , 22 , 37 gebel , w.c . 1968 , apj , 153 , 743 gerola , h. , seiden , p.e . , schulman , l.s . 1980 , apj , 242 , 517 filippenko , a.v . 1982 , pasp , 94 , 715 hippelein , h. , fried , j.w . 1984 , a&a , 141 , 49 hodge , p.w . 1983 , aj , 88 , 1323 hodge , p.w . , lee , m.g . 1990 , pasp , 102 , 26 hodge , p.w . , lee , m.g . , kennicutt , jr . 1989,pasp , 101 , 32 hopp , u. , schulte - ladbeck , r.e . 1991 , a&a , 248 , 1 hua , c.t . , grundseth , b. , nguyen - trong , t. 1987 , ap let . comm . , 25 , 187 huchra , j.p . 1977 , apjs , 35 , 171 hunter , d. a. 1982 , apj , 260 , 81 hunter , d.a . , gallagher , iii j.s . 1985 , apjs , 58 , 533 hunter , d.a . , gallagher , iii j.s . 1986 , pasp , 98 , 5 hunter , d.a . , thronson , jr . h.a . 1995 , apj , 452 , 238 ( ht95 ) kennicutt , jr . r.c . , edgar , b.k . , hodge , p.w . 1989 , apj , 337 , 761 klein , u. , grave , r. , wielebinski , r. 1983 , a&a , 117 , 332 kunth , d. , sevre , f. 1986 , star forming dwarf galaxies , eds . d.kunth , t.x.thuan , j.t.t.van , ( editions frontieres ) p.331 kunth , d. , sargent , w.l.w . 1986 , apj . , 300 , 496 kunth , d. , maurogordato , s. , vigroux , l. 1988 , a&a , 204 , 10 kunth , d. , lequeux , j. , sargent , w.l.w , viallefond , f. 1994 , a&a , 282 , 709 kunth , d. , matteucci , f. , marconi , g. 1995 , a&a , 297 , 634 laval , a. , boulesteix , j. , georgelin , y.p . , georgelin , y.m . , marcelin , m. 1987 , a&a . , 175 , 199 lequeux , j. , viallefond , f. 1980 , a&a , 91 , 269 lequeux , j. , peimbert , m. , rayo , j.f . , serrano , a. , torres- peimbert , s. 1979 , a&a , 80 , 155 lequeux , j. , maucherat - joubert , m. , deharveng , j.m . , kunth , d. 1981 , a&a , 103 , 305 lequeux , j. , kunth , d. , mas - hesse , j.m . , sargent , w.l.w . 1995 , a&a , 301 , 18 loose , h .- h . , thuan , t.x . 1986 , star forming dwarf galaxies , eds . d. kunth , t.x.thuan , j.t.t.van , ( editions frontieres ) , p. 73 martin , c. 1996 , apj , in press ( m96 ) mazzarella , j.m . , boroson , t.a . 1993 , apjs , 85 , 27 mazzarella , j.m . , bothun , g.d . , boroson , t.a . 1991 , aj , 101 , 2034 melnick , j. , terlevich , r. , moles , m. 1988 , mn , 235 , 297 ostlin , g. , bergvall , n. , ronnback , j. 1996 , in `` the interplay between massive star formation , the ism and galaxy evolution '' , ed.kunth d. , guiderdoni b.,heydari - malayeri t.x . thuan , ( editions frontieres ) , in press osterbrock , d.e . 1989 , astrophysics of gaseous nebulae and active galactic nuclei ( university science books : mill valley , ca ) pettini , m. , lipman , k. 1995 , a&a , 297 , l63 petrosian , a.r . , saakian , k.a . , khachikian , e.e . 1978 , afz , 14 , 69 rosa , m. , solf , j. 1984 , a&a , 130 , 29 roy , j .- r . , boulesteix , j. , joncas , g. , grundseth , b. 1991 , apj , 367 , 141 roy , j .- r . , aube , m. , mccall , m.l . , dufour , r.d . 1992 , apj , 386 , 498 saito , m. , sasaki , m. , ohta , k. , yamada , t. 1992 , pasj , 44 , 593 sargent , w.l.w . , searle , l. 1970 , apj , 162 , l155 searle , l. , sargent , w.l.w . 1972 , apj , 173 , 25 shostak , g.s . , skillman , e.d . 1989 , a&a , 214 , 33 skillman , e.d . , kennicutt , jr.r.c . 1993 , apj , 411 , 655 strobel , n.v . , hodge , p. , kennicutt , jr.r.c . 1991 , apj , 383 , 148 taylor , c.l . , skillman , e.d . , brinks , e. 1991 , baas , 23 , 920 terlevich , r. , melnick , j. 1981 , mn , 195 , 839 thuan , t.x . 1983 , apj , 268 , 667 thuan , t.x . 1987 , starbursts and galaxy evolution , eds . , t.montmerle , t.x.thuan , j.t.t.van , ( edition frontieres ) , p.129 thuan , t.x . , williams , t.b . , malumuth , e. 1987 , star bursts and galaxy evolution , eds.,t.montmerle , t.x.thuan , t.t.van,(editions frontieres ) , p.151 tomita , a. , ohta , k. , saito , m. 1993 , pasj , 45 , 693 tremaine , s. , gunn , j.e . 1979 , phys . rev.letters , 42 , 407 viallefond , f. , thuan , t.x . 1983 , apj , 269 , 444 viallefond , f. , lequeux , j. , comte , g. 1987 , star bursts and galaxy evolution , eds . t. montmerle , t.x . thuan , t .- t.van , ( editions frontieres ) , p.139 vigroux , l. , stasinska , g. , comte , g. 1986 , star forming dwarf galaxies , eds d. kunth , t.x . thuan , j.t.t . van , editions frontieres , p. 425 zwicky , f. 1966 , apj , 143 , 166

synfire chains are thought to underlie precisely - timed sequences of spikes observed in various brain regions and across species . how they are formed is not understood . here we analyze self - organization of synfire chains through the spike - timing dependent plasticity ( stdp ) of the synapses , axon remodeling , and potentiation decay of synaptic weights in networks of neurons driven by noisy external inputs and subject to dominant feedback inhibition . potentiation decay is the gradual , activity - independent reduction of synaptic weights over time . we show that potentiation decay enables a dynamic and statistically stable network connectivity when neurons spike spontaneously . periodic stimulation of a subset of neurons leads to formation of synfire chains through a random recruitment process , which terminates when the chain connects to itself and forms a loop . we demonstrate that chain length distributions depend on the potentiation decay . fast potentiation decay leads to long chains with wide distributions , while slow potentiation decay leads to short chains with narrow distributions . we suggest that the potentiation decay , which corresponds to the decay of early long - term potentiation of synapses ( e - ltp ) , is an important synaptic plasticity rule in regulating formation of neural circuity through stdp .

1307.7785

coevolution of the dynamics and topology of networks is widely observed in diverse systems from cellular biology to social networks @xcite . in the brain , the spiking dynamics of neurons depends on how they are connected . on the other hand , the connectivity can be modified by the spiking activity . the connections ( synapses ) between neurons in many brain areas are modified according to the spike - timing - dependent plasticity ( stdp ) rule @xcite . a most common stdp rule for excitatory neurons is as follows @xcite : the connection strength ( synaptic weight ) from neurons a to b is strengthened if a spikes before b ( long - term potentiation , or ltp ) , and weakened if a spikes after b ( long - term depression , or ltd ) . the amount of modification decreases with the time difference between the spikes of a and b. the connection between two neurons is lost if the synaptic weight is reduced below a threshold ; conversely , it can be established through consistent parings of the spikes of the neurons . studies that use stdp in spiking neural networks have shown a number of emergent properties @xcite . in this paper , we show that synfire chain connectivity @xcite , in which subsequent groups of neurons are connected into a feedforward network that supports sequential spiking of the neurons , emerges through coevolution of the spiking activity and the connectivity across many presentations of a training stimulus to a subset of neurons ( training neurons ) . sequential spiking of neurons is observed in a number of brain areas @xcite . some of the strongest experimental evidence for synfire chains producing spike sequences is from zebra finch premotor nucleus hvc ( proper name ) @xcite . projection neurons in hvc spike sequentially at precise times relative to the learned song @xcite . consistent with the synfire chain dynamics , cooling hvc uniformly slows down the song @xcite , and the sub - threshold membrane potentials of neurons rapidly depolarize 5 - 10 ms before they spike @xcite . it is well - established that synfire chains robustly produce spike sequences @xcite . however , how neurons are wired into synfire chains is not well - understood . an intriguing possibility is that synfire chains self - organize through activity - driven synaptic plasticity . earlier studies using stdp or similar hebbian rules resulted in short chains with a few groups @xcite . the most likely reason is that these rules are prone to producing unstable growth of connections @xcite . two recent studies introduced additional homeostatic synaptic rules to limit such instability , and showed that long synfire chains can form @xcite . the key idea behind both studies is to restrict the connectivity of the network after a certain amount of growth has occurred . fiete et al achieved this by limiting the total synaptic weight in and out of every neuron @xcite . however , a study using large scale simulations and mean - field analysis suggested that the regulation of the total synaptic weights does not prevent unstable network growth @xcite . in @xcite ( jun - jin model ) , we took a different approach , and imposed an axon remodeling rule that limits the number @xmath0 of strong connections , defined as those with synaptic weights above a threshold , that one neuron can maintain . reaching the limit leads to pruning of all weak connections from the neuron . there are two additional features of the jun - jin model . first , the model includes an gradual , activity - independent decay of synaptic weights , which we call potentiation decay . second , an activation threshold switches synapses on or off depending on the magnitude of the synaptic weights . this rule allows the active connections between neurons to form or disappear as the synaptic weights are modified . simulations of 1000 leaky integrate - and - fire neurons showed that synfire chains emerge from initially random active connections when 6 to 40 training neurons are intermittently activated by external inputs for many trials . the number of neurons in each group roughly equals to @xmath0 ( set to 10 in the simulations ) , and is not affected by the number of training neurons except for the first 2 - 3 groups @xcite . in this paper , we perform an in - depth analysis of the jun - jin model . we address unresolved fundamental questions such as what determines the lengths of the emergent chains and how the length distributions are influenced by the total number of the neurons ( network size ) . we establish that , when the network is randomly active , the synaptic plasticity rules in this model allow the network connectivity to fluctuate , but the synaptic weights remain in a statistically stationary distribution . this ensures that the chain formation does not depend on specific initial network connectivity . we demonstrate that in between training trials , sequential spikes can spontaneously emerge in the forming chain . this noise - induced re - activation of the chain creates connections from neurons outside of the chain to those in the chain , and plays a critical role in determining the length distributions of the final chains . most notably , there is an upper limit for the mean chain length as the network size becomes large . we show that slow potentiation decay leads to short chains with narrow length distributions , while fast potentiation decay leads to long chains with a wide length distributions . we compare the results of network simulations to a lottery - type stochastic process in which neurons are selected iteratively to enter a chain , and the chain stops growing when a loop is formed , either by selecting neurons already in the chain or by selecting neurons connected to the chain . the distribution of chain lengths from the network simulations fits well with distributions generated by the lottery process . the analysis of this simple growth model shows that the rate of potentiation decay influences the chain length distributions by controlling the emergence of the connections to the growing chain . these connections also lead to a finite limit for the mean chain length as the network size increases . we simulate synfire chain formation in recurrent networks of excitatory neurons . we model each neuron using a leaky integrate - and - fire ( lif ) model . the neurons interact via pulse conductances , and they receive dominant feedback inhibition . they are also driven by upstream regions that we do not simulate , but instead model as independent , fluctuating external inputs . synaptic weights between neurons are modified according to an stdp rule . the details of modeling can be found in the appendix . our model choices were dictated by necessity to simulate the network dynamics quickly @xcite . a large number of training trials ( @xmath1 to @xmath2 ) are required for synfire chain formation in our model , and many training sessions are needed to construct the chain length distributions for a range of model parameters . therefore it was necessary that our simulation algorithm be efficient . we modified a fast , event - driven algorithm @xcite that had been developed to generate activity of pulse - coupled neurons that are targeted by a fluctuating external input . when the external input is modeled by gaussian white noise ( gwn ) , one can numerically solve the fokker - planck equation @xcite , store particular solutions in `` lookup tables '' and sample them during the network simulation to generate spike times . the steps of the algorithm are detailed in the appendix . the computational advantage of using pulse - coupled neurons is that the response of the membrane potential is instantaneous and can be calculated exactly . the time - evolution of the membrane potential between spikes is calculated from the lookup tables . by our measurements , this algorithm is up to 150 times faster than simulating with 4th - order runge - kutta method @xcite . instead of scaling with number of timesteps , the simulation time scales with number of spikes , which results in increased simulation speed . two differences distinguish our simulation algorithm from that reported in @xcite . first , we simulate conductance - based neurons instead of current - based neurons . second , instead of a spike latency , we impose a time resolution on the arrival times of spikes , as suggested in @xcite . algorithmically , imposing a time resolution on the spike arrival times means that , instead of the neuron with the single earliest predicted time emitting a spike , any neurons that spike within an interval @xmath3 of the earliest predicted spike time effectively spike together . the arrival time of the spikes at synapses is picked to be at the end of the resolution interval . this method has the same effect as a random latency , allowing neurons to cooperate to excite common targets , but it requires no additional queuing of events , which can be computationally intensive and slows simulation considerably @xcite . the population of neurons we simulate make excitatory connections to each other . however , we assume a population of interneurons targets the excitatory population , and these neurons reliably spike immediately when the excitatory neurons spike . all of the neurons in the excitatory population are inhibited at the end of the resolution interval . near - global inhibition is observed in neocortical circuits @xcite and in the songbird premotor area hvc @xcite . we do not simulate interneuron activity in order to conserve computational resources . the inhibition is stronger when more neurons spike within the spike resolution window , but we put a constraint on it based on the assumption that there is a finite size to the interneuron population targeting the simulated neurons @xcite . details are left to the appendix . the synaptic weight between each pair of simulated neurons is modified based on the stdp rule ( details are in the appendix ) . three additional synaptic plasticity rules are implemented to deter unchecked synaptic strengthening that stdp alone can lead to @xcite , and to ensure stable synaptic weight distribution when the network is in the state of spontaneous activity with no training stimulations . _ activation threshold : _ silent synapses are those with no post synaptic ampa receptors @xcite . at physiological conditions , these synapses do not produce responses in the postsynaptic neuron , hence ltp can activate silent synapses to become functional synapses @xcite ; conversely , ltd can silence active synapses @xcite . abundant especially during the development , silent synapses allow the possibility of sculpting wide variety of neural circuits through neural activity . we model silencing and activation of synapses by thresholding the synaptic weights . if synaptic weight from neuron @xmath4 onto neuron @xmath5 , @xmath6 , grows larger than @xmath7 , then it is active and evokes a response from its target ; otherwise it is silent and behaves as if it has zero weight . for our simulations , we pick @xmath8 . regardless of whether a synapse is silent or active , it obeys the stdp rule . our results do not depend on synaptic depression acting on silent synapses . active synaptic connectivity _ in vivo _ is sparse with a single neuron connecting to less than 10% of its neighbors @xcite . the activation threshold directly avoids densely connected network states by deactivating all sufficiently weak connections . a synapse between any pair of neurons can be activated , so activity can drive the development of any possible synaptic connectivity . in other words , there is no _ a priori _ restriction on how neurons can be connected after training . parameters are selected such that the connectivity remains within the sparseness bounds observed experimentally . a common modeling approach for avoiding dense connectivity is to specify a sparse connectivity between neurons and allow only synaptic weights of these connections to change @xcite . in this strategy no new connections can form , and the effects of training on the connectivity is much more restricted than in our model . _ potentiation decay : _ in addition to the activation threshold , a potentiation decay is applied to the weights of all synapses , amounting to a slow memory leak within the system . the decay is activity - independent and is implemented as a rescaling of all synaptic weights @xmath9 , where @xmath10 , as in previous phenomenological synaptic growth models @xcite . long - term potentiation of synapses usually decays to baseline within three hours , which is called the early ltp ( e - ltp ) @xcite . we assume that the reduction of the synaptic weight during the trial time is insignificant given the typical three - hour time scale of potentiation decay ; therefore , weight rescaling is applied between consecutive two - second training trials , but not during the trial interval . implementing the rescaling during trials is computationally intensive and produces no observable differences . in this simplified model of potentiation decay , all synaptic weights are subjected to the weight rescaling between each trial , including weights of silent synapses . the decay of the silent synapses is important for our model . consider what happens to synaptic weights that become deactivated due to either potentiation decay or synaptic depression . if synaptic depression were the only mechanism that modifies the weights of silent synapses , then a deactivated weight may remain close to the activation threshold . this would lead to an accumulation of weights near the threshold that require a small increase in order to become active . on the other hand , if deactivated synapses have their weights immediately set to zero after deactivation , then a synapse that is consistently active , but does not evoke a spike from its target because of noisy fluctuations over a few consecutive training trials , is immediately destroyed . choosing to implement a decay of the silent synaptic weights is proposed as a balanced solution to these two scenarios . the decay of silent synapses can be related to gradual elimination of spines observed on dendrites @xcite . its biological mechanism is most likely different from the decay of e - ltp . we apply the same decay rule for both silent and active synapses for the sake of simplicity . the details of how silent synapses decay do not matter . the functional role of the potentiation decay in our growth simulations is to regulate runaway synaptic growth . we will demonstrate that , in combination , the activation threshold and the potentiation decay have a stabilizing effect on the growth of the network . _ axon remodeling : _ synaptic weights are clipped if they are strengthened above a threshold @xmath11 ( see the appendix ) . however , this does not limit the number of strong synapses that approach the strongest allowed weight . another mechanism , axon remodeling , regulates the number of strong synapses a neuron can maintain with limited resources available . limiting the number of strong synapses stabilizes an emerging synaptic topology of strong synapses . if axon remodeling were not imposed on each neuron s axonal tree , a well - connected neuron would continue to accrue targets . neurons _ in vivo _ are observed sending out many axons during development , then retracting most and maintaining only the strongest @xcite . a small number of strong synapses in a network have also been measured in experiments @xcite . a slower potentiation decay is also applied to the strongest synapses , resulting in further stabilization . axon remodeling is implemented with the following rules , which are nearly identical to those in @xcite . 1 . a second threshold , @xmath12 , in addition to the active threshold , is introduced within the range of allowed synaptic strengths . weights greater than this value characterize a strong active synapse , which we deem a _ supersynapse_. supersynapses elicit spikes reliably from a target despite the noisy fluctuations of the membrane potential . the supersynapse threshold is greater than the active threshold : @xmath13 . a limit , @xmath0 , is imposed on the number of neurons that a presynaptic neuron contacts along supersynapses . this is the maximum number of axons a neuron can maintain with its limited resources . when this number of supersynapses is attained , the neuron is said to be `` saturated . '' once a neuron is saturated , the stdp rule is only applied to its supersynapses . after saturation all synaptic weights continue to decay . the potentiation decay reduces the weights of non - supersynapses , and as a result they will eventually approach zero with no opportunity to be potentiated unless the neuron de - saturates . supersynapses are reinforced by repeated ltp ; without regular reinforcement , potentiation decay can cause de - saturation and all connections will undergo stdp again . if de - saturation occurs frequently , no stable synaptic structure emerges . high membrane potential variability reduces the frequency of ltp at a supersynapse because higher noise reduces reliability of a supersynapse to produce a spike from its target . in order to ensure ltp occurs frequently enough to overcome the potentiation decay , in all simulations , we apply a slower potentiation decay @xmath14 to supersynapses of a saturated neuron . this corresponds to the slower decay of the late phase ltp ( l - ltp ) compared to e - ltp @xcite . axon remodeling and the synaptic cap are non - essential to stability of the weight distribution of a network before training ; they are only necessary when the network is presented with a stimulus . in the next section , the roles of of axon remodeling will be further articulated where the training regimen is described . network training is broken into a series of identical trials . supersynapses can emerge within a network as it is presented repeatedly with a training stimulus . we model this stimulus with a short , strong excitatory current onto a small subset of training neurons . the training excitation originates in an upstream brain area , possibly one processing sensory stimuli . training continues until the number of supersynapses contained in the network stabilizes . a trial commences with the presentation of the training signal to the training neurons . the signal is modeled by a strong external drive biasing the training neurons to spike within several milliseconds of the beginning of the trial . after 8 milliseconds the driving current onto the training neurons returns to its baseline value . the spontaneous activity and synaptic weight dynamics are simulated for one second after the training signal is withdrawn . the trial ends after this specified trial time and an inter - trial interval commences , which we do not simulate . during this period , which is assumed much longer than one second , synaptic weights are reduced by the potentiation decay factor @xmath15 and the membrane potentials are randomized . training is repeated until the number of supersynapses reaches a stable value for 2,500 trials ; this may take as few as 5,000 trials up to 100,000 trials depending on the size of the network and learning scale factor ( see appendix ) . we will show in the next section that the training neurons form the seed for development of a synaptic chain of neurons connected by supersynapses . synfire chain growth in response to training is governed by stochastic selection of post synaptic targets until a loop forms and the growth stops . repeated stimulations of the training neurons change the synaptic weight distribution and produce strong , stable synaptic chain connectivity . chain growth emanates from the training neurons . neurons that spontaneously spike shortly after the training neurons may be targeted by the training neurons due to the stdp rule . since the training neurons spike synchronously , they make convergent connections to the same set of neurons . subsequent training strengthens these connections until the the synapses become supersynapses . once supersynapses develop , reliable spikes can be evoked in these targets on nearly every trial . when this is the case , we say that the targets have been _ recruited_. the cooperation via the convergent synapses is important for the targets to overcome membrane noise . axon remodeling restricts the number @xmath0 of supersynapses that one neuron can maintain . consequently , the number of recruited neurons is close to @xmath0 regardless of the number of the training neurons , although some fluctuations exist due to the noise in the recruiting process . the recruitment process continues as the second group accrues their own targets via the same cooperative process . new groups are recruited until previously - recruited neurons are recruited again , forming a closed loop . this stochastic , iterative process yields stable synfire topologies that produce long , stereotypical sequences of spikes @xcite . the chains consist of an introductory sequence that begins with the training neurons , which feeds a loop of strong synaptic connectivity , examples of which are displayed in fig . [ chain_sam ] . the network structure is clearly reflected in raster plots of the activity of the population during a typical trial after the chain is fully formed , as shown in fig . [ raster ] . the length of the chain formed by this process varies from trial to trial , and depends on the values of synaptic plasticity parameters and the size of the network ( fig . [ fig_synfire_sims ] ) . we find that the potentiation decay @xmath15 is the crucial model parameter that predicts the mean and variance of the distribution . we present a simple , analytically solvable , `` lottery '' chain - growth model to explain how the potentiation decay controls the characteristics of the length distributions . our lottery model predicts that the mean chain length approaches a finite value as network size is increased . this simple model reproduces what is observed in the full , simulated model . the mean chain length in the lottery model is controlled by a small parameter quantifying the likelihood that a neuron that is recruited by the chain already targets the chain . we observe these preferential connections from unrecruited neurons onto recruited neurons in the full simulations , and we describe why these connections appear . what these results show is that training alters not only the network topology among neurons recruited to the chain , but also the connections from all other neurons to the neurons in a chain . this `` global '' response of the connectivity to an excitation targeting only a small subset of the population is indicative of a synergistic relationship between the spike activity on the network and the underlying topology . chain growth is initiated by stimulating the training neurons . before training , the initial values of synaptic weights are drawn from a particular distribution , which we refer to as a _ dynamic ground state_. around 2 - 10% of the synapses are active . this connection probability of the active synapses is a generally accepted range for cortical networks @xcite . spontaneous activity occurs in the network due to the noise and the active connections . the rules that govern synaptic dynamics ( stdp , active threshold , potentiation decay , etc . ) yield a distribution of synaptic weights that is statistically stationary as the population is spontaneously spiking . the dynamic ground state is stationary due to the homeostatic effect of the potentiation decay and the activation threshold . if not for the interplay between these two plasticity rules , supersynapses would spontaneously emerge due to positive feedback across the strongest synapses @xcite . instead , a unimodal distribution of synaptic weights emerges . in fig . [ necessary_decay ] we compare a synaptic weight distribution when potentiation decay is acting on the synaptic weights ( fig . [ necessary_decay]a ) to when it is not ( fig . [ necessary_decay]b ) . a particular synaptic weight in a dynamic ground state takes a random walk with steps generated by the stdp rules and the potentiation decay . stability relies on potentiation decay that prevent synaptic weights from diffusing to large values . synapses driven below the activation threshold by the potentiation decay and ltd can be reactivated by ltp . in a dynamic ground state , any neuron is connected to @xmath16 of the other neurons via active synapses at any given moment . the distribution of synaptic weights in the dynamic ground state is obtained by letting the weights evolve while simulating spontaneous activity without a training signal over a sufficent number of trials . these `` initialization '' trials are identical to the training trials except that the training neurons are not subjected to the focused strong excitation . neurons are driven with noisy excitation over two - second trials resulting in spontaneous activity while the synaptic weights evolve according to the plasticity rules . after several hundred initialization trials , the stationary weight distribution emerges . we identify this network state when the number of active synapses in the network reaches a stable value . a dynamic ground state does not emerge for all sets of the synaptic plasticity parameters ( details in the appendix ) . for example , if the maximum possible potentiation of the synaptic weight is small compared to the activation threshold and the potentiation decay is fast , the stationary state may contain only a few , short - lived active synapses because newly activated synapses are driven below the threshold before they can be further strengthened . the opposite situation is also possible when the potentiation decay is too slow to admit a stable weight distribution in which 2 - 10% of the synapses are activated . finding the full parameter space for a stable dynamic ground state requires a parameter search . however , a working combination can be found by setting the maximum potentiation slightly larger than the activation threshold , and the potentiation decay rate fast enough to deactivate a newly activated synapse within 10 s of trials . these parameters produce a stationary distribution in which the number of active synapses is likely smaller than 2% . the number of the active synapses can be increased by decreasing the maximum potentiation strength and the potentiation decay rate from this point . when a training stimulus is presented repeatedly to a network in the dynamic ground state , the stationary distribution of synaptic strengths is disturbed . this response of the network to the training stimulus drives emergence of the synfire chain within the initially disordered ground state network . during the training , the neurons recruited into the emerging chain have different synaptic strength distributions compared to those unrecruited ( or `` pool '' ) neurons . to illustrate why this is so , consider specifically the training neurons as they contact neurons in the pool . because the network is initialized in the dynamic ground state before training , all neurons , including the training neurons , have the same initial distribution of synaptic strengths onto their targets . however , when training begins , the training neurons spike at the beginning of each trial with high probability , and the synaptic weights from the training neurons onto the pool neurons are more likely to increase because the training neurons spike reliably every trial . the weights approach a new equilibrium that has higher average weight than the dynamic ground state . this is shown in fig . [ seq_shift ] . the positive shift of the average strength of a synapse targeting the pool is a result of spiking with near - certainty every trial . as the distribution of weights of synapses from training neurons onto the pool shifts positive , the potentiation decay is not sufficient to deter rapid growth from positive feedback . consequently supersynapses emerge from the training neurons onto pool neurons . these synapses tend to be convergent since the convergence allows the training neurons to evoke reliably a spike from a shared target . furthermore , the training neurons that do not share the target are likely to develop connections onto a shared target since the shared target spikes frequently after the training neurons , which spike synchronously at the start of each trial . as the training progress , the training neurons accrue supersynapses onto shared targets , with their strengths capped at @xmath11 . when the number of supersynases from each training neuron hits the limit @xmath0 imposed by the axon remodeling rule , all weak synapses are pruned and decay away due to the potentiation decay . training neurons maintain only @xmath0 supersynapses . consequently , no more targets are recruited , and the second group is formed . the number of recruited neurons is close to @xmath0 because of the convergence . because of the strong , convergent connections from the training neurons , the second group spikes reliably in each trial . they accrue their own targets in the pool following the same process as the training neurons . the result is a positive shift of the distribution of synaptic weights away from the stationary distribution of the dynamic ground state . like the training neurons , the second group of neurons can eventually saturate by accruing shared targets within the pool until axon remodeling prevents further growth . the targets of the second group form a third group whose distribution of synaptic weights responds similarly . iterations of this recruitment process result in emergence of a synfire chain within the network . as the chain network develops , spikes propagate along the chain when it is initiated by the training signal , and the ordering and timing of the spikes is almost the same across trials . a sequence may also be ignited by spontaneous activity , which we call _ re - ignition_. this can be observed directly in raster plots of spontaneous activity in networks with developing chains . an example is shown in fig . [ reignition_prob]a . spontaneous activity can initiate spike propagation from a random point in the chain . to quantify this observation , we simulated spontaneous activity of a network in which a subset of neurons are wired into a synfire chain and all other connections are randomly set ( synaptic plasticity was suppressed ) . we measured the spontaneous firing rates of all neurons . as shown in fig . [ reignition_prob]b , the downstream neurons in the chain have higher firing rates than the upstream neurons . this is because spikes reliably propagate down the chain wherever the re - ignition starts . the linear increase of the firing rates down the chain suggests that the probability of starting re - ignition is uniform across the chain . re - ignition has direct impact on the distribution of synaptic weights of the network . after multiple re - ignitions , the number of neurons targeting the chain increases . this is shown in fig . [ reignition_prob]c . pool neurons that are spontaneously active immediately before chain re - ignition have increased likelihood of targeting the chain . once these synapses from pool to chain are activated there is a decreased likelihood of ltd events on these synapses , since the strong connectivity within the chain makes it more likely for activity to remain on the chain after chain neurons are spontaneously active . hence , pool neurons tend to connect to a developing synfire chain . this positive shift of the weights from pool neurons onto neurons in the chain plays a role in the closure of the chain . once these preferential connections from the pool to the chain become numerous , it becomes likely that the pool neurons newly recruited into the chain are already connected to the chain , forming a loop that stops the chain growth . in fig.[reignition_prob]c it is clear that for faster potentiation decay , the total synaptic strength targeting the chain is smaller , implying that the stronger decay is more effective at reducing the strengths of pool neurons targeting the chain . chains recruit more groups and produce longer sequences if there are fewer pool neurons preferentially targeting the chain , which can be facilitated by strong decay . the length distributions reflect this association ( fig . [ fig_synfire_sims]a ) : for slower potentiation decay , chains tend to be shorter with a smaller variance , compared to chains subject to stronger decay . to test this association , chain length distributions are generated by a simple lottery growth model . we model chain growth as a random process : neurons in the chain are drawn sequentially from a lottery of all neurons with equal probability . for simplicity , we assume that there is one neuron in each group in the chain , which is equivalent to setting @xmath17 and using one neuron in the training set . the number of training neuron is also 1 . at the @xmath18 iteration there are @xmath4 neurons in the chain out of the total network size @xmath19 . this simple model allows us to derive the chain length distribution analytically . we first consider the case that chain closes when a previously drawn neuron is re - drawn the second time , forming a loop in the chain and stopping its growth . the probability that the @xmath20 neuron is drawn from the pool neurons and the chain does not close at length @xmath4 is @xmath21 the probability that it is re - drawn from the neurons in the chain and the chain closes at this iteration is @xmath22 . using these conditional probabilities , the probability @xmath23 of a mature chain with length @xmath24 is given by @xmath25 \prod_{i=1}^{a-1 } p(i+1|i),\ ] ] which , plugging in eq . ( [ cond_p_rr ] ) , becomes @xmath26 after applying stirling s approximation . to calculate the mean chain length as a function of network size @xmath19 , we expand to lowest order in @xmath27 and approximate the sum as an integral to find @xmath28 as @xmath29 , the mean chain length is on the order of @xmath30 and is unbounded . this is because the chance of re - drawing neurons in the chain is zero when @xmath29 . we now consider the case that chain also closes when a pool neuron preferentially connected to the chain is drawn , in addition to re - drawing a neuron in the chain . as we have shown in the previous section , slower potentiation decay enhances the probability of preferential targeting of the chain and reduces mean chain length ( fig . [ fig_synfire_sims]a ) . to model this effect , we introduce a parameter @xmath31 , which is the probability that a pool neuron is preferentially targeting one neuron in the chain . the probability that the @xmath32 neuron is drawn from the pool and does not close the chain is @xmath33 there are two scenarios in which the chain ends with @xmath24 neurons . one , when the chain has length @xmath34 , it can recruit a neuron from the pool of @xmath35 neurons that has at least one connection onto a chain neuron . two , when the chain has length @xmath24 , it can recruit one of the @xmath34 neurons above it in the chain . therefore , the probability of chain closes at length @xmath24 has two terms : @xmath36 q(a-1|a-2) ... q(2|1 ) + \frac{a-1}{n-1 } q(a|a-1) ... q(2|1).\ ] ] in the first term , the quantity in the brackets is the probability of selecting a pool neuron that has at least one connection onto a chain neuron . equation ( [ eqn - q - a-2 ] ) can be re - written into the form of eq . ( [ gen_prob_a ] ) , with the conditional probability of chain not closing at length @xmath4 modified to @xmath37 given the above conditional probability , the probability distribution of chain lengths is then : @xmath38 \\ & \times p_a^{(rr ) } \frac{(1-p_0)^{\frac{1}{2 } a ( a-1)}}{a-1 } \end{split}\ ] ] where @xmath39 is eq . ( [ rand_recruit_stir ] ) , the probability of chain length assuming no preferential targeting . equation ( [ gen_prob_a2 ] ) can be simplified in the large @xmath19 limit and moments of this distributions can be computed @xcite . however , the expressions are too onerous to print here . a notable feature is that the mean of this distribution approaches a finite limit as @xmath29 . in this limit , @xmath40 and the probability of chain closing at length @xmath24 becomes @xmath41 according to eq . ( [ gen_prob_a ] ) . the mean chain length is @xmath42^{(k+1)k } = 1 + \frac{\vartheta_2(z=0,(1-p_0)^{1/2})}{2 ( 1-p_0)^{1/8}},\ ] ] where @xmath43 is the jacobi theta function @xcite . this is a finite number . since every neuron in the pool has a non - zero probability of preferential targeting the chain , the mean chain length does not diverge even for @xmath29 . in fig . [ fig_models]a we display several chain length distributions for different @xmath31 . as @xmath31 is increased , the distribution shifts toward shorter chains and becomes sharper , indicating that the chains close at an increasingly predictable length . this trend corresponds to the sharpening of the length distribution of synfire chains as the potentiation decay is slowed , shown in fig.[fig_synfire_sims]a . to confirm the model prediction that the mean chain length approaches an asymptotic value even as the network size grows very large , we performed a set of simulations with different network sizes . we set @xmath17 and the number of training neurons to 1 to make the simulations directly comparable to the model . since neurons can not cooperate , the variance of the gwn used in the simulations was reduced to @xmath44 . also the ltd time constant @xmath45 was set to @xmath46ms ( see below ) . the potentiation decay was kept constant . the mean chain lengths in the simulations are well fit with the model using a single value of @xmath31 , and show clear sign of saturation as the network size increases ( fig.[fig_models]b ) . this trend is also observed in the fully complex , cooperative simulations which produce synfire chains . figure [ fig_synfire_sims]b is indicative of an upper bound on the length of the emergent synfire chain as the network size is increased . an minor effect omitted from the lottery growth model that also contributes to the shape of the length distribution is the ltd window function ( see eq . ( [ ltd ] ) in the appendix ) . the width of the window controlled by @xmath47 sets a soft minimum on sequence length . this effect was mentioned in @xcite ; here we show more detailed measurements in fig . [ fig_ltd_shift ] . the effect can be attributed to reliable propagation of the training signal along a partially formed chain during each trial . a recruited neuron may target an upstream chain neuron directly or by targeting a pool neuron that is targeting the chain , contributing to the likelihood of chain closure . however , during each training trial upstream chain neurons spike before the newly - recruited neuron . therefore , the synapse onto the targeted neuron is weakened by ltd . if the temporal distance from the spike of the targeted neuron to that of the recruited neuron falls within the ltd window , the weight reduction quickly silences the synapse and any possibility of reconnection is eliminated . in the simulations that are used to validate the growth model ( fig . [ fig_models ] ) , we used @xmath48 ms to minimize the ltd effect . besides ltd , there are other simplifications assumed by our growth model . a time - independent model parameter @xmath31 describes the probability that a pool neuron targets one of the entrained sequence neurons and that the chain closes on itself by recruiting such a neuron . this assumption ignores the non - equilibrium dynamics of the weight distribution as the chain recruits additional neurons . re - ignition of the partial chain precedes development of connections that preferentially target the chain . this is a random event that occurs at finite intervals , implying that preferential connectivity has an associated time scale depending on the probability of a re - ignition event . the response time of preferential targeting can be seen directly in fig . [ reignition_prob ] , showing that the sum of weights targeting the chain , averaged over all members , saturates only after a number of groups have formed . the above discussion of the effect of the ltd window function also indicates that @xmath31 is not uniform over the length of the partial chain . in fact , it is effectively zero for neurons immediately upstream from the end of the chain . furthermore , because chain re - ignition is a random process driven by spontaneous activity , fluctuations in the strength and number of synapses targeting the partial chain may contribute to the probability of closure . a constant @xmath31 ignores such fluctuations . however , the model still gives reasonably accurate predictions . in large recurrent networks with stdp , axon - remodeling and an activity - independent potentiation decay of synapses , we observed emergence of long , stereotypical sequences of spikes . the sequences are produced by stable synfire chain topologies that self - organize via a stochastic growth process . we studied the distributions of synfire chain lengths and concluded that the rate of potentiation decay in our synaptic plasticity model primarily controls the shape of the distributions . the chains develop in response to a stimulus presented to the network in a dynamic ground state , in which the distribution of synaptic weights is invariant to synaptic modifications due to spontaneous activity on the network . this network state would not exist without the potentiation decay . synfire chain growth in our network model results from a global response of the connectivity among the neurons to a stimulus that targets only a small subset of the population , the training neurons . repeated stimulations of the training neurons leads to iterative growth of a synfire chain embedded in the network . this result was expected based on previous work @xcite . however , what was not expected , but what we observed , is global response of the connectivity as the chain develops . as the sequence begins to emerge , neurons in the pool are increasingly likely to target the neurons in the chain . we suggest this process of targeting the strongly connected neurons in the chain is loosely analogous to preferential attachment in other complex networks @xcite . in contrast to other systems with preferential attachment , a scale - free distribution does not emerge from training because of the topological constraint imposed on the network by axon remodeling . the complex response of synapses throughout the network illustrates co - evolution of spike activity ( the emerging sequence ) and synaptic topology ( preferential targeting ) . we expect this observation generalizes to other recurrent network models with stdp in which spike sequences emerge . since pre - post synaptic strengthening is a common feature of stdp models , other neurons will attach to sequence members when a sequence is initiated by spontaneous activity . we believe our insight may explain the observation of neuron clustering @xcite and small - world network degree distributions @xcite in other studies where the number of strong connections a neuron can make is unconstrained . the coevolution of the network activity and network connectivity in response to an external stimulus is reflected in the spectrum of length distributions of the synfire chains . when the potentiation decay is too slow to sufficiently reduce the weights of connections from pool neurons onto a partially formed chain , the variation of chain lengths is reduced . when potentiation decay is fast , the number of preferential connections is reduced and the synfire chain has an opportunity to grow longer . we contrast our mechanism for synfire chain development with other studies in which chains emerge in a recurrent network , such as in fiete , et . al . @xcite . in fiete s model , the synaptic plasticity rules are designed in such a way that each neuron ( or group of neurons receiving correlated external input ) must connect to one other neuron ( or neuron group ) that is not already targeted . the selection of target is random , which leads to multiple closed loops where every neuron ( or neuron group ) is incorporated into a loop . the distribution of chain lengths in this model follows a power law . hence , short chains are more numerous than long chains . in contrast , the distribution of chain lengths in our model is close to a skewed gaussian . there are typical chain lengths , and short chains are rare . in our model , not every neuron is part of the chain . we introduced a growth model that incorporates preferential targeting to confirm the general form of the length distributions of the chains . the growth model is verified with corresponding simulations of networks producing single - neuron chains . the model illustrates tuning of the length distribution through the potentiation decay rate . furthermore , it predicts that the mean chain length approaches a constant in the limit of large network size . simulations of the more complex process of synfire chain growth confirm the same saturation effect ( fig . [ fig_synfire_sims ] ) . this is in contrast to the case of @xmath49 , for which the mean length diverges as @xmath30 . any small preferential targeting probability @xmath31 limits the mean length as @xmath29 . this result indicates that chain size is bounded softly , even in the limit of very large networks . it would be interesting to confirm this plateau effect in recurrent networks larger than those we were able to simulate . in at least one case we know of @xcite , a much larger network has been simulated . however , chains did not emerge upon externally stimulating the network in this study . contrasting the result of this study with our own , we have validated the iterative recruitment of synfire chain groups using a power - law stdp rule @xcite instead of the additive ltp / multiplicative ltd model ( see eqs . ( [ ltp ] ) and ( [ ltd ] ) ) introduced by @xcite . additionally , we observe the growth process is unaffected when setting the number of allowed strong connections to larger values ( 50 instead of 5 used in the simulations presented in the results ) , demonstrated also in @xcite . key differences that may account for the emergence of chains in our model are , dually , the vast restructuring of the network connections allowed by imposing an activation threshold on each synaptic weight , and also restricting the influence of a single neuron by imposing the axon - remodeling rule . before training , networks are initialized to a dynamic ground state . the distribution of synaptic weights in a ground state network is stationary while neurons are spontaneously active . synapses are activated and silenced by random activity , and the average flux of weights across the active threshold is zero . our synaptic dynamics model is distinguished from others in two ways . one , we impose an activation threshold on a synaptic weight between every pair of neurons in the network . the picture that emerges is one in which neurons are actively connecting and disconnecting to other neurons in the population freely and on a relatively short time scale , minutes to hours . a number of imaging studies support this fast restructuring of network connectivity patterns @xcite . the time scale of emergence and subsequent withdrawl of dendritic spines can be as short as ten minutes and has been linked to synaptic activity @xcite . network rewiring is not permitted in all but a few network growth models that have been proposed @xcite . instead , it is much more common to select _ a priori _ the postsynaptic targets of each neuron . we argue that this modeling choice neglects an important feature of biological networks and places limits the emergent topology of the network . two , we subject all synapses to an activity - independent decay . we propose that this is related to the widely observed decay of the early phase long - term potentiation ( e - ltp ) . the time scale of the potentiation decay is several hours @xcite , much longer than the length of an individual training trial . the role of the potentiation decay is to avoid the accumulation of random potentiation of synaptic weights known to destabilize the network dynamics @xcite . the combination of these two rules yields a robust spectrum of stationary network states . as a final note , we compare our network rewiring rules with a similar approach taken by iglesias et al @xcite . in this study a recurrent network is initialized with all - to - all connectivity and network connections are eliminated via stdp . this modeling choice is also notable in that final network connectivity is not limited by only modifying weights between specific pairs of neurons . however , it is unclear from the results of this study whether sequences emerge after pruning . the stability of the ground state network indicates the rules of our model encode a homeostatic mechanism @xcite . several other models of homeostatic mechanisms have been proposed recently , including a sliding modification threshold based on post - synaptic firing history @xcite , a dependence on fluctuations of post - synaptic membrane potential @xcite and heterosynaptic plasticity that limits the total weight targeting a single neuron @xcite . an overlooked mechanism that we propose is activity - independent , multiplicative rescaling of weights , as we have implemented here , potentiation decay or decay of e - ltp . this form of ltp returns synaptic efficacies to the baseline within 3 hours @xcite and is independent of protein synthesis . only through repeated potentiation , e - ltp can turn into the late phase ltp ( l - ltp ) , which is maintained by protein synthesis and can last over days and weeks @xcite . in our model , consistently potentiated synapses turn into supersynapses whose decay is much weaker than other active synapses . the supersynapses can be considered in the l - ltp state . emergence of the synfire chain relies on stabilization of the small percentage of supersynapses , while there are many weaker , more transient synapses . this long - tail synaptic weight distribution is consistant with physiological observation @xcite and appear in other theoretical studies @xcite . the functional role of e - ltp decay is largely ignored in the ltp literature . our model suggests that the e - ltp decay may be crucial in stabilizing synaptic weight distribution against random accumulations of ltp through spontaneous activity . moreover , the e - ltp decay can be important in the formation of functional networks through stdp . the time scale of the potentiation decay in our model is congruent with time scales of e - ltp decay , which can be seen with an order of magnitude estimate . if we assume learning occurs on a scale of tens of days and @xmath1 training trials are necessary for synaptic chains to crystalize , this places the time scale corresponding to our decay parameter @xmath15 , which we vary from 0.9 to 0.99 , in a range of @xmath50 and @xmath1 seconds . it will be interesting to test these ideas by manipulating the e - ltp decay constants in developing or learning brains _ in vivo_. a natural extension of this work is to construct a growth model in such a way that more complex asymptotic synaptic topologies emerge . many learned motor behaviors can be complex . for example , stochastic ordering of distinct elements of a behavior is one kind of behavioral complexity . the song of the bengalese finch can be described by this type of stochastic process @xcite . a single synfire chain can not capture this complexity since they produce only a single spike sequence ; multiple chain or branching chains would be necessary . one possible scenario for growing multiple chains in the same network would be to have distinct sets of training neurons . however , it turns out that preferential targeting of sequence members within the network prohibits the development of distinct synaptic structures . we implemented two distinct training groups of 5 neurons in a network of 2500 neurons . a training neuron set that is excited at the beginning of each learning trial . which of the two sets is selected at random with equal probability . in fig . [ two_training ] we display the resulting growth . the chains develop several groups individually , but ultimately they merge to a single chain . initially , the training neurons seed two disjoint sequences that recruit targets iteratively . emergence of synfire groups embeds two disjoint sequences in the network . these sequences are occasionally activated by spontaneous activity . therefore , neurons in the pool will target the partial chains preferentially . in particular , neurons at the end of one of the chains may target a neurons in the other chain with elevated probability . once one chain reliably activates neurons in the other chain , they will merge . merging occurs reliably each time we simulated two training groups . other growth mechanisms must be present , or the chains must be encoded within distinct populations of neurons . this conclusion is consistent with other studies , such as @xcite . in a recurrent network of neurons driven by high - frequency noisy input and synapses governed by a set of plasticity rules which include stdp , a potentiation decay and axon remodeling , we showed that neurons cooperate via convergent synapses to self - organize into a synfire chain characterized by a precisely - timed sequence . the network is initialized in a state characterized by a statistically stationary distribution of synaptic weights , invariant to network spontaneous activity . the combination of a potentiation decay plus an activation threshold imposed on the synaptic weights provides a homeostatic mechanism within the network . a small subset of neurons stimulated by a strong excitation forms the seed for recursive synaptic growth of synfire groups . during repeated presentations of this stimulus and emergence of the chain , we observe a complex response of the network connectivity that is reflected in the distribution of asymptotic chain lengths . we have demonstrated a clear example of interplay between neural activity and emergent synaptic topology in a developing network . the simulated networks consist of @xmath19 excitatory , conductance - based , pulse - coupled leaky integrate - and - fire ( lif ) neurons . the state of the @xmath18 neuron is described by a single variable @xmath51 , its membrane potential , which obeys @xmath52 where @xmath53 is synaptic input to the membrane . the lif neuron requires several parameters : leak reversal potential @xmath54 , membrane time constant @xmath55 , spike threshold @xmath56 and reset potential @xmath57 . if @xmath58 , the neuron emits a spike and is instantaneously reset to @xmath59 . the synaptic input to the @xmath18 neuron @xmath53 consists of three sources : a noisy external drive , an excitatory conductance , and an inhibitory conductance : @xmath60 \\ & + g_i^{(i)}(t ) \left [ e^{(i ) } -v_i \right ] . \end{split}\ ] ] we choose the reversal potentials @xmath61 and @xmath62 . the drive includes a gaussian white noise ( gwn ) term @xmath63 , obeying @xmath64 and @xmath65 with all higher order correlations equal to zero . the noise is uncorrelated across individual neurons . driving current is @xmath66 and @xmath67 . training neurons ( see methods ) are subjected to larger driving current ( 100 mv ) for the first 8ms of each training trial . the external drive originates in upstream regions , which we do not simulate . gaussian white noise is commonly employed to model this input @xcite . the conductances @xmath68 and @xmath69 take the form of sums of @xmath70-functions centered on the spike times of neurons in the network . specifically , @xmath71 where @xmath72 is the excitatory synaptic weight from neuron @xmath5 onto @xmath4 and @xmath73 is the time of the @xmath74 spike of neuron @xmath5 . weight @xmath72 is zero if @xmath5 does not have a synapse onto @xmath4 . @xmath75 is in a range @xmath76 $ ] , expressed as a multiple of the neuron s leak conductance . also , we forbid self - synapses : @xmath77 . global inhibition , i.e. inhibition that targets all neurons , is concurrent with each spiking event in the network . the inhibitory conductance takes the form @xmath78 here @xmath79 is the inhibitory conductance on each excitatory neuron induced by @xmath80 nearly synchronous ( within the time resolution @xmath81 , methods , section a ) spikes of the excitatory neurons at any single iteration of the population dynamics . @xmath79 is computed under the following assumptions : 1 ) the feedback inhibition is mediated by @xmath82 inhibitory neurons ; 2 ) each excitatory neuron randomly connects to @xmath83 inhibitory neurons ; 3 ) when excited , an inhibitory neuron emits a single spike with certainty regardless the number of nearly synchronous spikes it receives . the maximum inhibition that can be provided to an excitatory neuron is @xmath84 , where @xmath85 is the weight of a single inhibitory synapse . we can derive that @xmath86 here @xmath87 is the average number of inhibitory neurons contacted by @xmath80 excitatory neurons . each term in the sum over @xmath4 in eq . ( [ inh_series ] ) is the average number of addition interneurons contacted by the @xmath88-th excitatory neuron . note that @xmath89 is the probability that an inhibitory neuron is _ not _ yet contacted by previous @xmath4 excitatory neurons . for all simulations , we pick @xmath90 and @xmath91 . the computational advantage of using pulse - coupled neurons is that the response of each neuron to a spike is instantaneous and can be calculated by plugging eq . ( [ ex_cond ] ) and eq . ( [ inh_cond ] ) into eq . ( [ lif_eq ] ) and integrating over the @xmath70-functions in an infinitesimal neighborhood @xmath92 around spike time @xmath93 @xcite . the result is @xmath94 \\ & + \frac{g^{(i ) } e^{(i)}}{g^{(e)}_{j , i}+g^{(i ) } } \end{split}\ ] ] as @xmath95 . since the interactions are instantaneous , between spikes the probability distribution @xmath96 describing the time - evolution of the membrane potential is the solution of a fokker - planck equation @xcite @xmath97 \\ & + \frac{\sigma_v^2}{\tau } \frac{\partial^2 \rho^{mp}}{\partial v^2}(v , t;w ) , \end{split}\ ] ] where the variance of the distribution in the large @xmath98 limit @xmath99 . @xmath100 is the initial value of the membrane potential . an fast event - driven algorithm @xcite has been developed to generate spike activity on networks of pulse - coupled neurons with gwn using `` lookup tables '' containing the solutions to eq . ( [ slif_fp ] ) , as well as distributions of spike threshold first passage times @xcite describing spike - timing distributions . the algorithm has four steps . 1 . calculate a predicted spike time for each neuron , given an initial value of the membrane potential @xmath100 , by sampling the first passage time distributions . 2 . identify the neuron with the minimal predicted spike time . also , identify any neurons that is predicted to spike within @xmath3 of the neuron with the minimal predicted spike time . these neurons will be next to spike . 3 . assign a membrane potential to each neuron that does not spike by sampling the membrane potential distributions at the spike time calculated in step 2 . reset the neurons that spike to @xmath101 . 4 . apply eq . ( [ mp_resp ] ) to determine the membrane potentials immediately after the spike is applied . calculate new values of synaptic weights according to stdp and other synaptic dynamics rules . these steps are iterated as long as desired . the weight of a synapse is updated according to an stdp rule during the last step of the simulation algorithm . the weight of a synapse from excitatory presynaptic neuron @xmath4 to postsynaptic neuron @xmath5 is @xmath6 . we implement the additive / multiplicative rule @xcite . we introduce also a tunable scale factor @xmath102 @xcite . the weight is modified using the following stdp window : @xmath103 @xmath104 with @xmath105 , the spike time difference , @xmath106 , @xmath107 , and unless otherwise specified @xmath108 . we impose @xmath109 to ensure non - negative weights . weights are clipped above @xmath110 . the long - term potentiation(p)/depression(d ) function is defined as @xmath111 this rule is identical to the rule proposed in the jun - jin model @xcite except for the introduction of an additional parameter @xmath112 . for nonzero @xmath112 , there is a net reduction in weight for @xmath113 @xcite , the magnitude of which is controlled by the parameter @xmath112 ; we pick @xmath114 . this rule has been shown to be equivalent to the effect of jitter on the arrival times of pre- and post- spikes at the synapse and compliments the time resolution we impose on the spike arrival times . functionally , it discourages connections between neurons that spike reliably within @xmath81 and effectively reduces any the weights of extraneous active synapses that may exist among neurons recruited to the same synfire group . this research was supported by the sloan research fellowship , the huck institute of life sciences at the pennsylvania state university , and nsf grant no . we thank jason wittenbach for his comments on the manuscript . 71ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ] + 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop _ _ ( , , ) @noop _ _ ( , ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop _ _ ( , , ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) _ _ , @noop ph.d . thesis , ( ) http://mathworld.wolfram.com/jacobithetafunctions.html [ `` , '' ] @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop _ _ ( , ) @noop * * , ( ) @noop * * , ( ) fig.1 . samples of asymptotic configurations of synapses in a 4000 neuron network generate long , stereotypical sequences of spikes . the chains above have 25 ( left ) and 40 ( right ) synfire groups . the time between firing of adjacent groups is @xmath115 ms . the number @xmath0 of super synapses one neuron can maintain is set to 5 . the number of training neurons is 5 . * ( a ) * blowup depicts regular synfire connectivity between groups . the regular structure observed is a result of neurons cooperating to excite targets . a neuron most effectively excites a target when the target is shared with a neuron in its group , so convergent synapses are favored for development into supersynapses . a group accumulates shared targets until the maximum allowed number of supersynapses is reached . there are five neurons per synfire group in this network because the maximum number of supersynapses allowed by axon remodeling is five per neuron . * ( b ) * the reconnection point is splayed across several groups . the connections that form first are to the group nearest the top of the chain . the downstream connections follow due to elevated probability of re - ignition at the initial connection point . the splayed connectivity allows spontaneous activity to restart when the excitation reaches this point because neurons across several synfire groups spike and sequential activity is most stable when a full synfire group fires . however , the synapses at the reconnection point potentiated often enough to remain stable . * ( c ) * defects sometimes appear as the chain emerges . small defects like the one depicted can remain stable . severe defects are not stable and never appear due to lack of a clear sequence that is consistently reinforced by stdp . a raster plot of the population activity after a synfire chain has self - organized . the neurons are labeled according to their time of first spike . some neurons in the chain do not spike during the first iteration around the loop due to fluctuations of the membrane potential , but they do spike during subsequent iterations . these neurons have the highest valued labels across the top of the plot . + fig.3 . * ( a ) * two simulated synfire chain length distributions for two different potentiation decay parameters . as the potentiation decay slowed , the resulting distribution of lengths narrows . sample size : 100 networks . @xmath116 neurons . * ( b ) * the mean synfire chain length as a function of network size . the mean length saturates as the network size increases . sample size of each data point : 100 networks . @xmath117 . the error bars denote the standard error of the mean . the potentiation decay suppresses runaway synaptic growth resulting from positive feedback along active synapses . two networks are compared by setting the activation threshold to @xmath118 ( `` active '' ) and @xmath119 ( `` silent '' ) , respectively . the `` silent '' network effectively has all synaptic conductance set to zero . in * ( a ) * , a potentiation decay is applied to all synapses after each 1 s interval of simulated time . in * ( b ) * , there is no potentiation decay . in both scenarios there is no positive feedback in the silent networks , so these distributions ( light gray ) are stable in both * ( a ) * and * ( b)*. however , in the networks with active synapses , only the distribution in the network with potentiation decay * ( a ) * remains stable indefinitely . in the network without potentiation decay * ( b ) * , synapses grow large resulting in runaway activity . histogram of synaptic weights in network with silent synapses . a group of training neurons spikes at the beginning of every trial . the distribution of weights onto the non - training neuron ( pool neurons ) stabilizes with a higher mean weight . the synapses in these simulated networks are silenced ( i.e. , @xmath120 ) in order to emphasize that the net strengthening is independent of interactions between individual neurons . when interactions are allowed , the strongest synapses may overcome the potentiation decay leading to development of strong synapses within the network . + fig.6 . * ( a ) * an example of re - ignition of a developing chain during the spontaneous activity period in a training trial . spike raster of a network of 400 neurons are shown . at @xmath121 ms , the training neurons are stimulated , and spikes propagate down the chain until around @xmath122 ms . spontaneous activity starts afterwards . the chain is re - ignited around @xmath123 ms , evident from the sequential spikes shown in the shaded area . the re - ignition starts from a random point in the chain . * ( b ) * spike probability for 1000 neurons is plotted against neuron label over @xmath2 ms of spontaneous activity in a network with synaptic weights held fixed ( no synaptic plasticity ) . the neurons are wired such that a synfire chain is embedded in an otherwise randomly connected network . neurons labeled 1 through 130 are connected into a synfire chain , with 1 - 5 forming the first group and 126 - 130 forming the last . synaptic weights of all other connections are drawn from the the synaptic weight distribution in the dynamical ground state ( fig.[necessary_decay]a ) . random spiking of neurons in the synfire chain often leads to re - ignition and propagation of spikes down the chain . this makes neurons at the end of the synfire chain have the highest probability of spiking . the dashed line is the uniform spike probability expected in the absence of the embedded synfire chain . * ( c ) * preferential targeting emerging during training is measured by averaging the sum of active synaptic weights targeting the chain over the length of the chain . this is plotted as a function of partial chain length over @xmath124 training trials for two different values of the potentiation decay . the `` enhancement '' is calculated as the sum of weights divided by the average sum of weights onto a pool neuron in the stable ground state . this measurement was repeated over 10 independent instances of the network . also measured and averaged ( dotted lines ) is the sum of weights targeting random pool neurons . pool neurons are more likely to target chain neurons than other pool neurons and the likelihood increases as the chain grows . chain length distributions from the lottery growth model . * ( a ) * chain length distributions are plotted for different probabilities @xmath31 from the pool neurons to the chain . as @xmath31 decrease , the mean and the variance of the distribution increase . for @xmath49 , there are no preferential targeting , and the mean length and variance are maximal . * ( b ) * comparison of the mean chain length as a function of the network size between the model and the simulations . the simulations were done for five network sizes . for each network size , 150 simulations were performed . the data points are the mean chain length and the error bars indicate the standard error of the mean . the model prediction is plotted as the solid line . the parameter @xmath31 in the model was picked such that the root - mean - square error between the predictions and the simulations at the five network sizes is minimized . the dotted line is the prediction for @xmath49 , for comparison . mean chain length is offset with the ltd window size @xmath125 . simulations of a 2500 neuron network with two training sets . preferential targeting causes the two chains to merge .

in the symmetric group on a set of size , let denote the conjugacy class of involutions with no fixed points ( equivalently , we refer to these as `` pairings '' , since each disjoint cycle has length ) . harer and zagier explicitly determined the distribution of the number of disjoint cycles in the product of a fixed cycle of length and the elements of . their famous result has been reproved many times , primarily because it can be interpreted as the genus distribution for-cell embeddings in an orientable surface , of a graph with a single vertex attached to loops . in this paper we give a new formula for the cycle distribution when a fixed permutation with two cycles ( say the lengths are , where ) is multiplied by the elements of . it can be interpreted as the genus distribution for-cell embeddings in an orientable surface , of a graph with two vertices , of degrees and . in terms of these graphs , the formula involves a parameter that allows us to specify , separately , the number of edges between the two vertices and the number of loops at each of the vertices . the proof is combinatorial , and uses a new algorithm that we introduce to create all rooted forests containing a given rooted forest .

0803.0707

let @xmath8=\{1,\ldots , p\}$ ] , and @xmath9 be the set of permutations of @xmath8 $ ] , for @xmath10 . when @xmath10 is even , let @xmath11 be the set of _ pairings _ on @xmath8 $ ] , which are partitions of the set @xmath8 $ ] into disjoint pairs ( subsets of size @xmath2 ) . we refer to the single element of @xmath12 as the _ empty _ pairing . where the context is appropriate , we shall also regard @xmath11 as the conjugacy class of involutions with no fixed points in @xmath9 . in this latter context , each pair becomes a disjoint cycle consisting of that pair of elements . of course , the number of pairings in @xmath11 is @xmath13 , with the empty product convention that @xmath14 . now , for @xmath15 and even , let @xmath16 , in disjoint cycle notation , and let @xmath17 . let @xmath18 be the number of permutations in @xmath19 with @xmath20 cycles in the disjoint cycle representation , for @xmath21 . the generating series for these numbers are given by @xmath22 . harer and zagier [ [ hz ] ] obtained the following result . [ hzthm ] _ ( harer and zagier [ [ hz ] ] ) _ for a positive , even integer @xmath6 , with @xmath23 , @xmath24 other proofs of theorem [ hzthm ] have been given by itzykson and zuber [ [ iz ] ] , jackson [ [ j ] ] , kerov [ [ ke ] ] , kontsevich [ [ k ] ] , lass [ [ l ] ] , penner [ [ p ] ] and zagier [ [ z ] ] ( see also the survey by zvonkin [ [ zv ] ] , section 3.2.7 of lando and zvonkin [ [ lz ] ] and the discussion in section 4 of the paper by haagerup and thorbjornsen [ [ ht ] ] ) . recently , goulden and nica [ [ gn ] ] gave a direct bijective proof of theorem [ hzthm ] . in the present paper , we consider a similar bijective approach to extend this important result of harer and zagier to the case in which the permutation @xmath25 is replaced by a fixed permutation with two cycles in its disjoint cycle representation . some additional notation is required . let @xmath26'=\{1',\ldots , q'\}$ ] , and let @xmath27 be the set of permutations of @xmath8\cup[q]'$ ] , for @xmath28 . let @xmath29 be the set of pairings on @xmath8\cup[q]'$ ] , for @xmath28 , where @xmath30 is even ( we refer to the single element of @xmath31 as the _ empty _ pairing ) . a pair in a pairing is called _ mixed _ if it consists of one element from @xmath8 $ ] and one element from @xmath26'$ ] . where the context is appropriate , we shall also regard @xmath29 as the conjugacy class of involutions with no fixed points in @xmath27 . for @xmath32 , we consider the permutation @xmath33 , and let @xmath34 , and @xmath35 be the number of permutations in @xmath36 with @xmath20 cycles in the disjoint cycle representation , for @xmath37 . consider the generating series @xmath38 the main result of this paper is the following expression for @xmath39 . [ mainthm ] for @xmath40 , with @xmath41 of the same odd - even parity and @xmath42 , we have @xmath43 where @xmath44 note that theorem [ mainthm ] gives a summation of nonnegative terms , since for all choices of summation indices @xmath45 with @xmath46 ( so that @xmath47 is nonzero ) , the difference @xmath48 is nonnegative . the proof of theorem [ mainthm ] is based on a combinatorial model that is developed in section [ sec2 ] . as a consequence , it is sufficient to enumerate a particular graphical object that we call a _ paired array_. we then give two combinatorial reductions , in sections [ sec3 ] and [ sec4 ] , in terms of a simpler class of paired arrays called _ vertical _ paired arrays . these are explicitly enumerated in section [ sec5 ] , which allows us to complete the proof of theorem [ mainthm ] . one of the combinatorial conditions on paired arrays is that two graphs associated with them must be acyclic . because of this , a key component of sections [ sec4 ] and [ sec5 ] is the enumeration of rooted forests which contain a given forest as a subgraph . thus in section [ sec35 ] we give a new bijection for this fundamental combinatorial problem . however , before we turn to our combinatorial model and subsequent reductions , we consider some consequences of theorem [ mainthm ] , and give some comparisons to results in the existing literature . a major reason that harer and zagier s result ( theorem [ hzthm ] ) is important ( as evidenced by so many published proofs ) is that it can be restated as an equivalent geometric problem in terms of maps . map _ is an embedding of a connected graph ( with loops and multiple edges allowed ) in an orientable surface , partitioning the surface into disjoint regions ( called the _ faces _ of the map ) that are homeomorphic to discs ( this is called a two - cell embedding ) . a _ rooted _ map is a map with a distinguished edge and incident vertex ( so , the map is `` rooted '' at that end of the distinguished edge ) . the well - known embedding theorem allows us to consider this as equivalent to a pair of permutations and their product ( see , e.g. , tutte [ [ t ] ] , where the terminology `` rotation system '' is used to describe this triple of permutations ) . from this point of view , the @xmath20th coefficient @xmath18 in the generating series @xmath49 evaluated in theorem [ hzthm ] is equal to the number of rooted maps with @xmath50 vertex , @xmath3 edges and @xmath20 faces ( where @xmath23 , as in theorem [ hzthm ] ) . denoting the genus of the surface in which such a map is embedded by @xmath51 , then the euler - poincar theorem implies that @xmath52 , or that @xmath53 . similarly , theorem [ mainthm ] has a geometric interpretation . let @xmath54 be the conjugacy class of @xmath55 in which there are two disjoint cycles , of lengths @xmath6 and @xmath7 . then the coefficient @xmath35 in the generating series @xmath39 is equal to @xmath56 times the number of rooted maps with @xmath2 vertices ( of degrees @xmath6 and @xmath7 ) , @xmath3 edges ( exactly @xmath57 of which join the two vertices together , plus @xmath58 that are loops at the vertex of degree @xmath6 , plus @xmath59 that are loops at the vertex of degree @xmath7 ) , and @xmath20 faces ( where @xmath42 , as in theorem [ mainthm ] ) . in this case , if we denote the genus of the surface in which such a map is embedded by @xmath51 , then we obtain @xmath60 . of course , since genus is a nonnegative integer , we must have @xmath61 , and indeed the coefficient of @xmath62 in the summation for @xmath39 given in theorem [ mainthm ] is zero , since the summand corresponding to @xmath63 ( which has @xmath64 as a factor ) is itself equal to zero . for the planar case , which corresponds to @xmath65 , the only nonzero summand that contributes to the coefficient of @xmath66 in the summation of theorem [ mainthm ] corresponds to @xmath67 , and this gives immediately that @xmath68 this checks with the straightforward computation that one can make to determine this value by elementary means there are @xmath57 edges between the two vertices ; between the ends of these edges at each vertex is an even number of vertices , joined by loops without crossings ( and there is catalan number of such arrangements for each such even interval ) . this explains the term `` genus '' in the title ; the term `` annular '' is adapted from its usage in mingo and nica [ [ mn ] ] . it refers to an equivalent embedding for a map with two vertices , in an annulus . the ends of the edges incident with one of the vertices ( say the one of degree @xmath6 ) are identified with @xmath6 points arranged around the disc on the exterior of the annulus , and the ends incident with the other vertex are identified with @xmath7 points arranged around the disc on the interior of the annulus . the points corresponding to the two ends of an edge are joined by an arc in the interior of the annulus . we have been able to find one relevant enumerative result ( jackson [ [ j ] ] ) in the literature about such maps , in which the total number of edges is specified , but not the exact number joining the two vertices together . to compare this result to our main result , we must sum over @xmath69 ( since the underlying graph must be connected , then @xmath57 , the number of edges joining the two vertices together , must be positive ) , and thus define @xmath70 then jackson [ [ j ] ] has considered the case @xmath71 , and obtained the following result , restated in terms of our notation ( by applying the proportionality constant @xmath72 ) . [ jint ] _ ( jackson [ [ j ] ] ) _ for @xmath73 , @xmath74 by slightly modifying jackson s [ [ j ] ] integration argument we are able to obtain the following expression for @xmath75 , with arbitrary @xmath4 of the same parity . [ gsint ] for @xmath76 , with @xmath30 even , and @xmath42 , @xmath77 we have checked computationally , with the help of maple , that theorems [ jint ] and [ gsint ] agree with theorem [ mainthm ] , summed over @xmath69 , for a wide range of values of @xmath4 . however , we have been unable to prove this for all @xmath4 , since we have not been able to show that the sum over @xmath69 of the result of theorem [ mainthm ] is equal to the result of theorem [ gsint ] . note that the summation in theorem [ gsint ] can be made symmetrical in @xmath4 ( so the ordering @xmath78 is not required ) by changing the summation variable @xmath20 to @xmath79 . the method employed in jackson [ [ j ] ] for theorem [ jint ] , and in many of the papers listed above that give proofs of theorem [ hzthm ] , is matrix integration . however , we do not see how to adapt the matrix integration methodology to prove our main result , theorem [ mainthm ] , since it does nt seem possible to specify that there are exactly @xmath57 edges joining the two vertices together in the matrix method . the simplicity of our result seems to suggest that an extended theory of matrix integration to allow a specified number of edges between particular vertices might be possible , and worth investigating . the simplicity of the result also suggests that there should be a more direct combinatorial proof than the one presented in this paper . the combinatorial model for our proof of theorem [ mainthm ] is based on a _ paired surjection _ , which has the following definition . [ shiftsydef ] for @xmath80 , with @xmath41 of the same odd - even parity , let @xmath81 be the set of ordered pairs @xmath82 , where @xmath83 has @xmath57 mixed pairs , and @xmath84 is a surjection from @xmath8\cup[q]'$ ] onto @xmath85 $ ] , satisfying the condition @xmath86\cup[q]'.\ ] ] such an ordered pair @xmath82 is called a _ paired surjection_. let @xmath87 . in the following result , the generating series @xmath39 evaluated in theorem [ mainthm ] is expressed in terms of the numbers @xmath88 of paired surjections . paired surjections are closely related to shift - symmetric partitions , that arose in goulden and nica [ [ gn ] ] . indeed , the proof of the following result is identical to the proof of proposition 1.3 in goulden and nica [ [ gn ] ] , and is hence omitted . [ fallfac ] for @xmath89 , with @xmath41 of the same odd - even parity , we have @xmath90 we consider @xmath91 , and construct various objects associated with @xmath92 . first let @xmath93 $ ] and @xmath94'$ ] , for @xmath95 $ ] . let @xmath96 , and @xmath97 , and let @xmath98 , @xmath99 , and @xmath100 , @xmath101 . suppose that @xmath50 is contained in @xmath102 , and that @xmath103 is contained in @xmath104 . define @xmath105 by @xmath106 when @xmath107 , and @xmath108 by @xmath109 when @xmath110 . now , if @xmath106 , then ( interpreting @xmath50 as @xmath111 ) condition ( [ muphicond ] ) means that @xmath112 , so we have @xmath113 . this implies that the functional digraph of @xmath114 ( the directed graph on vertex - set @xmath115 with an arc directed from @xmath116 to @xmath117 for each @xmath118 ) is actually a tree , in which all arcs are directed towards vertex @xmath119 ( which we consider as the root of this tree ) . we denote this rooted tree by @xmath120 . similarly , the functional digraph of @xmath121 , on vertex - set @xmath122 , is also a tree , with all arcs directed towards vertex @xmath123 ( which we consider as the root of this tree ) . we denote this rooted tree by @xmath124 . one condition that the paired surjection @xmath92 satisfies is that the number of mixed pairs containing an element of @xmath125 is equal to the number of mixed pairs containing an element of @xmath126 for all @xmath127 $ ] . ( for the reason that this necessary condition arises , see the discussion of `` unique label recovery '' in the next section . ) we call this the _ balance _ condition for @xmath92 . the fact that @xmath84 is a surjection is equivalent to @xmath128 , for @xmath127 $ ] , and we call this the _ nonempty _ condition for @xmath92 . the fact , established above , that the graphs of @xmath114 and @xmath121 are trees is called the _ tree _ condition for @xmath92 . now we consider a graphical representation for the paired surjection @xmath92 , called its _ labelled paired array_. this is an array of cells , arranged in @xmath20 columns , indexed @xmath129 from left to right , and two rows . in column @xmath116 of row @xmath50 , place an ordered list of @xmath130 vertices , labelled by the elements of @xmath125 from left to right ; in column @xmath116 of row @xmath2 , place an ordered list of @xmath131 vertices , labelled by the elements of @xmath126 from left to right . for each pair of @xmath132 draw an edge between the vertices whose labels are given by the pair . for example , when @xmath133 , consider @xmath134 , given by @xmath135 @xmath136 @xmath137 , and @xmath138 , @xmath139 , @xmath140 @xmath141 , @xmath142 . the corresponding labelled paired array is given in figure [ labarr ] , and the trees @xmath120 and @xmath124 are given in figure [ trees ] . now suppose that we mark the cells in column @xmath119 of row @xmath50 and in column @xmath123 of row @xmath2 ( by placing a small box in the top righthand corner of the marked cell in row 1 , and in the bottom righthand corner of the marked cell in row 2 ) , and remove the labels from all vertices call the resulting object the _ paired array _ of @xmath92 . the ordered list of vertices in each cell is now to be interpreted as a generic totally ordered set , with the given left to right order , and the pairing @xmath132 now acts on these ordered sets in the obvious way . for example , the paired array determined from the labelled paired array displayed in figure [ labarr ] is given in figure [ arr ] . what information have we lost when the labels are removed ? the answer , perhaps surprisingly , is that no information is lost , since we have _ unique label recovery _ by applying condition ( [ muphicond ] ) iteratively , as follows : for the first row , place label @xmath50 on the leftmost vertex in the marked cell of row @xmath50 ; for each @xmath116 from @xmath2 to @xmath6 , place label @xmath116 on the leftmost unlabelled vertex in column @xmath143 of row @xmath50 . the same process applied to the second row will place labels @xmath103 to @xmath144 on the vertices in row @xmath2 . ( the reader can apply this to the paired array in figure [ arr ] , to check that indeed the labelled paired array in figure [ labarr ] is recovered in this way . ) the proof that this process always works for a paired array satisfying the balance , nonempty and tree conditions ( and the proof that these conditions are necessary for this process to work ) requires only a slight modification of the results in section 3 of [ [ gn ] ] , and is not given here ( note that neither the functions @xmath114 and @xmath121 , nor the trees @xmath120 and @xmath124 , depend on the labels of the vertices , and the number of mixed pairs incident with the vertices in each cell of the paired array also does not depend on the labels , so the balance , nonempty and tree conditions can be checked on the paired array alone ) . this motivates us to define a _ paired array _ in the abstract ( and not as obtained by removing the labels from a labelled paired array ) , and in fact to extend it to a more general class of objects , in the following definition . [ pairarr ] for @xmath80 , with @xmath41 of the same odd - even parity , we define @xmath145 to be the set of arrays of cells , arranged in @xmath20 columns and @xmath2 rows , subject to the following conditions : * each cell contains an ordered list of vertices , so that there is a total of @xmath6 vertices in the first row , and @xmath7 vertices in the second row . the vertices are paired ( in the language of graph theory , there is a perfect matching on the vertices ) , so that @xmath57 pairs join a vertex in the first row to a vertex in the second row ( these are the _ mixed pairs _ ) . the number of mixed pairs containing a vertex in column @xmath116 of row @xmath50 is equal to the number of mixed pairs containing a vertex in column @xmath116 of row @xmath2 , for all @xmath146 ( this is called the _ balance _ condition ) . * there is at least one marked ( with a small box ) cell in row @xmath50 , and we denote the set of such columns by @xmath147 . there is at least one marked ( with a small box ) cell in row @xmath2 , and we denote the set of such columns by @xmath148 . there is at least one vertex in every column that is not contained in @xmath149 ( this is called the _ nonempty _ condition ) . * denote the set of columns in which there is at least one vertex in row @xmath50 by @xmath115 , and the set of columns in which there is at least one vertex in row @xmath2 by @xmath122 . define the function @xmath150 as follows : if the rightmost vertex in column @xmath116 of row @xmath50 is paired with a vertex in column @xmath151 , then @xmath106 . similarly , define @xmath152 as follows : if the rightmost vertex in column @xmath116 of row @xmath2 is paired with a vertex in column @xmath151 , then @xmath109 . the functional digraph of @xmath114 is a forest with @xmath153 components ( called the _ rightmost forest for row _ @xmath50 ) ; each component is a tree in which all edges are directed towards an element of @xmath147 ( and this is called the _ root _ of that tree ) . the functional digraph of @xmath121 is a forest with @xmath154 components ( called the _ rightmost forest for row _ @xmath2 ) ; each component is a tree in which all edges are directed towards an element of @xmath148 ( and this is called the _ root _ of that tree ) . together , these specify the _ forest _ condition . the elements of @xmath145 are called _ paired arrays_. a paired array is defined to be _ canonical _ if @xmath155 . define @xmath156 to be the set of canonical paired arrays in @xmath145 , and @xmath157 . the uniqueness of label recovery described in the previous section proves that there is a bijection ( via labelled paired arrays ) between the set @xmath81 of paired surjections and the set @xmath156 of canonical paired arrays , so we have @xmath158 ( it is straightforward to verify that the conditions for canonical paired arrays imply that _ every _ column is nonempty . ) in this paper we shall determine @xmath88 , and hence the generating series @xmath39 via proposition [ fallfac ] , by giving a combinatorial reduction for canonical paired arrays , thus directly determining @xmath159 . a _ redundant _ pair in a paired array is a vertex pair that is _ not _ mixed , and does _ not _ contain a vertex that is rightmost in an unmarked cell . minimal _ paired array is a paired array without redundant pairs . we define @xmath160 to be the set of minimal , canonical paired arrays in @xmath145 , and @xmath161 . in our next result , we remove redundant pairs from a canonical paired array , and thus show that the enumeration of canonical paired arrays can be reduced to the enumeration of minimal , canonical paired arrays . [ redundmin ] for @xmath80 , with @xmath41 of the same odd - even parity , we have @xmath162 for the proof , it is convenient to introduce some notation . a _ partial _ pairing on @xmath8 $ ] is a pairing on a set @xmath163 $ ] of even cardinality . if @xmath164 , then we also call it an @xmath116-partial pairing . for each of these partial pairings @xmath132 , we call @xmath165 the _ support _ , and denote this by @xmath166 . let @xmath167 be the set of @xmath116-partial pairings on @xmath8 $ ] . similarly , let @xmath168 be the set of @xmath116-partial pairings on @xmath26'$ ] . consider an arbitrary @xmath169 . we now describe a construction for three objects , @xmath170 , @xmath171 and @xmath172 , obtained from @xmath173 . we begin by attaching the numbers @xmath174 to the vertices and the small box in row @xmath50 of @xmath173 , from left to right ( under the interpretation that all vertices in column @xmath116 are to the left of all vertices in column @xmath151 for @xmath175 , and that the small box representing a marking is rightmost in its cell ) . let @xmath170 be the partial pairing consisting of pairs of numbers attached to the redundant pairs in row @xmath50 of @xmath173 . we follow the analogous procedure for row @xmath2 : we attach primed numbers @xmath176 to the vertices and small box in row @xmath2 , and let @xmath171 be the partial pairing consisting of the pairs of ( primed ) numbers attached to the redundant pairs in row @xmath2 of @xmath173 . third , we remove all redundant pairs ( both vertices and edges ) from @xmath173 , to get the paired array @xmath172 , with the same marked cells as @xmath173 . the vertices in each cell of @xmath172 have the same relative order as they did in @xmath173 . for example , if @xmath173 is the paired array in figure [ arr ] , then we have @xmath177 , @xmath178 , and @xmath172 is given in figure [ newminarr ] . now , the only vertex that can be numbered @xmath111 in row @xmath50 is the rightmost vertex of the rightmost nonempty cell in row @xmath50 ( if this cell is not marked ) , but this vertex can not appear in a redundant pair since it is rightmost in an unmarked cell . this implies that the numbers on redundant pairs in row @xmath50 all fall in the range @xmath179 , and so @xmath170 is a partial pairing on @xmath8 $ ] . similarly , @xmath171 is a partial pairing on @xmath26'$ ] . also , since the redundant pairs that were removed in the construction do not involve the rightmost vertex in any nonempty cell , @xmath172 has the same rightmost functions @xmath114 and @xmath121 as @xmath173 , and the same mixed pairs as @xmath173 , so it must satisfy the balance , nonempty and forest conditions , which implies that @xmath172 is a minimal paired array . thus we have a mapping @xmath180 we now prove that @xmath181 is a bijection . it is sufficient to describe the inverse mapping , so that we can uniquely recover @xmath173 from an arbitrary triple @xmath182 @xmath183 . given @xmath182 , let @xmath184 , @xmath185 , and @xmath186\setminus{\sigma}_1 $ ] , @xmath187'\setminus{\sigma}_2 $ ] . number the vertices and small box in row @xmath50 of @xmath172 with the elements of @xmath188 . then insert vertices numbered with the elements of @xmath189 , so that the numbers on all vertices and the small box in row @xmath50 increase from left to right , and so that the vertex numbered with @xmath190 is placed in the same cell as either the vertex or small box numbered with @xmath191 . inserting vertices numbered from @xmath192 in row @xmath2 using an analogous process , pairing the inserted vertices with @xmath170 and @xmath171 , and removing the numbers , we arrive at a paired array @xmath173 . it is straightforward to check that @xmath173 satisfies the balance , nonempty , and tree conditions , and that the process described above reverses the numbering scheme used in the mapping @xmath181 . thus , we have described the mapping @xmath193 , and our proof that @xmath181 is a bijection is complete . the result follows from the easily established facts that @xmath194 , @xmath195 . in this section , we detour to consider the basic combinatorial question of how many rooted forests with a given set of root vertices contain a given rooted forest . we give a bijection for this that differs from the standard ones in the literature , like the cycle lemma ( see , _ e.g. _ , [ [ stan ] ] , p. 67 ) or the prfer code ( see , _ e.g. _ , [ [ stan ] ] , p. 25 ) , because it is more convenient for our constructions involving paired arrays . suppose we have a rooted forest @xmath196 ( all edges directed towards a root vertex in each component ) on vertex - set @xmath85 $ ] , whose components are the rooted trees @xmath197 , @xmath198 . suppose the root vertex of @xmath199 is @xmath200 , @xmath201 , and the root vertex of @xmath202 is @xmath203 , @xmath204 . for convenience , we order the trees so that @xmath205 , @xmath206 , and we let @xmath207 denote the union of the sets of vertices in the trees @xmath208 . [ fct ] there is a bijection between @xmath85^{m-1}\times s$ ] and the set of rooted forests on vertex - set @xmath85 $ ] with root vertices @xmath209 that contain @xmath196 as a subforest . we describe such a mapping , which we call the `` forest completion algorithm '' ( fca ) . consider the @xmath210-tuple @xmath211^{m-1}\times s$ ] . we construct the forest corresponding to @xmath119 iteratively , in @xmath212 stages @xmath213 . at every stage , we have a forest @xmath214 containing @xmath196 as a subforest , a permutation @xmath215 of @xmath216 $ ] , and a sequence @xmath217 in @xmath85^{m}$ ] . initially , at stage @xmath218 , we have @xmath219 , @xmath215 is the identity permutation and @xmath220 . then , for @xmath221 : * if @xmath222 is in a different component of @xmath214 from @xmath223 , then add an arc directed from @xmath223 to @xmath222 in @xmath214 , and leave @xmath215 and @xmath123 unchanged ; * otherwise ( so @xmath222 is in the same component of @xmath214 as @xmath223 ) , add an arc directed from @xmath223 to @xmath224 in @xmath214 ( to obtain the new @xmath214 ) , switch @xmath225 and @xmath226 in @xmath215 , and switch @xmath222 and @xmath224 in @xmath123 . the forest corresponding to the @xmath210-tuple @xmath119 is the terminating forest @xmath214 . we call the terminating permutation @xmath215 the `` forest completion permutation '' ( fcp ) . the significance of the fcp is that it identifies precisely the arcs that are added to @xmath196 they are @xmath227 , @xmath221 . in our examples throughout the paper , we shall specify the second line in the two line representation of @xmath215 the list of images @xmath228 . in figure [ foralgo ] we give an example of the fca with @xmath229 , @xmath230 , @xmath231 . the trees @xmath232 , @xmath233 , @xmath234 , with @xmath235 , @xmath236 , @xmath237 , are given in the box at the top left ; the trees @xmath238 , @xmath239 , with @xmath240 , @xmath241 , are given in the box at the top right . then , corresponding to the triple @xmath242 , we construct the forest at the bottom of figure [ foralgo ] . the corresponding fcp is @xmath243 . in analyzing this mapping , it is convenient to use the term `` safe '' to describe a vertex in a component of a forest rooted at one of the vertices @xmath244 . thus , initially , @xmath224 is safe . it is trivial to prove by induction that , after stage @xmath116 , for @xmath245 , @xmath214 is a forest with root vertices @xmath246 , and that @xmath224 is safe for @xmath214 ( which implies that @xmath247 are in different components of @xmath214 from @xmath224 ) . thus , at stage @xmath210 , @xmath224 is indeed in a different component of @xmath214 from @xmath248 , so we successfully add the final arc from @xmath248 to @xmath224 , to obtain a terminating forest @xmath214 rooted at @xmath244 ( this explains our use of `` safe '' for extending our forest , it is always safe to add the new arc directed to @xmath224 ) . this proves that the fca does indeed produce a rooted forest on vertex - set @xmath85 $ ] with root vertices @xmath244 that contains @xmath196 as a subforest . to prove that the fca is a bijection , we describe its inverse . suppose that we are given a rooted forest @xmath249 on vertex - set @xmath85 $ ] , with root vertices @xmath244 , and that we wish to remove the arcs directed from @xmath223 to @xmath250 , @xmath221 , where @xmath251 are distinct non - root vertices , with the convention that @xmath205 . we proceed iteratively , through stages @xmath252 . at every stage we have a subforest @xmath214 of @xmath249 , a permutation @xmath253 of @xmath216 $ ] , and a sequence @xmath217 in @xmath85^m$ ] . initially , at stage @xmath210 , we have @xmath254 , @xmath253 is the identity permutation , and @xmath255 . then , for @xmath256 , remove the arc from @xmath257 to @xmath258 in @xmath214 ( to get the new @xmath214 ) , and : * if @xmath224 is safe for ( the new ) @xmath214 , leave @xmath253 and @xmath123 unchanged ; * otherwise ( so @xmath224 is not safe for @xmath214 ) , switch @xmath259 and @xmath260 in @xmath253 , and switch @xmath258 and @xmath224 in @xmath123 . we claim that the @xmath210-tuple corresponding to the forest @xmath249 is the terminating @xmath210-tuple @xmath123 , so that this mapping uniquely reverses the fca . in fact , it is easy to establish that this mapping uniquely reverses the fca stage by stage , since it is trivial to prove by induction that the values of @xmath123 and @xmath214 after stage @xmath116 of the above mapping are exactly the same as @xmath123 and @xmath214 after stage @xmath116 of the fca . it is also easy to prove that the fcp is given by @xmath261 for the terminating @xmath253 . the result follows , since the fca is a bijection between the required sets . of course , it is an immediate enumerative consequence of theorem [ fct ] that there are @xmath262 rooted forests on vertex - set @xmath85 $ ] with root vertices @xmath209 that contain @xmath196 as a subforest . in the special case @xmath263 ( so that @xmath196 has no edges ) this gives the classical result that there are @xmath264 rooted forests on vertex - set @xmath85 $ ] , with a prescribed set of @xmath3 root vertices ( see , _ e.g. _ , [ [ stan ] ] , p. 25 ) . a _ vertical _ paired array is a paired array in which all pairs are mixed . we define @xmath265 to be the set of vertical paired arrays in @xmath266 , in which there are @xmath267 marked cells in row @xmath50 and @xmath268 marked cells in row @xmath2 , for @xmath269 , and let @xmath270 . in our next result , we remove non - mixed pairs from a minimal , canonical paired array , and thus show that the enumeration of minimal , canonical paired arrays can be reduced to the enumeration of vertical paired arrays . we use the following notation . for a finite set @xmath271 , let @xmath272 denote the set of @xmath116-tuples consisting of @xmath116 distinct elements of @xmath271 . thus @xmath273 , where @xmath274 and @xmath275 is the _ falling factorial _ : for positive integers @xmath116 , @xmath276 ; for @xmath277 , @xmath278 ; otherwise @xmath279 . [ nonmixedprop ] for @xmath40 of the same odd - even parity , let @xmath280 and @xmath281 . then @xmath282 note that every element of @xmath160 has exactly @xmath116 and @xmath151 non - mixed pairs in the top and bottom rows , respectively . taking an arbitrary @xmath283 , we now describe a mapping that is initially identical to that used in the proof of theorem [ redundmin ] . we attach the numbers @xmath284 to the vertices and the small box in row @xmath50 , using the same left to right convention as in the proof of theorem [ redundmin ] . let the pairs of numbers attached to the non - mixed pairs in row @xmath50 be denoted by @xmath285 , where @xmath286 are attached to the rightmost vertices in these pairs ( with @xmath287 ) , and @xmath288 are attached to the other ( not rightmost in their cell ) vertices in these pairs . suppose that the marked cell in row @xmath50 is in column @xmath210 , and that the rightmost tree for row @xmath50 of @xmath173 is @xmath120 , so @xmath120 is rooted at vertex @xmath210 . now run the inverse of the fca on @xmath120 , to remove the arcs directed from @xmath289 to @xmath290 , @xmath291 , ( here , @xmath292 denotes the column in which the number @xmath293 appears ) and let @xmath294 be the corresponding fcp . let @xmath295 . we follow the analogous procedure for row @xmath2 : we attach primed numbers @xmath176 to the vertices and small box in row @xmath2 . let the pairs of numbers attached to the non - mixed pairs in row @xmath50 be denoted by @xmath296 , where @xmath297 are attached to the rightmost vertices in these pairs ( with @xmath298 ) , and @xmath299 are attached to the other ( not rightmost in their cell ) vertices in these pairs . suppose that the marked cell in row @xmath2 is in column @xmath3 , and that the rightmost tree for row @xmath2 of @xmath173 is @xmath124 , so @xmath124 is rooted at vertex @xmath3 . now run the inverse of the fca on @xmath124 , to remove the arcs directed from @xmath300 to @xmath301 , @xmath291 , and let @xmath302 be the corresponding fcp . let @xmath303 . finally , we mark the cells in row @xmath50 containing @xmath288 ( in addition to the existing marked cell in column @xmath210 ) , and the cells in row @xmath2 containing @xmath299 ( in addition to the existing marked cell in column @xmath3 ) , and then remove all non - mixed pairs ( both vertices and edges ) from @xmath173 , to get the vertical paired array @xmath172 . the vertices in each cell of @xmath172 have the same relative order as they did in @xmath173 . for example , suppose that @xmath173 is the paired array in figure [ minarr ] . then we have @xmath304 , with @xmath305 , @xmath306 , @xmath307 , @xmath308 , @xmath309 , @xmath310 , and @xmath311 , with @xmath312 , @xmath313 , @xmath314 , @xmath315 . we have @xmath230 , @xmath316 , and the trees @xmath120 and @xmath124 are given in figure [ newtrees ] . when we run the inverse of the fca on @xmath120 to remove the arcs @xmath317 , @xmath318 and @xmath319 ( for which we use dashed lines in figure [ newtrees ] ) , we obtain @xmath320 as the fcp , which gives @xmath321 . when we run the inverse of the fca on @xmath124 to remove the arcs @xmath322 and @xmath323 ( for which we use dashed lines in figure [ newtrees ] ) , we obtain @xmath324 as the fcp , which gives @xmath325 . the vertical paired array @xmath172 in this example is given in figure [ vertarr ] . now , for the same reasons as in the proof of theorem [ redundmin ] , we have @xmath326 for @xmath327 , so @xmath328,i}$ ] , and similarly , @xmath329',j}$ ] . thus we have a mapping @xmath330,i}\times{\mathcal{l}}_{[q]',j}\times{\mathcal{v}}^{(s)}_{k , i , j } : { \alpha}\mapsto ( { \kappa}_1,{\kappa}_2,{\beta}).\ ] ] in fact , @xmath331 is a bijection . before proving this , for @xmath332 , we first note the key dependencies between @xmath173 and @xmath333 that follow from the fca : let @xmath196 and @xmath334 denote the rightmost forests of @xmath172 for rows @xmath50 and @xmath2 , respectively . then , in @xmath173 with numbers attached as in the construction above , the column containing the vertex whose number is the last entry of @xmath335 is contained in the component of @xmath196 rooted at @xmath210 . similarly , the column containing the vertex whose number is the last entry of @xmath336 is contained in the component of @xmath334 rooted at @xmath3 . for example , for @xmath337 given in figures [ minarr][vertarr ] , the last entries of @xmath335 and @xmath336 are @xmath338 and @xmath339 , respectively , corresponding to vertices of @xmath173 in columns @xmath340 and @xmath341 , respectively . now , the forests @xmath196 and @xmath334 in this case are obtained from the trees @xmath120 and @xmath124 , respectively , by removing the dashed edges in figure [ newtrees ] . then , indeed , vertex @xmath340 is contained in the component of @xmath196 rooted at @xmath230 , and vertex @xmath341 is contained in the component of @xmath334 rooted at @xmath316 . we now prove that @xmath331 is a bijection by describing the inverse mapping , so that we can uniquely recover @xmath173 from an arbitrary triple @xmath342,i}\times{\mathcal{l}}_{[q]',j}\times{\mathcal{v}}^{(s)}_{k , i , j}$ ] . let @xmath196 denote the rightmost forest of @xmath172 for row @xmath50 . let @xmath343 denote the set consisting of the entries in @xmath335 , and let @xmath344 , @xmath345\setminus\{{\kappa}_1\}$ ] . next ( as a generalization of the procedure for the inverse of @xmath181 described in the proof of theorem [ redundmin ] ) , we number the vertices and small boxes in row @xmath50 of @xmath172 with the elements of @xmath346 . then insert vertices numbered with the elements of @xmath343 , so that the numbers on all vertices and small boxes increase from left to right ( with any small box regarded as the rightmost object in its cell ) and so that the the vertex numbered with @xmath347 is placed in the same cell as the object numbered with @xmath348 . now suppose that @xmath349 , and use the key dependency noted above : let the column containing the vertex numbered @xmath350 be contained in the component of @xmath196 rooted at vertex @xmath210 . let @xmath287 denote the numbers attached to the small boxes that are not in column @xmath210 ( the columns containing @xmath286 are the root vertices for the components of @xmath196 not rooted at @xmath210 ) . now , apply the fca on @xmath116-tuple @xmath351 , to give the tree @xmath120 rooted at @xmath352 , that contains the forest @xmath196 as a subforest . let @xmath294 be the fcp . finally , replace the small boxes numbered @xmath286 by rightmost vertices ( in the same cells ) numbered @xmath286 , pair the vertex numbered @xmath353 with the vertex numbered @xmath354 , @xmath327 , and remove the numbers from row @xmath50 . repeat the analogous process for row @xmath2 , and we arrive at a minimal paired array @xmath173 . it is straightforward to check that the process described above reverses @xmath331 . thus , we have described @xmath355 , and our proof that @xmath331 is a bijection is complete . the result follows immediately . for every column of a vertical paired array , the cells in rows @xmath50 and @xmath2 have the same number of vertices , because of the balance condition . full _ , vertical paired array is a vertical paired array with a positive number of vertices in every column . let @xmath356 be the set of full , vertical paired arrays in @xmath265 , and @xmath357 . in theorem [ nonemvert ] we shall give an explicit construction for the elements of @xmath356 , and thus obtain an explicit formula for @xmath358 . to help in the proof of this result , we first introduce some terminology and notation associated with an arbitrary @xmath359 . a vertex is said to be _ dependent _ if it is paired with the rightmost vertex of an unmarked cell ( in the other row ) . if the rightmost vertex of an unmarked cell in row @xmath50 , column @xmath116 is paired with the rightmost vertex of an unmarked cell in row @xmath2 , column @xmath151 , then we call this a _ shared pair _ of @xmath173 . in this case , the rightmost forest @xmath196 for row @xmath50 of @xmath173 contains arc @xmath360 and the rightmost forest @xmath334 for row @xmath2 of @xmath173 contains arc @xmath361 , and we call each of these a _ shared arc_. now , canonically number the vertices in row @xmath50 of @xmath173 @xmath362 , from left to right , and number the vertices in row @xmath2 of @xmath173 @xmath363 , from left to right . let @xmath364 be the subforest of @xmath196 with only the shared arcs of @xmath196 . suppose that @xmath196 has @xmath73 non - shared arcs , corresponding to pairs @xmath365 , where @xmath366 , and @xmath367 are rightmost vertices in their ( unmarked ) cells . run the inverse of the fca on the forest @xmath196 , to obtain the subforest @xmath364 by removing the non - shared arcs directed from @xmath368 to @xmath369 , @xmath370 , and let @xmath302 be the corresponding fcp . define @xmath371 . if all arcs of @xmath196 are shared , then let @xmath372 be the vertex in row @xmath2 that is paired with the rightmost non - dependent vertex in row @xmath50 ( we call this the _ non - fca option _ ) . in both cases , define @xmath373 , and let @xmath374 be the vertices on the unique directed path in @xmath364 from vertex @xmath375 to the root vertex ( @xmath376 ) of the component of @xmath364 containing @xmath375 . thus @xmath377 , and the cell in row @xmath2 , column @xmath378 is marked , and the cell in row @xmath2 , column @xmath379 is not marked , @xmath380 . also , the cell in row @xmath50 , column @xmath379 is not marked , @xmath381 . now define @xmath382 to be the subforest of @xmath334 containing the arcs @xmath383 , @xmath381 . suppose that @xmath334 has @xmath210 arcs that are not in @xmath382 , corresponding to pairs @xmath384 , where @xmath385 , and @xmath386 are rightmost vertices in their ( unmarked ) cells . run the inverse of the fca on the forest @xmath334 , to obtain the subforest @xmath382 by removing the arcs directed from @xmath387 to @xmath388 , @xmath389 , and let @xmath390 be the corresponding fcp . define @xmath391 . if @xmath392 , let @xmath123 be the vertex in row @xmath50 that is paired with the rightmost non - dependent vertex in row @xmath2 ( again , we call this the _ non - fca option _ ) . let @xmath393 , which we call the _ tail _ of @xmath173 . the tail length is @xmath394 . we say that @xmath123 is _ in the tail _ when the column containing vertex @xmath123 is one of @xmath395 . the _ type _ of @xmath173 is given by @xmath396 . if the cells in rows @xmath50 and @xmath2 of column @xmath293 in @xmath173 have @xmath397 vertices , @xmath398 , then we say that @xmath173 has _ shape _ note that @xmath400 , and that @xmath397 is positive for all @xmath398 , so @xmath401 is a _ composition _ of @xmath57 with @xmath20 parts . for example , suppose that @xmath173 is the full , vertical paired array given at the top of figure [ verteg ] , with @xmath402 , @xmath403 , @xmath404 , @xmath405 , shape @xmath406 , and rightmost forests @xmath196 and @xmath334 given at the bottom of figure [ verteg ] . the lines joining the pairs in @xmath173 are of various types ( and the same type of line is used in the rightmost forests when the pair corresponds to an edge in one or other of these forests ) : a thick solid line indicates a shared pair in the tail , a thick dashed line a shared pair not in the tail , a thick dashed and dotted line a pair contributing to @xmath196 only , a thin solid line a pair contributing to @xmath334 only , and a thin dashed line a pair that contributes to neither of @xmath407 . when we run the inverse of the fca to remove the arcs @xmath408 and @xmath409 from @xmath196 , we obtain the ordered pair @xmath408 , and hence obtain @xmath410 as indicated in figure [ verteg ] , contained in column @xmath2 . thus the tail is @xmath411 , of length @xmath412 . when we run the inverse of the fca to remove the arcs @xmath413 , @xmath414 , @xmath415 , @xmath416 and @xmath417 from @xmath334 , we obtain the @xmath418-tuple @xmath419 , and hence obtain @xmath420 as indicated in figure [ verteg ] , contained in column @xmath421 . thus we conclude that @xmath173 has type @xmath422 . [ nonemvert ] for @xmath269 , @xmath423 , we have @xmath424 each paired array in @xmath356 has a unique type and shape , and we can uniquely construct those of given type and shape as follows . for the given shape , we begin with a @xmath2 by @xmath20 array , with each cell containing an ordered set of vertices of prescribed size . then we pair these vertices ( all are mixed pairs ) in all possible ways for the given type in four stages . first , we pair the rightmost vertices as prescribed by the tail . second , we use the fca to pair the rightmost vertices in row @xmath2 ( note that @xmath123 is safe , by construction ) . third , we use the fca to pair the unpaired rightmost vertices in row @xmath50 ( note that @xmath372 is safe , by construction ) . fourth , we pair the remaining vertices arbitrarily . ( in addition , along the way , we have to choose the marked cells in a consistent fashion . ) in this way , for each composition @xmath401 of @xmath57 with @xmath20 parts and @xmath377 , we enumerate elements of @xmath356 with shape @xmath401 and tail of length @xmath394 . there are three cases : * ( vertex @xmath372 is not rightmost in its cell ) : * there are @xmath425 choices for @xmath372 , which then fixes @xmath378 . there are then @xmath426 choices for @xmath427 , and then @xmath394 of the pairs are determined . this leaves @xmath428 choices for @xmath123 ( any vertex not yet paired in row @xmath50 ) . now mark the cell in row @xmath50 , column @xmath376 . also , if @xmath123 is in the tail , mark the cell in row @xmath2 , column @xmath378 , or if @xmath123 is not in the tail , mark the cell in row @xmath2 of the column that contains @xmath123 ( @xmath123 is in row @xmath50 ) . choose , from the @xmath429 cells not in the tail or already marked , @xmath116 cells to mark in the top row , and @xmath151 cells to mark in the bottom row . now pair the unpaired rightmost vertices of unmarked cells in row @xmath2 with vertices in row @xmath50 to satisfy the forest condition , using vertex @xmath123 as the safe position . there are @xmath430 possible choices for this , from the fca . suppose that there are @xmath3 unmarked cells in row @xmath50 whose rightmost vertices are not yet paired . then pair these with vertices in row @xmath2 to satisfy the forest condition , using vertex @xmath372 as the safe position . there are @xmath431 possible choices for this , from the fca . finally , there are @xmath432 ways to pair the remaining vertices , arbitrarily . but @xmath433 , and we conclude that the number of elements in @xmath356 with shape @xmath401 and tail of length @xmath394 in this case is @xmath434 ( it is straightforward the check that these cardinalities are correct when we use the non - fca options also . ) * ( vertex @xmath372 is rightmost in its cell , and @xmath123 is in the tail ) : * the number of choices for @xmath123 and @xmath294 is @xmath435 . then @xmath372 , and @xmath394 of the pairs , are uniquely determined . the cell in row @xmath50 , column @xmath376 is marked . the cell in row @xmath2 , column @xmath378 is marked for two reasons because @xmath123 is in the tail , and because the cell contains @xmath372 , rightmost , so that @xmath372 will be paired with a non - dependent vertex in row @xmath50 . the rest of the enumeration proceeds as in case 1 , and we conclude that the number of elements in @xmath356 with shape @xmath401 and tail of length @xmath394 in this case is @xmath436 * ( vertex @xmath372 is rightmost in its cell , and @xmath123 is not in the tail ) : * the number of choices for @xmath123 and @xmath294 is @xmath437 , and then @xmath372 together with @xmath394 of the pairs are uniquely determined . in this case three different cells must now be marked : in row @xmath50 , column @xmath376 ; in row @xmath2 , column @xmath376 ; in row @xmath2 of the column that contains @xmath123 . there are then @xmath438 ways to choose which other cells are marked , and the rest of the enumeration proceeds as in cases 1 and 2 . we conclude that the number of elements in @xmath356 with shape @xmath401 and tail of length @xmath394 in this case is @xmath439 adding ( [ case1claim2 ] ) , ( [ case2claim2 ] ) , and ( [ case3claim2 ] ) , and simplifying , we obtain that the total number of elements in @xmath356 with shape @xmath401 and tail of length @xmath394 is @xmath440 and the result follows , by summing over @xmath377 and multiplying by @xmath441 , the number of choices for @xmath401 . in the next result , we give an explicit enumeration for vertical paired arrays , by applying theorem [ nonemvert ] . the proof is quite technical , involving generating functions and a hypergeometric summation . [ closedform ] for @xmath269 , @xmath423 , we have @xmath442\ ] ] if a column in a vertical paired array has no vertices , then at least one of the cells in rows @xmath50 and @xmath2 of that column must be marked , from the nonempty condition . thus , suppose that a vertical paired array with @xmath57 mixed pairs and @xmath20 columns , has @xmath443 columns with no vertices and @xmath210 with a positive number of vertices ( the same number in both rows of such columns ) . of the @xmath443 columns with no vertices , suppose that @xmath119 are marked in row @xmath50 only , @xmath123 are marked in row @xmath2 only , and that @xmath444 are marked in both row @xmath50 and @xmath2 ( we use this parameterization for convenience in determining the summations below ) . then we have @xmath445 where @xmath446 but , from theorem [ nonemvert ] , we have @xmath447 \sum_{s-1\ge m-1\ge l\ge 0 } l } \big(yz\big)^{l}\big ( ( 1+y)(1+z)\big)^{m-1-l } , \end{aligned}\ ] ] where we use the notation @xmath448b$ ] to denote the coefficient of @xmath375 in the expansion of @xmath449 . now @xmath450x^m(1-x)^{-m-1},\ ] ] which gives @xmath451g(x , y , z),\ ] ] where , summing over @xmath210 by the binomial theorem , we have @xmath452 but , changing variables in ( [ coeffsf ] ) , we have @xmath453x^k g(x,{\frac}{y}{x},{\frac}{z}{x}),\ ] ] so from ( [ tsum ] ) we obtain @xmath454x^k ( 1+y+z)^k g(x,{\frac}{y}{x},{\frac}{z}{x})\\ & = & s!\ , [ x^0y^{k}z^{k}]x^{k - i - j - s-1}y^{i+1}z^{j+1}{\frac}{(1+y+z)^{k-1}}{(1-x)^{s+1 } } \big(\big ( x(1+y+z)+yz\big ) ^s-\big ( yz\big ) ^s\big)\\ & = & r_1-r_2 , \end{aligned}\ ] ] where @xmath455{\frac}{(1+y+z)^{k-1}}{(1-x)^{s+1 } } ( yz)^s\\ & = & s!{2s+i+j - k+1\choose s}{\frac}{(k-1)!}{(k - s - i-1)!(k - s - j-1)!(2s+i+j - k+1)!}\\ & = & { \frac}{(s+i ) ! ( s+j)!}{(s+i+j)!}{s+i+j \choose k-1 } { k-1 \choose s+i}{k-1 \choose s+j},\\ r_1&=&s!\ , [ x^{i+j+s - k+1}y^{k - i-1}z^{k - j-1}]{\frac}{(1+y+z)^{k-1}}{(1-x)^{s+1 } } \big ( x(1+y+z)+yz\big ) ^s\\ & = & s!\ , [ x^{i+j+s - k+1}y^{k - i-1}z^{k - j-1}]\sum_{m\ge 0}{s\choose m } x^{s - m}(1+y+z)^{k+s - m-1}(yz)^{m}(1-x)^{-s-1}\\ & = & s!\sum_{m\ge 0}{s\choose m}{s+i+j+m - k+1\choose s } { \frac}{(s - m+k-1)!}{(k - m - i-1)!(k - m - j-1)!(s+i+j+m - k+1)!}.\end{aligned}\ ] ] now , for this latter sum over @xmath456 , we observe that the ratio of the @xmath212st term to the @xmath210th term is a rational function of @xmath210 , which implies that it is a hypergeometric sum . in particular , using the standard notation for hypergeometric series , we have @xmath457 where the second last equality follows from the pfaff - saalschtz theorem for @xmath458 hypergeometric summations ( see , _ e.g. _ , theorem 2.2.6 on page 69 of [ [ aar ] ] ) . the result follows immediately . we finish with the proof of our main result , now that we have completed all the intermediate results in our reduction . _ proof of theorem [ mainthm]_. from proposition [ fallfac ] , ( [ equivbe ] ) and theorems [ redundmin ] , [ nonmixedprop ] , we obtain @xmath459 but , simplifying , we obtain @xmath460 and the result follows from theorem [ closedform ] . this work was supported by a discovery grant from nserc ( ig ) and by a usra award from nserc ( ws ) . we are grateful to andu nica for helpful discussions and suggestions . [ aar ] g. e. andrews , r. askey and r. roy , `` special functions '' , encyclopedia of mathematics and its applications , vol . 71 , cambridge university press , 1999 . 2 . [ ht ] u. haagerup and s. thorbjornsen , _ random matrices with complex gaussian entries _ , expositiones math . , * 21 * ( 2003 ) , 293337 . [ gn ] i. p. goulden and a. nica , _ a direct bijection for the harer - zagier formula _ , j. comb . theory ( a ) , * 111 * ( 2005 ) , 224238 . [ hz ] j. harer and d. zagier , _ the euler characteristic of the moduli space of curves _ , inventiones mathematicae , * 85 * ( 1986 ) , 457486 . 5 . [ iz ] c. itzykson and j .- b . zuber , _ matrix integration and combinatorics of modular groups _ , phys . , * 134 * ( 1990 ) , 197207 . 6 . [ j ] d. m. jackson , _ on an integral representation for the genus series for @xmath2-cell embeddings _ , trans . soc . , * 344 * ( 1994 ) , 755772 . 7 . [ ke ] s. kerov , _ rook placements on ferrers boards , and matrix integrals _ ( english translation ) , j. math . ( new york ) , * 96 * ( 1999 ) , 35313536 . 8 . [ k ] m. kontsevich , _ intersection theory on the moduli space of curves and the matrix airy function _ , comm . phys . , * 147 * ( 1992 ) , 123 . [ lz ] s. k. lando and a. k. zvonkin , `` graphs on surfaces and their applications '' , springer - verlag , berlin , heidelberg , 2004 . 10 . [ l ] b. lass , _ dmonstration combinatoire de la formule de harer - zagier _ , c. r. acad . paris , * 333 * , srie i ( 2001 ) , 155160 . 11 . [ mn ] j. a. mingo and a. nica , _ annular noncrossing permutations and partitions , and second - order asymptotics for random matrices _ , international math . research notices , no . 28 ( 2004 ) , 1413 1460 . 12 . [ p ] r. c. penner , _ perturbative series and the moduli space of riemann surfaces _ , j. diff . geometry , * 27 * ( 1988 ) , 3553 . [ stan ] r. p. stanley , `` enumerative combinatorics '' , vol . 2 , cambridge university press , 1999 . 14 . [ t ] w. t. tutte , `` graph theory '' , addison - wesley , london , 1984 . 15 . [ z ] d. zagier , _ on the distribution of the number of cycles of elements in symmetric groups _ , nieuw archief voor wiskunde , * 13 * ( 1995 ) , 489495 . 16 . [ zv ] a. zvonkin , _ matrix integrals and map enumeration : an accessible introduction _ , combinatorics and physics ( marseilles , 1995 ) , math . modelling , * 26 * ( 1997 ) , 281304 .

based on recent first - principles computations in perovskite compounds , especially batio , we examine the significance of the born effective charge concept and contrast it with other atomic charge definitions , either static ( mulliken , bader ... ) or dynamical ( callen , szigeti ... ) . it is shown that static and dynamical charges are not driven by the same underlying parameters . a unified treatment of dynamical charges in periodic solids and large clusters is proposed . the origin of the difference between static and dynamical charges is discussed in terms of local polarizability and delocalized transfers of charge : local models succeed in reproducing anomalous effective charges thanks to large atomic polarizabilities but , in abo compounds , ab initio calculations favor the physical picture based upon transfer of charges . various results concerning barium and strontium titanates are presented . the origin of anomalous born effective charges is discussed thanks to a band - by - band decomposition which allows to identify the displacement of the wannier center of separated bands induced by an atomic displacement . the sensitivity of the born effective charges to microscopic and macroscopic strains is examined . finally , we estimate the spontaneous polarization in the four phases of barium titanate .

cond-mat9805013

for a long time , there has been a continuing interest in the definition of atomic charges in solid state physics as well as in chemistry @xcite . this interest lies essentially in the fact that the concept of atomic charge naturally arises in a large diversity of frameworks and is frequently helpful for a simple description of solids and molecules . the variety of contexts in which the charge is involved ( ir spectrum analysis , xps chemical shifts analysis , theory of ionic conductivity of oxides , determination of electrostatic potential , definition of oxidation states ... ) underlines its central role but also reveals a concomitant problem : inspired by various models or by the description of various physical phenomena , many different definitions have been proposed that , unfortunately , are not equivalent @xcite . following a distinction already made by cochran @xcite , it seems possible to classify the different concepts into static and dynamical charges . the _ static _ charge is an intuitive concept , usually based on a partitioning of the ground - state electronic density into contributions attributed to the different atoms . it is an ill - defined quantity that depends on the convention artificially chosen to affect a given electron to a particular ion @xcite . on the other hand , the _ dynamical _ charge is directly related to the change of polarization ( or dipole moment , for molecules ) created by an atomic displacement . this change of polarization is a quantity that can be experimentally measured , at least in principles , giving the dynamical charge a well - defined character . in order to clarify the concept of atomic charge , it was important to compare on practical examples the numerical results obtained from its different definitions . recent studies of the statistical correlation between various atomic charges using a principal component analysis have suggested that the different definitions are not independent but correspond to different scales driven by a unique underlying physical factor @xcite . if this assertion seems plausible as far as static charges are concerned , we will argue that the dynamical charge should not reduce to the same physical factor but should also depend on an additional parameter : the rate of transfer of charge , influenced by the bonding with the other atoms of the system and additionally , for large systems , by the condition imposed on the macroscopic electric field . the born effective charge tensor @xmath1 alias transverse charge , that is at the center of the present study , is a dynamical quantity introduced by born @xcite in 1933 . in solid state physics , it is since a long time considered as a fundamental quantity because it monitors the long - range coulomb interaction responsible of the splitting between transverse and longitudinal optic phonon modes @xcite . during the seventies , the born effective charges were already investigated and discussed within empirical approaches ( see for example harrison @xcite ) . more recently , they became accessible to first - principles calculations @xcite , and accurate values have been reported for a large variety of materials . for the case of abo@xmath0 compounds , old experimental data @xcite and empirical studies @xcite had suggested that the amplitude of the born effective charges should deviate substantially from the amplitude of the static atomic charge . surprisingly , this result remained in the dark until first - principles calculations confirmed that the components of @xmath1 are anomalously large in these oxides @xcite . it was observed that the components of @xmath1 can reach twice that of the nominal ionic charges . this result reopened the discussion on the physics of the born effective charges and different recent studies tried to clarify the microscopic processes monitoring the amplitude of @xmath1 . in this paper , we first clarify the relationship between various atomic charges . we then present results concerning batio@xmath0 and srtio@xmath0 in order to illustrate how a careful analysis of the born effective charges can teach us interesting physics concerning these compounds . it reveals the mixed ionic and covalent character of the bond @xcite . it allows to visualize the mechanism of polarization as electronic currents produced by dynamical changes of orbital hybridizations @xcite . it also clarifies the origin of the giant destabilizing dipole - dipole interaction producing the ferroelectric instability of these materials @xcite . in section ii and iii , we contrast the concepts of static and dynamical charges and we reintroduce the born effective charge that is at the center of the present discussion . in section iv , we compare various results obtained within different frameworks for the cubic phase of batio@xmath0 and srtio@xmath0 . the bond orbital model of harrison is explicitely applied to srtio@xmath0 ( appendix a ) . we also discuss the origin of the large anomalous contributions in terms of local electronic polarizability and dynamical changes of orbital hybridization . a decomposition of the role played by the different bands is reported in section v. section vi is devoted to the evolution of the born effective charges in the three ferroelectric phases of batio@xmath0 as well as in the cubic phase under hydrostatic pressure . this points out the role of the anisotropy of the atomic environment on the amplitude of @xmath1 . finally , in section vii , we report the evolution of the effective charges all along the path of atomic displacements from the cubic to the rhombohedral phase and we estimate the spontaneous polarization of the three ferroelectric phase of batio@xmath0 . intuitively , the atomic charge may first appear as a static concept . the charge associated to an isolated atom is a well - defined quantity . the purpose of defining static atomic charges is therefore to extend this notion to molecules and solids . for these cases , the challenge basically consists to replace the delocalized electronic density by localized point charges associated to each atom . this could _ a priori _ be performed from electronic density maps obtained experimentally or theoretically . however , as already mentioned by mulliken @xcite in 1935 , `` there are some difficulties of giving exact definition without arbitrariness for any atomic property '' . during the seventies , cochran @xcite similarly emphasized that the partition of the electronic distribution into atomic charges can only be done unambiguously when `` boundary can be drawn between the ions so as to pass through regions in which the electron density is small compared with the reciprocal of the volume inclosed '' . this is never the case in practice , and especially when there is appreciable covalent bonding . for most of the solids and molecules , there is consequently no _ absolute _ criterion to define the static atomic charge and a large variety of distinct definitions have been proposed that are not necessarily quantatively equivalent ( see for instance ref . @xcite ) . as an illustration , different approaches have been considered in order to evaluate the amplitude of the static atomic charges of barium titanate . some results are summarized in table [ zcubic ] , where different atomic charges are reported in comparison with the nominal charges expected in a purely ionic material ( + 2 for ba , + 4 for ti , 2 for o ) . some of them were obtained from empirical models ; others were deduced from first - principles . the static atomic charges of ref . @xcite were deduced by harrison within his bond orbital model using universal parameters and neglecting the interactions with the ba atom . the atomic charges reported by hewat @xcite were approximated from a model initially designed by cowley @xcite for srtio@xmath0 . the charges reported by khatib _ et al . _ @xcite have been obtained in a shell - model context . in two references , turik and khasabov @xcite estimated the charges from the madelung constant thanks to a fit of the crystal energy with shell - model parameters . michel - calendini _ et al . _ @xcite proposed charges from a population analysis of the x@xmath2 electronic distribution of a tio@xmath3 cluster , assuming a charge of @xmath4 on ba . cohen and krakauer @xcite deduced the atomic charges from a fit of the dft - lda electronic distribution by that of overlapping spherical ions ( generated according to the potential induced breathing model ) for different ionic configurations . _ et al . _ @xcite reported values deduced from a mulliken population analysis of a self - consistent olcao calculation @xcite . in another reference @xcite , xu _ et al . _ proposed different values by integrating the electronic charges in spheres centered on the ions , and partitioning rather arbitrarily the remaining charge outside the spheres following a method proposed in ref . @xcite . the results of table [ zcubic ] are not quantitatively identical and illustrate that there is no formal equivalence between the different procedures used to define the atomic charge . however , in agreement with an analysis reported by meister and schwartz @xcite for the case of molecules , we observe that the values of table [ zcubic ] have some common features , suggesting that the different charges are not independent but should correspond to different scales driven by a common factor . in particular , all the calculations reveal that the charge transfer from ti to o is not complete . if batio@xmath0 was a purely ionic crystal , the 3d and 4s electrons of ti would be entirely transferred to the oxygen atoms , yielding a charge of + 4 on titanium . however , due for instance to the partial hybridization between o 2p and ti 3d states @xcite , these electrons remain partly delocalized on the ti atom so that the static charges on the ti and o atoms are smaller than they would be in a purely ionic material . this delocalization is illustrated in figure [ fig.1 ] , where we have plotted the partial electronic density associated to the o 2p bands . for the ba atom , the situation is not so clear than for titanium but most of the studies suggest similarly that the 6s electrons are not fully transferred to the oxygen . from the previous examples , it is clear that , strictly speaking , the static charge does not give a quantitative information . in the study of mixed ionic - covalent compounds , it remains however a useful concept to discuss qualitatively the transfer of charges from one atom to the other . as a general rule , the partial covalence reduces the amplitude of the static charge . comparison of a specific static charge in the different phases of a given material or in different compounds can therefore give a relevant information on the evolution of the chemical bond @xcite . as emphasized by harrison @xcite , `` whenever an ambiguity arises about the definition of a concept such as the atomic charge , it can be removed by discussing only quantities that can be experimentally determined , at least in principles '' . if there are some ambiguities to determine the charge to be associated to a given atom , the charge carried by this atom when it is displaced is directly accessible from the induced change of polarization ( or dipole moment for molecules ) . as it is now discussed , the dynamical charges are defined by the change of polarization induced by an atomic displacement ; from the viewpoint of harrison , they appear therefore as more fondamental quantities . in molecules , the change of dipole moment in direction @xmath5 ( @xmath6 ) linearly induced by a small displacement of atom @xmath7 in direction @xmath2 ( @xmath8 ) is uniquely defined . the proportionality coefficient between the dipole moment and the atomic displacement has the dimentionality of a charge and is usually referred to as the _ atomic polar tensor _ ( apt ) : @xmath9 this concept was introduced by biarge , herranz and morcillo @xcite for the interpretation of infra - red intensities measurements . later , cioslowski @xcite introduced a scalar charge in connection with this tensor : it is the generalized atomic polar tensor ( gapt ) defined as one - third of the trace of the atomic polar tensor . in periodic systems , equivalent atoms appear in the different unit cells and the definition of the charge can be generalized . a dynamical charge tensor is conventionally defined as the coefficient of proportionality at the linear order between the _ macroscopic polarization _ per unit cell created in direction @xmath5 and a rigid displacement of the _ sublattice _ of atoms @xmath7 in direction @xmath2 , times the unit cell volume @xmath10 : @xmath11 we note that @xmath12 can be interpreted as a dipole moment per unit cell . as one @xmath7 atom is displaced in each unit cell , in the linear regime , this definition is equivalent to eq . ( [ zmol ] ) : it corresponds to the change of dipole moment induced by an isolated atomic displacement . however , contrary to the case of molecules , in macroscopic systems , the previous quantity is not uniquely defined . indeed , the change of polarization is also dependent on the boundary conditions fixing the macroscopic electric field @xmath13 throughout the sample . basically , we can write : @xmath14 as the electrostatics imposes a relationship between macroscopic polarization , electric and displacement fields : @xmath15 we can deduce the following equivalent expression : @xmath16 depending on the condition imposed on the macroscopic electric field , different concepts have historically been introduced @xcite . the _ born effective charge _ @xcite alias transverse charge , @xmath1 refers to the change of polarization that would be observed under the condition of zero macroscopic electric field , so that the second term appearing in the previous equation vanishes : @xmath17 the _ callen charge _ @xcite alias longitudinal charge , @xmath18 is defined from the change of polarization under the condition of zero macroscopic displacement field : @xmath19 introducing in eq . ( [ a5 ] ) the relationship between field @xmath13 and polarization @xmath20 , deduced from eq . ( [ a4 ] ) under the condition of vanishing displacement field , born and callen charges can be related to each other thanks to the knowledge of the optical dielectric tensor @xmath21 : @xmath22 so that finally @xcite : @xmath23 for the case of isotropic materials , we recover the well known equality : @xmath24 . even if they are both related to the change of polarization induced by an atomic displacement , born and callen charges appear as two distinct quantities and will be significantly different in materials where @xmath21 is different from unity . basically , an infinite number of charges could be defined corresponding to different boundary conditions , relating @xmath20 and @xmath13 . one of them is the _ szigeti charge _ @xcite @xmath25 , defined as the change of polarization under the condition of vanishing local field , @xmath26 , at the atomic site : @xmath27 contrary to born and callen charges , @xmath25 was sometimes considered as a model - dependent concept in the sense that the local field is not observable as the macroscopic field but require some assumptions to be estimated . in the particular case of an isotropic material , the condition of vanishing local field can be written as follows : @xmath28 introducing this condition in eq . ( [ a5 ] ) : @xmath29 so that we find : @xmath30 in calculations of the dynamical properties of crystals , the contribution from the long - range coulombic interaction to the atomic forces is usually restricted to dipolar forces and is included through a term : @xmath31 . from eq . ( [ eq4 ] ) , it can be checked that this force can be alternatively written in terms of local quantities : @xmath32 . in shell - model calculations , this second formulation is usually preferred . indeed , from its definition , @xmath25 only includes the effects of charge redistribution resulting from short - range interactions and it is therefore conveniently assimilated to the static charge @xcite . from the previous discussion , it appears that the amplitude of the dynamical charge in macroscopic bodies is sensitive to the condition imposed on the macroscopic electric field . considering finite clusters of increasing size , we deduce that the amplitude of the dynamical charge , reducing to the apt for a microscopic body , will tend to a different value when the macroscopic limit is taken , depending from the _ shape _ of the cluster . we investigate now this observation in more details , and provide a unified treatment of dynamical charges in periodic solids and clusters , sufficiently large for the macroscopic quantities ( @xmath33 ) to be defined . following the well - known practice for the study of dielectric bodies @xcite , we consider that the cluster has a macroscopic ellipsodal shape . in this case , the macroscopic field within the cluster present the practical advantage to be homogeneous . in absence of any applied external field , it reduces to the depolarizing field related to the macroscopic polarization thanks to the depolarization coefficients @xmath34 @xcite . if we assume in what follows that the principal axes of the ellipsod are aligned with the axes of coordinates , we have the following relationship : @xmath35 where the geometry imposes : @xmath36 . following the same procedure as previously , the dynamical charge @xmath37 of a given atom @xmath7 in an ellipsod of volume @xmath38 can be written as : @xmath39 and we have the general relationship : @xmath40 \ ; \ ; z^{*(e)}_{\kappa , \alpha j}\end{aligned}\ ] ] in this expression , the presence of the depolarization coefficients emphasizes the influence of the shape of the cluster on the amplitude of @xmath37 . the above - mentioned sum rule on the depolarization coefficients forbid to impose the condition of zero electric or displacement fields simultaneously in the three directions . however , we have the following three interesting cases . first , we consider an extremely oblate ellipsodal ( slab - like ) cluster and take the macroscopic limit . along the @xmath41 direction perpendicular to the surface , @xmath42 , while , along the two other directions , @xmath43 . the dynamical charge for the ellipsod is therefore related to the born effective charge through the following expression : @xmath44 for uniaxial systems with no off - diagonal terms in the dielectric tensor , we note that the cluster charge along the direction perpendicular to the slab becomes identified with the callen charge , while that in the slab plane reduces to the born effective charge . differently , for an extremely prolate ellipsodal ( needle - like ) cluster aligned along the @xmath41 direction ( for which @xmath45 and @xmath46 ) , we have the following relationship : @xmath47 here also , the charge along the @xmath41 direction will reduce to the born charge in uniaxial systems . finally , for a spherical cluster , the symmetry imposes @xmath48 , so that @xmath49 . for the case of an isotropic material , we recover therefore the condition of vanishing local field and @xmath50 becomes equivalent to @xmath51 . therefore , we obtain the interesting result that in isotropic compounds , the szigeti charge appears as a well - defined quantity and is simply the dynamical charge observed in a spherical cluster . to summarize , the concept of dynamical charge in macroscopic systems is not uniquely defined : it depends on the relationship between @xmath13 and @xmath20 . in each case , the charge was however expressed in terms of two basic concepts , @xmath1 and @xmath21 . in this section , we focused on the term that includes the dielectric constant , and that describes the part of the electronic charge redistribution induced by the presence of a macroscopic field . in the next section , we will discuss the physical processes responsible of the amplitude of @xmath1 . during the seventies , a large variety of semi - empirical models were proposed to investigate the underlying physical processes driving the amplitude of dynamical charges . without being exhaustive , let us mention the interesting treatments of lucovsky , martin and burnstein @xcite who decomposed @xmath1 in a local and a non - local contribution , of lucovsky and white @xcite discussing @xmath1 in connection with resonant bonding properties , or the bond charge model of hbner @xcite . the most popular and sophisticated of these approaches remains however that of harrison @xcite within his bond orbital model ( bom ) . a similar theory was developed independently by lannoo and decarpigny @xcite . the bom basically consists in a simplified tight - binding model , where the hamiltonian is limited to the on - site and nearest - neighbour terms . the on - site elements are identified to free atom terms value , while the interatomic elements are taken as universal constants times a particular distance dependence . among other things , these parameters determine the transfer of charge between the interacting atoms . as noted by dick and overhauser @xcite , the charge redistribution produced by the sensitivity of the overlap integrals on the atomic positions is at the origin of an `` exchange charge polarization '' . similarly , in the harrison model , the dependence of the parameter on the bond length are at the origin of dynamical transfer of charges and monitors the amplitude of @xmath1 that can become anomalously large as it is illustrated in the following examples . let us first consider a diatomic molecule xy , composed of two open shell atoms , where y has the largest electronegativity . the interatomic distance is @xmath52 and the dipole moment @xmath53 . these observables allow us to identify a convenient static charge @xmath54 , while the dynamical charge is defined as : @xmath55 in the last expression , @xmath56 appears composed of two terms . the first one is simply the static charge . the second corresponds to an additional _ dynamical _ contribution : it originates in the transfer of charge produced by the modification of the interatomic distance . within the bom , this last contribution is associated to off - site orbital hybridization changes and is deduced from the universal dependence of the orbital interaction parameters on the bond length , as illustrated on a practical example in appendix [ apdx-2 ] . we deduce that the difference between @xmath57 and @xmath58 will be large if @xmath57 changes rapidly with @xmath52 . it can even be non - negligible when @xmath59 is small , if the charge is transferred on a large distance @xmath52 . moreover , this simple model naturally predicts _ anomalous _ amplitude of the dynamical charges , i.e. a value of @xmath60 not only larger than the static charge @xmath57 but even larger than the `` nominal '' ionic charge . as the distance between x and y is modified from 0 to some @xmath61 , the distance corresponding to a _ complete _ transfer of electrons from x to y , the dipole moment evolves continuously from @xmath62 ( since there is no dipole for that case ) to @xmath63 . interestingly , @xmath64 = \overline{u } \ ; z(\overline{u } ) \\\ ] ] so that : @xmath65 from the last relationship the mean value of @xmath58 from 0 to @xmath61 is equal to @xmath66 ( the `` nominal '' ionic charge ) . consequently , if @xmath57 changes with @xmath52 , @xmath58 must be larger than @xmath66 for some @xmath52 between @xmath67 $ ] . the difference between @xmath58 and the nominal charge @xmath66 is usually referred to as the _ anomalous _ contribution @xcite . switching now from a molecule to a linear chain ... -y - x - y- ... , and displacing coherently the x atoms by @xmath68 , shortened and elongated bonds will alternate all along the chain . for harrison @xcite , the interaction parameters will be modified such that `` the covalent energy increases in the shorted bond , making it less polar by transferring electron to the positive atom '' . inversely , electronic charge will be transferred to the negative atom in the elongated bond . these transfers of charge will propagate all along the chain , so that even if the net charge on the atom is not modified , a _ current _ of electrons will be associated to the atomic displacement . _ the direction of this electronic current is opposite to that of the displacement of positive atoms , so that it reinforces the change of polarization associated to this displacement and may generate an anomalously large dynamical charge_. in our example , we have implicitely considered a truly periodic system under the condition of zero macroscopic electric field so that the associated dynamical charge is @xmath1 . under other conditions , the amplitude of the transfers of charge would be additionally influenced by the presence of the field as discussed in the previous section . we note that , contrary to what was observed for the static charge , consequences of the covalence effects are to increase the amplitude of @xmath1 . the previous model can finally be extended to three dimensional solids . for this case , however , the calculation of the dynamical contribution may become questionable when the identification of the charge transfers is restricted to some specific bonds @xcite . as it will be discussed in sections [ cubic ] and [ bbb ] the harrison model remains however a meaningful picture of practical interest to interpret more accurate results . up to now , we focused on a `` delocalized '' model within which the electronic charge redistribution induced by an atomic displacement is visualized by transfer of charge induced by _ off - site _ changes of hybridization . in the past , various shell - models have however also been developed to investigate the dynamical properties of crystals . in these calculations , an accurate description of @xmath1 was mandatory in order to reproduce correctly the splitting between longitudinal and transverse optic modes in the vicinity of the @xmath69 point . contrary to the bom , the shell - model is `` local '' and treats the charges within the clausius - mosotti limit . the previous dicussion in terms of a static and dynamical contribution to @xmath1 remains valid . however , the dynamical contribution results there simply from the relative displacement of the shell charge as a whole with respect to the atom . it is attributed to the polarizability of the electrons in the local field at the atomic site . in the language of the bom , such a displacement of the electronic cloud can be understood in terms of _ on - site _ changes of hybridizations . this approach contrasts with the model developped by harrison but can also yield plausible born effective charge amplitudes /citeghosez95b . it must be emphasized that it is not possible to discriminate _ a priori _ between localized and delocalized models . within the recent theory of polarization , it has been clarified that for the purpose of understanding polarization problems , `` the true quantum mechanical electronic system can be considered as an effective classical system of quantized point charges , located at the centers of gravity associated with the occupied wannier functions in each unit cell '' @xcite . consequently , the correct description of the born effective charges does not require to reproduce correctly all the features of the valence charge distribution but _ only _ the displacement of its wannier center ( see ref . @xcite ) . as schematized in fig . [ fig.2 ] , antagonist models can reproduce a similar displacement of the wannier center . in real materials , both local polarizability and transfers of charge do probably contribute to the charge reorganisation . it will be emphasized later how first - principles investigations can help to identify the dominant mechanism . in conclusion , this section has shown that @xmath56 is related to the static charge ( see eq . [ zmod ] ) but does not restrict to it : @xmath56 may also include an additional , important , dynamical contribution . whatever the mechanism of the charge redistribution ( localized or delocalized ) , the amplitude of the dynamical contribution can not be estimated from the inspection of the electronic density alone . so , we partly disagree with meister and schwarz @xcite who suggested that all the charges _ including _ the gapt are driven by the same underlying parameter . in what follows , based on firs - principles calculations , we illustrate on different examples that @xmath1 may become anomalously large and independent of the amplitude of the static charge @xmath70 . moreover , two atoms with similar @xmath70 can also exhibit strongly different @xmath1 . a brief review of the most commonly used first - principles approaches for computing the born effective charges has been reported recently @xcite . going beyond semi - empirical approaches , _ ab initio _ techniques allow accurate prediction of @xmath1 in materials where its amplitude is not necessarily directly accessible from the experiment . going further , the first - principles approaches are also offering a new opportunity to clarify the microscopic mechanism modulating the amplitude of @xmath1 without any preliminary hypothesis . as it will be illustrated in the following sections , it reveals particularly useful to understand the origin of anomalously large @xmath1 in abo@xmath0 compounds . the results of the present paper have been obtained in the framework of the density functional formalism @xcite . the exchange - correlation energy has been evaluated within the local density approximation , using a pad parametrization @xcite of ceperley - alder homogeneous electron gas data @xcite . integrals over the brillouin - zone were replaced by a sum on a mesh of @xmath71 special * k*-points @xcite ( 10 points in the irreducible brillouin zone ) . the `` all electron '' potentials were replaced by the same ab initio , separable , extended norm - conserving pseudopotentials as in ref . the wavefunctions were expanded in plane waves up to a kinetic energy cutoff of 35 hartree ( about 4100 plane waves ) . the born effective charges have been deduced from linear response calculations @xcite , using a variational formulation @xcite to the density functional perturbation theory . the decomposition of individual contributions from separate groups of occupied bands has been performed following the scheme described in ref . the parameters used for the calculations guarantee a convergency better than 0.5% on @xmath1 as well as on each of its band - by - band contributions . recently , the born effective charge tensors of perovskite abo@xmath0 compounds have been at the center of numerous investigations @xcite . in the cubic phase , they are fully characterized by a set of four independent numbers . the charge tensor of the a and b atoms is isotropic owing to the local spherical symmetry at the atomic site . for oxygen , the local environment is tetragonal and two independent elements o@xmath72 and o@xmath73 must be considered , referring respectively to the change of polarization induced by an atomic displacement parallel and perpendicular to the b - o bond . in table [ z*cubicba ] , we summarize the results obtained within different approaches for the cubic phase of batio@xmath0 . the first reliable estimation of @xmath1 in batio@xmath0 is probably due to axe @xcite , from empirical fitting to experimental mode oscillator strengths @xcite . in abo@xmath0 compounds , @xmath1 can not be determined unambiguously from the experiment . however , within some realistic hypothesis , axe identified the independent elements of the effective charges of batio@xmath0 and already pointed out their two essential features . first , the oxygen charge tensor is highly anisotropic . second , the charges on ti and o@xmath72 contain a large _ anomalous _ contribution ( i.e. an additional charge with respect to the nominal ionic value of + 2 for ba , + 4 for ti and -2 for o ) . both these characteristics are confirmed by the first - principles calculations . our ab initio results , computed from linear response , are also in excellent agreement with those of zhong _ et al . _ @xcite , obtained from finite differences of polarization . the charge neutrality sum rule , reflecting the numerical accuracy of our calculation , is fulfilled to within 0.02 . we note that the values of @xmath1 are also qualitatively reproduced from a shell - model calculation @xcite and accurately predicted within the scad model @xcite . the anomalous amplitude of the dynamical charge , reported in this section , is not a specific feature of batio@xmath0 . similar computations of @xmath1 were performed on different perovskite abo@xmath0 compounds and they all reproduce the same characteristics than in batio@xmath0 . a non exhaustive list of these results is reported in table [ z*cubicab ] . we observe that the choice of the a atom has a rather limited influence on @xmath74 and @xmath75 , which appear closely related to the b atom . while the nominal ionic charge of ti and zr is + 4 in these compounds , the born effective charge is between + @xmath76 and + @xmath77 for ti , and approximately equal to + @xmath78 for zr . for nb , the ionic charge is + 5 , while the born effective charge is between + @xmath79 and + @xmath80 . extending the investigations to wo@xmath0 in the reference cubic phase ( defect perovskite structure ) , the ionic charge on w is equal to + 6 , while the born effective charge reaches the much larger value of + @xmath81 . for the class of perovskite abo@xmath0compounds , it can be checked that @xmath82 evolves quasi linearly with the nominal charge of the b atom @xcite . for materials containing pb , the previous considerations remain valid but there are additional anomalies concerning @xmath83 and @xmath84 . this feature is due to the more covalent bonding of lead with oxygen that was illustrated in ref . @xcite . in what follows , we will not be concerned with these lead compounds . the approximate reciprocity between o@xmath72 and b anomalous contributions suggests that they should originate in a global transfer of charge between b and o atoms as described in section [ hybrid ] . in ref . @xcite , harrison had in fact already suggested the existence of giant born effective charges in perovskite materials . being unaware of the earlier results of axe , he had however no experimental evidence to corroborate his semi - empirical calculations . in appendix [ apdx-2 ] , we report results obtained within the harrison model ( it follows the method described for kcl in ref . @xcite , p. 334 . ) . for srtio@xmath0 , from the universal tight - binding parameters of harrison , we get a value of @xmath858.18 for @xmath75 , making plausible the giant anomalous effective charges only by focusing on the dynamical changes of hybridization between occupied o 2s o 2p states and the unoccupied metal d states . in batio@xmath0 , the hybridization between o 2p and ti 3d orbitals is a well known feature , confirmed by various sources ( experiments @xcite , lcao calculations @xcite and dft results @xcite ) . in this context , it was therefore realistic to focus on o 2p - b d hybridization changes to explain intuitively large anomalous contributions @xcite . at the opposite , it may therefore appear surprizing that model calculations which does _ not _ explicitly include transfer of charges are able to predict correctly the amplitude of the born effective charges . for instance , in table i , we observe that the values of @xmath1 are qualitatively reproduced by a shell - model calculation @xcite . a similar agreement between _ ab initio _ and shell model results was highlighted for knbo@xmath0 @xcite . in both cases , the calculation was performed within the `` polarizability model '' introduced by bilz _ et al . _ @xcite , which includes an anisotropic and non - linear polarizability of the o atoms . in the same spirit , at the level of the scad model , the born effective charges are accurately reproduced while there is no explicit transfer of electrons between the different atomic sites . as discussed in section [ dynamical ] , antagonist models can be invoked to explain the origin of anomalous contributions as soon as they globally reproduce a similar displacement of the wannier center of the valence charge distribution . what appear as a macroscopic current along the ti o chain within the bom shows itself as an unusual polarizability of the oxygen atoms within the shell model . it was not possible to discriminate unambiguously between localized and delocalized model until posternak _ et al . _ @xcite proposed a convincing proof of the crucial role of `` off - site '' hybridizations . based on first - principles calculations , they demonstrated for knbo@xmath0 that the anomalous contribution to the charge of nb and o@xmath72 disappears if the hybridization between o 2p and nb 4d orbitals is artificially suppressed . in a similar spirit , the inspection of the wannier functions of batio@xmath0 and the analysis of their deformation under an atomic displacement reported by marzari and vanderbilt @xcite confirm the predominant role played by the ti 3d orbitals and the explanation introduced by harrison . in the next section , we propose a band - by - band decomposition of the born effective charges @xcite . this technique appears as a tool of paramount importance to clarify the microscopic origin of anomalous contributions . identifying the dynamical transfer of charges without any preliminary hypothesis on the orbitals that interact , it will allow to generalize the basic mechanism that was proposed by harrison . in abo@xmath0 compounds , the electronic band structure is composed of well separated sets of bands . the hybridizations between the orbitals of the different atoms are relatively small and each band can be identified by the name of the main atomic orbital which contributes to this energy level in the solid . the born effective charge is defined by the change of polarization associated to a specific atomic displacement . our purpose will be to identify the contribution of each well separated set of bands to this change of polarization @xcite . in ref . @xcite , we have described how band - by - band contributions to @xmath1 can be separated from each others . moreover , it has been demonstrated that the contribution to @xmath86 from a single occupied band @xmath87 can be interpreted as a change of polarization @xmath88 where @xmath89 is the displacement in direction @xmath5 of the wannier center of band @xmath87 , induced by the unitary displacement of the sublattice of atoms @xmath7 in direction @xmath2 . in order to understand the origin of the displacement of the wannier center of each band , it is helpful to define a _ reference _ configuration that corresponds to what we would expect in a purely ionic material . in such fictitious material , each band would be composed of a single non - hybridized orbital and the wannier center of each band would be centered on a given atom . in absence of any hybridization , when displacing a given sublattice of atoms @xmath7 , the wannier center of the bands centered on the moving atoms would remain centered on it , while the position of the center of gravity of the other bands would remain unaffected . the contributions of these two kind of bands to @xmath90 would therefore be @xmath91 and 0 electrons , respectively . in the real material , the _ anomalous _ contribution of a particular band @xmath92 to a given atom @xmath7 is defined as the additional part with respect to the reference value expected in absence of any hybridization : it reflects how the center of the wannier center of band @xmath92 is displaced relatively to the atoms when the sublattice @xmath7 moves @xcite . considering each band as a combinaison of atomic orbitals , such a displacement of the wannier center of a band with respect to its reference position _ must _ be attributed to hybridization effects : it is associated to the admixture of a new orbital character to the band . when the orbitals which interact are located on different atoms ( `` off - site '' hybridization ) , the dynamical changes of hybridization can be visualized as transfers of charge . if the interacting orbitals are on the same atom ( `` on - site '' hybridization ) , the mechanism much looks like a polarizability . rigorously , our band - by - band decomposition is performed within dft and formally only concerns the kohn - sham particles . it seems however that the results are rather independent of the one - particle scheme @xcite used for the calculation so that the results presented here should give a good insight on the physics of the abo@xmath0 compounds . let us first applied the band - by - band decomposition to barium titanate . the band structure of batio@xmath0 is presented in fig . [ fig.b.ba ] . results of the decomposition of @xmath1 in the theoretical cubic structure of batio@xmath0 are reported in table [ z*bbbba ] . the first line ( @xmath93 ) brings together the charge of the nucleus and core electrons included in the pseudopotential . the other contributions come from the different valence electron levels . the sum of the band - by - band contributions on one atom is equal to its global effective charge while the sum of the contribution to a particular band from the different atoms is equal to @xmath91 ( within the accuracy of the calculation ) , the occupancy of this band . focusing first on the titanium charge , we observe that the ti 3s contribution ( @xmath94 ) is close to @xmath91 , confirming that the ti 3s electrons follow the ti atom when moving , independently from the change of its surrounding . this result _ a posteriori _ justifies the inclusion of deeper electronic levels as part of the ionic pseudopotentials . at the opposite , it is shown that the giant anomalous charge of titanium essentially comes from the o 2p bands ( @xmath95 ) . it corresponds to a displacement of the wannier center of the o 2p bands in opposite direction to the displacement of the ti atom . this observation is in perfect agreement with the harrison model : it can be understood by dynamical changes of hybridization between o 2p and ti 3d orbitals , producing a transfer of electron from o to ti when the ti - o distance shortens . this explanation was confirmed recently from the inspection of the o 2p wannier functions @xcite . beyond the previous observations , we note however that there are also small anomalous charges from the ti 3p , o 2s and ba 5p bands . these contributions are not negligible . the positive anomalous charges correspond to a displacement of the center of the wannier function of the o and ba bands in the direction of the closest ti when this atom has moved . some of these features go beyond the harrison model , within which anomalous contributions to @xmath96 in table [ z*bbbba ] would be restricted to the o 2p and o 2s bands . they suggest other kind of hybridization changes , that will be now more explicitly investigated . focusing on barium , the global anomalous effective charge ( @xmath97 ) is small compared to that of ti and this feature was first attributed to its more ionic character @xcite . this ionicity is inherent to the harrison model @xcite and was confirmed in some ab initio studies @xcite . surprisingly , our decomposition reveals however that the anomalous charges of the o 2s ( @xmath98 ) and o 2p ( @xmath99 ) bands are not small at all . they are nevertheless roughly compensated by another ba 5s ( @xmath100)and ba 5p ( @xmath101 ) anomalous contributions . this result suggests that there are dynamical changes of hybridization between ba and o orbitals as it was the case between o and ti , except that the mechanism is here restricted to _ occupied _ states . _ this basically corresponds to a unitary transform within the subspace of the occupied states which is unable to displace the global wannier center of the valence charge_. our result so supports the hybridization of ba orbitals , in agreement with experiment @xcite , lcao calculations @xcite and dft @xcite computations . similar compensating contributions were recently observed in zno which has conventional born effective charges @xcite and in a series of alkaline - earth oxides @xcite . we note that a confusion sometimes appears that should be removed : the _ amplitude _ of the anomalous contributions to @xmath1 is not related to the amplitude of the hybridizations but to the _ rate of change _ of these hybridizations under atomic displacements . it is clear that , in batio@xmath0 , the ba 5p contribution to the o 2p bands is smaller than the contribution from the ti 3d orbitals @xcite . however , the high sensitivity of this relatively weak covalent character under atomic positions is sufficient to produce large band by band anomalous contributions to @xmath1 . from that point of view , the born effective charge appears therefore as a sensitive tool to identify the presence of even small hybridizations . finally , concerning the oxygen , even if o@xmath102 and o@xmath103 are defined respectively for a displacement of o in the ti and ba direction , it seems only qualitative to associate @xmath104 with @xmath96 and @xmath105 with @xmath106 as suggested in ref . the o 2p anomalous contributions to ti and o@xmath102 do not exactly compensate . moreover , o 2p contribution to @xmath106 does not come from o@xmath103 only but has equivalent contributions from o@xmath102 . this seems to confirm the idea of bennetto and vanderbilt @xcite that in 3d materials , transfers of charges are not necessarily restricted to a particular bond , but is a rather complex mechanism that must be treated as a whole . to summarize , our study has clarified the mixed ionic - covalent character of batio@xmath0 : it clearly establishes that the covalent character is not restricted to the ti - o bond but also partly concerns the ba atom . moreover , it leads to a more general issue . it illustrates that _ the presence of a large anomalous charge requires a modification of the interactions between occupied and unoccupied electronic state . the contributions originating from the change of the interactions between two occupied states correspond to unitary transforms within the subspace of the valence charge : they compensate , and do not modify the global value of @xmath1_. the same analysis is now performed on srtio@xmath0 . its band structure ( fig . [ fig.b.sr ] ) is very similar to that of batio@xmath0 , except that the ti 3p and sr 4s bands are energetically very close to each others . consequently , they strongly mix and it should be relatively meaningless to separate their respective contributions . the sr 4p and o 2s states are in the same energy region but can be separated , contrary to what was observed in a study of sro @xcite . the result of the decomposition is very similar ( table [ z*bbbsr ] ) to that reported for batio@xmath0 . there is still a giant contribution to @xmath96 from the o 2p bands . on the other hand , while the ba 5p bands were approximately centered between o 2s and o 2p bands in batio@xmath0 , the sr 4p electrons are closer to the o 2s bands and mainly hybridize with them in srtio@xmath0 . this phenomenon produces large but compensating contributions from sr 4p and o 2s bands to @xmath107 . such an evolution is in agreement with the picture that anomalous contributions originate from off - site orbital hybridization changes . from the two previous results that concern two very similar materials , it might be suggested that not only the dynamical hybridization of the valence bands with unoccupied d - states but also the particular cubic perovskite structure of abo@xmath0 compounds plays a major role in determining @xmath1 . for instance , the anomalous charge could partly originate in the local fields at the atomic sites , known to be anomalously large in this structure @xcite it is interesting to observe that anomalous charges are not restricted to perovskite solids but were also detected in a series of alkaline - earth oxides of rocksalt structure ( cao , sro , bao ) @xcite or even al@xmath108ru @xcite , all examples where the unoccupied d - states seem to play a major role . interestingly , two materials belonging to the same structure can present completely different charges . this was illustrated for the case of tio@xmath108 rutile and sio@xmath108 stishovite @xcite : while relatively conventional charges were observed on si ( + 4.15 ) and o ( -2.46 ) along the si - o bond in stishovite , giant effective charges , similar to those of batio@xmath0 , were obtained on ti ( + 7.33 ) and o ( -4.98 ) along the ti - o bond in rutile . similarly , no anomalous charge was reported for mgo ( @xmath109 ) , presenting the same rocksalt structure than bao ( @xmath110 ) @xcite . in the same spirit , the same atom in different environments can present similar dynamical charge , as illustrated for @xmath96 in batio@xmath0 and tio@xmath108 @xcite , or for @xmath111 in bazro@xmath0 @xcite and zro@xmath108 @xcite . also , in the family of abo@xmath0 compounds , giant effective charges are observed on ti in catio@xmath0 ( @xmath112 , @xcite ) but not on si in casio@xmath0 ( @xmath113 , @xcite ) . we observe that the presence of partly hybridized d - states seems the only common feature between the materials presenting giant anomalous effective charges , listed up to date . this feature finds a basic justification within the bom of harrison : the interaction parameters involving d - states are indeed much more sensitive to the interatomic distance than those involving , for example , s and p orbitals @xcite : they will therefore be associated to larger dynamical transfer of charge and will generate higher @xmath1 . in the litterature , calculations of @xmath1 essentially focused on the cubic phase of abo@xmath0 compounds @xcite . on the basis of an early study of knbo@xmath0 @xcite , it was concluded that the born effective charges are independent of the ionic ferroelectric displacements ( i.e. they remain similar in the different phases ) . another investigation in the tetragonal phase of knbo@xmath0 and pbtio@xmath0 @xcite , seemed to confirm that @xmath1 are quite insensitive to structural details . these results were surprising if we remember that anomalous contributions to @xmath1 are closely related to orbital hybridizations , these in turn , well known to be strongly affected by the phase transitions @xcite . we will see in this section that , contrary to what was first expected , born effective charges in batio@xmath0 are strongly dependent of the structural features . we first investigate the sensitivity of the born effective charges to the ferroelectric atomic displacements @xcite . for that purpose , we compute @xmath1 in the three ferroelectric phases at the experimental unit cell parameters , with relaxed atomic positions as reported in ref . @xcite . table [ z*ferro - bto ] summarizes the results for a cartesian set of axis where the @xmath41-axis points in the ferroelectric direction . the ba and ti charge tensors are diagonal in each phase for this particular choice . in the case of o , we note the presence of a small asymmetric contribution for the lowest symmetry phases . the eigenvalues of the symmetric part of the tensor are also reported . in each phase , the eigenvector associated to the highest eigenvalue of o approximately points in the ti - o direction and allows to identify the highest contribution as o@xmath72 . the other eigenvalues can be referred to as o@xmath73 , by analogy with the cubic phase . although the charges of ba and o@xmath73 remain globally unchanged in the 4 phases , strong modifications are observed for ti and o@xmath72 : for example , changing the ti position by 0.076 ( 2% of the unit cell length ) when going from the cubic to the rhombohedral phase , reduces the _ anomalous _ part of @xmath96 by more than 50% along the ferroelectric axis ( table [ z*ferro - bto ] ) . equivalent evolutions are observed in the other ferroelectric phases . similar changes were detected in knbo@xmath0 @xcite . in the isotropic cubic structure , harrison had explained the large value of @xmath1 in terms of the ti - o bond length . for the anisotropic ferroelectric phases , it should be intuitively expected that the shortest ti - o distance @xmath114 in the structure will dominate the bonding properties . it is therefore tempting to transpose the harrison model to understand the evolution of @xmath1 in terms of the distance @xmath114 . in fig . [ fig.z*tio ] , the amplitude of @xmath96 in the direction of the shortest ti - o bond length of each phase is plotted with respect to @xmath114 . i similar graph can be obtained for o. for the different phases , at the experimental lattice parameters , we observe that the anomalous parts evolve quasi linearly with @xmath114 . independently from the previous calculations , we also investigated the evolution of @xmath1 under isotropic pressure ( table [ z*pressure ] ) . in contrast with the changes observed with respect to the atomic displacements , the charge appears essentially insensitive to isotropic compression . in particular , in the compressed cubic cell at 3.67 where the ti - o distance is _ the same _ that the shortest ti - o bond length in the tetragonal structure @xcite , @xmath96 remains very close to its value at the optimized volume . this new element clearly invalidates the expected dependence from @xmath1 to @xmath114 . the fundamental difference between the cubic and tetragonal structures lies in the fact that in the cubic phase every ti - o distance is equal to the others , while in the tetragonal phase , along the ferroelectric axis , a short ti - o bond length ( @xmath114 ) is followed by a larger one ( @xmath115 ) which _ breaks _ the ti - o chain in this direction . in order to verify that it is not this large ti - o distance which , alternatively to @xmath114 , is sufficient to inhibit the giant current associated to the anomalous charges , we also performed a calculation in an expanded cubic phase where @xmath116 : we observe however that the ti charge is even larger than in the optimized cubic phase . we conclude from the previous investigations that _ the amplitude of @xmath1 in batio@xmath0 is not dependent on a particular interatomic distance ( @xmath114 , @xmath115 ) but is more critically affected by the anisotropy of the ti environment along the ti o chains_. in agreement with this picture , wang _ et al . _ @xcite reported recently an insensitivity of @xmath1 to a tetragonal macroscopic strain in knbo@xmath0 . also , the charges reported by bellaiche _ _ @xcite in mixed a compound as pzt , where the ionic environment becomes anisotropic , seem to confirm our results . a band by band decomposition of @xmath96 ( table [ z*ti ] ) points out that the difference between the cubic and tetragonal phases is essentially localized at the level of the o 2p bands ( + 1.48 instead of + 2.86 ) while the other contributions remain very similar . this suggests an intuitive explanation . in the cubic phase the o 2p electrons are widely delocalized and dynamical transfer of charge can propagate along the ti - o chain as suggested by harrison . in the tetragonal phase , the ti - o chain behaves as a sequence of ti - o dimers for which the electrons are less polarizable . this smaller polarizability is confirmed by a similar reduction of the optical dielectric constant along the ferroelectric direction . this analysis seems plausible from the wannier function analysis reported recently by marzari and vanderbilt @xcite . finally , let us mention that if the evolution of @xmath1 is relatively weak under isotropic pressure , it would be wrong to consider that the dynamical properties of batio@xmath0 are insensitive to the volume : small changes are observed that are of the same order of magnitude than for other compounds like sic @xcite . the direction of the evolution is however different . moreover , the evolution of the different charges is even not identical : while the absolute value of @xmath106 and @xmath84 decreases with increasing volume , the inverse behaviour is observed for @xmath96 and @xmath117 . here also , the band by band decomposition ( table [ z*ba ] ) reveals some hidden features . in the compressed cubic phase , the anomalous part of the ba 5p , ba 5s and ti 3p bands are 50% larger than in the optimized cubic cell . this suggests an evolution of the interactions between occupied orbitals that is coherent with the modification of the interatomic short - range forces observed independently @xcite . at the opposite , in our expanded cubic phase , most of the anomalous contributions to @xmath106 and @xmath96 have disappeared in agreement with the picture of a more ionic material . the o 2p contribution , is the only one that remains surprisingly large . comparing to the value obtained for the cubic phase at the experimental volume , its evolution was even more important than the linear dependence upon the bond length , expected from the harrison model . the spontaneous polarization ( @xmath118 ) of the ferroelectric phases can be determined by integrating the change of polarization along the path of atomic displacement from the paraelectric cubic phase ( taken as reference ) to the considered ferroelectric structure . if the effective charges were roughly constant , this integration should be approximated by : @xmath119 however , we have seen , in the previous section , that the born effective charges are strongly affected by the atomic displacements . it is therefore important to investigate their evolution all along the path of atomic displacements from one structure to the other . we performed the calculation for a transformation from the cubic to the rhombohedral structure . the rhombohedral macroscopic strain is very small and was neglected @xcite : our calculation was performed by displacing the atoms to their theoretically optimized position in rhombohedral symmetry , when keeping the cubic lattice parameters . the result is reported in figure [ fig.z*.cr ] , for @xmath96 along the ferroelectric direction . a similar curve can be obtained for @xmath75 . we observe that the evolution of @xmath1 is approximately quadratic close to the cubic phase . however , it becomes rapidly linear , and remains linear for displacements even larger than those associated to the ferroelectric distortion . expecting a similar evolution of the dynamical charges for the tetragonal and orthorhombic displacements , an estimation of the spontaneous polarization in the ferroelectric phases can be found when using eq . ( [ eq.ps ] ) with a mean effective charge determined from its value in both phases . using a mean charge estimated from the values in the para- and ferro - electric phases , we obtain the spontaneous polarizations presented in table [ ps ] . our results are only in relative agreement with the experiment @xcite and suggest different comments . firstly , we would like to mention that part of the discrepancy must be assigned to the theoretical overestimation of the computed ferroelectric displacements , discussed in ref . @xcite : when using the experimental displacements of ref . @xcite , we recover a better estimation of @xmath118 as in ref . the dispersion of x - rays diffraction data makes however difficult the exact identification of the ferroelectric displacements . secondly , another part of the error could be due to the lack of polarization dependence of the lda @xcite . finally , we note that there is also some uncertainty on the experimental value of @xmath118 . in this paper , we first analyzed the links between different definitions of atomic charge . we have shown that , contrary to the static definitions , dynamical effective charges also depend on the electronic charge reorganisation induced by an atomic displacement . the amplitude of this dynamical contribution is monitored not only by the bonding with the other atoms but also , for large systems , by the condition imposed on the macroscopic electric field . a unified treatment of the concept of dynamical charge in molecules , large clusters , and truly periodic systems has been presented , in which the born effective charge and the optical dielectric constant appear as the two fundamental quantities . the microscopic origin of the dynamical contribution has been discussed in terms of local polarizability and delocalized transfers of electrons . based on various first - principles results , we have then emphasized that the born effective charges are anomalously large in the family of abo@xmath0 compounds : their amplitude can reach more than twice that of the nominal ionic charges . this feature was explained in terms of interatomic transfers of charge , produced by `` off - site '' dynamical changes of hybridization . for batio@xmath0 and srtio@xmath0 , we have brought to light complex dynamical changes of hybridization , concerning not only ti and o but also ba and sr orbitals . the hybridizations restricted to occupied states generate however compensating anomalous contributions so that , finally , the total value of @xmath1 is essentially affected by dynamical changes of hybridization between o 2p and ti 3d orbitals . as a more general issue , it appears that the existence of partial hybridizations between occupied and unoccupied states is an important feature for candidate to large anomalous born effective charges . moreover , the dynamical transfers of charge are expected to be larger when such a hybridization involve d states , for which the interactions parameters with other orbitals are particularly sensitive to the interatomic distance . investigating the evolution of @xmath1 to the structural features , we have shown that they are strongly affected by the ferroelectric atomic displacements and much less sensitive to isotropic pressure . the results have clarified that the amplitude of @xmath1 is not monitored by a particular interatomic distance but is dependent on the anisotropy of the ti environment along the ti - o chains . finally , the effective charges were used to estimate the spontaneous polarization in the ferroelectric phases of batio@xmath0 . for that purpose , their evolution was investigated all along the path of atomic displacement from the cubic to the rhombohedral structure . all along this work , we only focused on the _ microscopic _ mechanisms that govern the amplitude of the born effective charges . in independent studies , it was also emphasized that the anomalously large born effective charges produce a giant lo - to splitting in abo@xmath0 compounds , specially for the ferroelectric phonon mode @xcite . moreover , it was demonstrated that this feature is associated to the existence of an anomalously large destabilizing dipole - dipole interaction , sufficient to compensate the stabilizing short - range forces and induce the ferroelectric instability @xcite . in materials where polar modes play a major role , the born effective charge appears therefore also as a `` key concept '' to relate the electronic and structural properties @xcite . g. would like to thank r. resta for numerous illuminating discussions , ph . lambin for his help in shell - model calculations , l. l. boyer for clarifying some aspects of the scad model , and n. marzari and d. vanderbilt for providing an early copy of their recent results . x.g . is grateful to the national fund for scientific research ( fnrs - belgium ) for financial support . we acknowledge the use of ibm - rs 6000 work stations from common projects between ibm belgium , ucl - pcpm and fundp - scf , as well as the use of the maui high performance computing center ibm - sp2 . we thank corning incorporated for the availability of the planewave code , as well as j .- beuken for his kind and permanent computer assistance . we also acknowledge financial support from the pai - uiap p4/10 . in this appendix , we briefly describe the bond - orbital model proposed by harrison for the case of abo@xmath0 compounds @xcite . in particular , we pay a particular attention to the definition and the calculation of static and dynamical charges within this model . values are reported for srtio@xmath0 . the bond orbital model of harrison consists in a simplified tight - binding model , where the hamiltonian is limited to the on - site and nearest - neighbour terms . moreover , for abo@xmath0 compounds , it is assumed that the a atom has no other function than to provide electrons to the system , and is fully ionized . the only considered interactions involve b and o atom orbitals . the model includes o 2s , o 2p and b d orbitals , interacting through @xmath120 , @xmath121 and @xmath122 parameters . as the matrix elements @xmath120 and @xmath121 are nearly identical , it is suggested to construct two @xmath123 hybrids on the oxygen , @xmath124 . each of these hybrid will have a large matrix element ( @xmath125/ \sqrt 2 $ ] ) coupling it to the @xmath126 state on one side and a negligible matrix element ( @xmath127/ \sqrt 2 $ ] ) coupling it to the @xmath126 state on the other . in absence of orbital interactions , the static charges of srtio@xmath0 would be of @xmath4 on sr , @xmath128 on ti and @xmath91 on o. however , due to the ti - o orbital interactions , the transfer of electrons from ti to o is not complete . within the bom , the effective static charge on o can be estimated as : @xmath129 where @xmath130 } \biggr)^2\end{aligned}\ ] ] ( `` 4 '' because there are 2 hybrids composed of 2 electrons that each interacts mainly with one ti neighbour ) and @xmath131 } \biggr)^2\end{aligned}\ ] ] ( `` 8 '' because there are 4 electrons that are each partly delocalized on the 2 ti neighbours ) . following the discussion of section iii - b , the born effective charge can be obtained by adding to the static charge the dynamical contribution originating in the dependence of the matrix elements @xmath135 upon the interatomic bond length @xmath126 . from harrison s solid state table ( see also eq . ( 19 - 11 ) in ref . @xcite ) , the previous matrix elements have all the same interatomic dependence : @xmath136 so that @xmath137 when displacing the o atom along the ti - o direction , there will be an additional transfer of electron from o to the nearest ti that is equal to @xmath138 \frac{\delta d}{d } \\ \delta t_{\pi } & = & 4 [ 2 \ ; ( \frac{-7}{2 } ) ( \frac{v_{pd,\pi}}{\epsilon_d - \epsilon_p})^2 ] \frac{\delta d}{d}\end{aligned}\ ] ] on the other hand , there will be the same transfer of charge from the other neighbour ti atom to o , so that the previous electrons are globally transferred on a distance @xmath139 . the change of polarization associated to this transfer of charge is : @xmath140 the associated dynamic contribution to the born effective charge is ( @xmath141 ) : @xmath142 we note the unphysical charge on ba . it is the result of a negative mulliken population and reflects the inadequacy of the mulliken population analysis which assumes an equal share of overlap between each pair of atoms . besides the different definitions of dynamical charges given in eqs . 6 , 7 and 10 , let us note that another related charge is sometimes also introduced , which is defined as @xmath147 . this charge can reveal useful in the study of @xmath148 compounds in the sense that it is directly accessible from lo - to splitting data without any hypothesis on the amplitude of the dielectric constant . in more complex materials , as abo@xmath0 compounds , the extraction of the charges from the splitting is not straightforward but requires to introduce some hypothesis . this has been made by gervais _ et al . _ @xcite who estimated the amplitude of @xmath147 in batio@xmath0 . nominal and static charges may differ widely due to covalency effects . as the static charge is ill defined , one usually prefers to define the anomalous contribution in reference to the nominal charge . the difference between born effective charge and static charge is sometimes referred to as the dynamical contribution . let us mention that an early investigation of the born effective charges of batio@xmath0 was performed by last in 1957 @xcite , but without identifying any anomaly . another discussion was reported in ref . @xcite but without separating the respective values of @xmath1 . n. marzari and d. vanderbilt , in `` proceedings of the fifth williamsburg workshop on first - principles calculations for ferroelectrics '' , willamsburg , va , 1998 ( available at the url : _ http://www.physics.rutgers.edu/ dhv / preprints _ ) ; n. marzari and d. vanderbilt , unpublished . depending from the structure and from the interactions , the initial position of the electron wannier center is not necessarily on an atom . however , in the cubic perovskite structure , symmetry imposes to the wannier centers to be located on an atom so that the anomalous contribution exactly describes the displacement from such a centered position . l. bellaiche , j. padilla and d. vanderbilt , in `` proceedings of the fifth williamsburg workshop on first - principles calculations for ferroelectrics '' , willamsburg , va , 1998 ( available at this url : _ http://www.physics.rutgers.edu/dhv/preprints_ ) . in the tetragonal phase , shortened and elongated ti o bonds alternate along the ferroelectric axis . the shortened bond corresponds to an interatomic distance of 3.67 in our optimized tetragonal structure . the born effective charges obtained for the rhombohedral structure when neglecting the strain ( i. e. when keeping a cubic unit cell ) are the following : @xmath149 , @xmath150 , @xmath151 , @xmath152 , @xmath153 , @xmath154 . these values must be compared to those reported in table [ z*ferro - bto ] , where the rhombohedral strain was taken into accounts . it can be checked that the effect of this strain is negligible .

the combination of simple dynamics , small number of available decay channels , and extremely well controlled radiative and loop corrections , make charged pion decays a sensitive means for testing the underlying symmetries and the universality of weak fermion couplings , as well as for improving our understanding of pion structure and chiral dynamics . this paper reviews the current state of experimental study of the allowed rare decays of charged pions : ( a ) leptonic , , or , ( b ) radiative , , or , and , or , and ( c ) semileptonic , , or . taken together , the combined data set presents an internally consistent picture that also agrees well with standard model predictions . the internal consistency is illustrated well by the branching ratio of extracted in this work from the pibeta measurement of the decay and the current best value for the ckm matrix element . however , even after the great progress of the recent decades , experimental precision is lagging far behind that of the theoretical description for all above processes . we review the implications of the present state of knowledge and prospects for further improvement in the near term .

1407.2865

pi mesons hold a special place in both the weak and the strong interactions , and remain subjects of research interest ever since their discovery almost 70 years ago @xcite . historically , pion decay has provided an important testing ground for the weak interaction and radiative corrections from the earliest time of the development of modern particle theory . decays of the charged pion proceed via the weak interaction , and therefore closely reflect its properties and dynamics . in particular , the suppression of the direct electronic decay of the pion ( @xmath10 , or @xmath1 ) , manifested in the failure of initial searches to observe it , led to an early examination of the nature of the weak interaction governing pion decay @xcite . a low branching fraction of @xmath11 was predicted @xcite even before the decay s discovery @xcite , as a direct consequence of the @xmath12@xmath13@xmath14 nature of the weak interaction , through helicity suppression of the right - handed state of the electron . not only was this prediction confirmed by the early measurements @xcite , the size of the radiative corrections @xcite was soon proved correct as well @xcite . in more recent times pion decays have been theoretically described with extraordinary precision . thanks to the underlying symmetries and the associated conservation laws , the more complicated and thus more uncertain hadronic processes are suppressed . if measured with precision comparable to that of their theoretical description , pion decays offer an outstanding , clean testing ground of universality of lepton and quark couplings . any documented deviation from standard model expectations would indicate presence of dynamics not included in the sm , affecting pion decays through loop diagrams . in the hadronic sector the pion plays the role of the long - range exchange particle in effective nucleon - nucleon lagrangians . just as importantly , the pion also plays the role of the pseudo - goldstone boson in the breaking of chiral symmetry for hadrons @xcite . whereas early formulations of the phenomenon were focusing on nucleons as the fundamental hadronic fermions , the concept remains valid in the quark picture . while a massless pion would suffice to account for the spontaneous chiral symmetry breaking , the nonzero pion mass @xcite can be related directly to the light quark masses within the framework of the explicit chiral symmetry breaking , particularly in chiral perturbation theory ( chpt ) @xcite . although in principle the full theory of quantum chromodynamics ( qcd ) contains all of this physics , in practice qcd does not readily provide manageable calculations in the low energy , nonperturbative domain . the situation will improve in the near future as lattice qcd calculations continue to increase accuracy , reliability and reach . thus , for a long time chiral perturbation theory and other similar effective lagrangian approximate methods have provided the primary and the only practical systematic treatment of low energy hadronic processes based on qcd . basic properties of the pion , such as its charge radius and form factors , are closely related to the constants of the chiral lagrangian @xcite and are thus fundamental to nonperturbative qcd calculations . in the remainder of this work we review experimental studies of individual rare decay modes of the charged pion in the order of descending branching fraction , with a focus on recent and ongoing experiments . it is worth noting that almost all measurements to date have studied the positive pion . stopped @xmath15 mesons are captured with high probability in target nuclei ( and undergo strong interactions ) , leaving few , if any , pions to decay weakly . hence , in practice @xmath15 decays can only be studied in flight in vacuo , with typically a tiny fraction of the beam pions decaying in the apparatus , and the remaining pions escaping before decaying . positive pions can be easily stopped in a designated target , with a small fraction lost through prompt hadronic reactions with matter required to stop the beam . the stopped @xmath16 s decay with an exponential time distribution governed by the 26ns pion lifetime . the dominant decay mode is @xmath17 ( @xmath18 ) which will not be discussed here . the resulting 4.1mev muons are typically contained in the target and decay via the @xmath19 process with @xmath20s . the @xmath21 ( or , equivalently , @xmath22 ) decay connects a pseudoscalar @xmath23 state ( the pion ) to the vacuum . at the lowest , tree level , the ratio of the @xmath24 to @xmath25 decay widths is given by @xcite @xmath26 the first factor in the above expression , the ratio of squared lepton masses for the two decays , comes from the helicity suppression by the @xmath12@xmath13@xmath14 lepton-@xmath27 boson weak couplings . if , instead , the decay could proceed directly through the pseudoscalar current , the ratio @xmath28 would reduce to the second , phase - space factor , or approximately 5.5 . more complete treatment of the process includes @xmath29 , the radiative and loop corrections , and the possibility of lepton universality violation , i.e. , that @xmath30 and @xmath31 , the electron and muon couplings to the @xmath27 , respectively , may not be equal : @xmath32 where the `` @xmath33 '' indicates that radiative decays are fully included in the branching fractions . steady improvements of the theoretical description of the @xmath1 decay since the 1950 s have recently culminated in a series of calculations that have refined the standard model ( sm ) prediction to a precision of 8 parts in @xmath34 : @xmath35 a comparison with equation reveals that the radiative and loop corrections amount to almost 4% of @xmath36 . however , as discussed below , the experimental precision at this time lags behind the theoretical one by more than an order of magnitude . because of the large helicity suppression of the @xmath1 decay , its branching ratio is highly susceptible to small non-@xmath12@xmath13@xmath14 contributions from new physics , making this decay a particularly suitable subject of study , as discussed in , e.g. , refs . . this prospect provides the primary motivation for the ongoing pen@xcite and pienu@xcite experiments . of the possible `` new physics '' contributions in the lagrangian , @xmath1 is directly sensitive to the pseudoscalar one . at the precision of @xmath37 , @xmath28 probes the pseudoscalar and axial vector mass scales up to 1,000tev and 20tev , respectively@xcite . for comparison , cabibbo - kobayashi - maskawa ( ckm ) matrix unitarity and precise measurements of several superallowed nuclear beta decays constrain the non - sm vector contributions to @xmath38tev , and scalar to @xmath39tev @xcite . although scalar interactions do not directly contribute to @xmath28 , they can do so through loop diagrams , resulting in sensitivity to new scalar interactions up to 60tev @xcite . the subject was recently reviewed at length in ref . in addition , @xmath40 provides limits on masses of certain susy partners@xcite , and on neutrino sector anomalies@xcite . even though @xmath36 is small , of order @xmath41 , it would be relatively straightforward to measure precisely because of the large difference in decay ejectile energies , 69.79mev vs. 4.12mev for the @xmath42 and @xmath43 channels , respectively . however , the subsequent decay of the muon , @xmath44 , with the endpoint energy of 52.83mev , effectively brings the two processes kinematically closer , making the clean identification of the direct @xmath45 events challenging in a `` tail '' overlap region discussed below . this complication is partly mitigated by the @xmath46 times longer lifetime of the muon than that of the pion . selecting only early decays , say , within a couple of pion lifetimes , suppresses the sequential @xmath47 events by more than an order of magnitude , thus effectively enhancing the apparent @xmath48 branching ratio . nearly all experiments to date have used a variant of the stopped pion decay technique to measure @xmath36 . the technique based on detecting and normalizing @xmath49 to @xmath47 decays at rest in a nonmagnetic spectrometer is inherently independent of several important sources of systematic uncertainty that are shared by the two processes , and that consequently cancel in the ratio . prime examples of shared uncertainties that cancel include the relative placement of the target and the positron detector and the resulting acceptance . ( on the other hand , using a magnetic spectrometer removes this insensitivity , as it would introduce a strong dependence of the solid angle on @xmath50 momentum . ) further cancellations occur for the efficiency of positron detection ( largely independent of @xmath51 , although corrections at the @xmath37@xmath52 level are typically needed ) , various beam properties , efficiency of pion stop identification in the target , overall trigger efficiency etc . all of these quantities would have to be determined with higher precision in an experiment that would normalize the detected @xmath1 decay events to the corresponding number of beam pions stopped in the target . additional quantities such as the time displacement of the first detection interval from the arrival time of the pion and duration of the detection time gate , affect the extraction of @xmath36 in well understood ways . the first notable measurement of @xmath36 with 6% uncertainty ( rather than an upper limit or detection of a small number of @xmath1decay events ) , was performed by the university of chicago group with positive pions stopped inside a double - focusing magnetic spectrometer @xcite . this work was soon followed by a @xmath53% measurement @xcite at the columbia nevis synchrocyclotron laboratory ; decay positrons from @xmath16 s stopped in an active target were detected in a 23@xmath5424 cm nai(tl ) detector . given the nai detector size and acceptance for photons , the analysis included low-@xmath55 inner bremsstrahlung ( ib ) radiative @xmath56 events in the branching fraction , in accordance with the theoretical definition of @xmath36 , as given in . the nevis measurement brought to light an important experimental uncertainty , related to how accurately the nai(tl ) energy response function is known . positrons in the few to a hundred mev energy range generate electromagnetic showers that can not be completely contained in a finite - sized physical detector volume ; there is a non - zero probability that some energy will escape the volume , primarily in the form of photons , inducing a low - energy `` tail '' in the response function to a monoenergetic positron . ( use of a magnetic spectrometer would eliminate this problem , though at the cost of significantly more complicated acceptance systematics . ) this intrinsic instrumental tail coexists with the physical `` tail '' of radiative decay events for which the accompanying photon escaped detection , primarily due to limited detector solid angle . we note in passing that in a nearly hermetic detector , radiative decays will generate a high energy `` tail '' extending above the @xmath5769.8mev two - body decay positron energy ; this was not a feature of the early measurements . in a sufficiently segmented detector such events can be properly treated by analyzing @xmath58 rather than @xmath59 alone . the tail correction for @xmath10 events falling below the @xmath43 decay endpoint energy in the nevis experiment was 9.1% with a @xmath570.7% combined systematic and statistical uncertainty . while this correction diminishes for larger detectors , especially with external active enclosures , it remains a dominant source of systematic uncertainty for @xmath36 , especially as experiments approach the theoretical precision of equation . the triumf group of bryman and collaborators @xcite used a similar technique to that used in the nevis measurement , with improvements . tina , the triumf nai(tl ) detector was larger at 46 cm @xmath60 51 cm long , and subtended a solid angle of @xmath61% of @xmath62sr , with energy resolution of about 4% full width at half - maximum ( fwhm ) at 70mev , almost four times better than the nevis detector s @xmath5715% . the experiment collected approximately @xmath63 @xmath1 decay events . the resulting branching ratio had a 1.2% overall uncertainty . the single largest contribution to the uncertainty , at 0.75% , arose from the tail correction . tina was used again in a follow - up measurement ( ) by the same group @xcite , with numerous improvements in experimental design including an increased solid angle coverage of 2.9% of @xmath62sr and an order of magnitude more detected @xmath1 decay events . the measured branching fraction @xmath64 \times 10^{-4 } \ , , \label{eq : pi_e2_britt_92}\ ] ] is in excellent agreement with standard model predictions . at the paul scherrer institute ( psi ) a bern psi group performed a measurement of @xmath36 at a similar level of precision using a radically different apparatus @xcite , essentially simultaneously with the triumf measurement . psi setup is shown in . the key component of the apparatus was a nearly hermetic @xmath65sr calorimeter consisting of 132 bismuth germanium oxide ( bgo ) detectors in the shape of hexagonal prisms , each 20 cm or 18 radiation lengths long , and with inscribed radius of 5.5 cm . the detectors were arranged so that a narrow opening allowed the beam ( almost equal parts pions and muons ) to stop in a small cylindrical plastic scintillator ( active target ) , whose light was read out by a central calorimeter crystal . the resulting rms resolution was sub-2% at 70mev . because of the hermetic nature of the calorimeter , radiative muon decays presented a significant background under the @xmath66 signal peak in the energy spectrum . a total of @xmath67 @xmath49 , and @xmath68 @xmath69 events were recorded . the resulting branching fraction was @xmath70 \times 10^{-4 } \ , , \label{eq : pi_e2_bern - psi_93}\ ] ] with the systematic uncertainty coming primarily from corrections for photonuclear reactions , radiative @xmath43 decay background , and electromagnetic losses . summarizes the available experimental results on @xmath36 . @lccc experiment & @xmath1 events & @xmath71 & reference + u. chicago / efins & 1.2k & @xmath72 & @xcite + columbia / nevis & 10.8k & @xmath73 & @xcite + triumf / tina & 32k & @xmath74 & @xcite + triumf / tina & 190k & @xmath75 & @xcite + bern psi & 120k & @xmath76 & @xcite + pdg average & 342k & @xmath77 & @xcite + combined , the 1992/93 triumf and psi results define the present experimental precision for the decay @xcite , as the earlier measurements do not significantly affect the current global average : @xmath78 this experimental world average lags behind the theoretical precision of by more than an order of magnitude . the precision gap , already notable in 1993 , has only increased with the passage of time , and has motivated a new generation of experiments currently under way , and briefly discussed below . in 2006 a new measurement of @xmath36 was proposed at the paul scherrer institute by a collaboration of seven institutions from the us and europe @xcite , with the aim to reach @xmath79 the pen experiment uses an upgraded version of the pibeta detector system , described in detail in ref . @xcite , and previously used in a series of rare pion and muon decay measurements @xcite . the main component of the pen apparatus , shown in , is a spherical large - acceptance ( @xmath80sr ) electromagnetic shower calorimeter . the calorimeter consists of 240 truncated hexagonal and pentagonal pyramids of pure csi , 22 cm or 12 radiation lengths deep . the inner and outer diameters of the sphere are 52 cm and 96 cm , respectively . beam particles entering the apparatus with @xmath81mev/@xmath82 are first tagged in a thin upstream beam counter ( bc ) and refocused by a triplet of quadrupole magnets . after a @xmath83 m long flight path they pass through a 5 mm thick active degrader ( ad ) and a low - mass mini time projection chamber ( mtpc ) , finally to reach a 15 mm thick active target ( at ) where the beam pions stop . decay particles are tracked non - magnetically in a pair of concentric cylindrical multiwire proportional chambers ( mwpc1,2 ) and an array of twenty 4 mm thick plastic hodoscope detectors ( ph ) , all surrounding the active target . the bc , ad , at and ph detectors are all made of fast plastic scintillator material and read out by fast photomultiplier tubes ( pmts ) . signals from the beam detectors are sent to waveform digitizers , running at 2gs / s for bc , ad , and at , and at 250ms / s for the mtpc . as discussed in the preceding section , a key source of systematic uncertainty in previous @xmath1 experiments has been the hard to measure low energy tail of the detector response function , caused by electromagnetic shower leakage from the calorimeter , mostly in the form of photons . pen is no exception in this respect . other physical processes , if not properly identified and suppressed , also contribute events to the low energy part of the spectrum ; unlike shower leakage they can also produce high energy events . one process is ordinary pion decay into a muon in flight , before the pion is stopped , with the resulting muon decaying within the time gate accepted in the measurement . another is the unavoidable physical process of radiative decay . the latter is well measured and properly accounted for in the pen apparatus , as discussed below in the section on radiative decay . shower leakage and pion decays in flight can only be well characterized if the @xmath47 chain can be well separated from the direct @xmath45 decay in the target . therefore much effort has been devoted to digitization , filtering and analysis of the target waveforms @xcite , as illustrated in . the method used for separating the 2-peak ( @xmath1 ) and 3-peak ( @xmath47 ) events is illustrated and explained in . path length in the target , reconstructed from the observed @xmath16 and @xmath50 trajectories . shown are events with proper @xmath47 sequences for which the @xmath50 signal is well separated from other signals . right : difference in @xmath84 for the assumptions of a target waveform with / without a muon pulse present . the observable is normalized such that @xmath10 events peak at @xmath85 , and @xmath47 at @xmath86 . shown are events for two different combinations of @xmath50 energy and decay time resulting in almost pure samples of @xmath87 and @xmath47 , respectively . tiny admixtures of another process are readily identified and are of great help in reducing the systematic uncertainties . ] the key to the method is provided by the beam and mwpc detectors which are used to predict the pion and positron energy deposition in the target , and the times of their signals . once the predicted waveform is subtracted the net waveform is scanned for the presence of a 4.1mev muon peak . the difference between the minimum @xmath84 values with and without the muon peak is reported as @xmath88 , constructed so that clean 2-peak and 3-peak fits return values of @xmath85 and @xmath86 , respectively . the scan is fast and returns a @xmath88 value for every event , as illustrated in the figure . a particularly telling figure regarding the pen data quality are the decay time histograms of the @xmath10 decay and @xmath89 sequence , shown in for a subset of data recorded in 2010 . events and @xmath47 sequential decay events . the two processes are distinguished primarily by the total @xmath50 energy and by the absence or presence , respectively , of an extra 4.1mev ( muon ) in the target due to @xmath90 decay . the @xmath1 data are shown with a pion lifetime @xmath91ns exponential decay function superimposed . the @xmath47 data were prescaled by a factor of @xmath571/64 ; they are shown with the cut on the probability of @xmath92% for a second , pile - up muon to be present in the target at @xmath93 , the time of the nominal pion stop . the turquoise histogram gives the @xmath47 yield constructed entirely from the measured @xmath10 data folded with the @xmath43 decay rate , and corrected for random muons ; it perfectly matches the bold dark blue histogram . the two lower plots show the observed to predicted ratios for @xmath1 and @xmath47 events , respectively . the scatter in the ratio plots is statistical in nature . ] the @xmath10 data follow the exponential decay law over more than three orders of magnitude , and perfectly predict the measured @xmath47 sequential decay data once the latter are corrected for random ( pile - up ) events . both event ensembles were obtained with minimal requirements ( cuts ) on detector observables , none of which bias the selection in ways that would affect the branching ratio . the probability of random @xmath94 events originating in the target can be controlled in the data sample by making use of multihit time to digital converter ( tdc ) data that record early pion stop signals . with this information one can strongly suppress events in which an `` old '' muon was present in the target by the time of the pion stop that triggered the readout . the `` intrinsic '' low energy tail of the pen response function below @xmath57 50mev , due to shower losses for @xmath1 decay events for pions at rest , amounts to approximately 2% of the full yield . events with @xmath90 decays in flight , with subsequent ordinary michel decay of the stopped muon in the target , add a comparable contribution to the tail . the two contributions can be simulated accurately , with the respective detector responses independently verified through comparisons with measured data in appropriately selected processes and regions of phase space . these response functions are entered into the maximum likelihood analysis ( mla ) , used to describe all measured processes simultaneously . the quantity @xmath28 is evaluated in the mla as the ratio of the magnitudes found for the @xmath95 and @xmath47 processes . although verified through comparisons with monte carlo simulations , the intrinsic tail itself is not directly measurable at the required precision because of the statistical uncertainties arising in the tail data selection procedure . radiative decay processes are directly measurable and accounted for in the mla procedure . more information about the pen / pibeta detector response functions is given below , in the section on @xmath7 decays , and in @xcite . during the 2008 - 10 production runs the pen experiment accumulated some @xmath96 @xmath10 , and more than @xmath97 @xmath47 events , as well as significant numbers of pion and muon radiative decays . a comprehensive blinded maximum likelihood analysis is under way to extract a new experimental value of @xmath36 . as of this writing , there appear no obstacles that would prevent the pen collaboration to reach a precision of @xmath98 . the competing pienu experiment at triumf , discussed below , has a similar precision goal . the near to medium future will thus bring about a substantial improvement in the limits on @xmath42-@xmath43 lepton universality , and the attendant sm limits . the new pienu experiment at triumf builds on the earlier measurements at the same laboratory @xcite , aiming at a significant improvement in precision through refinements of the technique used . the branching fraction will again be obtained from the ratio of positron yields from the @xmath10 decay and from the @xmath47 decay chain . as in other experiments that detect decay positrons in a nonmagnetic spectrometer , many normalization factors , such as the solid angle of positron detection , cancel to first order in pienu , leaving only corrections for small energy - dependent effects . major improvements in precision in pienu over the earlier triumf tina measurement derive from improved geometry and beamline , a superior calorimeter , as well as high - speed digitizing of all detector signals . shows a schematic rendition of the pienu experimental apparatus . a 75mev/@xmath82 @xmath16 beam from the improved triumf m13 beam line @xcite is tracked in wire chambers , identified by plastic scintillators , and stopped in a 0.8 cm thick scintillator target . fine tracking near the target is provided by two sets of single - sided silicon strip detectors located immediately upstream and downstream of the target assembly . the positrons from @xmath10 and @xmath47 decays are detected in the positron telescope , which consists of a silicon strip counter , two thin plastic counters , and an acceptance - defining wire chamber that covers the front of the crystal calorimeter . the solid - angle acceptance of the telescope counters is 20% of 4@xmath99sr . the primary energy measurement is performed using bina , a cylindrical , single - crystal 48 cm @xmath60 48 cm long nai(tl ) detector @xcite . bina s energy resolution of approximately 2.2% fwhm for incident positrons at 70mev @xcite is better by a factor of approximately two than the resolution observed in the previous measurement with tina @xcite . the nai detector is surrounded by two layers consisting of 97 pure csi crystals , 8.5 cm thick and 2@xmath6025 cm long @xcite to capture electromagnetic shower leakage from bina , thus helping to suppress the instrumental low - energy `` tail '' . as in pen , analog signals from plastic scintillators , silicon strip , nai , and csi detectors are recorded as waveforms , using appropriately fast digitizers . to suppress background arising from old muon decay signals in the target region and to reduce possible distortions in the time spectrum due to pileup , the incident pion rate is kept at @xmath100/s . the pienu collaboration has accumulated upwards of @xmath101 @xmath10 events through 2012 . combining the corresponding statistical uncertainty with reduced systematic uncertainties , the collaboration expects to reach @xmath102% . the data acquisition phase of the experiment indeed ended in 2012 , and , as of this writing , the collaboration is working on data analysis . the decay @xmath2 proceeds via a combination of qed ( inner bremsstrahlung , @xmath103 ) and direct , structure - dependent ( @xmath104 ) amplitudes @xcite . under normal circumstances , as in @xmath105 decay , the direct amplitudes are hopelessly buried under an overwhelming inner bremsstrahlung background . however , the strong helicity suppression of the primary non - radiative process , @xmath10 , also suppresses the bremsstrahlung terms , making the direct structure dependent amplitudes measurable in certain regions of phase space @xcite . ( we recall that the same helicity suppression made possible sensitive searches for non-@xmath12@xmath13@xmath14 interaction terms in precision measurements of the primary @xmath10 decay , discussed in . ) this result is of great use to effective low - energy theories of the strong interaction , primarily chpt , which rely on the @xmath104 amplitudes to provide important input parameters . whereas the @xmath103 amplitude is completely described by qed , the structure - dependent amplitude can be parametrized in terms of the pion form factors . as indicated in the tree - level feynman diagrams in , standard @xmath12@xmath13@xmath14 electroweak theory requires only two pion form factors , @xmath106 , axial vector , and @xmath107 , vector ( or polar - vector ) , to describe the @xmath104 amplitude . ) and structure - dependent ( @xmath104 ) amplitudes determined by the vector and axial - vector form factors , @xmath107 and @xmath106 , respectively . a new interaction type , such as one mediated by a hypothetical tensor particle proposed and discussed in the literature through the 1990s , would add an @xmath104 amplitude defined by a corresponding form factor , i.e. , @xmath108 . ] the amplitudes ( form factors ) @xmath106 and @xmath107 in principle depend on the 4-momentum transfer to the @xmath42-@xmath109 pair ( or to the @xmath27 boson ) , @xmath110 . in @xmath111 decay @xmath112 remains low , @xmath113 , so that it is a good approximation to evaluate @xmath107 and @xmath106 at @xmath114 , often referred to as the soft pion limit . it is useful at this point to consider the tree - level double differential branching ratio for the @xmath111 decay which , in the usual parametrization , takes the form first worked out in detail by de baenst and pestieau @xcite : @xmath115 \right . \nonumber \\ + \left.\vphantom{\left(\frac{f_v m_\pi^2}{2 f_\pi m_e}\right)^2 } \left ( \frac{f_v m_\pi}{f_\pi } \right ) \left[\left ( 1+\gamma \right ) s_{\rm int}^+\left ( x , y \right ) + \left ( 1-\gamma \right ) s_{\rm int}^- \left ( x , y \right ) \right ] \right\}\ , , \label{eq : rpd_sm_rate}\end{aligned}\ ] ] where @xmath116 and @xmath117 are the appropriately normalized photon and electron ( positron ) energies , respectively , @xmath118 is the ratio of the axial and the vector pion form factors , @xmath119 is the familiar pion decay constant and `` int '' denotes interference terms between the @xmath103 and @xmath104 amplitudes . the functional dependence of the terms @xmath103 , @xmath120 , @xmath121 @xmath122 and @xmath123 on @xmath124 and @xmath125 is well established and is given in the literature , e.g. , in @xcite . we note that the @xmath120 and @xmath121 terms , which multiply the @xmath126 and @xmath127 form factor terms , respectively , map very different portions of the 3-body phase space in the final state . the @xmath120 term peaks for large values of positron energies , @xmath128 and moderate photon energies , @xmath129 , where there is relatively little background from the @xmath103 terms . the @xmath121 term , on the other hand , peaks for low values of @xmath130 , where the @xmath103 amplitude is comparatively strong , as shown in ; this kinematic region is also susceptible to background from muon radiative decays . + it is not surprising , then , that most of the measurements of the @xmath3 decay to date have focused on determining the @xmath120 amplitude . the conserved vector current ( cvc ) hypothesis @xcite relates the vector form factor @xmath107 to the @xmath131 decay amplitude @xcite : @xmath132 during a long period of time ending in 2012 , then current particle data group values for the @xmath133 lifetime and the @xmath131 branching fraction led to a cvc prediction of @xmath134 . as a rule , radiative pion decay measurements had access only to one structure dependent amplitude , the @xmath120 . treating the value of @xmath107 as known , predicted by the cvc hypothesis , allowed analyses of @xmath3 data to extract @xmath106 , or , more specifically the ratio @xmath118 . recent measurements of the @xmath133 lifetime , primarily that of the primex collaboration @xcite , have led to a change in the pdg accepted value of the pion lifetime , along with a reduced uncertainty . furthermore , bernstein and holstein have recently pointed out the necessary correction to the cvc expression in arising from isospin breaking effects @xcite . combined , the two effects lead to a new isospin corrected cvc prediction for the pion vector form factor : @xmath135 which modifies the previously reported values for @xmath106 or @xmath136 . before discussing individual measurements of the radiative @xmath56 decay , we note that the related decay @xmath137 , where the photon is virtual , adds @xmath138 , a second axial vector amplitude or form factor , to the @xmath104 terms . the @xmath138 form factor is proportional to @xmath139 , the electromagnetic ( charge ) radius of the pion squared @xcite . the first determination of the @xmath104 amplitudes in @xmath3 decay was performed by depommier and collaborators . in their apparatus pions were stopped in a plastic scintillator target , and the decay positrons and photons detected in opposing nai(tl ) and lead glass detectors , each preceded by a pair of thin plastic scintillator detectors in front @xcite . signals from all detectors were recorded on oscilloscope pictures which were subsequently analyzed . a total of @xmath140 @xmath3 events resulted after background subtraction ; some 110 were due to the @xmath104 emission . in their analysis , the authors used @xmath141 and obtained two solutions for @xmath142 , 0.4 and @xmath143 ; the data did not favor one over the other . the authors also reported a branching ratio for @xmath144mev with 17% relative uncertainty , where , however , the detector resolution had not been taken into account , so the result could not be compared with theoretical calculations . the berkeley - ucla experiment by stetz and collaborators @xcite detected the positrons in a magnetic spectrometer and the photons in a lead glass erenkov hodoscope . the experimenters observed @xmath145 events , and reported a branching ratio of @xmath146 for @xmath147mev and @xmath148 . to determine the @xmath104 terms , the authors used @xmath149 , based on the 1977 average value for the @xmath133 lifetime . they , too , found two solutions for @xmath142 : @xmath150 or @xmath151 , neither of which was clearly favored by the data . in the mid-1980 s at the swiss institute for nuclear research ( sin , now part of psi ) a lausanne zurich collaboration @xcite set up an improved version of the berkeley - ucla apparatus to study the @xmath3decay and determine @xmath106 . they used an intense pion beam with the stopping rate in the target of @xmath152 , an 8@xmath608 array of nai(tl ) crystals for photon detection , and a large - acceptance magnetic spectrometer for the positron . an mwpc near the target provided additional tracking information . data were taken in two detector geometries , centered around @xmath153 and @xmath154 , respectively . the former configuration accepts negligible contributions from the @xmath103 and @xmath121 amplitudes , while the latter accepts a sizable @xmath121 contribution still with minimal @xmath103 background . using this method , the collaborators collected @xmath155 @xmath3 events , as well as @xmath156 @xmath1 events for normalization . thanks to the expanded phase space coverage compared to previous experiments , bay were able to resolve the quadratic ambiguity inherent in the @xmath120 analysis only , and eliminate the negative - sign solution by a confidence factor of 8.5 . using the cvc value of @xmath157 , incidentally the same central value as the current best isospin - corrected cvc prediction in , the authors obtained @xmath158 , or @xmath159 . around the same time as the sin experiment , a group at the los alamos meson physics facility ( lampf ) made a measurement of @xmath160 @xcite using the crystal box detector which consisted of 396 nai(tl ) crystals , 36 plastic scintillation hodoscope counters , and a cylindrical drift chamber surrounding the stopping target @xcite . unlike the other experiments discussed above , the crystal box detector s single - particle acceptance for events originating in the target was about 45% of @xmath62sr . such a large acceptance enabled the experimenters to cover a broad portion of the decay phase space , collecting 2364 coincidence events that included both decay signals and random coincidences . however , only @xmath161 high - energy nearly back - to - back photon - positron pairs were used to determine @xmath142 , as they were largely background free and strongly dominated by the @xmath120 amplitude . the low statistics is reflected in the large relative uncertainty in @xmath162 ; however the remaining @xmath572300 coincident events were included in a maximum likelihood analysis which preferred the positive @xmath142 solution by a factor of 2175:1 . in all , the likelihood analysis found a total of @xmath163 @xmath164 events , in good agreement with the integral of the @xmath42-@xmath142 timing peak . thus , although a low statistics measurement , this experiment s main result is the strongly favored selection of the @xmath165 solution , removing the prior longstanding quadratic ambiguity . although the subject of this review are measurements of rare decays of the pion , it is worthwhile to consider the 1988 analysis of dominguez and sol @xcite who extracted a soft - pion value for @xmath166 from semileptonic tau lepton decays , @xmath167 . analysis of decays with odd ( even ) values of @xmath168 gives access to the vector ( axial - vector ) hadronic spectral functions up to the kinematical limit @xmath169 . dominguez and sol studied and refined such fits to existing @xmath170 decay data , and extracted the soft - pion value of @xmath171 , or @xmath172 , with @xmath173 . we note that this result is _ independent of pion radiative decay _ data . the lone measurement of @xmath111 decays in flight was performed by the moscow institute for nuclear research ( inr ) group , using the istra apparatus and a 17gev pion beam at the ihep protvino u70 accelerator @xcite . the experimental technique and systematics are radically different from all of the stopped pion decay measurements discussed so far . since decays occurred in flight , the experimenters were free to use negative pions and observe @xmath174 decays , making this the only experiment to date to do so . the wide acceptance of the apparatus enabled the experimenters to study the pion rest frame kinematical region defined by @xmath175 the results of this work , obtained in a maximum likelihood analysis , can be summarized as follows : ( a ) @xmath176 ( with no corresponding value for @xmath107 quoted explicitly ) , ( b ) negative solution for @xmath142 disfavored by a likelihood factor of @xmath177 or 6.7 standard deviations , ( c ) an independently determined @xmath178 , and ( d ) the branching ratio for the kinematic limits of equation , @xmath179 . although not explicitly given , the total number of @xmath164 decay events appears to be approximately 90 after background subtraction , which would explain the large quoted uncertainties . besides the strong preference for the @xmath165 solution , this work presented another notable result : a deficit of @xmath121 events , forcing a non - physical negative @xmath121 amplitude . the authors speculated that the deficit may be due to a destructive interference with a tensor term of the size @xmath180 . we will revisit the tensor hypothesis below . merits special mention , as it contributes information not accessible through the regular @xmath3 radiative decay channel . following unsuccessful attempts @xcite , the @xmath5 decay was first observed by the sindrum i collaboration in a 1985 experiment @xcite ; the same data were reanalyzed more carefully and the results reported in 1989 @xcite . the sindrum detector system , which became known as `` sindrum i '' after the construction of `` sindrum ii '' , was built primarily to search for the forbidden @xmath181 decay @xcite . the instrument , schematically depicted in , was capable of recording search @xcite : b , @xmath43 beam ; s , focusing solenoid ( not used with the pion beam ) ; t , target ( a different target was used for the @xmath5 measurement ) ; c , five cylindrical mwpc s ; h , hodoscope of 64 scintillators ; l , light guides for the hodoscope ; p , 128 photomultiplier tubes ; a , cathode strip preamplifiers and anode wire amplifier / discriminators ; m , magnet coil . for further details see @xcite ] the trajectories of electrons and positrons in a solenoidal magnetic field with the help of five concentric cylindrical mwpc s of very similar design and construction to those used later in the pibeta / pen experiments . decay particles were required to reach a 64 element scintillating hodoscope situated outside the tracking detectors , resulting in a lower threshold on transverse momentum around 17mev/@xmath82 . a cone - shaped 12-element segmented scintillating target was used , enabling a reduction of random coincidences between beam pions and decay particles by an order of magnitude . in their first paper egli @xcite reported merely the first observation of the decay , with 79 recorded events . the authors also reported values of form factor ratios @xmath182 and @xmath183 with large uncertainty limits . they were , however , able to exclude by a ratio of 9:1 the previously reported large negative value of @xmath184 from @xmath3 @xcite in favor of a smaller positive solution @xmath185 , thus anticipating the soon to follow @xmath3 results of the lausanne zurich @xcite and lampf @xcite groups . in a second article @xcite , the sindrum i collaboration reported results of a more careful and comprehensive analysis of the same 1985 data set . by applying a series of cuts designed to reduce the backgrounds , the authors arrived at a data set of 98 events with 1 background event , based on @xmath186 pions stopped in the target . using this event set , the detector acceptance , the cvc value @xmath187 , the bay value @xmath188 @xcite and the pcac value for the second axial form factor @xmath189 , the authors derive the branching ratio for the full phase space : @xmath190 where the second uncertainty is propagated from @xmath191 . of additional interest are the values for the three pion form factors derived in a likelihood analysis of the data : @xmath192 it is worth noting that fixing the vector form factor value to @xmath193 does not significantly alter the maximum likelihood values of @xmath106 and @xmath138 . a subsequent dubna experiment @xcite which detected only 7 events of @xmath5 decay did not improve the precision of the sindrum i branching ratio for the decay . the @xmath56 and @xmath194 results published prior to 2000 may be summarized as follows . 1 . the positive - sign solution for @xmath118 was established with high likelihood by three experiments , effectively ruling out a negative solution . 2 . the number of reconstructed events adds up to less than 1,200 events @xcite . the combined statistical and systematic uncertainties of the parameter @xmath195 extracted by the individual experiments range from 12% @xcite to 56% @xcite . the world average @xcite @xmath196 was assigned the confidence level of 0.175 by the pdg . there were only two , relatively low precision measurements of @xmath107 resulting in a world average with a 50% uncertainty : @xmath197 4 . there was only one , low precision measurement of the @xmath198 branching fraction , defined over a correlated kinematic region given in equation . the protvino experiment @xcite raised the possibility of a substantial tensor term , @xmath108 , subsequently discussed during more than a decade in a series of theoretical papers @xcite . study of the radiative pion decay has been a major component in the long term program of the pibeta / pen experiments at psi . firstly , the strong intrinsic physics significance for low - energy qcd and chpt as well as the sensitivity to non-@xmath12@xmath13@xmath14 interaction terms such as tensor , and its relatively poor experimental quantification prior to 2000 discussed in the preceding section , clearly placed radiative pion decay at high priority . secondly , @xmath199 events for which the positron annihilates externally create a significant background in the measurement of the pion beta decay @xmath200 , @xmath7 , the primary goal of the pibeta experiment , discussed in below . thirdly , @xmath7 events for which one of the photons in the subsequent @xmath133 decay converts into an asymmetric @xmath201 pair , with only one detected electromagnetic shower , create a background for the @xmath3 process . furthermore , a precise measurement of @xmath28 of equation is impossible without a precise knowledge of the radiative decay width . in fact , because of the overlapping nature of the instrumental response functions to the various decay processes , precision measurement of any one of them requires simultaneous detection and characterization of all pion and muon decay processes present in the data sample . for all these reasons , the pibeta collaboration @xcite performed a series of measurements at the paul scherrer institute , focused on improving the experimental precision of the @xmath3 branching ratio as well as the form factors @xmath106 and @xmath107 . the apparatus is essentially the same as in the pen configuration shown in . the only differences were the absence of the beam mini time projection chamber ( mtpc ) , the use of a segmented target in the early runs ( ) , a thicker active target and degrader detectors to accommodate the higher momentum of @xmath57114mev/@xmath82 , pion beam , and custom electronics for 1- and 2-arm triggers . the arrangement of the central detectors is shown in . the data were collected in two distinct sets of runs . the first pibeta radiative pion decay measuring period took place during 19992001 , in a configuration optimized for recording pion beta decay , @xmath202 , events with a high stopping rate of pions in the target , averaging approximately @xmath203 . the experiment trigger logic was based on shower energy summed over a cluster of adjacent csi calorimeter detector modules , with two energy thresholds : a high one of @xmath204mev , and a low one of @xmath205mev . the experiment recorded every event containing two showers , both with energy above the high threshold ( ht ) , and spatially separated by an opening angle @xmath206 . two - shower events with one exceeding the high and the other the low energy threshold were prescaled by a factor of at least an order of magnitude , while events with two low - threshold showers were prescaled even more . one - arm events , consisting of one shower were also recorded , with the purpose of collecting significant samples of @xmath207 and @xmath208 events for normalization and systematic studies . detector response to the @xmath209 single - arm events , shown in , best illustrates the intrinsic performance of the spectrometer . over @xmath210 @xmath3 events were collected and analyzed in this set of runs . the results , bringing about an order of magnitude improvement in precision of the branching ratio and @xmath160 , were published in @xcite . however , the authors found a deficit of events in one kinematic region , corresponding to high-@xmath55 and low-@xmath211 events , that was well outside the limits of the statistical uncertainties of the fit . the authors concluded that further study was needed in a dedicated measurement , as the first run had been optimized for measurement of @xmath202 decays . the second run , dedicated to radiative @xmath3 decays took place in 2004 , with the pion stopping rate lowered to @xmath212 and the trigger electronics modified to accept all high - threshold one - arm events . the first half of the run was carried out with the same detectors as in 1999 - 2001 , while in the second half the 9-element active target was replaced with a single unit . approximately @xmath213 additional events were collected over significantly broader kinematic regions . most importantly , the lower beam rate resulted in large improvements in the ratio of peak signal to accidental background ( p / b ) in @xmath214 spectra , and sufficient low - energy data to perform an independent energy calibration for the charged and neutral showers in the ht trigger . the combined data sets were carefully analyzed , and the results published in @xcite . the key results are shown in figures [ fig : pienug_regions ] , [ fig : slope-2013 ] and [ fig : f_a - f_v-2013 ] . distribution of the kinematic variable @xmath215 for 8 regions in the @xmath211-@xmath55 ( or @xmath125-@xmath124 ) plane . the layout of the regions in the @xmath124-@xmath125 plane is indicated in the lower right panel . solid red histograms : results of geant3 calculations for the best - fit values of @xmath107 , @xmath106 , and @xmath216 . ] slope analysis for @xmath217 in the pibeta @xmath3 measurement @xcite . black data points : best - fit values of @xmath217 for three momentum bins in the measurement . solid and dashed blue lines represent the central values and @xmath218 band , respectively , of the minimum @xmath84 fit . solid red square point and red lines : results of the resonance chpt calculation of mateu and portols @xcite . the current isospin - corrected cvc prediction @xmath219 is indicated by the empty circle offset slightly to @xmath220 for better visibility . ] for the minimum value @xmath221 ( solid red dot ) plus 1 , 2 , and 4 units , respectively , in the @xmath222 parameter plane , keeping the slope parameter @xmath216 fixed at the theoretical value @xmath223 @xcite . the range of the cvc prediction @xmath224 is indicated by the dashed vertical lines . ] with the improved calibrations , the previously observed event deficit vanished , as seen in , which shows the measured yield in 8 different kinematic regions , compared with the best - fit standard description of equation with the addition of radiative corrections by bystritsky @xcite . no statistically significant deviations were observed . the wider kinematic coverage of the new measurement enabled a much improved analysis , including an independent determination of the polar- and axial - vector form factors @xmath107 and @xmath106 , and , for the first time , a determination of the 4-momentum transfer @xmath225 dependence of the polar vector amplitude @xmath107 , : @xmath226 the uncertainties on the slope parameter are wide enough to accommodate the resonance chiral perturbation calculation of mateu and portols @xcite . even after the inclusion of the lower - energy phase space regions , the pibeta @xmath3 decay data are strongly dominated by the @xmath120 amplitude , leading to a stringent constraint on @xmath227 , and a lax one on @xmath228 . as in previous measurements , the probability of the negative solution for @xmath182 is disfavored by a large factor , about 880 . the narrow elliptically shaped locus of best - fit values for @xmath106 and @xmath107 is shown in . the corresponding equation of the major axis of the ellipse is given by @xmath229 we note that , for a fixed value of @xmath107 , the uncertainty in @xmath106 has been reduced 16-fold compared to the prior world average . in a fully unconstrained fit , the authors report @xmath230 the pibeta result for @xmath107 represents a 5-fold improvement over the prior average . the correlated nature of the pibeta form factor results allows for a posteriori improvement in precision of @xmath106 as the precision of the determination of the @xmath133 lifetime improves . finally , thanks to the experiment s broad kinematic coverage , the pibeta authors have reported a branching ratio @xmath231 with more than an order of magnitude improved precision over the previous result . the pen experiment will add new data to the existing @xmath3 data set . since the pen beam stopping rate is significantly lower than that used in the 2004 pibeta run , the impact on the statistical uncertainties will not be great . however , the cleaner running conditions of pen present greater access to the kinematic region @xmath232mev , previously strongly contaminated by muon decay background , and , thus , holding the prospect for a better constraint of the @xmath121 amplitude , which , in turn , would lead to improved precision in the direct determination of @xmath107 . these results will be forthcoming in the near future . unlike the @xmath233 decays discussed in the preceding sections , the extremely rare , @xmath234 , pion beta decay is not suppressed by any factor apart from the restricted phase space of final states due to the small difference between the charged and neutral pion masses , @xmath235 . as a pure vector @xmath236 transition , it is completely analogous to the superallowed fermi nuclear beta decays . in fact , it can be regarded as the simplest realization of the latter , fully free of complications arising from nuclear structure corrections . superallowed fermi decays have historically led to the formulation of the conserved vector current hypothesis , and have played a critical role in testing the unitarity of the cabibbo - kobayashi - maskawa quark mixing matrix through evaluations of the @xmath9 element@xcite . within the framework of the @xmath12@xmath13@xmath14 theory of the weak interactions , the pion beta decay rate can be expressed in terms of the leading - order width @xmath237 and the radiative and loop corrections @xmath238 as @xcite @xmath239 where @xmath240 is the fermi coupling constant and @xmath241 . finally , up to leading order in @xmath242 , the function @xmath243 has the value of @xmath244 . the overall uncertainty of the rate in is dominated by two comparable contributions , one from the @xmath245 factor , and the other from the @xmath238 radiative / loop corrections , each uncertainty contribution being in the range of 0.050.1% of the full rate @xcite . thus , the pion beta decay rate provides a direct means to determine the ckm matrix element @xmath246 . in fact , being free of nuclear structure corrections present in superallowed nuclear beta decays , and free of tree - level axial corrections present in neutron beta decay , @xmath7 decay offers the theoretically cleanest path to measuring @xmath9 and , hence , testing quark - lepton universality . however , the extremely low branching ratio for the process has so far limited the experimental accuracy . experimentally , @xmath202 decay is observed primarily through detection of the final decay particles produced in the near - instantaneous neutral pion decay ( @xmath247s ) . the positron is generally harder to detect with precise efficiency , except for @xmath16 decays in flight , since its kinetic energy ranges from 0 to only 4mev . the key @xmath133 decay modes and their branching ratios are : @xmath248 all experiments to date have focused on the @xmath249 channel rather than the @xmath250 , dalitz mode . due to the small charged to neutral pion mass difference , the maximum kinetic energy of the @xmath133 is low , about 75kev . thanks to the correspondingly low @xmath133 velocity , the directions of the two emitted photons deviate from @xmath251 by no more than @xmath252 , on average by @xmath253 . furthermore , the doppler broadening of the photon energies results in a narrow boxlike spectrum between @xmath254mev and @xmath255mev . this kinematics provides a robust signal that is additionally separated in time from the dominant prompt hadronic background events thanks to the 26ns pion lifetime . the first observation @xcite of the decay at cern in 1962 using a stopped @xmath16 beam and a combination of lead glass and nai(tl ) detectors , was followed by a quick succession of early measurements at cern @xcite , dubna ( total absorption pb glass erenkov counters ) @xcite , columbia university @xcite , lawrence berkeley lab @xcite and carnegie tech ( now carnegie mellon ) @xcite . the berkeley , columbia and carnegie tech experiments used combinations of spark chambers and scintillation detectors . the five measurements achieved approximately 20% uncertainties on the branching ratio , based on samples of between 30 and 50 events each . to date only three measurements of the @xmath202 decay branching ratio have been made with precision better than 10% , approximately one for every 20 years , which reflects the challenges of the task . the cern group of depommier @xcite was first to break the 10% uncertainty level in 1967 , using a lead glass erenkov detector array along with plastic scintillator detectors , as shown in . stop . candidate pion beta decay events are recorded as narrow coincidences of a pair of electromagnetic showers in the lead glass detectors a through h , delayed with respect to the @xmath16 stop pulse . detectors 4 and 5 veto charged particle events in the calorimeter array . for details see @xcite . ] the cern apparatus had an acceptance of @xmath25622.4% , and enabled the experimenters to record @xmath257 pion beta decays from @xmath258 pion stops in the target . the reported branching ratio value was @xmath259 in very good agreement with cvc theory predictions , but not precise enough to test the radiative corrections at @xmath260 . the authors increased the lower - side uncertainty of their result in order to account for the possible influence of nuclear reaction products in their measurement of the @xmath133 detection efficiency of the lead glass detector array . in the early 1980 s a temple los alamos group performed a radically different measurement of the pion beta decay rate at the los alamos meson physics facility ( lampf ) @xcite . unlike previous experiments which detected decays of @xmath16 s brought to rest in a target , mcfarlane and collaborators detected decays of an intense pion beam in flight in the apparatus shown in . target was used for calibrations and removed during decay measurements . for further details see @xcite . ] at @xmath261 pions / s and with a momentum of @xmath262 , the lampf pion beam was orders of magnitude more intense and an order of magnitude higher in energy than in the previous experiments . the decay volume was evacuated to @xmath263torr in order to limit the pion hadronic interactions with material in the path , primarily single charge exchange reactions which would be hard to distinguish from pion beta decay events . because of the low recoil in the @xmath16 rest frame , the @xmath133 had transverse momentum of less than 5mev/@xmath82 , or 1% of the longitudinal momentum . the average @xmath133 momentum of about 505mev/@xmath82 implied a short mean @xmath133 laboratory flight length of about 100 nm before decay . the acceptance was defined by the combination of an upstream collimator , minimum opening angle of the two photons , and the solid angle of the active area of the two photon detectors . the latter were the two arms of the lampf @xmath133 spectrometer @xcite , appropriately modified @xcite for the pion beta decay measurement . each photon detector arm consisted of a front plastic scintillator veto detector followed by three successive lead glass converter layers ( 0.56 radiation length thick , each ) . each converter layer was followed by two scintillation hodoscopes for position and time measurement . finally , behind the forward crate containing the converter - hodoscope layers was a rear crate with an array of 15 lead glass blocks , forming a 14 radiation lengths thick calorimeter . the cross sectional area of the detector package as seen by the photons was @xmath264 . the arms were well matched to detecting the final - state photons that ranged between 175 and 350mev in energy . at @xmath265ps per arm , the time resolution was excellent for such large detectors . at @xmath266 ( in mev ) , the energy resolution was adequate for the task . this measurement apparatus has several advantages over the stopped decay technique . first , the hadronic background can be effectively eliminated in an evacuated decay region . second , the high pion beam momentum focuses the decay photons into a solid angle significantly smaller than @xmath62sr . finally , in principle , the measurement can be done with both charged pion states , although in practice the intensity of @xmath15 beams is significantly lower than that of @xmath16 . however , there are also significant disadvantages compared to the stopped decay measurement . the proper time that beam pions spend in the decay region is very short , about @xmath37 pion lifetimes , requiring enormously higher beam intensities to achieve high statistical precision for the rare pion beta decays . normalization , i.e. , counting the number of pions passing through the apparatus is very challenging , as the readily available processes such as the @xmath18 decay have radically different kinematics and detection systematics , and require separate detectors . finally , detection efficiency ( including detector acceptance ) , as well as trigger and event selection efficiency must be determined in absolute terms with high precision . given all the above considerations , the temple lampf group reported the experimental partial rate , @xmath267 , and branching ratio , @xmath268 : @xmath269 based on @xmath270 pion beta decay events , and @xmath271 beam pions . the overall 3.8% uncertainty breaks down to 3.1% statistical and 2.0% systematic contributions . the largest systematic uncertainty contributions arose from the number of beam pions , time in decay region , and the various efficiencies . the authors also extracted a new value of @xmath272 for the cosine of the cabibbo angle , or @xmath9 . while not as precise as the contemporary values of @xmath9 derived from superallowed fermi nuclear beta decays , it was consistent with them . the primary goal of the pibeta experiment conducted at psi and introduced in sections [ sec : pi_e2 ] and [ sec : rpd ] , was to improve by an order of magnitude the existing experimental precision of the pion beta decay branching ratio . the apparatus is described schematically in , while the central detector region is shown in . a @xmath57114mev/@xmath82 pion beam of the psi @xmath99e1 beamline was focused onto the segmented nine - element active target which stopped an average of @xmath273 pions / s . the pibeta approach was essentially similar to the 1967 cern experiment of depommier @xcite , but with larger acceptance , much improved energy resolution of the electromagnetic shower calorimeter , mwpc tracking of charged particles between the central region and the calorimeter , a more intense pion beam , and faster electronics and data acquisition system . the key detector response functions to pion beta decay events are shown in figures [ fig : pb_ang12 ] , [ fig : pb_energ ] and [ fig : pb_time ] . photon opening angle in the pibeta experiment for the pion beta decay event sample . data points reflect measurements , while the red histogram indicates results of a realistic geant3 monte carlo simulation including the full effects of the pibeta detector resolution . radiative @xmath3 decay events , if misidentified as @xmath7 events through positron annihilation in flight , have a significantly different @xmath274 signature . ] energy in the pibeta experiment . data points reflect measurements , while the red histogram indicates results of a realistic geant3 monte carlo simulation including the full effects of the pibeta detector resolution . ] event times with respect to the time @xmath93 of the active degrader ( ad ) signal , reflecting the beam pion stop time , in the pibeta experiment . the curve is a best - fit exponential decay with the decay parameter in good agreement with 26.03ns , the pion lifetime . the time range of @xmath275ns around @xmath93 , the pion stop time , was vetoed in the decay event trigger in order to suppress prompt hadronic backgrounds which had a @xmath276 times higher rate than the @xmath7 events . a prescaled sample of the prompt hadronic events was collected for systematic studies and calibrations . ] the corresponding spectra for the normalizing @xmath1 decay , and for the @xmath3 decay which can contribute background events for the pion beta decay , are shown in sections [ sec : pen ] and [ sec : pibeta - rpd ] . since a subset of pion beta decays can also be misidentified as radiative @xmath3 decays , it was essential that all three processes be simultaneously accounted for and understood at the few parts per thousand level . in fact , ordinary ( `` michel '' ) and radiative muon decays , though not discussed here , were included as well in the comprehensive pibeta analysis , since they comprise the majority of actual events occurring in the apparatus , and dominate the accidental background to all pion decay channels . figures [ fig : pb_ang12][fig : pb_time ] demonstrate that the pion beta decay event sample was clean and well described in terms of the relevant instrumental resolutions . during three runs in 1999 , 2000 and 2001 , the pibeta collaboration acquired over 64,000 pion beta decay events , which led to an improvement of the experimental precision of the @xmath7 branching ratio of about an order of magnitude @xcite . the pibeta result was evaluated in two ways , ( a ) normalizing to the average experimental @xmath1 branching ratio of @xmath277 ( `` exp . norm . '' ) , and ( b ) normalizing to the established theoretical value of , @xmath278 ( `` theo . norm . '' ) . the resulting values were : @xmath279 \times 10^{-8 } , \\ b^{\rm theo.\ norm.}_{\pi\beta } = \left[1.040 \pm 0.004({\rm stat } ) \pm 0.004({\rm syst})\right ] \times 10^{-8}\,.\end{aligned}\ ] ] the external experimental uncertainty included the branching ratios @xmath280 , @xmath281 and that of @xmath282 , the pion lifetime , and were dominated by the former . the combined internal systematic uncertainty was dominated by the uncertainties related to the ratio of the `` pion gate '' live fractions for the @xmath7 and @xmath1 processes , and the number of the @xmath1 events used for normalization , followed by the ratio of acceptances for the same two processes . this result represents the most stringent test of cvc and cabibbo universality in a meson , as well as of the proper treatment of the radiative corrections , which , combined , predict @xmath283 at 90% confidence limits @xcite . excluding radiative corrections gives the range @xmath284 , a @xmath285 discrepancy from the pibeta result . the pibeta collaborators also extracted @xmath286 these values are in excellent agreement with the pdg recommended value for @xmath9 , although five times less precise . following in the same vein , it is tempting to go one step further and turn the pibeta determination of the @xmath7 branching ratio around , using it instead to evaluate @xmath28 . this is accomplished by fixing @xmath9 to its extraordinarily precise pdg 2013 recommended value of @xmath287 and adjusting @xmath28 until the extracted value of @xmath288 agrees . this intriguing exercise yields @xmath289 in good agreement with direct measurements reviewed in . appropriately averaging this value of @xmath28 with those listed in , one obtains a slightly higher value than the current pdg average : @xmath290 . given the advantage in lower theoretical uncertainties compared to nuclear decays , there was every incentive to pursue a higher precision result in the pion beta decay rate or branching ratio . however , the urgency of a further improvement was considerably reduced by the brookhaven national laboratory experiment bnl e865 result@xcite and the subsequent renormalization of @xmath291 that removed a longstanding 23@xmath292 shortfall in ckm matrix unitarity . in light of the considerable experimental challenges that an improved measurement would pose , the present full agreement of available data on @xmath9 with ckm unitarity means that a major new project on improving the precision of pion beta decay is not currently planned . we note , though , that the status of neutron beta decay is in considerable flux , with major discrepancies in the available data sets on both the beta asymmetry and the neutron lifetime @xcite . at present there are a number of projects addressing these deficiencies . while not as theoretically clean as the pion beta decay , neutron decay is free of the nuclear structure and coulomb corrections that affect nuclear beta decays , including the superallowed fermi @xmath293 transitions @xcite . a close look at the recent record of study of the decays of the charged pion reveals a great deal of activity and continued strong relevance . the extraordinary precision of the theoretical description of the @xmath294 , @xmath295 , @xmath296 and @xmath7 decays remains unmatched by the available experimental results . there have been impressive advances in the precision of the @xmath7 and @xmath3 decays in the past decade . as of this writing , we are on the verge of a significant improvement of the @xmath1 decay precision , as the pen and pienu experiments complete their analyses . even subsequent to that development , there will remain considerable room for improvement of experimental precision with high payoff in terms of limits on physics not included in the present standard model . this work remains relevant and complementary to the direct searches on the energy frontier currently under way at particle colliders , with considerable theoretical significance . this work has been supported by grants from the us national science foundation ( most recently phy-1307328 ) and the paul scherrer institute . 99 bartlett d , devons s , meyer s l and rosen j l 1964 _ _ * 136 * b1452 bacastow r , chesquire c , wiegand c and larsen r r 1965 _ _ * 139 * b407 bertram w k , meyer d i , carrigan , jr , r a and nadelhaft i 1965 _ _ * 139 * b617 depommier p , duclos j , heintze j , kleinknecht k , riesenberg h and soergel v 1968 _ _ * b4 * 189

the relations between smooth and peaked soliton solutions are reviewed for the camassa - holm ( ch ) shallow water wave equation in one spatial dimension . the canonical hamiltonian formulation of the ch equation in action - angle variables is expressed for solitons by using the scattering data for its associated isospectral eigenvalue problem , rephrased as a riemann - hilbert problem . the momentum map from the action - angle scattering variables to the flow momentum ( ) provides the eulerian representation of the-soliton solution of ch in terms of the scattering data and squared eigenfunctions of its isospectral eigenvalue problem . the dispersionless limit of the ch equation and its resulting peakon solutions are examined by using an asymptotic expansion in the dispersion parameter . the peakon solutions of the dispersionless ch equation in one dimension are shown to generalize in higher dimensions to peakon wave - front solutions of the epdiff equation whose associated momentum is supported on smoothly embedded subspaces . the eulerian representations of the singular solutions of both ch and epdiff are given by the ( cotangent - lift ) momentum maps arising from the left action of the diffeomorphisms on smoothly embedded subspaces . pacs numbers : 47.10.+g , 52.65.kj , 45.20.jj , 47.65.+a , 02.40.yy , 11.10.ef , 05.45.-a [ [ keywords ] ] keywords : + + + + + + + + + solitons , peakons , hamilton s principle , integrable hamiltonian systems , inverse spectral transform ( ist ) , singular solutions , shallow water waves , euler s equations , geodesic motion , diffeomorphisms , sobolev norms , momentum maps , compactons

1003.1338

euler s equations for irrotational incompressible ideal fluid motion under gravity with a free surface have an asymptotic expansion for shallow water waves that contains two small parameters , @xmath3 and @xmath4 , with ordering @xmath5 . these small parameters are @xmath6 ( the ratio of wave amplitude to mean depth ) and @xmath7 ( the squared ratio of mean depth to horizontal length , or wavelength ) . euler s equations are made non - dimensional by introducing @xmath8 for horizontal position , @xmath9 for vertical position , @xmath10 for time , @xmath11 for surface elevation and @xmath12 for velocity potential , where @xmath13 is the mean wave speed and @xmath14 is the constant gravity . the quantity @xmath15 is the dimensionless bond number , in which @xmath16 is the mass density of the fluid and @xmath17 is its surface tension , both of which are taken to be constants . after dropping primes , this asymptotic expansion yields the nondimensional korteweg - de vries ( kdv ) equation for the horizontal velocity variable @xmath18 at _ linear _ order in the small dimensionless ratios @xmath3 and @xmath4 , as the left hand side of @xmath19 here , partial derivatives are denoted using subscripts , and boundary conditions are @xmath20 and @xmath21 at spatial infinity on the real line . the famous @xmath22 traveling wave solutions ( the solitons ) for kdv ( [ kdv - eqn ] ) arise in a balance between its ( weakly ) nonlinear steepening and its third - order linear dispersion , when the quadratic terms in @xmath3 and @xmath4 on its right hand side are neglected . on the right hand side of equation ( [ kdv - eqn ] ) , a normal form transformation due to kodama @xcite has been used to remove the other possible quadratic terms of order @xmath23 and @xmath24 . the remaining quadratic correction terms in the kdv equation ( [ kdv - eqn ] ) may be collected at order @xmath25 . these terms may be expressed , after introducing a `` momentum variable , '' @xmath26 and neglecting terms of _ cubic _ order in @xmath3 and @xmath4 , as @xmath27 in the momentum variable @xmath28 , the parameter @xmath29 is given by @xcite @xmath30 thus , the effects of @xmath31dispersion also enter the nonlinear terms . after restoring dimensions in equation ( [ b - eqn ] ) and rescaling velocity @xmath32 by @xmath33 , the following @xmath34-equation " emerges , @xmath35 where @xmath36 is the dimensional momentum variable , and the constants @xmath37 and @xmath38 are squares of length scales . when @xmath39 , one recovers kdv from the @xmath34-equation ( [ dim - b - eqn ] ) , up to a rescaling of velocity . any value of the parameter @xmath40 may be achieved in equation ( [ dim - b - eqn ] ) by an appropriate kodama transformation @xcite . as we have emphasized , the values of the coefficients in the asymptotic analysis of shallow water waves at quadratic order in their two small parameters only hold , modulo the kodama normal - form transformations . hence , these transformations may be used to advance the analysis and thereby gain insight , by optimizing the choices of these coefficients . the freedom introduced by the kodama transformations among asymptotically equivalent equations at quadratic order in @xmath3 and @xmath4 also helps to answer the perennial question , `` why are integrable equations so ubiquitous when one uses asymptotics in modeling ? '' namely , there may be special values of the free parameters in the normal - form transformations of the asymptotics that allow integrability . [ [ integrable - cases - of - the - b - equation - dim - b - eqn . ] ] integrable cases of the @xmath34-equation ( [ dim - b - eqn ] ) . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the cases @xmath41 and @xmath42 are special values . for these values , the @xmath34-equation becomes completely integrable as a hamiltonian system . for @xmath41 , equation ( [ dim - b - eqn ] ) specializes to the integrable ch equation of camassa and holm @xcite . the case @xmath42 in ( [ dim - b - eqn ] ) recovers the integrable dp equation of degasperis and procesi @xcite . these two cases exhaust the integrable candidates for ( [ dim - b - eqn ] ) , as was shown using painlev analysis . the @xmath43family of equations ( [ dim - b - eqn ] ) was also shown in @xcite to admit the symmetry conditions necessary for integrability , only in the cases @xmath44 for ch and @xmath45 for dp . the @xmath34-equation ( [ dim - b - eqn ] ) with @xmath41 was first derived in camassa and holm @xcite by using asymptotic expansions directly in the hamiltonian for euler s equations governing inviscid incompressible flow in the shallow water regime . in this analysis , the ch equation was shown to be bi - hamiltonian and thereby was found to be completely integrable by the inverse scattering transform ( ist ) on the real line . this development of ist for ch equation ( [ dim - b - eqn ] ) with @xmath41 is discussed in section [ ist - sec ] . camassa and holm @xcite also discovered the remarkable peaked soliton ( peakon ) solutions of ( [ singlepeakon - soln],[peakontrain - soln ] ) for the ch equation on the real line , given by ( [ dim - b - eqn ] ) in the case @xmath41 . the peakons arise as solutions of ( [ dim - b - eqn ] ) , when @xmath46 and @xmath47 in the absence of linear dispersion . peakons move at a speed equal to their maximum height , at which they have a sharp peak ( jump in derivative ) . the single peakon solution is @xmath48 unlike the kdv soliton , the peakon speed is independent of its width ( @xmath49 ) . periodic peakon solutions of ch were treated in alber _ there , the sharp peaks of periodic peakons were associated with billiards reflecting at the boundary of an elliptical domain . these billiard solutions for the periodic peakons arise from geodesic motion on a tri - axial ellipsoid , in the limit that one of its axes shrinks to zero length . the ch equation was found after its derivation as a shallow water equation in @xcite to fit into a class of integrable equations derived previously by using hereditary symmetries in fokas and fuchssteiner @xcite . see fuchssteiner @xcite for an insightful history of how the shallow water equation ( [ dim - b - eqn ] ) in the integrable case with @xmath41 relates to the mathematical theory of hereditary symmetries . equation ( [ dim - b - eqn ] ) with @xmath41 was recently re - derived as a shallow water equation by using asymptotic methods in three different approaches in dullin _ @xcite , in fokas and liu @xcite and also in johnson @xcite . these three derivations all used different variants of the method of asymptotic expansions for shallow water waves in the absence of surface tension . only the derivation in dullin _ @xcite took advantage of the parametric freedom in the kodama normal - form transformations of the asymptotic expansion results at quadratic order . the effects of the parameter @xmath34 on the solutions of equation ( [ dim - b - eqn ] ) were investigated in holm and staley @xcite , where @xmath34 was treated as a bifurcation parameter , in the limiting case when the linear dispersion coefficients are set to @xmath46 and @xmath47 . this limiting case allows a variety of special solutions for different ranges of the values of @xmath34 , in which the two nonlinear terms in equation ( [ dim - b - eqn ] ) balance each other in the _ absence _ of linear dispersion . since its first appearance in @xcite , the ch equation has been the centre of a confluence of scientific endeavors that includes water waves , integrable systems , pde analysis , asymptotics , geometry and lie groups . this confluence has led to a continued interest and opportunities for contributions from many different fields in mathematics . the interest in ch solutions may be measured by noticing that the entry ` peakon ' has acquired many thousands of _ google hits_. this paper aims to review some of the geometric highlights of recent work on the ch equation . it is certainly not exhaustive . it mainly focuses on comparing the soliton theory for smooth ch solutions with the peakon theory for its singular solutions in the dispersionless limit . section [ ist - sec ] briefly explains the application of the method of the inverse scattering transform ( ist ) for obtaining the soliton solutions of ch equation . the set of scattering data is introduced and the formulation of the inverse scattering as a riemann - hilbert problem is outlined . the solution is expressed via the scattering data in a form that admits the peakon limit in the sections that follow . in section [ mommap - sec ] the map between the action - angle variables ( expressed via the scattering data ) and the momentum of the ch solution is formulated as a momentum map from the symplectic action - angle variables to the dual of the lie algebra of smooth vector fields on the real line . this is a poisson map , but the noncanonical bi - hamiltonian structure of the ch equation which led to the discovery of its isospectral problem in @xcite and its geometrical significance as geodesic flow on the diffeomorphism group are not discussed here . in section [ peakon - sec ] we introduce the peakons as singular solutions that appear in dispersionless limit . the @xmath2-peakon solution is governed by a finite - dimensional integrable dynamical system . section [ peakon - limit - sec ] presents the multi - peakon solution as a limiting case of the ch multi - soliton solution and points out the similarity between the dynamics of the peakon system and the well known toda lattice . in section [ superint - sec ] we comment briefly on the existence of additional integrals of motion of the peakon system , a property known as _ superintegrability_. the two - peakon system is analysed explicitly . section [ other - sing - sol - sec ] presents the compacton and pulson solutions , which are singular solutions , similar to peakons that can also be represented by the singular momentum map for peakons discussed in @xcite . in section [ ep - higher - sec ] we deal with generalizations of the ch equation in higher dimensions . these also allow for singular solutions . although such multidimensional generalizations are unlikely to be integrable , numerical studies show that their solutions are stable and interact elastically in collisions . in section [ conclus - sec ] we comment briefly on some limitations of our present discussion and point out three open problems for further research . one of the most significant results in the theory of nonlinear partial differential equations was the development of gardner , greene , kruskal and miura @xcite of a method for the exact solution of the initial - value problem for the kdv equation . prior to their work the only known exact solutions of kdv were the travelling wave solutions . the idea is based on a representation of the equation ( @xmath50 ) as a compatibility condition of two eigenvalue problems with a time - independent spectral parameter @xmath51 : [ kdv - isospec ] _ xx&+ & u= + _ t&=&(u_x+)-(4 + 2u)_x where @xmath52 is an arbitrary constant . the method is conceptually analogous in many ways to the fourier transform method for solving linear equations . this method is now known as inverse scattering transform ( ist ) . it was recast by lax @xcite in a general framework that allows the ist to be used for solving other nonlinear pdes . an important consequence is the relationship between the discrete eigenvalues and the characteristics of the solitons that emerge asymptotically . another feature of soliton collisions is the preservation of soliton identities after the interaction ( asymptotically in time ) . for example , collisions of kdv solitons only result in a phase shift from the positions where they would have been without the interaction . the kdv was formulated as a completely integrable hamiltonian system in a work by faddeev and zakharov @xcite , the hamiltonian form was also noted by gardner @xcite . reviews of ist may be found , for example , in ablowitz et al . @xcite , dubrovin _ et al . _ @xcite , novikov _ et al . _ @xcite . for discussions of other related bi - hamiltonian equations , see @xcite . in this section we outline the application of the ist for the ch equation . we use the form with a linear dispersion term @xmath53 where @xmath54 is a real constant . the equation in this form also appears as a model of unidirectional propagation of shallow water waves over a flat bottom @xcite as well as that of axially symmetric waves in a hyperelastic rod @xcite . it can be obtained from the @xmath34-equation ( [ dim - b - eqn ] ) with @xmath41 via galilean transformation that removes the @xmath55 term . equation ( [ eq1 ] ) admits a lax pair formulation @xcite [ eq3 ] _ xx&=&(1-(m+ ) ) + [ eq4 ] _ t&=&-(+u)_x++ where @xmath52 is an arbitrary constant . we will use this freedom for a proper normalization of the eigenfunctions . in our further considerations @xmath56 will be a schwartz class function , @xmath57 and @xmath58 . then @xmath59 for all @xmath60 @xcite . let us introduce a new spectral parameter @xmath61 , such that [ lambda ] ( k)= ( 1 + 4k^2 ) . the spectrum of the problem ( [ eq3 ] ) under these conditions is described in @xcite . the continuous spectrum in terms of @xmath61 corresponds to @xmath61 real . the discrete spectrum ( in the upper half plane ) consists of finitely many points @xmath62 , @xmath63 where @xmath64 is real and @xmath65 . for all real @xmath66 a basis in the space of solutions of ( [ eq3 ] ) can be introduced , @xmath67 and @xmath68 , fixed by its asymptotic when @xmath69 @xcite : [ eq5 ] ( x , k)&=&e^-ikx+o(1 ) , x another basis can be introduced , @xmath70 and @xmath71 fixed by its asymptotic when @xmath72 : [ eq5a ] ( x , k)&=&e^-ikx+o(1 ) , x - and the relation between the two bases is @xcite [ eq8 ] ( x , k)=a(k)(x , k)+b(k)|(x , k ) . where [ eq10 ] |a(k)|^2-|b(k)|^2=1 . the quantity @xmath73 is called the reflection coefficient , without any danger of confusing @xmath74 with the bifurcation parameter @xmath34 introduced in ( [ dim - b - eqn ] ) . all of the required information about the scattering , i.e. @xmath75 and @xmath74 , is provided by @xmath76 for @xmath77 only @xcite . it is sufficient to know @xmath76 only on the half line @xmath77 , since @xmath78 , @xmath79 and thus @xmath80 . the constant @xmath52 in ( [ eq4 ] ) can be chosen for each eigenfunction in such a way that @xmath75 does not depend on @xmath60 and is a generating function of the integrals of motion @xcite . at the points of the discrete spectrum , @xmath75 has simple zeroes @xcite , therefore @xmath81 and @xmath82 are linearly dependent ( [ eq8 ] ) : @xmath83 in other words , the discrete spectrum is simple , there is only one ( real ) eigenfunction @xmath84 , corresponding to each eigenvalue @xmath85 , and we can take this eigenfunction to be [ eq201]^(n)(x)(x , i_n ) the asymptotic behaviour of @xmath86 , according to ( [ eq5a ] ) , ( [ eq200 ] ) is [ eq203 ] ^(n)(x)&=&e^_n x+o(e^_n x ) , x- ; + [ eq204 ] ^(n)(x)&= & b_n e^-_n x+o(e^-_n x ) , x. the sign of @xmath87 obviously depends on the number of the zeroes of @xmath86 . suppose that [ eqn ] 0<_1<_2< <_n<1/2.then from the oscillation theorem for the sturm - liouville problem @xcite , @xmath86 has exactly @xmath88 zeroes . therefore [ eq205 ] b_n= ( -1)^n-1|b_n| . the set [ eq206 ] \ { ( k)(k>0),_n , r_n = b_n / ia(i_n),n=1, n } is called the _ scattering data_. the hamiltonians for the ch equation in terms of the scattering data are presented in @xcite . the time evolution of the scattering data is @xcite : [ eq19 ] ( k , t)&=&(k,0)e^-t . + [ eq207 ] r_n(t)&=&r_n(0)(t ) , where @xmath89 is always a positive quantity @xcite and @xmath90 . the scattering coefficient @xmath75 is analytic for @xmath91 with asymptotic @xcite @xmath92 where @xmath93 is a constant ( integral of motion ) : @xmath94 the asymptotics of the jost solutions ( or , rather the following quantities , depending on the jost solutions ) for @xmath95 have the form @xcite [ eq20 ] ( x , k ) & & ( x , k)e^=x_0(x)+++ , + ( x , k ) & & ( x , k)e^ik ( + ) = x_0(x)+++ [eq21 ] where @xmath96^{1/4}>0 $ ] , [ eq22 ] y(x)=. moreover , ( [ eq20 ] ) is analytic for @xmath97 , ( [ eq21 ] ) is analytic for @xmath91 @xcite . from ( [ eq8 ] ) with ( [ eq20 ] ) and ( [ eq21 ] ) we obtain [ eq27 ] = ( x , k)+ ( k)|(x , k)e^2iky(x)/. the function @xmath98 is analytic for @xmath91 , @xmath99 is analytic for @xmath100 . thus , ( [ eq27 ] ) represents an additive * _ _ * riemann - hilbert problem ( rhp ) with a jump on the real line , given by @xmath101 . the solution of this rhp follows a standard pattern , see @xcite for more details . the solution simplifies considerably in the case @xmath102 . this is the @xmath2-soliton solution : [ solk- ] ( x , k)&=&x_0(x)-i_n=1 ^n , k<0 , + & = & x_0(x)-i_n=1 ^n , k>0 . [ solk+ ] from ( [ solk- ] ) one has a linear system for the quantities @xmath103 with a solution [ eq43a ] ( x ,- i_n , t)=x_0(x)_p=1^n a^-1_np[y , t ] , n=1, ,n . where [ eq43 ] a_pn[y , t]_pn+ . taking @xmath104 in ( [ solk- ] ) produces e^-(x- ) & & ( x ,- i/2)=x_0(x)(1-_n , p=1 ^n a^-1_np[y , t ] ) . + [ -i/2 ] the substitution @xmath105 in ( [ solk+ ] ) with @xmath106 gives e^(x- ) & & = x_0(x)(1-_n , p=1 ^n a^-1_np[y , t ] ) . + [ i/2 ] from ( [ -i/2 ] ) and ( [ i/2 ] ) there follows a parametric representation x&=&x(y , t)+ , [ x = f / f ] + f _ & & 1-_n , p=1 ^n a^-1_np[y , t ] . [ f pm ] from ( [ eq22 ] ) and ( [ x = f / f ] ) one can compute the solution in parametric form [ eq u ] u(x(y , t),t)=x_t(y , t),x = x(y , t ) , [ sol ] where @xmath107 is given in terms of the scattering data in ( [ x = f / f ] ) , ( [ f pm ] ) . upon introducing new notations j&=&2_j ( -++x_j0 ) + x_j0&= & [ x_0j ] + _ j & = & [ phi_j ] + _ ij&= & ( ) ^2 [ gamma ij ] one can rewrite the expression for @xmath108 ( [ f pm ] ) in the form @xcite f _ & & _ = 0,1 . the solution for @xmath56 can be obtained from ( [ sol ] ) . first we notice that @xmath109 and also @xmath110 thus @xmath111 . @xmath112 and @xmath113 m(x , t)&=&_-^p(y , t)(x - x(y , t ) ) y [ m(x , t ) ] + & & p(y , t)=x_t x_y- ( ) _ y [ p(y , t ) ] , + u(x , t)&=&_-^p(y , t)(-|x - x(y , t)| ) y.[u(x , t ) ] as we shall see in section [ ep - higher - sec ] , this representation is also useful in the study of multidimensional solutions . the spectral problem ( [ eq3 ] ) is gauge equivalent to a standard sturm - liouville problem , well known from the kdv hierarchy , cf . ( [ kdv - isospec ] ) , with short notation @xmath114 : -_yy+u(y)&= & , = - , + ( y)&=&q^1/4 , = , [ chvar ] + u(y)&=&+-- . [ ep ] note that ( [ chvar ] ) leads to two possible expressions for the change of the variables in the liouville transformation : y&= & x+_-^x ( -)dx+ , [ y1 ] + y&= & x+_^x ( -)dx+ . [ y2 ] these two possibilities , ( [ y1 ] ) , ( [ y2 ] ) are only consistent iff _ -^ ( -)dx= , which is always the case , since the integral under question is ( up to a multiplier ) the casimir function @xmath93 ( [ eqi8 ] ) . the matching of the ch hierarchy to kdv hierarchy requires solving the ermakov - pinney equation ( [ ep ] ) @xcite , which is not straightforward and leads to the same solution ( [ sol ] ) in parametric form . the canonical poisson brackets for the scattering data of the ch equation are computed in @xcite where also the action - angle variables are expressed in terms of the scattering data . let us consider the action variable for the n - soliton solution . ( these considerations can be extended easily to the variables of the continuous spectrum . ) the angle variable is @xmath115 ; it is linear in time @xmath60 and @xmath116 let us introduce @xmath117 into hamilton s principle @xmath118 , with @xmath119 = \int\bigg ( \ell[u]+\sum_{n=1}^{n}\pi_n\big(\dot{\phi}_n -\lambda_n [ u ] \big ) \bigg ) dt.\ ] ] here the lagrange multiplier @xmath120 enforces the action - angle relation for the ch scattering data as a @xmath121 shift of the angles @xmath122 at constant angular frequencies @xmath123 , with @xmath124 . the stationary variation is @xmath125 \big)\delta \pi_n - \sum_{n=1}^{n}\dot{\pi}_n \delta \phi_n \right ) dt.\ ] ] since by definition @xmath126 is the momentum , @xmath118 implies the eulerian representation [ mommap1 ] m==_n=1^n_n , + _ n = _ n [ u ] _ n=0 . [ toralnaction ] relation ( [ mommap1 ] ) is the _ momentum map _ @xmath127 for the toral @xmath121 action ( [ toralnaction ] ) on the angles @xmath122 at constant angular frequencies @xmath123 . this momentum map from the action - angle scattering variables @xmath0 to the flow momentum @xmath128 ( dual to the smooth vector fields @xmath129 on the real line ) provides the eulerian representation of the @xmath2-soliton solution of ch in terms of the scattering data and squared eigenfunctions of its isospectral eigenvalue problem . momentum maps for hamiltonian dynamics are reviewed in @xcite , for example . by using the spectral quantities of the @xmath2-soliton solution ( recall : @xmath130 ) one may express the variation of the spectrum with respect to the ch solution in terms of the squared - eigenfunctions of the isospectral problem as @xcite @xmath131 ^ 2,\ ] ] in which @xmath132 is the eigenfunction that belongs to eigenvalue @xmath133 , see ( [ eq200 ] ) . on the other hand , the expansion of @xmath134 over squares of eigenfunctions is given by @xcite @xmath135 ^ 2 .\ ] ] consequently , @xmath136 ^ 2,\ ] ] or @xmath137 where @xmath120 and @xmath138 denote explicitly _ n&= & = , + j_n ( x , t ) & & + & = & r_n(t)(1-^2)[|(x ,- i_n , t)]^2 . thus , the momentum map ( [ mommap1 ] ) from the action - angle variables under going dynamics ( [ toralnaction ] ) to the eulerian representation of the momentum for the ch solution is expressed in terms of the scattering data and squared eigenfunctions of its @xmath2-soliton isospectral eigenvalue problem . perhaps not unexpectedly , this momentum map may be applied to the action - angle representation of the solution of any integrable hamiltonian pde . camassa and holm @xcite discovered the `` peakon '' solitary traveling wave solution for a shallow water wave , @xmath139 whose fluid velocity @xmath32 is a function of position @xmath140 on the real line and time @xmath60 . the peakon traveling wave moves at a speed equal to its maximum height , at which it has a sharp peak ( jump in derivative ) . peakons are an emergent phenomenon , solving the initial value problem for a partial differential equation derived by an asymptotic expansion of euler s equations using the small parameters of shallow water dynamics . peakons are _ nonanalytic _ solitons , which superpose as @xmath141 for sets @xmath142 and @xmath143 satisfying canonical hamiltonian dynamics . peakons arise for shallow water waves in the limit of zero linear dispersion in one dimension . peakons satisfy a partial differential equation ( pde ) arising from hamilton s principle for geodesic motion on the smooth invertible maps ( diffeomorphisms ) with respect to the @xmath144 sobolev norm of the fluid velocity . peakons generalize to higher dimensions , as well . we explain how peakons were derived in the context of shallow water asymptotics and describe some of their remarkable mathematical properties . peakons were first found as singular soliton solutions of the completely integrable ch equation . this is equation ( [ dim - b - eqn ] ) with @xmath41 , now rewritten in terms of the velocity , as @xmath145 peakons were found in @xcite to arise in the absence of linear dispersion . that is , they arise when @xmath146 and @xmath147 in ch ( [ ch - u - eqn ] ) . specifically , peakons are the individual terms in the peaked @xmath148soliton solution of ch ( [ ch - u - eqn ] ) for its velocity , in the absence of linear dispersion . each term in the sum ( [ peakontrain - soln ] ) is a solition with a sharp peak at its maximum . hence , the name `` peakon . '' expressed using its momentum , @xmath149 , the peakon velocity solution ( [ peakontrain - soln ] ) of dispersionless ch becomes a sum over a delta functions , supported on a set of points moving on the real line . namely , the peakon velocity solution ( [ peakontrain - soln ] ) implies @xmath150 because of the relation @xmath151 . these solutions satisfy the @xmath34-equation ( [ dim - b - eqn ] ) for any value of @xmath34 , provided @xmath46 and @xmath47 . as we shall discuss later , the peakon momentum relation ( [ chpeakon - m - soln ] ) is again a momentum map . thus , peakons are _ singular momentum solutions _ of the dispersionless @xmath34-equation , although they are not stable for every value of @xmath34 . from numerical simulations @xcite , peakons are conjectured to be stable for @xmath152 . in the integrable cases @xmath44 for ch and @xmath45 for dp , peakons are stable singular _ soliton _ solutions . the spatial velocity profile @xmath153 of each separate peakon in ( [ peakontrain - soln ] ) is the green s function for the helmholtz operator on the real line , with vanishing boundary conditions at spatial infinity . unlike the kdv soliton , whose speed and width are related , the width of the peakon profile is set by its green s function , independently of its speed . substituting the peakon solution ansatz ( [ peakontrain - soln ] ) and ( [ chpeakon - m - soln ] ) into the dispersionless ch equation , @xmath154 yields _ hamilton s canonical equations _ for the dynamics of the discrete set of peakon parameters @xmath155 and @xmath156 , @xmath157 for @xmath158 , with hamiltonian given by @xcite , @xmath159 or explicitly , _ a&=&_b=1^np_b e^-|q_a - q_b|/ [ q_a ] + _ a&=&_b=1^n p_b e^-|q_a - q_b|/(q_a - q_b ) [ p_a ] thus , one finds that the points @xmath160 in the peakon solution ( [ peakontrain - soln ] ) move with the flow of the fluid velocity @xmath32 at those points , since @xmath161 . this means the @xmath156 are _ lagrangian _ coordinates . moreover , the singular momentum solution ( [ chpeakon - m - soln ] ) is the lagrange - to - euler map for an invariant manifold of the dispersionless ch equation ( [ dchmom - eqn ] ) . on this finite - dimensional invariant manifold for the partial differential equation ( [ dchmom - eqn ] ) , the dynamics is canonically hamiltonian . with hamiltonian ( [ h - peakon - ansatz ] ) , the canonical equations ( [ ham - peakon - eqn ] ) for the @xmath162 canonically conjugate peakon parameters @xmath155 and @xmath156 were interpreted in @xcite as describing _ geodesic motion _ on the @xmath148dimensional riemannian manifold whose co - metric is @xmath163 . moreover , the canonical geodesic equations arising from hamiltonian ( [ h - peakon - ansatz ] ) comprise an integrable system for any number of peakons @xmath2 . this integrable system was studied in @xcite for solutions on the real line , and in @xcite and references therein , for spatially periodic solutions . the integrals generated by the action variables in terms of the coordinates can be recovered as tr@xmath164 , where @xmath165 is the lax operator for the peakon system @xcite . being a completely integrable hamiltonian soliton equation , the continuum ch equation ( [ ch - u - eqn ] ) has an associated isospectral eigenvalue problem , discovered in @xcite for any values of its dispersion parameters @xmath166 and @xmath167 . remarkably , when @xmath46 and @xmath47 , this isospectral eigenvalue problem has a purely _ discrete _ spectrum . moreover , in this case , each discrete eigenvalue corresponds precisely to the time - asymptotic velocity of a peakon . this discreteness of the ch isospectrum in the absence of linear dispersion implies that _ only _ the singular peakon solutions ( [ chpeakon - m - soln ] ) emerge asymptotically in time , in the solution of the initial value problem for the dispersionless ch equation ( [ dchmom - eqn ] ) . this is borne out in numerical simulations of the dispersionless ch equation ( [ dchmom - eqn ] ) , starting from a smooth initial distribution of velocity @xcite . figure [ peakon_figure ] shows the emergence of peakons from an initially gaussian velocity distribution and their subsequent elastic collisions in a periodic one - dimensional domain . this figure demonstrates that singular solutions dominate the initial value problem and , thus , that it is imperative to go beyond smooth solutions for the ch equation ; the situation is similar for the epdiff equation . [ [ peakons - as - mechanical - systems . ] ] peakons as mechanical systems . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + governed by canonical hamiltonian equations , each @xmath148peakon solution can be associated with a mechanical system of moving particles . calogero et al . @xcite further extended the class of mechanical systems of this type . the r - matrix approach was applied to the lax pair formulation of the @xmath148peakon system for ch by ragnisco and bruschi @xcite , who also pointed out the connection of this system with the classical toda lattice . a discrete version of the adler - kostant - symes factorization method was used by suris @xcite to study a discretization of the peakon lattice , realized as a discrete integrable system on a certain poisson submanifold of @xmath168 equipped with an r - matrix poisson bracket . @xcite used the stieltjes theorem on continued fractions and the classical moment problem for studying multi - peakon solutions of the ch equation . generalized peakon systems are described for any simple lie algebra by alber _ the limit @xmath169 in the @xmath2-soliton solution @xmath134 produces the @xmath2-peakon solution ( [ peakontrain - soln ] ) . the limiting procedure is described in detail in @xcite . due to ( [ lambda ] ) one can write for the discrete eigenvalues @xmath170 2_j=(1-_j)^1/2= 1-_j+ . [ kappa - limit ] the solution ( [ x = f / f ] ) depends explicitly on @xmath171 ( [ kappa - limit ] ) and the limit can be computed with ( [ kappa - limit ] ) by taking @xmath169 and keeping the eigenvalue @xmath172 constant . the result is the expression ( [ peakontrain - soln ] ) with p_i&=&4d_n - i+1^(0)d_n - i^(2)d_n - i+1^(1)d_n - i^(1),(i=1 , 2 , ... , n),[m_i ] + q_i&=&,(i=1 , 2 , ... , n ) , [ x_i ] where d_n^(m)=_1i_1<i_2< ... <i_nn_n(i_1,i_2 , ... , i_n)(_i_1_i_2 ... _i_n)^m r_i_1r_i_2 ... r_i_n , + n=1 , 2 , ... , n , _ n(i_1,i_2 , ... , i_n)&=&_1l < mn(_i_l-_i_m)^2,(n2 ) , + r_i(t)&=&r_i(0)e^t,(x_i0=r_i(0 ) ) . the quantities @xmath173 are called * _ _ * hankel determinants . by definition @xmath174 . in general , hankel determinant is a determinant of @xmath175 matrix of the form @xmath176 where @xmath177 are elements of a sequence , i.e. @xmath178 in this particular case a_l=_i=1^n_i^l r_i(t ) . [ [ similarity - of - peakon - lattice - and - toda - lattice ] ] similarity of peakon lattice and toda lattice + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + hankel determinants appear in the solutions of other integrable systems , e.g. toda lattice , e.g. see @xcite . the toda equation @xcite & = & e^q_n+1-q_n - e^q_n - q_n-1 , + & = & p_n , n is one of the most important integrable systems . we have a finite chain with @xmath2 nodes under the fixed ends boundary conditions @xmath179 . the solution is in the form q_n(t)&=&q_1(0)+ , + q_1(0)&=&-d^(0)_n(0 ) .the hankel determinants are obtained by a similar sequence , a_l&=&_i=1^n_i^l r_i(t ) , + r_i(t)&=&r_i(0)e^-_it , where @xmath180 are @xmath2 different constants ( eigenvalues of the lax matrix ) and the quantities @xmath181 , @xmath182 represent another set @xmath2 of constants . due to its simple form , the @xmath2-peakon solution can be used as an approximation of the @xmath2-soliton ch solution when the dispersion term is small ( and the term @xmath183 can be neglected ) . similarly , the toda chain with complex dynamical variables ( the so - called complex toda chain ctc ) provides an approximation for the n - soliton solution of the nonlinear schredinger equation ( nls ) @xmath184 see @xcite for more details . such an approximation is called adiabatic approximation and means that the @xmath2-soliton solution consists of @xmath185 well separated solitons @xmath186}}{\cosh ( 2\nu_k(x-\xi_k(t))},\ ] ] i.e. the overlap between the solitons is small . the variables @xmath187 of the ctc are related to the nls solitons parameters by @xmath188 where @xmath189 , @xmath190 , @xmath191 and @xmath192 characterize the center - of - mass position , the phase , velocity and amplitude respectively of the @xmath193-th soliton in the chain ; @xmath194 and @xmath195 are the average amplitude and velocity of the soliton train . @xmath189 and @xmath190 can be obtained as the real and imaginary parts of @xmath196 . such soliton trains and their asymptotic behavior appear to be important for the needs of soliton based fiber optics communications . suppose we have an integrable system with @xmath162- dimensional phase space which in terms of the action - angle ( canonical ) variables can be represented as _ n=0,_n=_n , n=1,2 , , n , or , if there exists a bracket such that \{_n , _ l } = _ nl , \{_n , _ l}=\{_n , _ l}=0,[brackets]the system is hamiltonian with a hamiltonian h_n=_1 ^ 2+ _n^2 . [ phamilt]the integrals @xmath123 , @xmath197 are clearly in involution , which guarantees the integrability of the system . there is however another set of integrals : i_j=(_j-_j+1)(_1+ +_n)-(_j-_j+1 ) ( _ 1+ _n ) + j=1,2, ,n-1.[i_j ] if the action variables @xmath123 are all different , the set ( [ i_j ] ) is functionally independent from the set @xmath123 , @xmath197 . in addition , the integrals ( [ i_j ] ) form another set of @xmath2-integrals in involution together with @xmath198 . due to the existence of two sets of functionally independent integrals in involution such systems are termed superintegrable . an example of such system is toda lattice , see the discussion in @xcite . the peakon system is also superintegrable . the canonical variables in terms of the scattering data for ch equation can be used in the peakon limit : @xmath199 , @xmath200 , the hamiltonian ( [ phamilt ] ) is also a peakon limit ( @xmath169 ) of the @xmath2-soliton hamiltonian @xcite h_n ( ) = ^2_n=1^n(+ ) .the poisson bracket for the ch peakon solution is @xmath201 and the scattering data satisfy ( [ brackets ] ) with respect to ( [ pb ] ) , see @xcite for the details . more interesting are the integrals ( [ i_j ] ) . from ( [ x_i ] ) and ( [ m_i ] ) one can recover these integrals in coordinate form . for example , when @xmath202 we have i_1 & = & + ( + ) , + & & j=(p_1-p_2)^2 + 4p_1p_2e^-|q_1-q_2|/ . note that @xmath203 depends on both combinations @xmath204 and @xmath205 as well as the momentum variables . the hamiltonian @xmath206 ( which depends only on @xmath205 and the momentum variables ) and @xmath203 form a complete system of integrals in involution . the integration of the 2-peakon system with these integrals can be performed as follows . first , one can express @xmath207 and @xmath208 in terms of @xmath203 , @xmath206 and the momentum variables : @xmath209 . next , _ i=+.[alg]the substitution of @xmath210 , @xmath211 from ( [ q_a ] ) ( [ p_a ] ) to ( [ alg ] ) produces an algebraic equation that gives , say @xmath212 as a function of @xmath213 . then ( [ p_a ] ) is an ode for @xmath213 of the form @xmath214 . clearly , from practical viewpoint it is much more convenient to work with the other system of integrals in involution : @xmath206 and the conserved momentum @xmath215 . note that @xmath216 is itself an integral . [ [ pulsons - generalizing - the - peakon - solutions - of - the - dispersionless - b - equation - for - other - greens - functions . ] ] pulsons : generalizing the peakon solutions of the dispersionless @xmath34-equation for other green s functions . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the hamiltonian @xmath217 in equation ( [ h - peakon - ansatz ] ) depends on the green s function for the relation between velocity @xmath32 and momentum @xmath56 . however , the singular momentum solution ansatz ( [ chpeakon - m - soln ] ) is _ independent _ of this green s function . thus , as discovered in fringer and holm @xcite , _ the singular momentum solution ansatz ( [ chpeakon - m - soln ] ) for the dispersionless equation , @xmath218 provides an invariant manifold on which canonical hamiltonian dynamics occurs , for _ any choice _ of the green s function @xmath14 relating velocity @xmath32 and momentum @xmath56 by the convolution @xmath219 . _ the fluid velocity solutions corresponding to the singular momentum ansatz ( [ chpeakon - m - soln ] ) for equation ( [ fhmom - eqn ] ) are the _ pulsons_. pulsons are given by the sum over @xmath2 velocity profiles determined by the green s function @xmath14 , as @xmath220 again for ( [ fhmom - eqn ] ) , the singular momentum ansatz ( [ chpeakon - m - soln ] ) results in a finite - dimensional invariant manifold of solutions , whose dynamics is canonically hamiltonian . the hamiltonian for the canonical dynamics of the @xmath162 parameters @xmath155 and @xmath156 in the `` pulson '' solutions ( [ pulson - u - soln ] ) of equation ( [ fhmom - eqn ] ) is @xmath221 again for the pulsons , the canonical equations for the invariant manifold of singular momentum solutions provide a phase - space description of geodesic motion , this time with respect to the co - metric given by the green s function @xmath14 . mathematical analysis and numerical results for the dynamics of these pulson solutions are given in @xcite . these results describe how the collisions of pulsons ( [ pulson - u - soln ] ) depend upon their shape . [ [ compactons - in - the-1alpha2to0-limit - of - ch . ] ] compactons in the @xmath222 limit of ch . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + as mentioned earlier , in the limit that @xmath39 , the ch equation ( [ ch - u - eqn ] ) becomes the kdv equation . in the opposite limit that @xmath222 ch becomes the hunter - zheng equation @xcite @xmath223 this equation has `` compacton '' solutions , whose collision dynamics was studied numerically and put into the present context in @xcite . the corresponding green s function satisfies @xmath224 , so it has the triangular shape , @xmath225 for @xmath226 , and vanishes otherwise , for @xmath227 . that is , the green s function in this case has compact support ; hence , the name `` compactons '' for these pulson solutions , which as a limit of the integrable ch equations are true solitons , solvable by ist . [ [ pulson - solutions - of - the - dispersionless - b - equation . ] ] pulson solutions of the dispersionless @xmath34-equation . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + holm and staley @xcite give the pulson solutions of the traveling wave problem and their elastic collision properties for the dispersionless @xmath34-equation , @xmath228 with any ( symmetric ) green s function @xmath14 and for any value of the parameter @xmath34 . numerically , pulsons and peakons are both found to be stable for @xmath152 , @xcite . the reduction to _ noncanonical _ hamiltonian dynamics for the invariant manifold of singular momentum solutions ( [ chpeakon - m - soln ] ) of the other integrable case @xmath42 with peakon green s function @xmath229 is found in @xcite . [ [ generalizing - the - peakon - solutions - of - the - ch - equation - to - higher - dimensions . ] ] generalizing the peakon solutions of the ch equation to higher dimensions . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + in @xcite , weakly nonlinear analysis and the assumption of columnar motion in the variational principle for euler s equations were found to produce the two - dimensional generalization of the dispersionless ch equation ( [ dchmom - eqn ] ) . this generalization is the euler - poincar ( ep ) equation @xcite for the lagrangian consisting of the kinetic energy , @xmath230dxdy \ , , \label{ke - def}\ ] ] in which the fluid velocity @xmath231 is a two - dimensional vector . evolution generated by kinetic energy in hamilton s principle results in geodesic motion , with respect to the velocity norm @xmath232 , which is provided by the kinetic energy lagrangian . for ideal incompressible fluids governed by euler s equations , the importance of geodesic flow was recognized by arnold @xcite for the @xmath233 norm of the fluid velocity . the ep equation generated by any choice of kinetic energy norm without imposing incompressibility is called `` epdiff , '' for `` euler - poincar equation for geodesic motion on the diffeomorphisms . '' epdiff is given by @xcite @xmath234 with momentum density @xmath235 where @xmath236 is given by the kinetic energy , which defines a norm in the fluid velocity @xmath232 , yet to be determined . by design , this equation has no contribution from either potential energy , or pressure . it conserves the velocity norm @xmath232 given by the kinetic energy . its evolution describes geodesic motion on the diffeomorphisms with respect to this norm @xcite . an alternative way of writing the epdiff equation ( [ epdiff - eqn ] ) in either two , or three dimensions is , @xmath237 this form of epdiff involves all three differential operators , curl , gradient and divergence . for the kinetic energy lagrangian @xmath238 given in ( [ ke - def ] ) , which is a norm for _ irrotational _ flow ( with @xmath239 ) , we have the epdiff equation ( [ epdiff - eqn ] ) with momentum @xmath240 . epdiff ( [ epdiff - eqn ] ) may also be written intrinsically as @xmath241 where ad@xmath242 is the @xmath233 dual of the ad - operation ( commutator ) for vector fields . see @xcite for additional discussions of the beautiful geometry underlying this equation . [ [ reduction - to - the - dispersionless - ch - equation - in-1d . ] ] reduction to the dispersionless ch equation in 1d . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + in one dimension , the epdiff equation ( [ epdiff - eqn]-[epdiffeo - eqn ] ) with lagrangian @xmath238 given in ( [ ke - def ] ) simplifies to the dispersionless ch equation ( [ dchmom - eqn ] ) . the dispersionless limit of the ch equation appears , because we have ignored potential energy and pressure . [ [ strengthening - the - kinetic - energy - norm - to - allow - for - circulation . ] ] strengthening the kinetic energy norm to allow for circulation . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the kinetic energy lagrangian ( [ ke - def ] ) is a norm for irrotational flow , with @xmath239 . however , inclusion of rotational flow requires the kinetic energy norm to be strengthened to the @xmath243 norm of the velocity , defined as @xmath244dxdy \nonumber\\ & = & \frac{1}{2 } \int \bigg[|{\mathbf{u}}|^2 + \alpha^2|\nabla{\mathbf{u}}|^2 \bigg]dxdy = \frac{1}{2 } \|{\mathbf{u}}\|^2_{h_\alpha^1 } \ , . \label{h1-eqn}\end{aligned}\ ] ] here we assume boundary conditions that give no contributions upon integrating by parts . the corresponding epdiff equation is ( [ epdiff - eqn ] ) with @xmath245 this expression involves inversion of the familiar _ helmholtz _ operator in the ( nonlocal ) relation between fluid velocity and momentum density . the @xmath243 norm @xmath246 for the kinetic energy ( [ h1-eqn ] ) also arises in three dimensions for turbulence modeling based on lagrangian averaging and using taylor s hypothesis that the turbulent fluctuations are `` frozen '' into the lagrangian mean flow @xcite . [ [ generalizing - the - ch - peakon - solutions - to - n - dimensions . ] ] generalizing the ch peakon solutions to @xmath247 dimensions . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + building on the peakon solutions ( [ peakontrain - soln ] ) for the ch equation and the pulsons ( [ pulson - u - soln ] ) for its generalization to other traveling - wave shapes in @xcite , holm and staley @xcite introduced the following measure - valued singular momentum solution ansatz for the @xmath248dimensional solutions of the epdiff equation ( [ epdiff - eqn ] ) : with @xmath249 . in contrast , the similar expression ( [ m(x , t ) ] ) for the soliton solutions represent smooth functions @xmath250 . ] @xmath251 these singular momentum solutions , called `` diffeons , '' are vector density functions supported in @xmath252 on a set of @xmath2 surfaces ( or curves ) of codimension @xmath253 for @xmath254 with @xmath249 . they may , for example , be supported on sets of points ( vector peakons , @xmath255 ) , one - dimensional filaments ( strings , @xmath256 ) , or two - dimensional surfaces ( sheets , @xmath257 ) in three dimensions . figure [ 2dstrip_plate ] shows the results for the epdiff equation when a straight peakon segment of finite length is created initially moving rightward ( east ) . because of propagation along the segment in adjusting to the condition of zero speed at its ends and finite speed in its interior , the initially straight segment expands outward as it propagates and curves into a peakon `` bubble . '' figure [ 2dstrip_plate - multi ] shows an initially straight segment whose velocity distribution is exponential in the transverse direction , but is wider than @xmath49 for the peakon solution . this initial velocity distribution evolves under epdiff to separate into a train of curved peakon `` bubbles , '' each of width @xmath49 . this example illustrates the emergent property of the peakon solutions in two dimensions . this phenomenon is observed in nature , for example , as trains of internal wave fronts in the south china sea @xcite . for the peakon solution will break up into a train of curved peakon `` bubbles , '' each of width @xmath49 . this example illustrates the emergent property of the peakon solutions in two dimensions . ] substitution of the singular momentum solution ansatz ( [ m - ansatz ] ) into the epdiff equation ( [ epdiff - eqn ] ) implies the following integro - partial - differential equations ( ipdes ) for the evolution of the parameters @xmath258 and @xmath259 , @xmath260 importantly for the interpretation of these solutions , the coordinates @xmath254 turn out to be _ lagrangian _ coordinates . the velocity field corresponding to the momentum solution ansatz ( [ m - ansatz ] ) is given by @xmath261 for @xmath262 . when evaluated along the curve @xmath263 , this velocity satisfies , @xmath264 consequently , the lower - dimensional support sets defined on @xmath263 and parameterized by coordinates @xmath265 move with the fluid velocity . this means the @xmath265 are lagrangian coordinates . moreover , equations ( [ intdiffeqn - q ] ) for the evolution of these support sets are canonical hamiltonian equations , @xmath266 the corresponding hamiltonian function @xmath267 is , @xmath268 this is the hamiltonian for geodesic motion on the cotangent bundle of a set of curves @xmath269 with respect to the metric given by @xmath270 . this dynamics was investigated numerically in @xcite to which we refer for more details of the solution properties . one important result found _ numerically _ in @xcite is that only codimension - one singular momentum solutions appear to be stable under the evolution of the epdiff equation . thus , we have _ stability for codimension - one : the singular momentum solutions of epdiff are stable , as points on the line ( peakons ) , as curves in the plane ( filaments , or wave fronts ) , or as surfaces in space ( sheets ) . _ proving this stability result analytically remains an outstanding problem . the stability of peakons on the real line is proven in @xcite . [ [ reconnections - in - oblique - overtaking - collisions - of - peakon - wave - fronts . ] ] reconnections in oblique overtaking collisions of peakon wave fronts . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + figures [ 2dstrip_skew ] shows results of oblique wave front collisions producing reconnections for the epdiff equation in two dimensions . figure [ 2dstrip_skew ] shows a single oblique overtaking collision , as a faster expanding peakon wave front overtakes a slower one and reconnects with it at the collision point . [ [ the - peakon - solution - ansatz - is - a - momentum - map . ] ] the peakon solution ansatz is a momentum map . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + as shown in @xcite , the solution expressions ( [ chpeakon - m - soln ] ) in one dimension and ( [ m - ansatz ] ) in higher dimensions may be interpreted as equivariant momentum maps , from the cotangent bundle of the smooth embeddings of lower dimensional sets @xmath271 to the dual of the lie algebra of vector fields defined on these sets . ( momentum maps for hamiltonian dynamics are reviewed in @xcite , for example . ) the result that the singular solution ansatz is a momentum map helps to organize the theory , to explain previous results and to suggest new avenues of exploration . this geometric feature underlies the remarkable reduction properties of the ch and epdiff equations , and _ explains _ why they must be lie - poisson hamiltonian equations . this is because of the general fact that equivariant momentum maps are poisson maps . this geometric feature also underlies the singular momentum solution ( [ m - ansatz ] ) and its associated velocity ( [ u - ansatz ] ) which generalize the peakon solutions , both to higher dimensions and to arbitrary kinetic energy metrics . as we saw in section [ mommap - sec ] , the soliton solution ( [ mommap1 ] ) is also a momentum map . this soliton momentum map may be expected to apply in the action - angle representation of the solution of any integrable hamiltonian pde . its further properties will be studied in detail elsewhere . ( 1 . ) throughout this discussion the solutions @xmath134 were confined to be functions in the schwartz class , @xmath272 . the situation when the condition @xmath273 on the initial data does not hold is more complicated and requires separate analysis @xcite . in general , it leads to wave - breaking @xcite . an attempt at developing the inverse scattering theory for this case has been made by kaup @xcite , who suggested applying the inverse scattering approach separately in each interval where @xmath274 is of the same sign . the problem however is how to join solutions that are valid in different intervals . the peakon solution ( [ m_i ] ) was obtained from the soliton solution under the assumption @xmath273 . thus , all @xmath275 are of the same sign , since all the eigenvalues @xmath133 in this case are positive . however , one can formally use the same solution with eigenvalues of various signs to model @xmath275 of various signs ( mixture of peakons and ` anti - peakons ' ) and thus to study peakon anti - peakon interactions , see e.g. @xcite . the result is that the multi - peakon interaction ( including anti - peakons ) in general decomposes into a sequence of pairwise collisions @xcite . the collision of a single peakon - anti - peakon pair was studied already in @xcite . when the eigenvalues are of mixed signs , the hankel determinants in the denominator of ( [ m_i ] ) may develop singularities for finite values of @xmath60 . this ` peakon - breaking ' phenomenon is apparently the analog of the wave - breaking mentioned earlier when @xmath276 . this needs to be investigated further . ddh is grateful to r. camassa , j. e. marsden , t. s. ratiu and a. weinstein for their collaboration , help and inspiring discussions over the years . ddh gratefully acknowledges partial support by the royal society of london s wolfson award scheme . rii acknowledges funding from a marie curie intra - european fellowship . both authors thank j. percival and m. f. staley for providing figures . alber m s , camassa r , fedorov y , holm d d and marsden j e 1999 _ phys . a _ * 264 * 171 ; + ibid 2001 _ comm . phys . _ * 221 * 197 ; + alber m s , camassa r , holm d d and marsden j e 1994 _ lett . math . phys . _ * 32 * 137 + alber m s , camassa r and gekhtman m 2000 _ crm proc . and lecture notes _ * 25 * 1 degasperis a and procesi m 1999 asymptotic integrability , in:_symmetry and perturbation theory _ , ed a degasperis and g gaeta ( singapore : world scientific ) p 23 ; + degasperis a , holm d d and hone a n w 2002 _ theoret . and math . phys . _ * 133 * 1463 dubrovin b a 1981 _ russ . surv . _ * 36 * 11 ; + dubrovin b a , novikov s p and krichever i m 1985 integrable systems . i in : _ itogi nauki i tekhniki . probl . mat . naprav . _ * 4 * ( moscow : viniti ) ( russian ) ; 1989 _ encyclopaedia of mathematical sciences _ * 4 * ( berlin : springer ) ( engl . transl . ) holm d d and marsden j e 2004 momentum maps and measure - valued solutions ( peakons , filaments and sheets ) for the epdiff equation . _ in : the breadth of symplectic and poisson geometry _ , _ progr _ * 232 * ed j e marsden and t s ratiu ( boston : birkhuser ) p 203

this study reports an unusual heterogeneity in [co]/[co ] abundance ratios of carbon monoxide observed in the gas phase toward seven solar - mass ysos and three dense foreground clouds in the nearby star - forming regions , ophiuchus , corona australis , orion , vela and an isolated core , l43 . robust isotope ratios were derived using infrared absorption spectroscopy of the 4.7 m fundamental and 2.3 m overtone rovibrational bands of co at very high resolution (/ ) , observed with the crires spectrograph on the very large telescope . we find [co]/[co ] values ranging from 85 to 165 , significantly higher than those of the local interstellar medium ( to 69 ) . these observations are evidence for isotopic heterogeneity in carbon reservoirs in solar - type yso environments , and encourage the need for refined galactic chemical evolution models to explain thec/c discrepancy between the solar system and local ism . the oxygen isotope ratios are consistent with isotopologue - specific photodissociation by co self - shielding toward the disks , vv cra n and hl tau , further substantiating models predicting co self - shielding on disk surfaces . however , we find that co self - shielding is an unlikely general explanation for the high [co]/[co ] ratios observed in this study . comparison of the solid co against gas - phase [co]/[co ] suggests that interactions between co ice and gas reservoirs need to be further investigated as at least a partial explanation for the unusually high [co]/[co ] observed .

1509.06607

observations of carbon monoxide ( co ) provide a robust window into circumstellar chemistry . due to the large percentage of carbon ( @xmath9 to @xmath10 ) and oxygen ( @xmath11 ) tied up in co in the early stages of young stellar object ( yso ) formation @xcite , co is a valuable tracer of chemical inheritance from the envelope to the disk in solar - type ysos . the use of high - resolution spectroscopy to observe co at infrared wavelengths enables high - precision investigations of isotope fractionation in both oxygen and carbon toward ysos in embedded and disk - phase evolutionary stages , and is consequently an important tool for connecting the solar system to currently forming protoplanetary systems . examples include observational studies toward diffuse clouds and inferring isotopologue - specific photodissociation by co self - shielding as a potentially important mechanism of carbon isotope fractionation in co in the interstellar medium @xcite . further , high - resolution near - infrared observations of the co rovibrational band toward protoplanetary disks @xcite and protostellar envelopes have recently revealed signatures in co isotopologue ratios that are consistent with co self - shielding , with precision sufficiently high for comparison to meteoritic data @xcite . observations of isotope ratios toward ysos and , where possible , their parent clouds , are valuable in assessing theorized similarities between circumstellar envelopes , disks and dense clouds local to their origin . this is important not only for evaluating predictions of galactic chemical evolution ( gce ) models that are constrained by solar system abundances , but also for characterizing the range of carbon isotope ratios toward ysos varying with evolutionary stage , and establishing accurate astrochemical constraints on protoplanetary environments . observational studies can help demonstrate whether there are significant isotopic differences or trends between diffuse clouds , dense molecular clouds , and ysos . these in turn help construct explanations for apparent chemical anomalies , such as the @xmath0c/@xmath2c solar system - interstellar medium ( ism ) discrepancy and solar system oxygen isotope anomaly . this paper focuses on the carbon isotopic systems in ysos as traced by high - resolution observations of co isotopologues , with additional mention of significant findings in oxygen isotopes . current measurements of the @xmath0c/@xmath2c ratio for the solar system are @xmath12 to 89 , derived from photospheric models applied to the solar co spectrum @xcite , and bulk solar system materials @xcite . values of @xmath0c/@xmath2c for the solar system have been consistently higher than those of @xmath13 to 69 observed toward the local ism , measured using millimeter to sub - millimeter rotational transitions of various carbon - bearing molecules , co , cn , ch@xmath14 , and h@xmath15co @xcite . the apparent discrepancy in @xmath0c/@xmath2c between the solar system and ism has been attributed to gce leading to enrichment of @xmath2c relative to @xmath0c over the @xmath16 gyr lifetime of the solar system ( e.g. * ? ? ? * ; * ? ? ? an important property of gce models is the assumption that the solar system records the chemical state of the molecular cloud material @xmath16gyr ago . that is , the assumption that the solar nebula was isotopically representative of its parent cloud at its galactic radius at the time of formation should also hold for present - day solar - type ysos . the challenge for observational comparisons of solar system isotope ratios in general is the need for high precision in molecular abundance measurements . while techniques using strong rotational lines in the millimeter / sub - millimeter range often involve model - dependent comparisons between optically thick and thin lines , high - resolution infrared spectrographs have enabled significant breakthroughs in observations of nearby ysos . infrared absorption spectroscopy of optically thin co lines allows for precise column density measurements along pencil - beam lines - of - sight , enabling the derivation of precise column densities and isotope ratios from the line profiles of as many as four currently observable co isotopologues ( @xmath0c@xmath1o , @xmath2c@xmath1o , @xmath0c@xmath17o , @xmath0c@xmath18o ) . robust isotope ratios so derived can be compared with statistical significance to sub - millimeter ism values , solar abundances , and meteoritic oxygen . a high degree of precision ( within @xmath19 ) on the absolute values of the column densities has been demonstrated for the four most common co isotopologues @xcite . the study reported here builds on this work . our targets span a range in @xmath0c@xmath1o ice optical depth , from nearly saturated to undetected , providing an opportunity to investigate a potential observational connection between the co ice and gas reservoirs by comparing the gas - phase [ @xmath0c@xmath1o]/[@xmath2c@xmath1o ] to the co ice column along a single line - of - sight . since solids tend to concentrate the heavy isotope , relative ice - gas partioning may lead to a paucity of @xmath2c ( or excess of @xmath0c ) in the gas phase , with the opposite signature sequestered in the ice . data for all targets were obtained as part of a european southern observatory large program @xcite to observe @xmath20 ysos and protoplanetary disks with the cryogenic infrared echelle spectrograph ( crires ) at the very large telescope ( vlt ) in chile . the crires instrument is an adaptive optics ( ao)-assisted spectrometer that operates at very high resolving power ( @xmath5/@xmath6@xmath5 @xmath21 95,000 ) . llll[ht ] doar24e s , n & oph & 125 & stage ii / ir companion + irs 43 & oph & 125 & stage i + irs 51 & oph & 125 & extincted stage ii + irs 63 & oph & 125 & stage i + rno 91 & l43 & 125 & stage i + vv cra n , s & cra & 130 & stage ii / ir companion + hl tau & tau & 140 & stage i + reipurth 50 & ori & 470 & stage i / fu ori + lln 19 & vel & 700 & stage i [ target_table ] targets tracing high column densities of molecular gas were selected based on evidence of deep co absorption lines at 4.7@xmath4 m in spectra obtained with vlt - isaac @xcite and keck - nirspec ( e.g. , * ? ? ? selected targets were observed by crires in the co overtone ( @xmath22 ) band at 2.3 @xmath4 m and the fundamental ( @xmath23 ) band at 4.7 @xmath4 m . because the overtone transitions are weaker than those of the fundamental band by a factor of @xmath24 , the former are typically optically thin for @xmath0c@xmath1o , thereby enabling the precise measurement of column densities for this molecule . the spectra of the fundamental rovibrational bands were obtained using the @xmath25 slit , resulting in a resolving power of @xmath26 ( corresponding to 3.2kms@xmath27 ) . several settings in the fundamental band were obtained for all targets in order to probe a range of rotational levels for @xmath2c@xmath1o , @xmath0c@xmath17o and @xmath0c@xmath18o , spanning @xmath28 to at least @xmath29 . observations were timed such that the doppler shifts of the targets relative to the earth s atmosphere were at least 20kms@xmath27 due to a combination of the intrinsic target velocity and reflex motion of the earth . as a consequence , telluric co lines are typically not affecting the absorption lines of the targets . lllll[ht ] doar24e s , n & 16 26 23.36 & -24 20 59.8 & @xmath30 & 2007 - 09 - 02 + & & & @xmath31 & 2010 - 03 - 18 + irs 43 & 16 27 26.94 & -24 40 50.8 & @xmath32 & 2008 - 08 - 05 + & & & @xmath31 & 2008 - 08 - 04 + irs 51 & 16 27 39.83 & -24 43 15.1 & @xmath33 & 2008 - 08 - 02 + & & & @xmath34 & 2008 - 08 - 03 + irs 63 & 16 31 35.65 & -24 01 29.3 & @xmath35 & 2007 - 04 - 25 + & & & @xmath31 & 2008 - 08 - 07 + rno 91 & 16 34 29.28 & -15 47 01.5 & @xmath36 & 2010 - 03 - 18 + & & & @xmath31 & 2010 - 04 - 08 + vv cra n , s & 19 03 06.74 & -37 12 49.7 & @xmath37 & 2007 - 08 - 31 + & & & @xmath31 & 2007 - 08 - 31 + hl tau & 04 31 38.44 & 18 13 57.7 & @xmath38 & 2010 - 01 - 30 + & & & @xmath39 & 2010 - 01 - 28 + reipurth 50 & 05 40 27.70 & -07 27 28.0 & @xmath40 & 2007 - 10 - 11 + & & & @xmath31 & 2007 - 10 - 17 + lln 19 & 08 48 48.00 & -43 32 25.0 & @xmath41 & 2007 - 04 - 25 + & & & @xmath42 & 2007 - 04 - 25 [ obs_table ] table [ target_table ] summarizes the region , distance and evolutionary stage of the observed targets , listed in order of increasing distance from the local solar neighborhood ( @xmath43kpc ) . specific observing parameters are listed in table [ obs_table ] . selected targets probe a range of circumstellar environments , as noted . because the lines - of - sight traced by infrared absorption may pass through the upper layers of an inclined disk as well as through envelope material and a foreground molecular cloud unrelated to the yso , some portion of each of these environments may be represented in the integrated spectra , depending on the target . for instance , near - infrared absorption spectra observed toward the extincted protoplanetary disk , irs 51 , will be an integrated sight line consisting mostly of the foreground ophiuchus ridge and a relatively small contribution of the protoplanetary disk . this geometry has been predicted by interferometric observations @xcite and is supported by our observations that reveal cold - only gas lines . ) . spectra for doar24e n , s show few co absorption lines , while vv cra n shows a rich co forest of narrow rovibrational absorption lines . sample lines are marked . the low - level emission from hot co gas toward doar24 s and vv cra n is possibly representative of inner disk disk material . the weak emission line at @xmath44 m for doar24e n is hi pfund@xmath45 . the main portions of the spectra used for the line analyses are presented.,width=321 ] the crires spectra were reduced using standard procedures for infrared echelle spectroscopy , including flat - field correction , adjustments to account for detector non - linearity , and linearization of the spectral traces in dispersion- and cross - dispersion directions . the spectra were optimally extracted , wavelength - calibrated using the telluric absorption lines referenced to an atmospheric model spectrum , and transformed to the local standard - of - rest frame . relative flux calibration was carried out by dividing the targets by observations of early - type telluric standards observed immediately surrounding observations of the science target at the same airmass . the division was optimized by modifying the telluric spectra to account for any difference in grating angle , airmass , and spectral resolving power . to aid in the measurement of any intrinsic line broadening , the resolving powers for each science spectrum were measured from telluric spectra to within an accuracy of @xmath46kms@xmath27 , by fitting the `` reference forward model '' ( rfm , see also section [ sec : derivation ] ) . further details on crires data processing is described in @xcite . ) . hl tau and lln 19 have a rich co forest of narrow rovibrational absorption lines due to the various co isotopologues present . sample lines are marked . few lines of the rarer isotopologues ( @xmath0c@xmath17o and @xmath0c@xmath18o ) are seen toward the more deeply embedded object , lln 19 . low - level emission toward vv cra s and hl tau are possibly representative of hot co gas of the inner disk.,width=321 ] 1).the rich co forest of narrow rovibrational absorption lines are due to the various co isotopologues present . sample lines are marked . rno 91 shows low - level emission possibly representative of the inner disk.,width=321 ] transitions . sample lines are marked.,width=321 ] the fundamental @xmath48 and first - overtone @xmath49 rovibrational bands are shown for all targets in figures [ specm1a ] through [ speck2 ] , in ascending order of co ice optical depths at @xmath50 m . a forest of narrow co rovibrational absorption lines is seen to varying degrees in all spectra . a magnified spectral region of hl tau is shown as a representative example of individual line profiles ( figure [ spec_zoom ] ) . broad , complex @xmath48 co emission is often seen in the spectra , and likely originates in the innermost ( @xmath51au ) regions of the disks @xcite . however , the emission does not interfere with the analysis , as it is much broader than the absorption , and is generally weak or absent in the rare isotopologue transitions . the potentially affected @xmath48 @xmath0c@xmath1o lines are not used for column density measurements . broad absorption observed toward several objects at @xmath49 ( figures [ speck1 ] , [ speck2 ] ) likely originates in the stellar photosphere and is clearly separated from the narrow interstellar co lines . temporal variation in [ @xmath0co]/[@xmath2co ] was found toward vv cra s , with a [ @xmath0co]/[@xmath2co ] abundance ratio of @xmath52 when using a @xmath48 observation four months prior to the first - overtone @xmath49 , as compared to @xmath53 when coeval observations were used ( errors are statistical 1@xmath54 values ) . the possibility for temporal spectral variability should therefore be considered for doar24e and irs 63 , targets for which only non - coeval observations are available . total abundances for each isotopologue were derived by measuring the column density of co in each observed rovibrational level of the ground vibrational state which , in turn , was obtained from the intrinsic optical depths at line center @xmath55 derived from a best - fit absorption line profile to each rovibrational transition . individual column densities were used to make a best fit to a boltzmann population , then summed over all levels . this method is described in detail in @xcite . llllllll[ht ] doar24e s & 2.79 & 3.12 & @xmath56 & @xmath57 & @xmath58 & @xmath59 & + doar24e n & 2.79 & 3.12 & @xmath56 & @xmath60 & @xmath61 & @xmath62 & + vv cra n & 2.82 & 3.49 & @xmath63 & @xmath64 & @xmath65 & @xmath66 & @xmath67 + vv cra s & 2.82 & 3.49 & @xmath68 & @xmath69 & @xmath70 & @xmath71 & @xmath72 + hl tau & 3.26 & 3.43 & @xmath73 & @xmath74 & @xmath75 & @xmath76 & @xmath77 + lln 19 & 3.22 & 4.68 & @xmath78 & @xmath79 & @xmath80 & @xmath81 & + rno 91 & 3.26 & 3.72 & @xmath82 & @xmath83 & @xmath84 & @xmath85 & @xmath86 + reipurth 50 & 3.44 & 4.54 & @xmath87 & @xmath88 & @xmath89 & @xmath90 & @xmath91 + irs 63 & 3.57 & 3.39 & @xmath92 & @xmath93 & @xmath94 & @xmath95 & @xmath96 + irs 43 & 3.03 & 3.45 & @xmath97 & @xmath98 & @xmath99 & @xmath100 & @xmath101 + irs 51 & 3.03 & 3.37 & @xmath102 & @xmath103 & @xmath104 & & [ tau_table ] the technique of using the line profiles to directly determine column densities takes advantage of the high spectral resolution of crires to spectrally resolve the velocity dispersion of the individual co lines . this in turn leads to greater precision than using unresolved lines which require a curve - of - growth analysis to derive intrinsic line widths @xcite . with crires , the resolving power is sufficiently high to partially resolve the intrinsic line profiles ( fwhm@xmath105 ) , albeit marginally in some cases : vv cra s , rno 91 , and irs 43 . therefore , it is possible that the derived column densities toward these targets may be more uncertain . llll[ht ] @xmath0c@xmath1o & @xmath106 & @xmath107 & @xmath108 + @xmath0c@xmath1o & @xmath109 & @xmath110 & @xmath111 + @xmath0c@xmath1o & @xmath112 & @xmath113 & @xmath114 + @xmath0c@xmath1o & @xmath115 & @xmath116 & @xmath117 + @xmath0c@xmath1o & @xmath118 & @xmath119 & @xmath120 + @xmath0c@xmath1o & @xmath121 & @xmath122 & @xmath123 + @xmath0c@xmath1o & @xmath124 & @xmath125 & @xmath126 [ line_tab ] the measured instrumental broadening , intrinsic velocity dispersions and optical depth ranges for each observed isotopologue are shown in table [ tau_table ] . line - by - line optical depths and doppler shifts for observed targets are available in the online version of this paper , with a sample partial line list shown in table [ line_tab ] . complete data tables are available in the online version of the paper . instrumental broadening values were modeled individually for each spectrum in its entirety , following the method in @xcite for using an atmospheric model for telluric spectra . the reference forward model ( rfm ; http://www.atm.ox.ac.uk/rfm/ ) code was used to derive a detailed fit to the individual telluric standard star spectra , with consideration of meteorological conditions at the time of observation . the atmospheric molecular parameters were taken from the hitran 2008 database @xcite . there are small relative differences in resolving power between the standard stars and science targets due to differences in seeing or adaptive optics ( ao ) correction , but these differences were accurately measured and corrected for by minimizing the continuum residuals after telluric division . optimized intrinsic velocity dispersions were derived from the weighted mean of dispersions measured in individual @xmath0c@xmath17o lines , assuming that both the intrinsic and instrumental profiles are well - approximated by gaussian profiles . the @xmath0c@xmath17o lines were used because this rare isotopologue is always optically thin , and therefore is not subjected to saturation line broadening , which may affect the more common @xmath2co isotopologue . optical depths of the optically - thin @xmath0c@xmath1o @xmath49 lines were similarly computed using the derived intrinsic broadening from the @xmath48 @xmath0c@xmath17o lines . it can be noted that , even when the resolving power of the two spectral bands differed for a single target , we obtained good fits to lines in both bands under the reasonable assumption that the intrinsic profile width is independent of transition and isotopologue . further , it is reasonable to use the same velocity dispersion parameter for a target with a two - temperature distribution of the data , since thermal contributions are very small contributors to the total intrinsic dispersion ( 0.077 and 0.242 km s@xmath27 for @xmath0c@xmath1o at 20 and 200 k , respectively ) . the rotational analyses for all targets are shown in figures [ roplots1 ] and [ roplots2 ] . sub - level column densities , @xmath127 , derived from the fits to the absorption lines , are plotted against the energy , @xmath128 , of the rotational transition state , @xmath129 . for targets where cold gas dominates , a single - temperature model is sufficient to fit the data , while targets with absorption that includes higher-@xmath129 transitions are best fit with a two - temperature model , representing cold and warm environments along the line - of - sight , and allowing for different excitation temperatures for each isotopologue along a given line - of - sight . derived isotopologue temperatures from the fits are shown in table [ temp_table ] . high temperatures ( @xmath130 to 250 k ) are roughly consistent with those of the inner parts of typical yso envelopes , while low temperatures ( @xmath131 to 30 k ) may trace outer envelopes or molecular cloud material in front of the ysos . for lines - of - sight with no warm component , it is likely that the gas is from a foreground cloud not related to the yso itself ( doar24e n and s ) , vv cra s , and irs 51 . total column densities and isotopologue ratios are shown in table [ ratio_table ] . due to the dominance of higher-@xmath129 lines in targets with both low- and high - temperature lines , the higher - temperature ratios are generally better constrained within two - temperature environments @xcite . statistical errors on the column densities are generally found to be small , reflecting the very high signal - to - noise ratio of the data . for example , average signal - to - noise per pixel in the fundamental bands is @xmath132 . with an average crires fwhm of 2.0 pixels , we retrieve average line errors of @xmath133 ( 1/(@xmath134 ) ) . first - overtone bands average @xmath135 . we therefore expect to know the column densities to a fraction of a percent , which is represented in our results . these errors do not take into account any potential systematic errors in extracting the line strengths and profiles due to inherent non - gaussianity of the line - fitting procedure , which we can not quantify with certainty . however , to roughly estimate the magnitude of systematic retrieval errors , we carried out a small number of forward models with 2-dimensional radiative transfer models ( radlite , * ? ? ? * ) simulating synthetic co isotopologue absorption spectra using a basic one - dimensional infalling envelope from @xcite . the synthetic spectra were convolved with the appropriate instrumental resolution and isotopologue ratios retrieved from the synthetic spectra using the same methods as for the observed data . in each case , we were able to reproduce the input ratios to better than 10% . we therefore estimate that model - dependent systematics add at most a 10% error to the retrieved isotopologue ratios , in addition to the statistical error . llcccc[ht ] doar24e s & single & @xmath136 & @xmath137 & @xmath138 & + doar24e n & single & @xmath139 & @xmath140 & @xmath141 & + vv cra n & high & @xmath142 & @xmath143 & @xmath144 & @xmath145 + vv cra n & low & @xmath146 & @xmath146 & @xmath147 & @xmath148 + vv cra s & single & @xmath146 & @xmath149 & @xmath150 & @xmath151 + hl tau & high & @xmath152 & @xmath153 & @xmath154 & @xmath155 + hl tau & low & @xmath156 & @xmath157 & @xmath158 & @xmath159 + lln 19 & single & @xmath160 & @xmath161 & @xmath162 & + rno 91 & high & @xmath163 & @xmath164 & @xmath165 & + rno 91 & low & @xmath166 & @xmath167 & @xmath168 & @xmath169 + reipurth 50 & high & @xmath170 & @xmath171 & @xmath172 & @xmath173 + reipurth 50 & low & @xmath174 & @xmath137 & @xmath175 & @xmath146 + irs 63 & high & @xmath176 & @xmath177 & @xmath178 & @xmath179 + irs 63 & low & @xmath180 & @xmath140 & & + irs 43 & high & @xmath181 & @xmath182 & @xmath183 & @xmath184 + irs 43 & low & & @xmath185 & @xmath167 & @xmath186 + irs 51 & single & @xmath187 & @xmath137 & & [ temp_table ] the doar24e system is a hierarchical multiple for which we have spectra of two components , with the rotational analyses nearly identical for both components . this lack of significant spatial variability strongly suggests that we are tracing spatially extensive and homogenous foreground material , that in turn is the dominant contributor smoothly distributed on scales of at least a few 100 au . the lack of isotopic variability between the doar24e components gives us further confidence on our retrieval method for deriving precise isotope ratios . due to high extinction leading to a faint @xmath188 band continuum for the embedded yso irs 43 , we detect high-@xmath129 lines for all co isotopologues in @xmath189 , but only low-@xmath129 lines for @xmath0c@xmath1o in the @xmath188 band . given this disparate set of temperature ranges , our data for irs 43 are therefore not useful for measuring co isotopologue abundance ratios relative to @xmath0c@xmath1o . toward vv cra s , we derive a @xmath0c@xmath1o column density of @xmath190@xmath191 . this is nearly three orders of magnitude higher than the value of @xmath192@xmath191 derived in @xcite . our rotational temperature ( @xmath193k ) is also significantly lower than their value ( @xmath194k ) . since @xcite used the @xmath48 spectra from the archive , these differences could be explained by strong saturation effects resulting from their use of the highly optically thick @xmath189 @xmath0c@xmath1o lines . this difference reinforces the necessity of using @xmath188 data for deriving precise @xmath0c@xmath1o column densities . @xcite proposed that the cold gas component observed toward vv cra s may be interpreted as either molecular cloud material in front of the vv cra binary system , or disk gas toward the primary component as suggested by @xcite . we point out that the former is more likely because the presence of both warm and cold components are expected for a disk , while a foreground cloud will typically only show a single cold component , such as that seen towards vv cra n. taken together , the likely geometry for the vv cra system and the spectroscopic results render it unlikely that the vv cra s spectra represent outer disk . vv cra n observations most likely represent warm and cold disk gas . llccccc[ht ] doar24e s & single & @xmath195 & @xmath196 & : @xmath197 & & + doar24e n & single & @xmath198 & @xmath199 & : @xmath200 & & + vv cra n & high & @xmath201 & @xmath163 & @xmath202 & @xmath203 & @xmath204 + vv cra n & low & @xmath205 & @xmath206 & @xmath207 & @xmath208 & @xmath209 + vv cra s & single & @xmath210 & @xmath211 & : @xmath212 & : @xmath213 & @xmath214 + hl tau & high & @xmath215 & @xmath216 & @xmath217 & @xmath218 & @xmath219 + hl tau & low & @xmath220 & @xmath221 & @xmath222 & @xmath223 & @xmath224 + lln 19 & single & @xmath225 & @xmath226 & : @xmath227 & & + rno 91 & high & @xmath228 & @xmath229 & @xmath230 & @xmath231 & @xmath232 + rno 91 & low & @xmath233 & @xmath234 & @xmath235 & @xmath236 & @xmath237 + reipurth 50 & high & @xmath238 & @xmath239 & @xmath240 & @xmath241 & @xmath242 + reipurth 50 & low & @xmath243 & @xmath244 & @xmath245 & @xmath246 & @xmath247 + irs 63 & high & @xmath248 & @xmath249 & @xmath250 & @xmath251 & @xmath252 + irs 63 & low & @xmath253 & @xmath254 & & & + irs 43 & high & @xmath255 & : @xmath256 & : @xmath257 & : @xmath258 & @xmath259 + irs 43 & low & & : @xmath260 & : @xmath261 & : @xmath262 & : @xmath263 + irs 51 & single & @xmath264 & @xmath265 & & & + local ism & & & @xmath266 & @xmath267 & @xmath268 & @xmath269 [ ratio_table ] derived [ @xmath0c@xmath1o]/[@xmath2c@xmath1o ] ratios are plotted against galactocentric radius ( r@xmath270 ) in figure [ gc_co_plot ] . the solar system and local ism values of @xmath271 @xcite and @xmath272 @xcite are plotted at 8 kpc , respectively . derived [ @xmath0c@xmath1o]/[@xmath2c@xmath1o ] toward our targets range from @xmath273 to 165 , values inconsistent with those toward most ysos and molecular clouds , or from gce @xcite and regression models @xcite . the distribution of data from the literature approximates the general trend of increasing metallicity toward the galactic center , with the corresponding increase in @xmath0c/@xmath2c with greater r@xmath270 . c/@xmath2c ratios vs. r@xmath270 ( kpc ) . gray symbols show @xmath0c/@xmath2c values from the literature , derived from different carbon - bearing molecules , co , cn , ch@xmath14 and h@xmath15co . data from our study are labeled for yso targets ( filled circles ) and foreground clouds ( unfilled circles ) . colors connect these data by molecular cloud with those in figure [ delta_delta ] . for targets where there is both warm and cold gas along a line - of - site , ratios are shown for the more robustly determined warm gas , unless otherwise noted . local ism ( @xmath274 ; @xcite ) and solar system values ( @xmath275 ; @xcite ) are noted . the black arrows indicate high @xmath0c/@xmath2c in diffuse regions of ophiuchus @xcite . yso data for monr2 irs3 , lkh @xmath276101 , afgl 490 @xcite and hl tau ( lower value ; @xcite ) are marked ( dark purple).,width=321 ] previous studies have discussed the exchange between co solid- and gas - phase reservoirs in disks and envelopes ( e.g. * ? ? ? * ; * ? ? ? * ) . with a range of co ice column densities centered at @xmath277 m ( figures [ specm1a ] through [ specm2b ] ) , our data set provided an opportunity to investigate relationships between the gas - phase isotopologue ratios and the co ice fraction along single lines - of - sight . co ice column densities were calculated from optical depth analyses in @xcite and @xcite . of the broad co ice profile , the pure , cold co component is most relevant with respect to interactions with co gas , as it is the solid state reservoir that would most likely directly exchange with the gas in cold , dense clouds . we used optical depths for this component to derive total co ice column densities for each target . using the phenomenological decomposition analyses in @xcite , the pure component corresponds to the middle optical depth ( @xmath278 ) profile measurement ; @xcite was used for lln 19 . for targets with no ( vv cra , doar24e ) or very weak ( hl tau ) spectral ice features , we used upper limits from the crires fundamental bands . total column densities of pure co in the ice phase @xmath279 were computed using the relation from @xcite , @xmath280 where @xmath281 is the optical depth at line center of the pure co component , @xmath282 is the band strength of @xmath0co , and the numerical factor , 6.03cm@xmath27 , considers the effect of grain shape on the band strength . in this study , we used the continuous distribution of ellipsoids ( cde ) model @xcite in computing solid co column densities . optical depths , column densities , and total co ice fractions are listed in table [ ice_table ] , in ascending order of co ice optical depth . although it is assumed that cold co gas will more likely exchange with co ice , for targets where co gas was observed in both cold and warm environments , values for co ice fractions are shown separately for each environment . the relation between the co ice fraction and [ @xmath0c@xmath1o]/[@xmath2c@xmath1o ] gas - phase ratios is shown in figure [ ice_vs_ratio ] , with co ice fractions plotted against derived [ @xmath0co]/[@xmath2co ] for cold ( @xmath284 to 30k ) and warm ( @xmath285 to 250k ) regimes , the latter only for targets where warm gas was observed . we distinguish between targets where only cold co gas is observed , and targets where both warm and cold temperatures is found . in the latter cases , cold - gas measurements are likely more uncertain due to the contribution of the high - temperature on the low - temperature lines . the optical depth for the @xmath2co ice toward the foreground cloud traced by irs 51 provides the only solid [ @xmath0c@xmath1o]/[@xmath2c@xmath1o ] ratio ( @xmath288 ) that can be compared to a gas - phase ratio along the same line - of - sight in this study . the ice - phase ratio is significantly lower than our derived gas - phase ratio of @xmath289 . the [ @xmath0c@xmath1o]/[@xmath2c@xmath1o ] value in ice for irs 51 is roughly consistent with other published [ @xmath0c@xmath1o]/[@xmath2c@xmath1o ] values for ice in yso targets : @xcite derive a [ @xmath0c@xmath1o]/[@xmath2c@xmath1o ] ratio of @xmath290 for the ice observed toward the massive yso ngc 7538 irs 9 , using keck - nirspec , and a lower ratio of @xmath291 was observed toward the crbr 2422.8 - 3423 edge - on disk using vlt - isaac @xcite . llllll[ht ] doar24e s & single & @xmath292 & @xmath293 & @xmath294 & @xmath293 + doar24e n & single & @xmath292 & @xmath293 & @xmath295 & @xmath293 + vv cra n & low & @xmath296 & @xmath297 & @xmath298 & @xmath292 + vv cra n & high & @xmath296 & @xmath297 & @xmath299 & @xmath300 + vv cra s & single & @xmath296 & @xmath297 & @xmath301 & @xmath297 + hl tau & low & @xmath302 & @xmath296 & @xmath303 & @xmath304 + hl tau & high & @xmath302 & @xmath296 & @xmath305 & @xmath306 + lln 19 & single & @xmath307 & @xmath308 & @xmath309 & @xmath310 + rno 91 & low & @xmath311 & @xmath312 & @xmath313 & @xmath314 + rno 91 & high & @xmath311 & @xmath312 & @xmath315 & @xmath316 + reipurth 50 & low & @xmath317 & @xmath318 & @xmath319 & @xmath320 + reipurth 50 & high & @xmath317 & @xmath318 & @xmath321 & @xmath307 + irs 63 & low & @xmath322 & @xmath323 & @xmath324 & @xmath325 + irs 63 & high & @xmath322 & @xmath323 & @xmath326 & @xmath327 + irs 51 & single & @xmath328 & @xmath329 & @xmath330 & @xmath331 [ ice_table ] c@xmath1o ice fraction with the [ @xmath0c@xmath1o]/[@xmath2c@xmath1o ] gas - phase ratios for targets observed in this study . for targets where both warm and cold gas was observed , co ice fractions are plotted with red diamonds when warm gas was used in computing the fraction , and blue stars when cold gas was used . for targets where only cold gas was observed , co ice fractions are shown with over - plotted blue ovals . errors are @xmath332 propagated from errors in ice and gas optical depths measurements . @xmath333 upper limits ( down - facing arrows ) are shown for doar24e n , s , vv cra n , s , and hl tau.,width=321 ] we find significant heterogeneity in gas - phase [ @xmath0c@xmath1o]/[@xmath2c@xmath1o ] toward our targets ( figures [ gc_co_plot ] and [ ice_vs_ratio ] ) , with most values falling above local ism values of @xmath334 , as determined by sub - millimeter observations @xcite . the large spread in [ @xmath0c@xmath1o]/[@xmath2c@xmath1o ] suggests either heterogeneity in the carbon isotopic composition in our local solar neighborhood ( e.g. r@xmath270@xmath335 kpc ) , and/or one or more new fractionation mechanisms for co operating in the dense envelopes of young stars . the distribution of co isotopologue ratios derived from our crires data do not match the general galactic trend , which suggests that the carbon reservoirs of young stellar envelopes do not necessarily isotopically represent their parent clouds , as is often assumed in gce models @xcite . we are not the first to observe unusual [ @xmath0c@xmath1o]/[@xmath2c@xmath1o ] abundance ratios in ysos and molecular clouds ( figure [ gc_co_plot ] ) . in ultraviolet absorption , @xcite and @xcite found high [ @xmath0c@xmath1o]/[@xmath2c@xmath1o ] values toward diffuse regions in the ophiuchus cloud , with values ranging from @xmath336 to 167 . these results have been attributed to selective co photodissociation by external ultraviolet radiation @xcite . we discuss connections with co photodissociation in section [ photodissociation ] . unusually high [ @xmath0c@xmath1o]/[@xmath2c@xmath1o ] values were also detected in rovibrational absorption using subaru ( @xmath337 toward the massive young stars lkh@xmath276101 and monr2 irs3 , with values of @xmath338 and 158 , respectively @xcite . @xcite suggested that photon - trapping in the rotational lines of the most abundant isotope as a possible explanation for the apparently elevated carbon isotope ratios , although the potential effect was not quantified in their study . a possible mechanism for the observed co fractionation is mass - dependent exchange between gas and ice . compared to single - dish , beam - averaged millimeter measurements of co gas - phase isotopologue ratios in molecular clouds , rovibrational absorption through protostellar envelopes is likely to probe denser gas . dense , cold gas is known to lead to efficient co freeze - out onto dust grains @xcite . if the adsorption / desorption rates of co are mass - dependent , isotopologue fractionation may occur . further , if the heavier isotopologue ( @xmath2co ) is more tightly bound to the dust grains , we would expect a increase in the [ @xmath0co]/[@xmath2co ] ratio in the gas , which is consistent with the detection of large columns of co ice along the observed lines - of - sight . here we discuss whether ice - gas partitioning is a possible explanation for the observed high co isotopologue ratios in the gas . the relative fraction of co sequestered in ice , averaged along the line - of - sight , is plotted against the [ @xmath0c@xmath1o]/[@xmath2c@xmath1o ] abundance ratios in the gas ( figure [ ice_vs_ratio ] ) . while we might expect to see increased values of the co abundance ratio for lines - of - sight with a high ice / gas fraction if @xmath2co is preferentially bound in the ice mantles , we instead find significant scatter , much larger than the individual errors on each data point , with no simple apparent relation between these quantities . a correlation is particularly absent between cold - gas reservoirs and co ice fraction , although these two quantities are likely cospatial and trace the same total co reservoir . however , the line - of - sight with the largest ice fraction ( irs 51 ) also has the highest co isotopologue ratio in the gas . therefore , while other factors are likely equally important , we can not rule out that ice - gas partitioning could play a significant role in the observed high gas - phase [ @xmath0c@xmath1o]/[@xmath2c@xmath1o ] . with respect to other factors affecting the gas - phase ratios , we note that high [ @xmath0c@xmath1o]/[@xmath2c@xmath1o ] ratios are also observed in the warm gas in some lines - of - sight ( figure [ ice_vs_ratio ] , red diamonds ) , which is unexpected if ice - gas partitioning is the only fractionation mechanism , since environments warm enough to fully sublimate the co ice should reset the gas isotopologue fraction back to the canonical value of @xmath8 . in order to reconcile this with ice - gas partitioning , there must be a mechanism to effectively remove co from the ice during the cold phase of envelope evolution . chemical pathways removing co from the ice include the successive hydrogenation of co to formaldehyde ( h@xmath15co ) and methanol ( ch@xmath341oh ) @xcite , or reactions with oh or o to form co@xmath342 @xcite . further , we would expect the column densities of ice to be comparable or larger than those of the gas in order for the ice to be a significant reservoir that could increase the gas - phase [ @xmath0c@xmath1o]/[@xmath2c@xmath1o ] . for irs 51 , the large discrepancy in [ @xmath0c@xmath1o]/[@xmath2c@xmath1o ] between the ice ( @xmath272 ) and the gas ( @xmath343 ) , along with a seemingly low co ice fraction ( @xmath344 ) , suggests instead that open - system chemistry may be occurring via co ice reactions . while the data are inconclusive with regard to carbon isotopic fractionation between reservoirs of co ice and gas , potential trends in cold- and warm - gas reservoirs can be further investigated with high - resolution spectroscopy for a larger sample of protostellar lines - of - sight . further , models considering chemical evolution in protoplanetary environments including ysos could consider interactions between ice and gas reservoirs as a potential influence on isotope partitioning . to estimate whether there is an expectation of ice - gas partitioning for co , we briefly review the theoretical basis for a potential mass - dependence in the ice - gas balance . following a standard approach for ice - gas partitioning ( e.g. * ? ? ? * ) , co separations between the solid- and gas - phase reservoirs can be investigated with respect to mass - dependent kinetics and the balance between the rates of @xmath0c@xmath1o and @xmath2c@xmath1o gas adsorption and desorption , as given by the relation , @xmath345 where @xmath346 is the number density of ice , @xmath347exp@xmath348@xmath349 @xmath350@xmath351 , and @xmath352 are the desorption and adsorption rates , respectively . the factor , @xmath353 , is the binding energy of co. the factor @xmath45 accounts for co desorbing only from the top monolayer of the ice , @xmath354 is the grain radius , and @xmath355 is the sticking coefficient . the mass of co @xmath356 is an obvious mass - dependency of this relation , and @xmath357 is the pre - exponential factor . the number densities of the co gas and dust particles are given by @xmath358 and @xmath359 , respectively @xcite . in addition to the mass itself , the binding energy and pre - exponential factors are likely both mass - dependent , although to unknown degrees . co]/[@xmath2co ] ratio predicted by the simple model described by eq . [ eq : rate ] , assuming a difference in binding energy of 10k between the two isotopologues , as allowed by the experimental uncertainty , and a pre - exponential that scales with @xmath360 . the model is a snapshot at @xmath361yr and the individual curves are created by varying the gas density from @xmath362 to @xmath363 and temperature from 9.1 to 10.0k , as indicated.,width=321 ] some experimental work exists for the mass dependence of thermal desorption of co. in an ultra - high vacuum , temperature - programmed desorption ( tpd ) experiment , @xcite measured binding energies and pre - exponentials for desorption from pure @xmath0co and @xmath2co ices . the parameters were derived using an empirical kinetic model . however , no difference between the @xmath0co and @xmath2co binding energies could be measured within the experimental uncertainties : @xmath364 k and @xmath365 k for @xmath0c@xmath1o and @xmath2c@xmath1o , respectively . to estimate the degree of ice - gas fractionation allowed by the boundaries imposed by the experimental evidence , we solved for the ice abundance in eq . [ eq : rate ] , assuming a 10k difference in binding energy between the two co isotopologues and a pre - exponential that scales as @xmath360 @xcite . in figure [ ratio_model ] it is seen that , under these assumptions , the [ @xmath0co]/[@xmath2c ] gas ratio may reach values as high as 170 , but only for a narrow range of temperatures close to those of co desorption ( above which no ice exists ) . essentially , a dust temperature exists at which @xmath2co can efficiently remain bound to the dust surface , while @xmath0co can not . nearly all of this effect is due to the assumed difference of only 10k in the binding energies . compared to what is currently available , significantly more precise laboratory measurements of mass - dependent binding energies would be needed to rule out binding - energy differences as a significant cause of ice - gas fractionation . while such temperature - dependent fractionation may be consistent with the observed spread in isotopologue ratios , it remains unclear to which degree the physical conditions of real molecular clouds and envelopes must be fine - tuned to lead to significant ice - gas fractionation . more realistic modeling of ice - gas fractionation of co is potentially complicated by several factors . for example , investigations of the effects of co entrapment in the ice on the desorption and adsorption ( both potential mass - dependent processes ) in astrophysical environments have shown that diffusion of co into the porous h@xmath15o ice leads to competing co entrapment and diffusion effects on the desorption of co , the degree to which depends on the ice temperature @xcite . further , a recent study has shown that three - phase modeling of h@xmath15o : co@xmath15 ice desorption depends on the ice thickness and mixing ratios , which may have significant effects on ice desorption @xcite . isotope - selective photodissociation of co due to self - shielding refers to the differential shielding of co isotopologues from photodissociation by far ultraviolet ( fuv ) radiation in proportion to their column densities . with regard to oxygen isotopes , this leads to spatial regions characterized by more rapid destruction of @xmath0c@xmath18o and @xmath0c@xmath17o compared to the much more abundant @xmath0c@xmath1o molecule . oxygen liberated during this process is thought to eventually end up in h@xmath15o , providing an explanation for the overabundance of h@xmath366o and h@xmath367o relative to h@xmath368o in the early solar system . isotope - selective photodissociation of co due to self - shielding has been shown to be a relevant process in the interstellar medium @xcite . the solar system oxygen isotope anomaly is characterized by the mass - independent correlation between [ @xmath1o]/[@xmath17o ] and [ @xmath1o]/[@xmath18o ] found in meteoritic materials as compared to the mass - dependent distribution known in terrestrial rocks . mass independence refers to relative differences in [ @xmath1o]/[@xmath17o ] and [ @xmath1o]/[@xmath18o ] among primitive rocky objects that are nearly identical , suggesting changes in [ @xmath1o ] relative to both [ @xmath18o ] and [ @xmath17o ] , rather than the expected mass - dependent trend in which relative changes in [ @xmath1o]/[@xmath18o ] are about half those in [ @xmath1o]/[@xmath17o ] on a log scale . the solar system oxygen anomaly has defied conclusive explanation since its discovery @xcite , and has been investigated observationally by @xcite and @xcite , the latter study finding mass - independent signatures in oxygen isotopes observed in co toward the vv cra n disk , consistent with isotope - specific photodissociation by co self - shielding . the compiled oxygen data for this study , which also includes refined analyses of vv cra n and reipurth 50 from @xcite , are shown in table [ ratio_table ] and figure [ delta_delta ] . the mass - dependent lines in figure [ delta_delta ] are derived by estimating the oxygen isotope ratios for the ism as a function of distance from the galactic center . these values can be estimated from the 4 kpc and 8 kpc values @xcite , such that for @xmath369/[^{18}o]$ ] : @xmath370/[^{18}o]}_{r_{\rm gc } } = ( 57.5 \pm 10)\ { r_{\rm gc } } + 97 , \label{eq:5}\ ] ] where the @xmath332 uncertainty in the slope reflects the uncertainty in the standard errors reported in table 4 of @xcite . for @xmath369/[^{17}o]$ ] , equation [ eq:5 ] is combined with values for @xmath371/[^{17}o]$ ] in the ism , which vary from 3.5 @xcite to 4.1 @xcite ( figure [ delta_delta ] ) . we find a mass - independent signature in the oxygen isotopes relative to solar system abundances , consistent with co self - shielding , toward the hl tau disk , and the same mass - independent signature is retrieved in the refined results for vv cra n ( as compared to @xcite ) . these findings lend further support for models considering co self - shielding on disk surfaces @xcite . it is particularly interesting to note that the recent suggestions of planet formation in the hl tau disk based on new alma imaging @xcite would imply that co self - shielding may be a process that occurs and/or persists later in the evolution of ysos than previously thought . we would not expect uv photolysis to be a significant process in the foreground cloud traced by vv cra s , or in the relatively embedded objects , rno 91 , reipurth 50 , and irs 63 . nevertheless , due to the errors in the rare isotopologue , @xmath0c@xmath18o , a mass - dependent effect can not be distinguished from mass - independence in oxygen fractionation for rno 91 and irs 63 . additional high - quality co data for ysos would help in evaluating mass - independence in oxygen for objects in various stages of evolution . with regard to the rare co isotopolgues , with the exception of low - temperature gas in reipurth 50 , the [ @xmath0c@xmath17o]/[@xmath0c@xmath18o ] ratios are within @xmath372 of molecular cloud radio measurements for local ism , further strengthening the claim that the solar system is anomalous in @xmath17o/@xmath18o with respect to the local solar neighborhood @xcite . this consistency in our data is observational support for @xmath0c@xmath17o and @xmath0c@xmath18o being similarly affected in young stellar environments . while critical evaluation of mass - independence requires three isotopes , we qualitatively considered if isotope - selective photodissociation by co self - shielding could be affecting the derived [ @xmath0c@xmath1o]/[@xmath2c@xmath1o ] toward our targets by comparing the oxygen and carbon isotope ratios where this was possible : vv cra n , vv cra s , hl tau , rno 91 , reipurth 50 , and irs 63 . these are plotted in figure [ oxygen_carbon ] , with the qualitative relative trend for co self - shielding overplotted on the data . the derived [ @xmath1o]/[@xmath18o ] changes by nearly a factor of two relative to [ @xmath0c@xmath1o]/[@xmath2c@xmath1o ] , while the [ @xmath0c@xmath1o]/[@xmath0c@xmath17o ] varies by roughly 1.5 times with [ @xmath0c]/[@xmath2c ] . if co self - shielding were the dominant mechanism , we would expect at least positive , direct changes in the carbon with the oxygen isotopes along an integrated line - of - sight , as well as similar variability in [ @xmath0c@xmath17o ] and [ @xmath0c@xmath18o ] compared with [ @xmath0c@xmath1o]/[@xmath2c@xmath1o ] due to their similar rates of photodissociation ( e.g. * ? ? ? * ; * ? ? ? therefore , while we can not be certain that photodissociation of co has no affect on the [ @xmath0c@xmath1o]/[@xmath2c@xmath1o ] , we can not claim that it is a dominant process affecting this carbon reservoir . o]/[@xmath18o ] and [ @xmath1o]/[@xmath17o ] from local ism values are shown in per mil using the linearized form of the delta notation . here , @xmath373 and @xmath374 = [ c@xmath17o]/[c@xmath1o ] . ratios for the ysos are derived from the best - constrained ( high- or single - t ) values from the rotational analyses ( table [ ratio_table ] ) . the solid and dashed lines show mass - independent fractionation ( slope @xmath375 ) lines with different assumptions for the @xmath17o/@xmath18o ratio , as indicated . the mass - dependent line ( slope @xmath376 ) is also shown for comparison . ellipses represent 65% confidence limits ( standard deviation ) derived from @xmath332 errors from the @xmath377 for each line . targets are colored according to parent cloud ( see also figure [ gc_co_plot]).,width=321 ] c@xmath1o]/[@xmath0c@xmath18o ] ratio is shown in purple , and [ @xmath0c@xmath1o]/[@xmath0c@xmath17o ] in red . the derived [ @xmath1o]/[@xmath18o ] changes by nearly a factor of two relative to [ @xmath0c@xmath1o]/[@xmath2c@xmath1o ] , while the [ @xmath1o]/[@xmath17o ] varies by roughly 1.5 times with [ @xmath0c@xmath1o]/[@xmath2c@xmath1o ] . the arrows indicate the relative trends we would expect if isotopologue - specific photodissociation by co self - shielding were the predominant process affecting the oxygen and carbon isotope ratios.,width=321 ] models for low - mass ysos predict that [ @xmath0c@xmath1o]/[@xmath2c@xmath1o ] should not increase significantly due to photodissociation ( e.g. * ? ? ? * ; * ? ? ? atomic carbon liberated by co photodissociation is photoionized largely to c@xmath378 , and c@xmath378 reacts rapidly to exchange carbon with co by the reaction @xmath2c@xmath378 + @xmath379o @xmath380 @xmath0c@xmath378 + @xmath2co + 35 k @xcite , driving @xmath2c back into co and diminishing the signature of selective photodissociation . however , very high [ @xmath0c@xmath1o]/[@xmath2c@xmath1o ] ratios derived for diffuse clouds have nevertheless been attributed to co self - shielding . values of @xmath336 , @xmath381 @xcite , and @xmath382 @xcite have been observed toward diffuse ophiuchus clouds that are thought to be subjected to high - uv fields , and where co photodissociation is thought to dominate over the carbon exchange reaction that would otherwise erase a photochemical signature . while the upper range in [ @xmath0c@xmath1o]/[@xmath2c@xmath1o ] in our sample is consistent with the ophiuchus data ( figure [ gc_co_plot ] ) , we do not find the expected mass - independent signatures in oxygen isotopes consistent with co self - shielding . further , as compared to diffuse clouds , we would not expect the molecular gas deep inside embedded yso envelopes to be subjected to such high uv fields given their relatively high values of visual extinction ( @xmath383 ( e.g. * ? ? ? * ) ) . heterogenous [ @xmath0c@xmath1o]/[@xmath2c@xmath1o ] gas - phase abundance ratios , with several unusually high values , have been found using infrared absorption spectroscopy at very high resolution toward a sample of ysos in nearby star - forming clouds . a possible partial explanation for this trend may be co isotopologue partitioning between the ice and gas reservoirs . the degree of partitioning and the preservation of the relation in warm regions where the co ice has sublimated would require the action of a process that can remove co from the ice phase without returning it to the gas . additional high - quality data could help substantiate these findings . signatures consistent with isotopologue - selective photodissociation by co self - shielding in oxygen were observed in the oxygen isotopes toward the disks , hl tau and vv cra . however , our results do not support co self - shielding as a dominant mechanism for explaining the trends in the gas - phase [ @xmath0c@xmath1o]/[@xmath2c@xmath1o ] heterogeneity . we conclude that chemical evolution models for young stellar environments should consider the interaction between co ice and gas reservoirs as a potential fractionation mechanism affecting the carbon isotope budget and influencing carbon isotopic heterogeneity in the gas . interactions between these reservoirs could also contribute to isotopic heterogeneity between solar - type ysos and their parent clouds . finally , isotopologue - selective photodissociation by co self - shielding should be considered an important mechanism affecting oxygen isotopes in protoplanetary disks , potentially persisting in the later stages of planet formation . coplen , t. b. , bhlke , j. k. , de bivre , p. , ding , t. , holden , n. e. , hopple , j. a. , krouse , h. r. , lamberty , a. , peiser , h. s. , rvsz , k. , rieder , s. e. , rosman , k. j. r. , roth , e. , taylor , p. d. p. , vocke , jr . , r. d. , & xiao y.k . 2002 , pure appl . , 74 ( 10 ) , 1987 hasegawa , t. i. , herbst , e. , & leung , c. m. 1992 , , 82 , 167 + herczeg , g. j. , brown , j. m. , van dishoeck , e. f. , & pontoppidan , k. m. 2011 , , 533 , a112 + hidaka , h. , watanabe , n. , shiraki , t. , nagaoka , a. , & kouchi , a. 2004 , , 614 , 1124 van dishoeck , e. f. , helmich , f. p. , de graauw , t. , black , j. h. , boogert , a. c. a. , ehrenfreund , p. , gerakines , p. a. , lacy , j. h. , millar , t. j. , schutte , w. a. , tielens , a. g. g. m. , whittet , d. c. b. , boxhoorn , d. r. , kester , d. j. m. , leech , k. , roelfsema , p. r. , salama a. , & vandenbussche , b. 1996 , , 315 , l349

we present a new method to remove the impact of random and small - scale non - circular motions from hi velocity fields in ( dwarf ) galaxies in order to better constrain the dark matter properties for these objects . this method extracts the circularly rotating velocity components from the hi data cube and condenses them into a so - called bulk velocity field . we derive high - resolution ( kpc ) rotation curves of ic 2574 and ngc 2366 based on bulk velocity fields derived from the hi nearby galaxy survey ( things ) obtained at the vla . we compare the bulk velocity field rotation curves with those derived from the traditional intensity - weighted mean velocity fields and find significant differences . the bulk velocity field rotation curves are significantly less affected by non - circular motions and constrain the dark matter distribution in our galaxies , allowing us to address the discrepancy between the inferred and predicted dark matter distribution in galaxies ( the `` cusp / core '' problem ) . _ spitzer _ infrared nearby galaxies survey ( sings ) 3.6 m data , which are largely unaffected by dust in these systems , as well as ancillary optical information , are used to separate the contribution of the baryons from the total matter content . using stellar population synthesis models , assuming various sets of metallicity and star formation histories , we compute stellar mass - to - light ratios for the 3.6 m and 4.5 m bands . using our predicted value for the 3.6 m stellar mass - to - light ratio , we find that the observed dark matter distributions of ic 2574 and ngc 2366 are inconsistent with the cuspy dark matter halo predicted by cold dark matter models , even after corrections for non - circular motions . this result also holds for other assumptions about the stellar mass - to - light ratio . the distribution of dark matter within our sample galaxies is best described by models with a kpc - sized constant - density core . # 1 # 1#1

0810.2119

cosmological cold dark matter ( cdm ) simulations have been very successful in describing the observed large - scale structures in the universe ( spergel et al . 2003 ; primack 2003 ) . they have , however , been less successful in describing the observed dark matter density profiles of galaxies at small radii . the most commonly used models ( navarro , frenk & white 1996 , hereafter nfw ; see also moore et al . 1998 ) predict that the dark matter density profile increases towards the center as a power law @xmath3 with @xmath4 to @xmath5 , giving rise to a `` cusp '' feature in the centers of galaxies . however , most observations do not confirm this nfw profile , instead preferring a sizeable central constant density - core with @xmath6 ( de blok et al . 2001 ; de blok & bosma 2002 ) . the `` cusp / core '' problem is one aspect of the small - scale crisis in @xmath2 cold dark matter ( @xmath2cdm ) cosmology ; the other two being the missing dwarf galaxies and the angular momentum problem ( klypin et al . 1999 ; moore et al . 1999 ; navarro et al . 1995 ) . considerable research has been devoted to this problem , and the apparent inability of standard @xmath2cdm simulations to produce dark matter density profiles that match the observed profiles is now well - known ( flores & primack 1994 ; moore 1994 ; de blok et al . 2001 ; de blok & bosma 2002 ; weldrake et al . 2003 ; simon et al . 2003 ; gentile et al . 2004 ; see also swaters et al . 2003 ) . it has been argued that several observational systematic effects , such as beam smearing , pointing offsets , and non - circular motions , could affect the measured inner slope of the density profile , and may therefore `` hide '' the signatures of cusps in the central parts of galaxies ( van den bosch et al . 2000 ; swaters et al . 2003 ; hayashi & navarro 2006 ) . the early rotation curves used to study the dark matter density profiles in low surface brightness ( lsb ) galaxies were mainly based on hi data with large beam - sizes , and beam - smearing could potentially have flattened the observed slopes of the rotation curves . however , high - resolution h@xmath7 rotation curves show consistency with the hi rotation curves ( de blok et al . 2001 ; mcgaugh et al . 2001 ; kuzio de narray et al . 2006 ) . from this , it seems that beam smearing effects are not significant enough to erase the signature of the cusp . when using one - dimensional long - slit spectra , potential cusps may also be `` hidden '' by telescope pointing offsets ( i.e. , slit offsets ) with respect to the galaxy centers . however , de blok et al . ( 2003 ) show that rotation curves observed by independent observers on different telescopes agree within the given errors and that pointing offsets are unlikely to exceed @xmath8 ( see also marchesini et al . 2002 ; de blok & bosma 2002 ; de blok 2004 ; gentile et al . 2007 ) . the insignificance of beam smearing and pointing offsets is also confirmed by high - resolution 2d optical velocity fields of lsb galaxies ( kuzio de narray et al . 2006 ) . one of the fundamental assumptions of most rotation curve studies is that the tracers that are typically used for galaxy dynamics , such as hi or h@xmath7 , travel on circular orbits . therefore , any significant non - circularity or random motions will affect the results derived by such studies . non - circular motions are not only thought to be due to star formation processes , but also due to bars , spiral density waves and non - circular halo potentials . for one dwarf lsb galaxy ( ddo 47 ) , gentile et al . ( 2005 ) have quantified the non - circular motions and found them to be @xmath9 @xmath10 , i.e. , too small by about an order of magnitude to explain the @xmath2cdm discrepancy . an analysis of non - circular motions in 19 things galaxies ( trachternach et al . 2008 ) also finds similar results : especially in the low - luminosity galaxies , non - circular motions are too small to explain the observed discrepancies . a different problem is that in order to isolate the dynamical contribution of the dark matter , the mass distribution of the baryons ( gas and stars ) needs to be determined first . for the neutral gas this can be directly derived from the integrated hi map , but for the stars this is less trivial . it requires knowledge of the stellar mass - to - light ratio ( hereafter @xmath11 ) which depends on several factors that are not well constrained , such as the amount of extinction from dust , the star formation history , and the stellar initial mass function . moreover , these factors are interdependent , making @xmath11 one of the parameters in galaxy mass modeling with the largest uncertainty . therefore , in many studies , minimum and maximum disk assumptions are often used ( van albada & sancisi 1986 ) : the minimum disk assumes that the observed rotation curve is entirely due to dark matter and thus gives a hard upper limit to the dark matter properties in galaxies . the maximum disk hypothesis maximizes the rotation contribution of the stellar disk and thus provides an upper limit on @xmath11 or , equivalently , a lower limit on the contribution of dark matter . however , these assumptions , though they provide useful limits , are not able to determine the exact amount of dark matter in galaxies . in this paper , we investigate the distribution of dark matter in two nearby dwarf galaxies , ic 2574 and ngc 2366 , using observations obtained as part of the hi nearby galaxy survey ( things ; walter et al . dwarf galaxies are dark matter dominated and the highly resolved things observations provide an opportunity to trace in detail the overall dynamics and constrain the dark matter distribution . the distorted velocity contours in ic 2574 and ngc 2366 indicate the presence of non - circular motions ( see walter et al . 2008 ; de blok et al . 2008 ; trachternach et al . a key element of our analysis is a new approach based on a gaussian fit algorithm to decompose the hi profiles . this method reduces and quantifies the effects of random non - circular motions or localized distortions due to , for example , hi shells and thus helps to extract the velocity component representing the underlying `` undisturbed '' kinematics . in this paper , we only focus on the small - scale non - circular motions due to , e.g. , star formation events and which are not related to large - scale features such as non - axisymmetric shapes of potentials etc . ( see trachternach et al . 2008 for a more extensive discussion on these large - scale non - circular motions ) . the adopted distances of ic 2574 and ngc 2366 are 4.0 kpc and 3.4 kpc , respectively ( walter et al . 2008 ) . we also constrain the mass of the stellar component of ic 2574 by using _ spitzer _ infrared array camera ( irac ) 3.6 @xmath1 m data from the _ spitzer _ infrared nearby galaxies survey ( sings ; kennicutt et al . ngc 2366 is not part of the sings sample and for this galaxy we retrieved 3.6 @xmath1 m data from the _ spitzer _ archives . we combine these data with population synthesis models derived using the `` galaxev '' package of bruzual & charlot ( 2003 ) . we compare mass models derived using our best 3.6 @xmath1 m @xmath11 values with models derived under different assumptions , such as maximum or minimum disk . the organization of this paper is as follows : in section 2 , we give a general description of data used in this paper . we derive the hi rotation curves of ic 2574 and ngc 2366 using a new gaussian decomposition method in section 3 . we present our determinations of the masses of the stellar components of ic 2574 and ngc 2366 in section 4 . the derived mass models of ic 2574 and ngc 2366 are presented in section 5 and fitted using a combination of halo models and @xmath11 assumptions . in section 6 , we measure the inner slopes of the mass density profiles of ic 2574 and ngc 2366 and the main results of this paper are summarized in section 7 . the hi nearby galaxy survey ( things ) is one of the largest h i survey programs undertaken with the nrao very large array ( vla ) and comprises observations of 34 nearby galaxies . it has high spatial ( @xmath12@xmath13 ) and spectral ( @xmath14 5.2 @xmath10 ) resolutions ( walter et al . in our analysis we use the natural - weighted cubes . in order to preserve the noise characteristics of the data , no residual scaling , primary beam correction or blanking was applied . the channel separation in the two galaxies discussed here , @xmath15 , is 2.6 @xmath10 . to constrain the contribution of the stellar component to the total kinematics we use 3.6 @xmath1 m data from the _ spitzer _ infrared nearby galaxies survey ( sings ) ( kennicutt et al . 2003 ) for ic 2574 and from the _ spitzer _ archives for ngc 2366 . these can be used as a proxy for the distribution of the stellar population . the resolution of @xmath16 of the 3.6 @xmath1 m images is comparable to the things resolution . in addition , we use ancillary optical @xmath17 , @xmath18 , and @xmath19 images taken with the 2.1 m telescope at kitt peak national observatory ( kpno ) as part of the sings survey . we show the data for ic 2574 and ngc 2366 in fig there are several ways to define a velocity field . these include first - moment maps , single gaussian , multiple gaussian , or hermite polynomial fits , as well as peak velocity fields ( see de blok et al.2008 for a detailed discussion ) . some , such as first - moment maps , gaussian fits , and peak velocity fields , work well for single , symmetrical hi profiles , but fail to properly take into account the presence of multiple velocity components . in the upper - left panels of fig . 2 and 3 we show the intensity - weighted mean first - moment ( hereafter iwm ) velocity fields of ic 2574 and ngc 2366 ( the `` standard '' velocity fields ) . from the distorted contours in these velocity fields , it is clear that there are strong non - circular motions in both galaxies . it is also clear that the iwm velocity fields are not optimal for deriving a rotation curve and/or correcting for the non - circular motions . one can try and capture these non - circular motions by fitting multiple gaussian components to the profiles . however , even though it is possible to model a non - gaussian profile using multiple gaussian components , the standard method often implemented in astronomical software packages will frequently fail as it tries to fit these components simultaneously . this usually requires too many free parameters ( i.e. , amplitude , dispersion , and central position for each component used ) , quickly exceeding the typical number of data points in an hi profile . in addition , the presence of noise can result in the fit becoming sensitive to the values of the initial estimates . however , even if profiles are decomposed perfectly , we are still left with the fundamental problem that a decomposition into gaussian components does not give much insight as to which component is more representative of underlying circular rotation of the disk and which one represents non - circular motion or additional velocity components along the line of sight . in order to address this problem and extract the undisturbed `` bulk motion '' ( i.e. , the velocity most representative of the undisturbed rotation ) from the hi data cube of a given galaxy , we devised an alternative gaussian decomposition method . this method minimizes the effects of localized non - circular motions ( such as those caused by star formation processes ) and extracts the circularly rotating components from the hi data cube . the method consists of the following steps . as an initial step , an approximate rotation curve is derived from a major axis position - velocity diagram . for this we need to derive the position angle ( pa ) of the major axis , but also the center coordinates ( xpos , ypos ) , inclination ( incl ) and systemic velocity ( vsys ) . initial estimates for xpos , ypos , pa , and incl can be derived from ellipse fits to isophotes in the hi surface density - map ( @xmath20 moment map ) , optical , or infrared images . this gives approximate but reasonable initial values . in this paper , we used ellipse fits to the 3.6 @xmath1 m images of ic 2574 and ngc 2366 at a level of 0.03 mjy sr@xmath21 to determine these initial values . we show the ellipse fits to the 3.6 @xmath1 m images for ic 2574 and ngc 2366 in the middle panels of fig . an estimate for vsys is determined by averaging the velocity values found near the center position . using the obtained geometric parameters , we perform single gaussian fits to the profiles found along the major axis in the hi data cube , and extract the rotation velocity ( vrot ) . if multiple peaks with different velocities exist at a given spatial position along the major axis , we select one of them as the initial estimate for the bulk motion , as determined by the average velocity at neighboring positions . we emphasize that at this stage we are only determining a first approximation to the rotation velocity and any errors introduced by an incorrect selection of a velocity component will be corrected in subsequent steps . we now create an artificial velocity field using the previously obtained geometrical parameters . as we are at this stage only interested in the overall shape of the rotation curve , we approximate the values of vrot obtained above with a fifth - order polynomial . the model velocity field will be used as initial estimates for extracting the bulk motions . the model fields for ic 2574 and ngc 2366 are shown in the top - right panels of figs . 2 and 3 , respectively . again , we stress that these model velocity fields are just first approximations . they are merely used to set the initial values for carrying out the multiple gaussian decomposition described in the following step . we now perform single gaussian fits to extract the typical velocities of all profiles in the data cube , taking into account the presence of possible secondary components . for this , we developed a program written in c. as is illustrated in the middle panel of fig . 4 , if a profile consists of two gaussian components , the primary ( i.e. , strongest intensity ) gaussian component is extracted as shown , taking into account the effect of the secondary component on the fit . to do this , the program only uses data points that are least affected by the secondary component ( as shown by the filled circles in the middle panel of fig . 4 ) . we can judge the effect of the secondary component by making gaussian fits to both halves of the primary gaussian component separately ( as defined with respect to the position of the peak flux ) . the fit to the side containing the secondary component will result in a larger dispersion compared to the side with only a single component . this difference can be used to indicate the approximate location of the secondary component in a profile . in the last step , we assess whether the extracted velocities of the profiles in @xmath22 are acceptable as representative of the underlying circular rotation of the disk . for this , we compare those velocities with the model velocity fields derived in @xmath23 . if at some position the velocity difference ( @xmath24 ) is less than a prescribed limit ( here we use @xmath25 ) , we select the velocity derived in @xmath26 for the primary component as the velocity of the bulk motion at that position . however , if the difference is larger , this suggests that a better velocity can be found by considering a secondary component , and we proceed as follows . the primary gaussian component of the profile in @xmath26 is subtracted from the original profile and subsequently a single gaussian fit is made to the residual profile , yielding a second gaussian component . this is illustrated in the right - most panel of fig . 4 . when performing this second gaussian fit , the model velocity field created in @xmath27 is used to determine an initial fit value . we compare the velocity derived for the secondary component with the model velocity field derived . likewise , we choose this velocity as the bulk velocity at this position , if the velocity difference @xmath24 is less than a certain limit ( where we again use @xmath25 ) . in cases where the velocity difference is larger , and no satisfactory primary or secondary component is found , we simply put a blank value in that position . we also note that the extracted fits that satisfy our @xmath28 criterion , are only defined as the `` bulk velocity '' component if they have a significant flux . we demand a peak flux @xmath29 , where @xmath30 is the @xmath31 noise in the cube . we thus substitute a blank value into the bulk velocity field at positions where the extracted velocity shows strong deviations or where no significant emission can be found . this means that our method does not `` invent '' data where there are none . using these procedures we have now extracted the first approximate bulk velocity field from the hi data cube . we now repeat the entire process and go back to @xmath32 . however , rather than using the major axis estimates , we now fit a full `` tilted ring model '' ( begeman 1989 ) to the extracted bulk velocity field obtained in @xmath33 and derive a second approximation to the rotation curve parameters ( xpos , ypos , vsys , pa , incl , and vrot ) . using these newly determined tilted ring parameters , we construct a second , improved model velocity field for the bulk motion , which is less affected by additional components along the line of sight , and then proceed to @xmath34 , @xmath35 , and @xmath36 . in this way , we iterate the above procedures until the mean difference ( after one round of @xmath37 outlier rejection ) between successive velocity fields is less than @xmath25 . we have tested this method using different initial conditions ( model velocity fields ) and find that the result is independent of these conditions . empirically it was determined that three or four iterations lead to a stable result . the procedure is summarized in fig . the bulk velocity method works particularly well for galaxies in which non - circular motions form a significant part of the rotation velocity . for galaxies with high rotation velocities , the method produces results very similar to the hermite polynomial method for constructing velocity fields described in de blok et al . ( 2008 ) . we applied this method to ic 2574 and ngc 2366 . the final extracted bulk velocity fields are shown in the bottom - left panels of figs . 2 and 3 , respectively . most disturbances seen in the iwm velocity fields of both galaxies are removed , especially in the central regions of ic 2574 and the north - western part of ngc 2366 . in general a warp significantly changes the position angle of the kinematical major axis as a function of radius and is found in the outer parts of a galaxy ( bosma 1978 ) . no significant changes of the derived position angles , as will be shown in section 3.3 , of the tilted ring models of ic 2574 and ngc 2366 are shown in regions where most disturbances are seen . therefore , these disturbances are not likely to be from a warp . as a bonus , we also obtain the velocity fields of the strong non - circular motions , which are shown in the bottom - right panels of figs . 2 and 3 . these contain the velocities of the primary components at the positions where these primary components were found to track the non - circular motions , i.e. , this is not a residual velocity field . in order to visualize how well the extracted bulk velocity field traces the rotation of the gas , we use azimuthal position - velocity diagrams measured along ellipses defined using the tilted ring models . in a quiescent galaxy with only circular motions , one expects the observed velocities to follow a cosine as a function of azimuth . any regions containing non - circular motions deviate from the cosine curves and we can therefore use the position - velocity diagrams to assess whether the bulk - motion extraction method is able to properly decompose the multiple profiles and identify the bulk motion component properly ( see stil & israel 2002 for an analysis which uses these azimuthal position - velocity diagrams to quantify non - circular motions ) . in figs . 6 and 7 , we show the azimuthal position - velocity diagrams of ic 2574 and ngc 2366 , respectively . in fig . 6 , strong non - circular motions in ic 2574 are easily identified at @xmath38 at most radii . for example , in the region at @xmath39 and @xmath38 , two clearly separated profiles are visible . this is where the iwm velocities are strongly affected by non - circular motions . these non - circular motions are caused by the effects of the supergiant shell in ic 2574 ( discussed in walter et al . in contrast , the bulk velocities extracted by our method follow a sinusoidal line tracing the underlying bulk motion of ic 2574 . similarly , for ngc 2366 in fig . 7 we can see the presence of strong non - circular motions at @xmath40 at most radii . the iwm velocities , affected by non - circular motions , display a non - symmetric shape in the position - velocity diagram in regions with @xmath41 and @xmath42 . the extracted bulk velocities again have a sinusoidal - shape , indicating that the bulk velocity field for ngc 2366 indeed traces the underlying rotation . we derived the rotation curves of ic 2574 and ngc 2366 using the bulk velocity fields obtained in section 3.2 . in order to quantify the difference with the iwm velocity field , we also derive the rotation curves from the latter velocity fields . we used the gipsy task rotcur to make tilted ring fits to the bulk and iwm velocity fields . the initial estimates of the ring parameters , such as xpos , ypos , pa , and incl were determined by an ellipse fit to the irac 3.6 @xmath1 m images for ic 2574 and ngc 2366 ( except vsys which was derived from the hi global profile ) . after fixing the position of the dynamical center ( xpos , ypos ) and vsys , we run rotcur with pa and incl as free parameters . we made low - order polynomial fits ( usually of fifth order ) to the pa and incl distributions to describe the large - scale variations with radius . we fixed pa and incl to these fit values and ran rotcur again with xpos , ypos , and vsys as free parameters . from this , we radially average the xpos , ypos , and vsys and obtain more fine - tuned mean values . these values were fixed again and rotcur was run once more leaving the pa and incl free . we iterated the rotcur task in this way until all ring parameters converged . for the last step , we fixed all ring parameters except vrot and derived the final rotation curves of ic 2574 and ngc 2366 . the derived rotation curves are shown in figs . 8 and 9 and the tilted ring parameters are summarized in table 1 . in the case of ic 2574 we found that the maximum difference between rotation velocities as derived from the iwm and bulk velocity fields is about @xmath1214 @xmath10 at a radius of around 7 kpc , as shown in fig . 10 and fig . 11 and this is a significant difference . note also the decreased scatter in the tilted ring parameters , compared to the iwm tilted ring model presented in fig . 10 . this can be considered as circumstantial evidence that the bulk velocity field is less affected by non - circular motions and therefore traces the underlying rotation of ic 2574 more accurately than the iwm velocity field . we note the large difference between the bulk velocity field rotation curve and the rotation curve derived by martimbeau et al . ( 1994 ) in an earlier study ( fig . martimbeau et al . ( 1994 ) used an iwm velocity field and adopted a large value for the inclination ( @xmath12@xmath43 ) , both of which are responsible for this difference . we compare the rotation curve of ngc 2366 derived from the bulk velocity field with the one derived from the iwm velocity field in fig . the derived geometrical tilted ring parameters from both bulk and iwm velocity fields are consistent with each other , except for small differences in the position angle . we do find a significant velocity difference in the outer parts . larger non - circular motions thus exist in the outer parts of ngc 2366 confirming the impression given by the bottom - right panel of fig.3 . 13 compares our bulk rotation curve of ngc 2366 with previous determinations by hunter et al . ( 2001 ) and swaters ( 1999 ) . the latter curve agrees with the current data out to 5.5 kpc . the hunter et al . ( 2001 ) curve is systematically lower than our bulk curve . a partial explanation is that they used a higher inclination of @xmath44 , but this difference is not large enough to explain the entire discrepancy . though hunter et al . ( 2001 ) do not explicitly state how their velocity field was constructed , it is likely that they used an iwm map , rather than one based on profile fits . declining rotation curves at the outer parts of a galaxy have been often considered as an indication that the end of the dark matter distribution is reached ( carignan & puche 1990 ; casertano & van gorkom 1991 ; ryder et al . 1997 ; carignan & purton 1998 ) . however , as shown in fig . 13 , the possibility that a declining rotation curve at the outer parts of a galaxy is simply mimicked by non - circular motions can not be ruled out . the comparisons between iwm and bulk velocity results given above have to some degree been qualitative . to really quantify the presence of non - circular motions in both types of velocity fields and gauge the effectiveness of the bulk velocity method in removing these effects , we show here harmonic decompositions of the iwm and bulk velocity fields of ic 2574 . we refer to trachternach et al . ( 2008 ) for an extensive discussion of the method . in summary , we used the gipsy task reswri to decompose the velocity fields into sine and cosine terms and only kept the center position fixed during the harmonic expansion . we only included terms up to third order . the line - of - sight velocity then has the following form : @xmath45 + { \rm resid.},\ ] ] where @xmath46 represents the systemic velocity , and @xmath47 the rotation velocity , the @xmath48 , and @xmath49 components quantify non - circular motions . for us , the total amplitudes of the non - circular motions are of most interest . the median absolute amplitudes @xmath50 of each component were calculated by taking the median of @xmath51 , where @xmath52 for @xmath53 , and @xmath54 for @xmath55 ( @xmath47 is the rotation velocity ) . as can be seen in the top - left panel of fig . 14 , the radially averaged , median amplitudes for the @xmath56 , and @xmath57 components are small for both velocity fields . for the second order ( @xmath58 ) , however , the harmonic decomposition of the bulk velocity field results in a much smaller amplitude than that of the iwm velocity field . the effect of the bulk velocity method can be seen even more clearly in the top - center panel of fig . , we plot the absolute amplitude of the non - circular motions using the following equation : @xmath59 the harmonic components found in the iwm velocity field show large amplitudes between @xmath60 , whereas the results from the bulk velocity field are consistent with regular circular rotation . the rotation curves obtained from the bulk velocity field reflect the total ( baryonic + dark matter ) kinematics of galaxies and in order to say anything about the dark matter , we thus need to quantify the contribution by the baryons . to derive the contribution of the gas component to the total kinematics the radial hi surface density distribution is required . this is obtained from the integrated hi map ( natural - weighted ) ( walter et al . 2008 ) using the derived tilted ring model parameters . the h i radial surface density is then used to derive the corresponding contribution to the rotation velocity after correcting the surface density for the presence of helium and metals ( i.e. , after scaling by a factor of 1.4 ; de blok et al . we use the task rotmod implemented in gipsy and assume an infinitely thin disk . the rotation velocities for the gas component of ic 2574 and ngc 2366 are presented in the bottom - left panels of figs . 19 and 20 , respectively . after removing bright foreground stars in the vicinity of ic 2574 and ngc 2366 manually , the 3.6 @xmath1 m surface brightness profiles are determined using the ellint task in gipsy and the tilted ring fit parameters derived in section 3.3 . we used the standard irac calibration which provides flux values in units of mjy sr@xmath21 . to convert this to surface brightness units of @xmath61 we use : @xmath62,\ ] ] where @xmath63 is the flux value of the 3.6 @xmath1 m band in units of mjy sr@xmath21 . @xmath64 is the irac zero magnitude flux density in jy and has 280.9 ( reach et al . the final surface brightness profiles for ic 2574 and ngc 2366 in the 3.6 @xmath1 m band are shown in the top - left panel of fig . 15 and fig . 16 , respectively . in general , the @xmath11 values at optical wavelengths are affected by many factors , including dust , age , metallicity , imf , and recent star formation . these give rise to large uncertainties in the inferred @xmath11values with corresponding uncertainties in dark matter halo parameters . to circumvent this problem usually some assumptions for the value of @xmath11 , such as minimum disk or maximum disk are used when performing disk - halo decompositions ( van albada & sancisi 1986 ; see discussion in de blok et al . however , the derived dark matter halo properties derived using these minimum / maximum disk assumptions are , as discussed above , only upper / lower limits of the properties on the dark matter halo . optical colors of the disk are also often used to put further constraints on @xmath11 . a relation between optical colors ( e.g. , @xmath65 , @xmath66 ) and @xmath11 has been found in earlier work ( e.g. , bell & de jong 2001 ) . however these previous studies do not provide the @xmath11 values for the irac bands , so we can not use them here . we have therefore calculated the @xmath11 values in the 3.6 @xmath1 m and 4.5 @xmath1 m bands . for this calculation we constructed stellar population synthesis models , with various sets of metallicity and star formation histories using the `` galaxev '' package of bruzual & charlot ( 2003 ) . assuming an age of 12 gyrs ( as found for the local group in whiting 1999 ) , we find a well - defined relation between @xmath67 and @xmath11 in the 3.6 @xmath1 m and 4.5 @xmath1 m bands as shown in fig . these can be parameterized as follows , @xmath68 for the @xmath69 band and , @xmath70 for the @xmath71 @xmath1 m band with coefficients , @xmath72 , @xmath73 , @xmath74 , and @xmath75 as given in table 2 . bell & de jong ( 2001 ; their table 4 ) give the relations between the @xmath67 and optical colors and we have : @xmath76 where @xmath77 and @xmath78 are also given in table 2 for the appropriate colors . combining eq . 6 with eq . 8 , adopting 20% solar metallicity ( miller & hodge 1996 ) and a scaled salpeter imf cutting off the stars less massive than @xmath79 @xmath80(bell & de jong 2001 ) , we calculated @xmath11 for ic 2574 in the 3.6 @xmath1 m band . as mentioned in section 2 , we use the ancillary optical @xmath81 , @xmath82 and @xmath83 images to derive the radial color distribution of ic 2574 . in the outer parts of the galaxy , we extrapolate the surface brightness profile using exponential fits . from these fits we extract optical colors ( @xmath84 and @xmath66 ) of ic 2574 as shown in the top - right panel of fig . the colors become bluer with increasing radius , consistent with earlier results found for spiral galaxies ( de jong 1996 ) . in general , these color gradients can be explained by different star formation histories and thus different present - day stellar populations , with relatively older populations in the inner parts than in the outer parts . this population change therefore also implies radial changes in @xmath11 in galaxies . in order to reflect this @xmath11 variation , we take the color gradient into account when determining @xmath11 . the resulting trends are shown in the bottom - left panel of fig . 15 . to determine our final @xmath11values , we simply average the @xmath67 values as derived from @xmath65 and @xmath66 , respectively . the average @xmath85 used for the final mass model of the stellar component of ic 2574 is shown in the bottom - right panel of fig . 15 ( gray solid line ) . similarly , we calculated @xmath11 values in the 3.6 @xmath1 m band for ngc 2366 using eqs . 6 and 8 . for ngc 2366 , we used a 10% solar metallicity ( hunter et al . 2006 ) and a scaled salpeter imf . we used the optical color ( @xmath66 ) of ngc 2366 given in hunter et al . ( 2006 ) and found the coefficients of @xmath77 and @xmath78 as shown in table 2 . we used a constant average color , ( @xmath66)=0.31 , since the radial @xmath66 color distribution given in hunter et al . ( 2006 ) is nearly constant ( except where the super - giant hii region of ngc 2363 is located at a radius @xmath121 kpc ) . from this we obtained @xmath85=0.33 for ngc 2366 as given in table 2 . see also de blok et al . ( 2008 ) for a comparison between the stellar disk masses of a number of things galaxies derived using our method and the approach adopted by leroy et al . ( 2008 ) . the 3.6 @xmath1 m emission may contain contributions from pah features but , in the case of the dwarfs , the contribution of pahs is not likely to be very high ( see walter et al . 2007 for a study of the sings dwarfs ) . the 3.6 @xmath1 m emission is likely to be a good proxy for the old stellar population even if there can be contributions from the intermediate - age population like agb stars . although the 3.6 @xmath1 m images thus provide a virtually dust - free picture of the stellar component of a galaxy , the optical colors that are implicitly used to determine @xmath85 are possibly affected by dust . however , bell & de jong ( 2001 ) show that @xmath67 is only weakly dependent on optical color , in contrast with @xmath86 . this is illustrated by the low values of @xmath78 in the ic 2574 @xmath67 relation ( @xmath87 for @xmath66 and @xmath88 for @xmath65 ) , especially when compared with the much higher @xmath89 values listed in bell & de jong ( 2001 ) . the precise value of @xmath85 is thus not very sensitive to variations in @xmath66 or @xmath84 . the 3.6 @xmath1 m surface brightness profiles derived in section 4.1 are in units of @xmath61 , and must , using the values of @xmath11 derived in section 4.2 , still be converted to a mass density profile in units of @xmath90 . the conversion to mass surface density is calculated with the following formula : @xmath91 = { \ensuremath{\upsilon_{\star}^{3.6}}}\times 10^{-0.4\times(\mu_{3.6\mu\rm m } - c^{3.6})},\ ] ] where @xmath92 is a constant for converting @xmath61 to @xmath93 . we have : @xmath94 where @xmath95 is the absolute solar magnitude in the 3.6 @xmath1 m band . to calculate this , we perform the following steps . + + * i. the apparent magnitude of the sun * + + the apparent magnitude of the sun , @xmath96 , in the 3.6 @xmath1 m or 4.5 @xmath1 m band is given by : @xmath97 + 2.5\log_{10}\biggl[\int_{0}^{\infty}{\rm d}\lambda r(\lambda)f_{\alpha\,\rm lyr}^{\lambda}\biggr],\ ] ] where @xmath98 is the filter response function and @xmath99 and @xmath100 are the spectral energy distributions of the sun and @xmath7 lyr respectively ( fukugita et al . 1995 ) . the kurucz model ( 1992 ) for @xmath7 lyr has @xmath101=9400k , @xmath102=3.90 , and @xmath103 . the convolved spectral energy distributions of the sun and @xmath7 lyr using the irac filter response functions in the 3.6 @xmath1 m and 4.5 @xmath1 m bands are shown in fig . we compute the areas below these convolved spectral energy distributions of the sun and @xmath7 lyr . 11 , we then obtain the apparent magnitude of the sun as @xmath104 and @xmath105 in the 3.6 @xmath1 m and 4.5 @xmath1 m bands , respectively . + + * ii . the absolute magnitude of the sun in the irac bands * + + we obtain the absolute magnitude of the sun as follows , @xmath106 where @xmath107 pc ( distance to the sun ) . from this , we find @xmath108 @xmath109 from eq . 10 , @xmath110 , and using eq . 9 we then convert the 3.6 @xmath1 m surface brightness to a mass density in units of @xmath90 . the final mass models for ic 2574 and ngc 2366 using the mass density for the stellar component are constructed with the rotmod task in gipsy . these are illustrated in fig . 19 and fig . 20 . for the stellar disk , we assume a vertical @xmath111 scale - height distribution with @xmath112 ( van der kruit & searle 1981 ) where @xmath113 is the radial scale - length and @xmath114 is the scale - height of disk in kpc . this @xmath112 ratio as determined in van der kruit & searle ( 1981 ) is based on a small number of galaxies but has , for a much larger sample , been confirmed by kregel et al . ( 2002 ) , who find @xmath115 . from exponential disk fits to the 3.6 @xmath1 m surface brightness distributions ( top - left panel of fig . 15 and fig . 16 ) we obtain scale - height values of ic 2574 and ngc 2366 of @xmath116 and 0.34 kpc , respectively . the properties of the dark matter halo of a galaxy are usually quantified by using dark matter halo models . the residuals obtained by subtracting from the derived rotation curves those contributions corresponding to the stellar and gas components , are assumed to be due to the dark matter halo . in this paper , we explore two models : the @xmath2cdmnfw cusp - dominated halo and the pseudo - isothermal core - dominated halo . their properties are given below . navarro , frenk & white ( 1996 , 1997 ; nfw ) give a prescription for the dark matter distribution found in numerical simulations , based on the cdmparadigm . this so - called `` universal density profile '' has a cusp feature towards the galaxy center . the profile has the form @xmath117 where @xmath118 is the initial density of the universe at the time of collapse of the halo and @xmath119 is the characteristic radius of the dark matter halo . the rotation velocity corresponding to the nfw halo density is given as @xmath120}},\ ] ] where @xmath121 is the concentration parameter and defined as @xmath122 . @xmath123 is the rotation velocity at radius @xmath124 where the density contrast exceeds 200 and @xmath125 is defined as @xmath126 . this universal density profile can be approximated using two power laws : @xmath127 at small radii and @xmath128 at large radii . see de blok et al . ( 2008 ) for a more extensive description . the spherical pseudo - isothermal halo model which is used in most of the early rotation curve studies is observationally motivated and has a core - like constant density profile which can be described as @xmath129 towards the galaxy center and @xmath130 for large @xmath19 . it has the following form , @xmath131 where @xmath132 and @xmath133 are the core - density and core - radius of the dark matter halo , respectively . this density profile gives the rotation velocity , @xmath134}.\ ] ] this approaches the asymptotic velocity at large radii given by @xmath135 we now construct mass models using the rotation curves derived from the bulk velocity fields of ic 2574 and ngc 2366 taking into account the distribution of baryons , and use these to fit nfw and pseudo - isothermal halo models . given the proximity of ic 2574 and ngc 2366 , the things ( natural - weighted ) resolution of @xmath136 corresponds to @xmath12 200 pc , comparable to that of the simulations of dwarf dark matter halos presented in navarro et al . as discussed earlier , differences between nfw and pseudo - isothermal dark matter models are the most distinct in the inner parts of galaxies . also , as the tilted rings in the outer parts of our galaxies are only partly filled with emission , we estimate the effects of these outer asymmetric features by presenting fits to both the entire radial ranges of the rotation curves , as well as fits to the inner parts only ( i.e. , @xmath137 7.5 kpc for ic 2574 and @xmath137 6 kpc for ngc 2366 ) . when performing the fits we also explore classical @xmath11 assumptions , such as maximum disk , minimum disk , minimum disk + gas ( the stellar component is ignored , but the contribution of the gas is taken into account ) , in addition to the model @xmath85 values we calculated . we also attempted a fit with @xmath85 as a free parameter . the results are given in figs . 21 and 22 , and tables 3 and 4 . the best - fitting unconstrained values for @xmath11 differ somewhat from those predicted by the population synthesis modeling described earlier ( except for the ic 2574 pseudo - isothermal fit ) . the negative values of @xmath11 that the curves seem to prefer are obviously unphysical . an extensive discussion of this behavior is given in de blok et al.(2008 ) . we find that the pseudo - isothermal halo provides a better fit than the nfw halo . nfw models resulted in inferior fits , independent of which value for @xmath11 was adopted . the preferred @xmath121 values ( concentration parameter ) are all negative with extremely large uncertainties . when making the final fits , we fixed these values to be @xmath138 . see de blok et al . ( 2008 ) for a discussion on the @xmath121-values found within the things sample . in contrast , the pseudo - isothermal halo models provide reasonable halo parameters , except for the ic 2574 maximum disk case . however , we found that the fit is very sensitive to the choice of @xmath139 for the maximum disk and sensible parameters can be obtained by slightly lowering @xmath139 . these results show that the nfw distribution is not an appropriate fitting function for our galaxies and that the implied dark matter distributions of the two dwarf galaxies , ic 2574 and ngc 2366 , instead show a sizeable central constant - density core , which can be well approximated by the pseudo - isothermal halo model . we now compare the hi rotation curves and mass modeling based on the bulk velocity fields with a similar analysis done using the iwm velocity fields to examine the effect of non - circular motions on the dynamics of galaxies . first , for convenience , we introduce some notation . we use @xmath140 to indicate the rotation curve derived from tilted ring fits using the bulk velocity field , as described in section 3.3 . we use @xmath141 to describe the rotation curve derived assuming the geometrical parameters from the bulk velocity field tilted ring model but applied to the iwm velocity field . finally , we use @xmath142 to describe the rotation curve derived using a tilted ring model as derived from and applied to the iwm velocity field . for a galaxy affected by non - circular motions , the traditional iwm velocity field will be distorted . in general , non - circular motions as present in galaxies , are likely to make the observed rotation velocity fall below the circular velocity ( rhee et al . this is indeed happening in our galaxies as indicated in fig . 23 , where @xmath142 is significantly lower than @xmath140 . the velocity difference between @xmath140 and @xmath141 can only be attributed to differences in the bulk and iwm velocity fields , as we use otherwise identical geometrical parameters . in the case of ic 2574 , a distinct velocity difference is seen in the central region where the iwm velocity field is distorted by non - circular motions ( fig . 23 , left column ) . such a velocity difference is also found in ngc 2366 , particularly in the outer parts ( see the fig . 23 , right column ) . in practice , this underestimate of the rotation velocity results in a decreased dynamical contribution of the dark matter to the total dynamics , as the contribution of stellar and gas components are fixed for any given @xmath11 assumption . consequently , non - circular motions decrease the role of dark matter with respect to the visible matter . in the bottom - left panel of fig . 23 the rotation velocity due to the stellar component using the model @xmath85 values for ic 2574 already exceeds @xmath141 and @xmath142 in the inner parts even without the dynamical contribution of the gas component . the @xmath140 rotation curve , however , can in the inner parts not only accommodate the rotation of the stellar component , but also that of the gas component . an intuitive way to illustrate the dark matter distribution of a galaxy is to calculate the mass density that corresponds to the observed rotation velocity . the observed rotation velocity is converted to mass density by assuming a spherical mass distribution ( i.e. , @xmath143 = @xmath144 , @xmath145 ) . for the inversion , we use the following formula ( de blok et al . 2001 ) , @xmath146 where @xmath18 is the rotation velocity observed at radius @xmath19 . this direct conversion is only valid under the assumption that the observed rotation velocity is entirely due to the dark matter component , as is the case with a minimum disk ( i.e. , maximum halo ) . in general the minimum disk assumption is a good description for dwarf galaxies and lsb galaxies ( de blok et al . 2002 ) , and we therefore also apply this assumption to ic 2574 and ngc 2366 . while one could , in principle , make an explicit correction for the rotational contribution of the baryons , for the crucial innermost parts this involves using the difference between two small numbers , which can lead to wildly fluctuating values of the derivative in eq . 20 . the minimum disk assumption thus yields a safe and robust upper limit on the dark matter properties and in particular to the steepness of the inner slope ( de blok & mcgaugh 1997 ; de blok et al . 2001 ) . we follow the method described in de blok et al . ( 2002 ) to determine the slope of the inner component of the mass density profile . we measure the slopes of the inner parts ( @xmath147 kpc ) of ic 2574 and ngc 2366 using a least squares fit and find the values of the slopes to be @xmath148 for ic 2574 and @xmath149 for ngc 2366 , respectively . these are shown in fig . 24 and are in good agreement with the earlier result of @xmath150 ( de blok et al . 2001 , de blok & bosma 2002 ) for a larger sample of lsb galaxies . these flat slopes thus imply that the dark matter distributions of ic 2574 and ngc 2366 are well characterized by a sizeable constant - density core . the mass density profiles for the best - fit minimum disk nfw and pseudo - isothermal models are also over - plotted in fig . the pseudo - isothermal halo follows the observed mass density profile most closely . this is in sharp contrast with the steep slope predicted by the nfw profile . we plot the value of the inner - slope @xmath7 of the mass - density profile against the observed radius of the innermost point @xmath151 in fig . the result is consistent with the earlier results by de blok et al . ( 2002 ) . in conclusion , the observed dark matter distributions of ic 2574 and ngc 2366 are both best described by the pseudo - isothermal halo model with a constant - density core . we have presented mass models for the nearby dwarf galaxies , ic 2574 and ngc 2366 , derived using the high - resolution data from things . these high - resolution data do not suffer from beam smearing , have a well - defined dynamical center and enable us to examine in detail the dark matter distribution of these galaxies . to minimize the effects of random non - circular motions on the derived kinematics of a galaxy , we developed a new gaussian decomposition method and used this to construct a `` bulk '' velocity field of ic 2574 and ngc 2366 , showing the underlying undisturbed rotation . the random , non - circular motions of ic 2574 and ngc 2366 , visible as distortions in the velocity contours of the traditional iwm velocity fields , were largely removed in the newly constructed bulk velocity fields . comparing the hi rotation curves derived from the bulk and iwm velocity fields , we find that the rotation velocities derived from the iwm velocity fields are significantly lower than those from the bulk velocity fields . in addition , non - circular motions of ngc 2366 are likely to be responsible for the declining rotation velocities derived from the iwm velocity fields in the outer parts . combining optical and sings 3.6 @xmath1 m data , we quantify the dynamical contribution of the stellar component to the total kinematics . for this we compute @xmath11 values based on the bruzual and charlot ( 2003 ) population synthesis models for the 3.6 @xmath1 m and 4.5 @xmath1 m bands . we have fitted nfw and pseudo - isothermal dark matter halos to the derived rotation curves , taking into account the contributions due to stars and gas . we found that the pseudo - isothermal halo provides a better fit to the observations than the nfw halo . we use the derived mass density profile to determine the value of the inner slope . the measured slopes are @xmath148 for ic 2574 and @xmath149 for ngc 2366 , compared to the nfw model which predicts @xmath152 . the dark matter distributions of ic 2574 and ngc 2366 are well described by the pseudo - isothermal model ( @xmath153 ) with a sizeable central constant - density core . these results are not affected by systematic effects due to lack of resolution or pointing offsets , take into account the effects of non - circular motions and use a well - constrained model for @xmath11 . the work of wjgdb is based upon research supported by the south african research chairs initiative of the department of science and technology and national research foundation . eb gratefully acknowledges financial support through an eu marie curie international reintegration grant ( contract no . mirg - ct-6 - 2005 - 013556 ) . this research has made use of the nasa / ipac extragalactic database ( ned ) which is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration . 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 . begeman , k. 1989 , , 223 , 47 bell , e. f. , & de jong , r. s. 2001 , , 550 , 212 bosma , a. 1978 , phd thesis , university of groningen brinks , e. , & walter , f. 2003 , ap&ss , 284 , 2 bruzual , g. 1983 , , 273 , 105 bruzual , g. , & charlot , s. 2003 , , 344 , 1000 carignan , c. , & puche , d. 1990 , , 100 , 394 carignan , c. , & purton , c. 1998 , , 506 , 125 casertano , s. , & van gorkom , j. h. 1991 , , 101 , 1231 de blok , w. j. g. , & mcgaugh , s. s. 1997 , , 290 , 533 de blok , w. j. g. , mcgaugh , s. s. , bosma , a. , & rubin , v. c. 2001 , , 552 , 23 de blok , w. j. g. , & bosma , a. 2002 , , 385 , 816 de blok , w. j. g. , bosma , a. , & mcgaugh , s. s. 2003 , , 340 , 657 de blok , w. j. g. 2004 , dark matter in galaxies , iau symposium 220 , 69 de blok , w. j. g. , walter , f. , brinks , e. , trachternach , c. , oh , s - h . , & kennicutt , r. c. 2008 , , submitted de jong , r. s. 1996 , , 313 , 377 flores , r. a. , primack , j. r. 1994 , , 427 , l1 fukugita , m. , shimasaku , k. , & ichikawa , t. 1995 , , 107 , 945 gentile , g. , salucci , p. , klein , u. , vergani , d. , & kalberla , p. 2004 , , 351 , 903 gentile , g. , burkert , a. , salucci , p. , klein , u. , & walter , f. 2005 , , 634 , l145 gentile , g. , tonini , c. , salucci , p. 2007 , , 467 , 925 hayashi , e. , & navarro , j. f. 2006 , , 373 , 1117 hunter , d.a . , elmegreen , b.g . , van woerden , h. 2001 , , 556 , 773 hunter , d.a . , elmegreen , b. g. 2006 , apjs , 162 , 49 kennicutt , r. c. et al . 2003 , , 115 , 928 klypin , a. , kravtsov , a. v. , valenzuela , o. , prada , f. 1999 , , 522 , 82 kregel , m. , van der kruit , p. c. , de grijs , r. 2002 , , 334 , 646 kurucz , r. l. 1992 , the stellar populations of galaxies , iau symposium 149 , 225 kuzio de naray , r. , mcgaugh , s. s. , de blok , w. j. g. , bosma , a. 2006 , , 165 , 461 leroy , a. et al . 2008 , , submitted martimbeau , n. , carignan , c. , & roy , j .- 1994 , , 107 , 543 mcgaugh , s. s. , rubin , v. c. , & de blok , w. j. g. 2001 , , 122 , 2381 miller , b. w. & hodge , p. 1996 , , 458 , 467 moore , b. 1994 , nature , 370 , 620 moore , b , ghigna , s. , governato , f. , lake , g. , quinn , t. , stadel , j. , tozzi , p. 1999 , , 524 , 19 moore , b. , governato , f. , quinn , t. , stadel , j. , & lake , g. 1998 , , 499 , 5 navarro , j. f. , frenk , c. s. , & white , s. d. m. 1995 , , 275 , 56 navarro , j. f. , frenk , c. s. , & white , s. d. m. 1996 , , 462 , 563 navarro , j. f. , frenk , c. s. , & white , s. d. m. 1997 , , 490 , 493 navarro , j. f. , hayashi , e. , power , c. , jenkins , a. r. , frenk , c. s. , white , s. d. m. , springel , v. , stadel , j. , & quinn , t. r. 2004 , , 349 , 1039 primack , j. r. 2003 , nuclear physics b proceedings supplements , 124 , 3 reach , william t. , et al . 2005 , , 117 , 978 rhee , g. , valenzuela , o. , klypin , a. , holtzman , j. , & moorthy , b. 2004 , , 617 , 1059 ryder , s. d. , zasov , a. , mcintyre , v. , & walsh , w. 1997 , , 14 , 85 salpeter , e. e. 1955 , , 123 , 666 simon , j. d. , bolatto , a. d. , leroy , a. , & blitz , l. 2003 , , 596 , 957 spergel , d. n. , et al . 2003 , , 148 , 175 stil , j. m. , & israel , f. p. 2002 , , 392 , 473 swaters , r. a. , 1999 , phd thesis , university of groningen swaters , r. a. , madore , b. f. , van den bosch , f. c. , & balcells , m. 2003 , , 583 , 732 trachternach , c. , de blok , w. j. g. , brinks , e. , walter , f. , & kennicutt , r. c. 2008 , , submitted van albada , t. s. , & sancisi , r. 1986 , philos . london a , 320 , 447 van den bosch , f. c. , robertson , b. e. , dalcanton , j. j. , & de blok , w. j. g. 2000 , , 119 , 1579 van der kruit , p. c. , & searle , l. 1981 , , 95 , 105 walter , f. , kerp , j. , duric , n. , brinks , e. , & klein , u. 1998 , , 502 , 143 walter , f. , brinks , e. , de blok , w. j. g. , bigiel , f. , thornley , m. , & kennicutt , r. c. 2005 , aas , 207 , 6409 walter , f. et al . 2007 , , 661 , 102 walter , f. , brinks , e. , de blok , w. j. g. , bigiel , f. , kennicutt , r. c. , & thornley , m. 2008 , , submitted weldrake , d. t. f. , de blok , w. j. g. , & walter , f. 2003 , , 340 , 12 whiting , a. b. 1999 , the stellar content of local group galaxies , iau symposium 192 , 420 cccccccc [ ml_param ] 3.6 @xmath1 m & & @xmath154 & & & & 0.92 & + 4.5 @xmath1 m & & @xmath155 & & & & 0.91 & + & & & + color & @xmath77 & @xmath78 & @xmath156 & & @xmath77 & @xmath78 & @xmath156 + & ( 3 ) & ( 4 ) & ( 5 ) & & ( 6 ) & ( 7 ) & ( 8) + @xmath157 & @xmath158 & 0.60 & 0.44 & & @xmath159 & 0.72 & 0.33 + @xmath160 & @xmath161 & 0.42 & & & & & + @lrclccccrc & & & + + & & & & & & & & & + & & & & & & & & & + min . disk & 0.00 & @xmath162 & 674.6 @xmath163 18.3 & 2.88 & & 0.00 & @xmath162 & 1213.6 @xmath163 ... & 3.39 + min . disk+gas & 0.00 & @xmath162 & 524.3 @xmath163 51.7 & 1.65 & & 0.00 & @xmath162 & 1005.5 @xmath163 ... & 2.32 + max . disk & 0.93 & @xmath162 & 634.4 @xmath163 ... & 2.33 & & 0.93 & @xmath162 & 353.8 @xmath163 ... & 1.63 + model @xmath164 disk & 0.44 & @xmath162 & 873.9 @xmath163 ... & 1.81 & & 0.44 & @xmath162 & 700.5 @xmath163 ... & 1.96 + @xmath165disk & @xmath166 & @xmath162 & 1107.7 @xmath163 ... & 1.66 & & & & & + + & & & + + & & & & & & & & & + & & & & & & & & & + min . disk & 0.00 & 5.77 @xmath163 0.16 & 7.8 @xmath163 0.2 & 0.25 & & 0.00 & 5.69 @xmath163 0.35 & 7.8 @xmath163 0.3 & 0.26 + min . disk+gas & 0.00 & 4.61 @xmath163 0.12 & 7.6 @xmath163 0.2 & 0.16 & & 0.00 & 3.88 @xmath163 0.16 & 8.7 @xmath163 0.3 & 0.13 + max . disk & 0.93 & 27.52 @xmath163 10.22 & 1.8 @xmath163 0.1 & 0.30 & & 0.93 & @xmath167 & 1.7 @xmath163 0.9 & 0.39 + model @xmath164 disk & 0.44 & 7.23 @xmath163 0.30 & 4.1 @xmath163 0.1 & 0.17 & & 0.44 & 5.87 @xmath163 0.55 & 4.5 @xmath163 0.3 & 0.19 + @xmath165disk & 0.10 & 4.99 @xmath163 0.34 & 6.7 @xmath163 0.7 & 0.16 & & & & + + * note.@xmath168 * * ( 1)(10 ) : * the stellar mass - to - light ratio @xmath11 assumptions . `` model @xmath85 disk '' uses the values derived from the population synthesis models in section 4.2 . @xmath165 has @xmath85 as a free parameter . * ( 2)(6)(11)(15 ) : * average @xmath11 in the 3.6 @xmath1 m band ( @xmath80/@xmath169 ) . * ( 3)(7 ) : * concentration parameter c of nfw halo model ( nfw 1996 , 1997 ) . * ( 4)(8 ) : * the rotation velocity ( @xmath10)at radius @xmath124 where the density constrast exceeds 200 ( navarro et al . 1996 ) . * ( 5)(9)(14)(18 ) : * reduced @xmath170 value . * ( 12)(16 ) : * fitted core - radius of pseudo - isothermal halo model ( kpc ) . * ( 13)(17 ) : * fitted core - density of pseudo - isothermal halo model ( @xmath171 @xmath172 ) . * ( @xmath167 ) : * blank due to unphysically large value or not well - constrained uncertainties . @lrclccccrc & & & + + & & & & & & & & & + & & & & & & & & & + min . disk & 0.00 & @xmath162 & 901.5 @xmath163 478.4 & 1.72 & & 0.00 & @xmath162 & 1600.5 @xmath163 @xmath167 & 2.35 + min . disk+gas & 0.00 & @xmath162 & 727.8 @xmath163 @xmath167 & 1.08 & & 0.00 & @xmath162 & 1136.6 @xmath163 @xmath167 & 1.48 + max . disk & 0.88 & @xmath162 & 936.1 @xmath163 @xmath167 & 0.89 & & 0.88 & @xmath162 & 954.8 @xmath163 @xmath167 & 1.26 + model @xmath164 disk & 0.33 & @xmath162 & 630.7 @xmath163 @xmath167 & 0.98 & & 0.33 & @xmath162 & 1143.6 @xmath163 @xmath167 & 1.37 + @xmath165 disk & 1.15 & @xmath162 & 665.7 @xmath163 @xmath167 & 1.26 & & & & & + + & & & + + & & & & & & & & & + & & & & & & & & & + min . disk & 0.00 & 1.47 @xmath163 0.06 & 44.6 @xmath163 2.2 & 0.16 & & 0.00 & 1.49 @xmath163 0.07 & 44.1 @xmath163 2.6 & 0.21 + min . disk+gas & 0.00 & 1.25 @xmath163 0.05 & 43.8 @xmath163 2.4 & 0.13 & & 0.00 & 1.25 @xmath163 0.06 & 43.7 @xmath163 2.8 & 0.18 + max . disk & 0.88 & 1.61 @xmath163 0.15 & 21.8 @xmath163 2.5 & 0.25 & & 0.88 & 1.62 @xmath163 0.18 & 21.8 @xmath163 3.0 & 0.34 + model @xmath164 disk & 0.33 & 1.36 @xmath163 0.07 & 34.8 @xmath163 2.4 & 0.17 & & 0.33 & 1.36 @xmath163 0.09 & 34.7 @xmath163 2.9 & 0.23 + @xmath165 disk & @xmath173 & 0.98 @xmath163 0.05 & 87.1 @xmath163 14.0 & 0.14 & & & & + + * note.@xmath168 * * ( 1)(10 ) : * the stellar mass - to - light ratio @xmath11 assumptions . `` model @xmath85 disk '' uses the values derived from the population synthesis models in section 4.2 . @xmath165 has @xmath85 as a free parameter . * ( 2)(6)(11)(15 ) : * average @xmath11 in the 3.6 @xmath1 m band ( @xmath80/@xmath169 ) . * ( 3)(7 ) : * concentration parameter c of nfw halo model ( nfw 1996 , 1997 ) . * ( 4)(8 ) : * the rotation velocity ( @xmath10)at radius @xmath124 where the density constrast exceeds 200 ( navarro et al . * ( 5)(9)(14)(18 ) : * reduced @xmath170 value . * ( 12)(16 ) : * fitted core - radius of pseudo - isothermal halo model ( kpc ) . * ( 13)(17 ) : * fitted core - density of pseudo - isothermal halo model ( @xmath171 @xmath172 ) . * ( @xmath167 ) : * blank due to unphysically large value or not well - constrained uncertainties .

high - resolution , high signal - to - noise spectral data are presented for four young b - type stars lying towards the galactic centre . determination of their atmospheric parameters from their absorption line profiles , and photometric measurement of the continua indicate that they are massive objects lying slightly out of the plane , and were probably born in the disk between 2.5kpc from the centre . we have carried out a detailed absolute and differential line - by - line abundance analyses of the four stars compared to two stars with very similar atmospheric parameters in the solar neighbourhood . the stars appear to be rich in all the well sampled chemical elements ( c , n , si , mg , s , al ) , _ except for oxygen_. oxygen abundances derived in the atmospheres of these four stars are very similar to that in the solar neighbourhood . if the photospheric composition of these young stars is reflective of the gaseous ism in the inner galaxy , then the values derived for the enhanced metals are in excellent agreement with the extrapolation of the galactic abundance gradients previously derived by rolleston et al . ( 2000 ) and others . however , the data for oxygen suggests that the inner galaxy may not be richer than normal in this element , and the physical reasons for such a scenario are unclear .

astro-ph0009157

the current metallicity of the galactic disk represents a fossil record of chemical enrichment since the birth of the milky way and during its subsequent evolution . as stars continually enrich the interstellar medium with their nucleosynthetic products , the metallicity of the disk increases over time . by investigating the relative abundance of chemical elements as a function of time and position within the galaxy , it is possible to constrain models of galaxy formation and evolution . much work has been done on tracking the chemical evolution of the solar neighbourhood as a function of time using age - metallicity relations and element abundance ratios in cool f and g - type stars . the seminal work by edvardsson et al . ( @xcite ) showed the potential of using samples of nearby stars with an age spread to trace the temporal evolution of the disk and the halo , and such work has been extended by , for example , fuhrmann ( @xcite ) and israelian et al . ( @xcite ) . these types of metallicity ratios have been modelled by many authors ( e.g. carraro et al . 1998 , pilyugin & edmunds 1996 and references therein ) , in order to constrain our galaxy s history . the photospheric abundances of massive stars ( viz . stars with zams @xmath2 10 m@xmath3 ) do not have the potential to provide us with temporal information as they are by nature short - lived objects ( with typical lifetimes of the order of @xmath410 - 50myrs ) . however , their high intrinsic luminosities mean that we can spectroscopically observe them at large distances from the solar neighbourhood ; ob - type main - sequence stars are typically 8@xmath5 visually brighter than early g - type dwarfs . their photospheres are representative of the interstellar material from which they were born due to their relative youth . hence , we can spatially and radially sample the metallicity of the galactic disk by determining abundances in distant , luminous blue stars . the radial variation of different elements provides a further key constraint on the formation and evolutionary mechanisms of the galactic disk ( matteucci & francois @xcite , matteucci & chiappini @xcite , prantzos & aubert @xcite , portinari & chiosi @xcite , pagel & tautvaiin @xcite ) . these models require further observational constraints to restrict the possible influence of their many free parameters such as initial mass function , gas fraction variations , star formation rates , infall and outflow rates and metallicity dependent stellar yields . previous work has shown that the metallicity of the galactic disk decreases as galactocentric distance increases . this metallicity gradient has been reproduced quantitatively using a number of different methods . for example , extensive studies of optical emission lines in hii regions by shaver et al . ( @xcite ) and fich & silkey ( @xcite ) have derived an oxygen abundance gradient of @xmath6dexkpc@xmath7 ; within the region @xmath8kpc ( where @xmath9 refers to galactocentric distance throughout this paper ) . good agreement was also found by maciel & kppen ( @xcite ) and maciel & quireza ( @xcite ) from pn studies , and by afflerbach et al . ( @xcite ) in a study of ultra - compact hii regions . several studies of b - type main - sequence stars throughout the disk had given discrepant results , and a much shallower gradient than that derived from the nebular studies ( e.g fitzsimmons et al . @xcite , and references therein ) . however smartt & rolleston ( @xcite ) , in a consistent re - analysis of an extensive data - set of b - type stars covering @xmath10kpc , showed that previous b - type star studies produced spurious gradient determinations , and derived an oxygen abundance gradient of @xmath6dexkpc@xmath7 in excellent agreement with the nebular studies . as a follow up to this , rolleston et al . ( @xcite ; hereafter rsdr ) carried out a more detailed and exhaustive survey of the large data - set gathered by the group at queen s university and have confirmed this result , while also producing abundance ratios of different elements along the disk . further work by gummersbach et al . ( @xcite ) also indicates that the nebular and stellar results are in agreement within the errors of measurement . however , most studies of the chemical composition of the galactic disk have concentrated on the solar neighbourhood , and the anti - centre direction . optical extinction rises steeply in the direction toward the galactic centre , and restricts the optical observation of disk stars more than a few kiloparsecs away . in the anti - centre direction we have previously probed out to @xmath1118kpc , experiencing a@xmath123 ( smartt et al . @xcite ) , but have only managed to observe open clusters towards the centre approximately 2@xmath13kpc away ( rsdr ) with a reddening of a@xmath122 . the metallicity within the solar galactic radius is poorly sampled , with only a few ir studies available . rudolph et al . ( @xcite ) and afflerbach et al . ( @xcite ) have used similar observational methods for far - ir emission lines in hii regions ( of differing sizes ) as ism abundance probes . this far - ir technique has allowed observation of some inner galactic regions . these studies suggest that the abundance gradients for both nitrogen and sulphur do tend to increase towards the galactic centre , but both indicate that their oxygen abundance derivations may suffer some uncertainties ( see section[discussion_i ] ) . carr et al . ( @xcite ) have achieved a first determination of stellar abundances ( in an m - type supergiant ) directly at the galactic centre in the spectacular star forming region within the central few hundred parsecs . they find an fe abundance close to that of the solar neighbourhood , which at first may seems surprising given the wealth of evidence of a strong abundance gradient in the milky way ( and also in other spirals ) . however as we discuss in section[dist_discuss ] , this region may be unrelated to the starforming galactic disk ( @xmath13kpc ) as a whole , and its evolution may have proceeded independently . the metallicity of the inner disk therefore has not been investigated in any great detail , and lacks information on a variety of elemental abundances . spectra of b - type stars provide the potential to sample many more chemical elements than is possible through emission line studies . as discussed above , massive b - type main - sequence stars are excellent probes of the current interstellar medium and sample particularly well the @xmath14-processed elements ( o , mg , and si ) as well as c and n ( smartt et al . 1996 ) . this paper describes observations of metal - rich massive stars lying towards the galactic centre together with detailed model atmosphere and abundance analyses . this allows the comparison of abundance ratios in this interesting regime with those in the rest of the disk , as a key probe of the evolution of our galaxy . the main difficulty in optically probing the inner regions of the galactic disk is of course the high extinction encountered . this section describes how we identified targets for optical photometry and spectroscopy which are in regions of low extinction , but which have probably originated in the inner disk region . the catalogue of reed ( @xcite ) provides a database of published ubv@xmath15 photoelectric photometry for objects in the stephenson and sanduleak catalogue _ luminous stars in the southern milky way _ ( @xcite ) . this catalogue is a fairly complete record of ob - type stars and a- & f - type supergiants and giants within 10@xmath16 of the galactic plane and with longitude limits @xmath17 , down to a limiting photographic magnitude of approximately 12.5 . in order to identify ob - type stars near the galactic centre , which also have moderate galactic latitudes and hence suffer from low extinction , stars from the reed catalogue were selected which satisfied all of the criteria listed below: 1 . stars within @xmath18 ; hence restricting the sample to objects towards the galactic centre . stars which are more than 4@xmath16 above or below the plane ; in order to limit the effects of interstellar extinction . stars which are fainter than v = 11.0 , to sample the more distant objects . stars with a colour factor between @xmath10.90 @xmath19 q @xmath19 @xmath10.66 ; appropriate for early b - type stars ; + . stars with a b v of less than 0.5 @xmath1 which implies a reddening of @xmath20 ( for early b - type intrinsic colours ) . the combination of these criteria should produce targets which are potentially normal early b - type dwarfs ( or _ near _ main - sequence objects ) lying within @xmath44kpc of the galactic centre and twelve such stars were found . the database of reed & beatty ( @xcite ) contains published spectral types for a subset of the objects in reed ( @xcite ) . however , the fainter objects of the earlier study ( which were the most promising distant b - type stellar candidates ) did not have an entry in reed & beatty ( 1995 ) . as a result we had to rely on colours as a primary selector . the spectroscopic data presented in this paper were taken on two different runs . all twelve candidates were observed at the anglo - australian telescope in august 1994 at intermediate resolution . the rgo spectrograph along with the 1200b diffraction grating , 82 cm camera and the tek1k ccd were used to give a resolution of approximately 0.7 ( fwhm ) . two grating positions were employed with central wavelengths of 4020 and 4580 , each giving a wavelength coverage of 240 . the primary motivation of these observations was to identify normal stars for further , higher resolution observations at a higher signal - to - noise ratio , and to provide reliable hi profiles for the stars which did turn out to be normal . six of the stars observed appeared to be normal main - sequence early - type b - stars . however one of these was a fast rotator with a lack of metal line features , while another appeared to be a double - lined spectroscopic binary ( composed of two early - type components ) with the metal lines severely blended . the nature of the other six peculiar stars is discussed in venn et al . ( @xcite ) . the four apparently normal stars ( see table [ photometry_tab1 ] ) were further observed with the eso 3.6 m telescope using the caspec chelle spectrograph during july 1995 . the central wavelength was 4500 with a wavelength coverage of near 1300 ( 3920 5240 ) at a resolution of @xmath40.1 . .photometric details of galactic centre b - type main - sequence stars . our estimated spectral types are quoted [ cols="<,^,^,^,^,^,^",options="header " , ] in order to constrain the regions of the disk where these stars were formed , we follow the arguments of smartt et al . ( @xcite ) . for each of the stellar spectra we have determined a radial velocity , and we compare this velocity to that expected if the stars were co - rotating with the disk at their current projected positions ( assuming a flat rotation curve and r@xmath3 = 8.5kpc , @xmath21 km s@xmath7 from kerr & lynden - bell @xcite ) . we calculate a peculiar radial velocity ( prv ) which is the difference between these two , as a representative measure of the dynamics of the stars with respect to the galaxy rotation . this prv is only one component of the total peculiar space velocity , with the other two components unknown ( directed along axes mutually perpendicular to the prv ) . we assume that the second component of velocity is perpendicular to the galactic plane ( which , given the small galactic latitudes of these stars is a reasonable assumption ) . house & kilkenny ( @xcite ) have given an expression for the force field normal to the galactic plane ( @xmath22 ) , but this is probably only valid for @xmath23kpc . given that these stars have origins closer to the galactic centre , we have assumed a simple relation which is an approximation to the shape of the house & kilkenny form of @xmath22 , but which is always greater in magnitude ( with a peak of twice the house & kilkenny maximum ) . we assume that within @xmath24pc the force field @xmath22 is approximated by @xmath25 ( with @xmath26cms@xmath27pc@xmath7 ) and for @xmath28pc then @xmath29 . analytical expressions for the initial velocity ( v@xmath30 ) and time required for the stars to reach their present @xmath31 distance , could then be deduced . the initial ejection velocity perpendicular to the galactic plane needs to be no more than 100 km s@xmath7 ( see table[distance_data ] ) for these stars to reach their present positions , and significantly less for three of the objects . velocities of this order are well within those plausible for common ejection methods ( leonard @xcite ) . the prv of each star is also relatively low , indicating that the initial space velocities of these stars were probably directed mostly in the @xmath31-direction . however without proper motion information we can not definitively constrain their space velocity . if we assume that the mean ejection velocity in the sample ( v@xmath32 60 km s@xmath7 ) is indicative of the maximum value of any one velocity component , then we can calculate a circular locus within which the star probably formed . in 20myrs , a star could have moved 1.7kpc within ( or parallel to ) the galactic disk , and this is a reasonable error on the birth sites of the stars . [ galactic_pos ] figure [ galactic_pos ] illustrates the estimated stellar birth sites and also schematically draws attention to features of galactic structure in this region . bertelli et al . ( @xcite ) have discussed the possibility that a ` stellar ring ' exists , which surrounds the galactic centre at a distance of 3.5kpc , with a 0.5kpc half - width . the nature of this ring is still somewhat uncertain , but bertelli et al . claim that assuming it to be a region of increased stellar disk density would produce synthetic colour - magnitude diagrams closely matching those they have observed in fields towards the galactic centre . they suggest that it is associated with active star formation , and that the cross - section through this ring is at a maximum when viewed at @xmath33 . they further suggest that this is consistent with an observed peak in oh maser sources ( blommaert et al . @xcite ) around @xmath34 , a signature of relatively recent star formation . whether this is a complete ring surrounding the inner galaxy , or part of an inner spiral arm , is still debatable ( ortiz & lpine @xcite ) . clemens et al . ( @xcite ) have mapped the distribution of molecular gas in the first quadrant of the galaxy , and found an h@xmath35 ring at approximately 4kpc from the centre . this would appear to correlate with the stellar ring found by bertelli et al . , but the latter suggest that the stellar and molecular ring are not necessarily co - incident . rather that the molecular ring maybe the result of a `` bow - shock of propagating star formation '' . although some galaxy models ( bahcall & soneira 1980 ; bahcall 1986 ) adopt a relation for the number density of disk stars which falls off exponentially with increasing distance from the galactic centre , it appears that at r@xmath362.5kpc from the galactic centre there is a sharp decrease in the stellar disk density ( bertelli et al 1995 , paczynski et al . this has been interpreted as the transition between the disk and the bulge . figure [ galactic_pos ] suggests that the four b - type stars have their origins in this stellar ring , or where the molecular ring dynamics are causing current star formation . indeed if there is a disk / bulge transition , early - type stars may be relatively rare within this inner region , although current and spectacular star formation is certainly on - going in the inner few hundred parsecs ( serabyn et al . 1998 ; najarro et al . the above indicates that our b - type stars probably formed around r@xmath373.5kpc in the galactic disk and were subsequently ejected . carr et al . ( @xcite ) have presented ir spectra of a red m - type supergiant ( irs7 ) lying in the central cluster , and compared the metal line spectra with that of a similar star in the solar neighbourhood . a differential abundance analysis of the star indicates that it has an fe abundance very close to solar , and certainly does not appear metal enhanced in any significant way ( although it does appear to show cno - abundances typical of severe dredge - up after core h - burning ) . at first this may seem a surprising result given that one might expect a linear galactic abundance gradient in the thin disk to keep rising towards the centre , and that the central regions would thus show super - metal rich abundances . however there is lack of molecular gas , young stars and hii regions within the inner 3.5kpc , until one gets to the central cluster . hence , whatever the evolutionary history of the massive stellar clusters at the centre , it seems to have occurred independently of the milky way disk ( @xmath38kpc ) . there is no reason therefore too assume that the disk and central cluster are chemically linked , and one should consider them as separate entities in evolutionary scenarios . as discussed in sections[ls5130_dis][ls5127_dis ] , all four stars have relatively normal oxygen abundances . if the oxygen abundance gradients derived in the solar neighbourhood and outer disk are extrapolated inwards ( smartt & rolleston 1997 , rsdr ; @xmath10.07 dexkpc@xmath7 ) their o abundances would be 0.3@xmath10.4dex above the solar neighbourhood value . ls5130 and ls4419 show no evidence at all of oxygen - rich photospheres , and both sample the oii line spectra extremely well , with 43 and 28 features measured respectively . if the statistical errors are normally distributed around the mean , then the random error in our mean differential abundance ( @xmath39 ; where @xmath40 is the standard deviation quoted in tables[ls5130_abundances ] and [ ls4419_abundances ] ) may be as small as @xmath410.04 for both these stars . furthermore , a systematic error in the atmospheric parameters significantly greater than that quoted in table[typical_errors ] would be required to make these stars o - rich which would have major consequences for the other elements and for the continuum fitting of the @xmath0 indices . the large number of oii lines well observed in these stars and the possible error sources in table[typical_errors ] make this a robust result . the other two stars ls4784 and ls5127 show marginal evidence for oxygen abundances higher than their comparison stars , with standard errors of the mean @xmath410.1dex . however again these results appear to be incompatible with an o abundance that is 0.3@xmath10.4dex above solar . a number of previous authors have speculated that the oxygen abundance gradient ( in particular ) may steepen towards the centre ( vlchez & esteban @xcite , shaver et al . @xcite ) . however , we can find no evidence that this is the case . figure[osimg_gradients ] shows the oxygen abundances of our stars together with those of the large galactic b - star sample of rsdr . the low metallicity stars in the outer galaxy ( @xmath42kpc ) are clearly evident , but between 6@xmath19kpc there is little evidence of o abundance variations . the four new stellar points toward the centre accentuate this relation , and it would appear that the o abundance gradient in the galaxy flattens off at around 9kpc and does not significantly increase at smaller galactocentric distances . afflerbach et al . ( @xcite ) have carried out an extensive study of the n , s and o abundances in hii regions with @xmath43kpc . in fig.[on_hii_comp ] , their o abundances are plotted together with those for our stars and the stellar sample of rsdr with @xmath43kpc . the results from the hii regions and the rsdr data - set in the outer galaxy are in excellent agreement ( see rsdr and smartt @xcite for further details ) . however differences appear at smaller galactocentric distances . in particular while the hii region data is consistent with a linear gradient @xmath10.06dexkpc@xmath7 , no significant variations are apparant for the b - stars . one possible explanation is that the zero - points for the two studies is somewhat different . for example , if the afflerbach et al . data were shifted upwards by @xmath40.2dex ( the difference in the mean results within @xmath44kpc ) , then good agreement would be found within the solar neighbourhood , but the b - stars towards the centre would not reflect the hii region results . the cause of this discrepancy is unclear and appears to be limited to oxygen ( see sect . [ discussion_ii ] and [ discussion_iii ] ) . we note that afflerbach et al . is the only hii region study of the inner galaxy to quote oxygen abundances . rudolph et al . ( 1997 ) were unable to independently determine s and o , while shaver et al . ( 1983 ) point out that their two inner galaxy objects ( @xmath11 4kpc ) are somewhat peculiar , and that their radio and optical sources may not be the same . afflerbach et al . have recalculated n abundances from the data of simpson et al . ( @xcite ) and find differences of a factor of @xmath42 in some of the highest metallicity regions . additionally , on average there seems to be a 50% discrepancy in the sulphur abundances . therefore it appears that the nebular abundances in the inner galaxy require further study to constrain their absolute values , while a more extensive sample of b - stars is required . smartt et al . ( @xcite ) have discussed the problem of determining reliable absolute nebular abundances in regions of solar abundance and above in m31 , and find significant discrepancies across different analysis methods . in terms of effective radius ( where @xmath45kpc , from de vaucouleurs & pence @xcite ) , the break in the observed o abundance gradient occurs at @xmath46 . the spiral galaxy ngc2403 has been studied by garnett et al . ( @xcite ) who estimated hii region abundances at 12 points across the disk . in fig[mw_ngc2403_grads ] the oxygen abundance is plotted as a function of effective radius for both these galaxies . a similar gradient change may be present in ngc2403 , although the sampling is relatively sparse . this would imply that a two component model for the radial variation of oxygen is not exclusive to our own galaxy and suggests indirect evidence that it is a real effect . certainly the data suggests that for both galaxies , at @xmath47 , there is little evidence for increasing oxygen abundances . however , the well studied galaxy m33 ( garnett et al . @xcite , henry & howard @xcite , monteverde et al . @xcite ) which is similar in size , mass and metallicity to ngc2403 does not appear to show such a break . additionally , m101 and ngc4303 are other examples of extensively studied galaxies ( kennicutt & garnett @xcite , martin & roy @xcite ) , with 30@xmath180 hii region data points sampled in each . neither show evidence for anything other than a linear relation across the whole disk . the magnesium and silicon abundances for the galactic centre stars and the corresponding rsdr dataset are plotted in fig[osimg_gradients ] . for oxygen and silicon we note that the absolute abundances derived in our two solar neighbourhood standard stars are similar to the mean of the rsdr data set within 1kpc of the sun . hence the absolute results of the galactic centre stars are on a consistent scale to the rest of the data . however with mg , the abundance derived in @xmath48 peg and @xmath49 her is 7.2dex , whereas in rsdr the mean local value is 7.85 dex ( again estimated from the mean of stellar values within 1kpc of the sun ) . the stars analysed in this paper have effective temperatures in the range @xmath50k , while the rsdr sample have a mean temperature of 25275k and it is possible that the modelling of this line is temperature dependent . it is important to use homogeneous datasets when looking for abundance trends and that is indeed what was done in rsdr . they showed that although the zero - point of any particular abundance may be in the error , any gradient biasing is removed by using suitable sub - samples of stars . hence to put the four galactic centre stars onto the same scale as rsdr , we have added 7.85dex to their _ differential _ mg abundances . this assumes that @xmath48 peg , @xmath49 her and the stars within 1kpc of the sun in rsdr have similar abundances and that any offset is due to temperature dependent modelling inaccuracies . it appears that the galactic centre stars are richer in mg than their local counterparts ( see also sect[ls5127_mgii_line ] ) . from table[typical_errors ] an underestimate of the microturbulence ( @xmath51 ) in each star could reduce this mg overabundance as it is based on a relatively strong line . however an increase in @xmath51 by 5 km s@xmath7 or more in each star would be required . while observational uncertainties might cause a particular determination of @xmath51 to have such an error , there is no evidence that we have _ systematically _ underestimated the microturbulence in each star . the mg abundance gradient in fig . [ osimg_gradients ] hence appears to steadily increase towards the centre . a value of @xmath52dexkpc@xmath7is derived by fitting all points , which is slightly larger the rsdr result , although they agree within the uncertainties . the silicon abundances of the galactic centre stars are also plotted in fig.[osimg_gradients ] , and again appear compatible with a steadily increasing abundance gradient towards the centre . a fit through all the si points gives a gradient of @xmath53dexkpc@xmath7 , again in good agreement with the rsdr value for the outer galaxy only . the differential abundances derived for each star are between 0.3@xmath10.6dex higher than their respective standards , strongly supporting the idea that these stars are richer in si than solar neighbourhood material . in fig.[fig_sal_grads ] the al abundances of the galactic centre stars are plotted again with the data from rsdr . aluminium appears to increase steadily toward the centre , and a gradient of @xmath54 dexkpc@xmath7 is derived from the whole data set , similar to that produced by rsdr for the anti - centre direction . sulphur abundances for our stars are also available , although we have no similar data at greater radii from rsdr . in fig[fig_sal_grads ] our stellar abundances are compared with with those from afflerbach et al . reasonable agreement is found with these stars appearing to be s - rich and having abundances similar to those found in nebular studies towards the centre ( although the caveat regarding the nebular absolute values discussed in the previous section may be important ) . the correlation of of o , mg , and si abundances found in the outer galaxy ( see rsdr ) and in the low metallicity magellanic clouds ( rolleston et al . @xcite ) is theoretically expected as these elements are all produced and returned to the ism through supernovae typeii . one would expect that this trend should continue towards the inner galaxy , and it is surprising that our o abundances show no sign of increase , while mg and si appear significantly higher than their comparison stars . the s abundances in these stars are also higher than normal , although the dataset is somewhat less robust given the weakness of the absorption features of this element . prantzos et al . ( @xcite ) have investigated the effect of metallicity dependent yields ( taken from the model calculations of maeder @xcite ) on the galactic chemical evolution of c and o and concluded that oxygen yields from massive stars tend to decrease towards higher metallicities . this is due to mass loss in low metallicity stars , resulting in most of the star s mass being retained up until the final explosion . however , at high metallicities significantly more mass is ejected in the stellar wind ( which is rich in helium and carbon ) , and leaves a smaller mass fraction which can be converted into oxygen . the relatively high carbon and low oxygen abundances derived here are in qualitative agreement with the situation predicted by these models . a full study of the evolution of elements heavier than oxygen has not been undertaken using metallicity dependent yields . however maeder ( 1992 ) has discussed how the combined yield of `` heavy elements '' ( defined as ne , mg , si , s , ca & fe ) from massive stars varies with original metal composition . interestingly , he finds that the `` heavy element '' contributions of the stellar models at high metallicity ( @xmath55 ) are not as low as is the case for oxygen . the situation is further complicated by the fact that chemical yields are dependent on the conditions for black hole formation since the final yield of a pre - supernova massive star ( pictured in the conventional onion - skin model ) depends on what fraction of the stellar layers are retained during core collapse . the observational findings presented here should provide a stimulus to investigations of the evolution of the galaxy at high metallicities , given the probability that stellar yields are highly variable with metallicity , and dependent on the limits of black hole formation . furthermore , we should attempt to determine the true quantitative variation in the ratios of the @xmath14-elements in the inner galaxy to investigate if this trend is real . the recent chemical evolution models of portinari & chiosi ( @xcite , @xcite ) do indicate that the o abundance gradient may indeed flatten off in the region @xmath56kpc , which is encouraging . however their sn typeii yields would suggest that the other @xmath14-processed elements ( mg and si ) should follow a similar trend , which we do not observe . boissier & prantzos ( @xcite ) also suggest that their evolutionary models for the milky way predict a flattening of the metallicity gradient with time , and that saturation is reached in the inner galaxy . this is in reasonable quantitative agreement with what we find for oxygen abundances , but does not explain the behaviour of the other elements . each of the four stars do appear to have nitrogen abundances significantly higher than their solar neighbourhood counterparts . the result for ls5130 is more marginal than the others , but with 19 features measured an error in the mean ( @xmath57 ) of 0.03dex would suggest that the enrichment is real although small . figure4(b ) shows the nitrogen abundances of the galactic centre stars plotted along with a sample of stars from rsdr , and the hii regions from afflerbach et al . the gradient found in rsdr appears to continue towards the centre ; certainly the abundance estimates do not appear to flatten as for oxygen . reasonably good agreement with the absolute values of the hii region analyses is found , and we calculate quite a significant gradient for @xmath58n / o . the full sample of rsdr yielded a gradient in n / o of @xmath59dexkpc@xmath7 , and this is increased slightly by the inclusion of the four galactic centre stars . the latter effect is clearly due to the stars showing enhanced nitrogen in their atmospheres , but normal oxygen . we have previously discussed the possibility of n - rich core gas contaminating the surface of young massive stars , and hence polluting the natal photospheric material ( rsdr ) . however due to the lack of a strong anti - correlation of c and n , rsdr found no evidence for such contamination and the same argument holds for the four stars analysed here . an explanation of the n / o gradient was discussed in rsdr , based on the original ideas of vila - costas & edmunds ( @xcite ) . at metallicities higher than @xmath60 o / h @xmath48.3 , @xmath58 n / o would be proportional to @xmath58 o / h , because n enrichment would be dominated by secondary production , i.e. from c and o seed nuclei in the natal interstellar material through the cno - cycle . primary production would play a minimal role in contributing to the ism abundance of n , and simple closed - box models in which the production of n is dominated by secondary production do predict a linear trend . at low metallicities the secondary component becomes less important ( due to its inherent dependence on metallicity ) , and the dominant mechanism for the production of nitrogen is primary . the metallicity at which this happens is around 8.3dex , close to the value found in the outer most regions of the galactic disk . these ideas stem from the fact that the n / o ratios in low metallicity dwarf galaxies show a very large scatter , with little discernable trend . as primary nitrogen orginates in low - intermediate mass stars there is a time delay between is enrichment and that of primary oxygen ( from short - lived massive stars ) , hence the scatter at low metallicitities is often interpreted as time dependent delays between oxygen and nitrogen enrichments . however at the high metallicities we are dealing with , it is likely that secondary production ( in massive stars is dominant ) , and there would thus be no delay between n and o production . our n / o ratios toward the galactic centre tend to support the conclusions of rsdr that the n / o gradient in the galaxy is real , and due to secondary n production dominating across the disk . in fig.[fig_co_grads ] carbon abundances for the four galactic centre stars and from the rsdr compilation are plotted . as three of our new points are c - rich , it is not surprising that the abundances appears to steadily increase toward the centre . as discussed for nitrogen , because the stars are not rich in oxygen the c / o ratio is greater than normal , hence producing a larger ( and statistically more significant ) gradient than that given in rsdr . as discussed by maeder ( 1992 ) this would be consistent with c being produced at a greater rate than o in metal - rich stars ( section[discussion_ii ] ) , due to the increased mass - loss in the post he - burning phase . the analysis of high - quality data for four blue stars lying towards the galactic centre , and a comparison with young stars throughout the galactic disk , leads to the following conclusions 1 . the four stars appear to be normal , young , massive , early b - type stars . they lie slightly out of the galactic plane and their calculated distances indicate that they are at distances from the disk of @xmath61kpc . a kinematic analysis indicates that they were probably born in the disk within @xmath62kpc of the galactic centre and then subsequently ejected . comparison with galactic structure models suggests they may have originated in the stellar or molecular ring at @xmath63kpc . 2 . a differential line - by - line abundance analysis of the four stars with two bright stars near the sun indicates a surprising abundance pattern : * two of the stars ( ls5130 and ls4419 ) have oxygen abundances very similar to stars in the solar neighbourhood . the other two ( ls4784 and ls5130 ) show only marginal evidence for enhanced oxygen . in all cases any enhancement is significantly below the @xmath64dex that would be expected given the galactic position of the stars and the gradient from rsdr . oxygen is a very reliably determined abundance in these particular types of stars . * the relatively well measured abundances of si , al , mg and s however do indicate that these stars have enhanced abundances in most other elements . however , the physical cause of the natal gas having a `` normal '' oxygen content and enriched @xmath14-processed elements is unclear . the si and mg abundances are consistent with the continuation of the linear gradient derived by rsdr for stars in the solar neighbourhood and the outer galaxy . * the four stars also appear rich in both c and n , again in agreement with a continuation of the rsdr gradients towards the centre . * we discuss the possible errors in our analysis , and conclude that it is unlikely that these stars have higher oxygen abundances than those typically found in the solar neighbourhood . it is also very likely they are rich in all the other elements studied ( c , n , mg , si , s , al ) although the physical pocesses that produces such a pattern is unclear . * appendix a : table9 and 10 are the lists of equivalent widths of all metal lines measured in the spectra of the galactic centre stars and the standards @xmath49her , and @xmath48peg . and is available electronically * afflerbach a. , churchwell e. , werner m.w . , 1997 478 , 190 anders e. , grevesse n. , 1989 , geochim . acta 53 , 197 bahcall j.n . , 1986 , ara&a 24 , 577 bahcall j.n . , soneira r.m . , 1980 , apjs 44 , 73 baldwin j.a . , ferland g.a . , martin p.g . , et al . 1991 , apj 374 , 580 bertelli g. , bressan a. , chiosi c. , ng y.k . , ortolani s. , 1995 , a&a 301 , 381 blommaert j.a.d.l . , langevelde van h.j . , michiels w. , 1994 , a&a 287 , 479 brown p.j.f . , 1987 , phd thesis , queens university of belfast boissier s. , prantzos n. , 1999 , mnras 307 , 857 cardelli j.a . , meyer d.m . , jura m. , savage b.d . , 1996 , apj 467 , 334 carr j.s . , sellgren k. , balachandran s.c . , 2000 , apj 530 , 307 carraro g. , ng y.k . , portinari l. , 1998 , mnras 296 , 1054 clemens d.p . , sanders d.b . , scoville n.z . , 1998 , apj 327 , 139 cunha k. , lambert d. , 1994 , apj 426 , 170 davis l.e . , 1994 , a reference guide to the iraf / daophot package , noao laboratory deutschman w.a . , davis r.j . , schild r.e . , 1976 , apjs 30 , 97 de vaucouleurs g. , pence w. d. , 1978 , aj 83 , 1163 edvardsson b. , andersen j. , gustafsson b. , et al . 1993 , a&a 75 , 101 fich m. , silkey m. , 1991 , apj 366 , 107 fitzsimmons a. , dufton p.l . , rolleston w.r.j . , 1992 , mnras 259 , 489 fuhrmann k. , 1998 , a&a 338 , 161 garnett d.r . , shields g.a . skillman e.d . , sagan s.p . , dufour r.j . , 1997 , apj 489 , 63 gies d.r . , lambert d.l . , 1992 , apj 387 , 673 grevesse n. , noels a. , 1993 in origin and evolution of the elements , ed . n. prantzos , e. vangioni - flam , & m. casse ( cambridge : cambridge univ . press ) , 14 gummersbach c.a . , kaufer a. , shfer d.r . , szeifert t. , wolf b. , 1998 , a&a 338 , 881 henry r. b. c. , howard j. w. , 1995 , apj 438 , 170 heger a. , langer n. , 2000 , apj in press , ( astro - ph/0005110 ) . house f. , kilkenny d. , 1980 , a&a 81 , 251 howarth i.d . , murray j. , mills d. , berry d.s . , 1998 starlink user notice , no.50.21 israelian g. , garcia lopez r.j . , rebolo r. , 1998 , apj 507 , 805 jeffery c.s . , 1991 , in : newsletter on analysis of astronomical spectra no . 16 , pg . 17 kennicutt r.c . , garnett d.r . , 1996 , apj 456 , 504 kerr f.j . , lynden - bell d. , 1986 , mnras 221 , 1023 kilian j. , 1994 , a&a 282 , 867 kilkenny d. , balona l.a . , carter d.b . , ellis d.t . , woodhouse g.f.w . , 1998 , mnassa 47 , 69 kurucz r.l . , 1979 , apjs 40 , 1 kurucz r.l . , 1991 , in precision photometry : astrophysics of the galaxy , ed . davis - philip , a.r . upgren , & p.l . janes ( schenectady : davis ) leonard p. j. t. , 1993 , in luminous high - latitude stars , ed . d. sasselov ( san francisco : asp ) lester j.b . , gray r.o . , kurucz r.l . , 1986 , apjs 61 , 509 maciel w. j. , quireza 1999 , a&a 345 , 629 maciel w.j . , kppen j. , 1994 , a&a 282 , 43 maeder a. , 1992 , a&a 264 , 105 maeder a. , & meynet g. , 2000 , astro - ph/0006405 martin p. , roy j .- , 1992 , apj 397 , 63 massey p. , 1992 , a user s guide to ccd reductions with iraf , noao laboratory matteucci f. , francois p. , 1989 , mnras 239 , 885 matteucci f. & chiappini c. , 1999 , in proc . eso conf . on chemical evolution from zero to high redshift , springer conference series , eds . j. walsh & m. rosa meyer d. m. , jura m. , cardelli j. a. , 1998 , apj 493 , 222 monteverde i. , herrero a.h . , lennon d.j . , 2000 , apj , submitted mooney c.j . , rolleston w.r.j . , keenan f.p . , et al . 2000 , a&a 357 , 553 najarro f. , krabbe a. , genzel r. , et al . 1997 , a&a 325 , 700 napiwotzki r. , schonberner d. , wenske v. , 1993 , a&a 268 , 653 ortiz r. , lpine j.r.d . , 1993 , a&a 279 , 90 paczynski b. , stanek k.z . , udalski a. , et al . 1994 , aj 107 , 2060 pagel b.e.j . , tautvaiin g. , 1995 , mnras 276 , 505 peters g.j . , 1976 , apjs 30 , 551 pilyugin l.s . , edmunds m.g . , 1996 a&a 313 , 792 portinari l. , chiosi c. , 1999 , a&a 350 , 827 portinari l. , chiosi c. , 2000 , a&a 355 , 929 prantzos n. , vangioni - flam e. , chauveau s. , 1994 , a&a 285 , 132 prantzos n. , aubert o. , 1995 , a&a 302 , 69 reed b.c . , beatty a.e . , 1995 , apjs 97 , 189 reed b.c . , 1993 , apjs , 87 , 367 rolleston w.r.j . , smartt s.j . , dufton p.l . , ryans r.s.i . , 2000 , a&a , in press rolleston w.r.j . , hambly n.c . , keenan f.p . , dufton p.l . , saffer r.a . , 1999 , a&a 347 , 69 rolleston w.r.j . , hambly n.h , dufton p.l . , et al . 1997 , mnras 290 , 422 rolleston w.r.j . , brown p.j.f . , dufton p.l . , howarth i.d . , 1996 , a&a 315 , 95 ryans r.s.i , hambly n.c . , dufton p.l . , keenan f.p . , 1996 , mnras 278 , 132 rudolph a.l . , simpson j.p . , haas m.r . , erickson e.f . , fich m. , 1997 , apj 489 , 94 seaton m.j . , 1979 , mnras 187 , 73 serabyn e. , shupe d. , figer d. , 1998 , nature 394 , 448 schaller g. , schaerer d. , meynet g. , maeder a. , 1992 , a&as 96 , 269 shaver p.a . , mcgee r.x . , newton m.p . , danks a.c . , pottasch s.r . , 1983 , mnras 204 , 53 shortridge k. , meyerdierks h. , currie m. , clayton m. , lockley j. , 1997 , starlink user note 86.14 simpson j. p. , colgan s. w. j. , rubin r. h. , erickson e. f. , haas m. r. , 1995 , apj 444 , 721 smartt s.j . , rolleston w.r.j . , 1997 , apj 481 , l47 smartt s.j . , dufton p.l . , lennon d.j . , 1997 , a&a 326 , 763 smartt s.j . , dufton p.l . , rolleston w.r.j . 1996 , a&a 305 , 164 smartt s.j . , 2000 , in the chemical evolution of the milky way : stars vs. clusters , rome - trieste workshop , es f. matteucci , f. giovanelli smartt s.j . , crowther p.a . , dufton p.l . , et al . , 2000 mnras , to be submitted snow t.p . , witt a.n . , 1996 , apj 468 , l65 stephenson c.b . , sanduleak n. , 1971 , publ . warner & swasey obs . , 1 , 1 venn k.a , smartt s.j . , lennon d.j . , dufton p.l . , 1998 , a&a 334 , 987 vlchez j. m. , esteban c. , 1996 , mnras 280 , 720 vila - costas m. b. , edmunds m. g. , 1993 , mnras 265 , 199 rcccc species and line & ls5130 & ls4419 & ls4784 & @xmath48peg + oii 3911.96 & 57a & 96c & 76b & 50a + + cii 3918.98 & 138a & 130a & 261a@xmath65 & + nii 3919.01 & & & & + oii 3919.28 & & & & + + cii 3920.69 & 127a & 113a & & + + oii 3945.04 & 30b & 63b & 58c & + + oii 3954.37 & 51a & 63b & 74b & 48a + + hei 3964.73 & 177a & 187a & 188a & 155a + + oii 3982.72 & 29c & 72b@xmath66 & 60a@xmath66 & 31a + + siii 3983.77 & 28c & & & 14a + + nii 3995.00 & 102a & 104a & 93a & 78a + + nii 4035.08 & 35b & 46a & 42c & 25a + + nii 4041.31 & 44b & 51a & 60b & 37a + + oii 4069.62 & 95a & 115a & 97b & 94 + oii 4069.89 & & & & + + oii 4072.16 & 62a & 75a & 43c & 65a + + cii 4074.48 & 48b & 147a@xmath67 & 141b@xmath67 & 42a + cii 4074.85 & & & & + + cii 4075.85 & & & & + oii 4075.86 & 124a & & & + cii 4076.53 & & & & + nii 4076.91 & & & & + + oii 4078.84 & 24c & 19c & & 26a + + oii 4085.11 & 32c & 39b & 34b & 26a + + oii 4087.15 & 22c & & & 12a + + siiv 4088.85 & 37c & 44b & 61a & + oii 4089.28 & & & & + + siiv 4116.10 & 20c & & & + + oii 4119.22 & 62a & & & + + oii 4120.28 & 291a & & & + hei 4120.81 & & & & + hei 4120.99 & & & & + oii 4121.46 & & & & + + siii 4128.07 & 53a & 38c & 65b & 33a + + siii 4130.89 & 54a & 86c@xmath68 & 70c@xmath68 & 34a + + arii 4131.71 & 38b & & & 17a + + oii 4132.80 & 33b & & 54c@xmath68 & 30a + + aliii 4149.50 & & 56b & 64a & + aliii 4150.14 & & & & + + sii 4153.10 & 60a & 125b & 105a & + oii 4153.30 & & & & + + sii 4162.70 & 47b & & & + + feiii 4164.79 & 45c & 32b & & + + feiii 4166.86 & 26c & & & 13a + + hei 4168.97 & 99a & 99a & 104a & + oii 4169.22 & & & & + + sii 4174.24 & 40b & 28c & & + + nii 4179.67 & 14c & 21c & & + + oii 4185.45 & 36b & 43b & & 30a + + oii 4189.79 & 46b & 54b & 65b & + + piii 4222.15 & 21c & 20c & & 21a + + nii 4227.74 & 30b & 55b & 61b & 18a + + nii 4236.86 & 51a & 55a & 66c & 28a + nii 4236.98 & & & & + + nii 4241.78 & 49a & 86a & 105b & + + rcccc species and line & ls5130 & ls4419 & ls4784 & @xmath48peg + siii 4253.59 & & 118a & & + oii 4253.90 & & & & + oii 4254.13 & & & & + + cii 4267.02 & 262a & 245a & 305a & + cii 4267.27 & & & & + oii 4267.71 & & & & + + oii 4275.56 & & 97b & 123a@xmath69 & + + oii 4277.40 & & 72b & & 16a + + siii 4284.99 & & & 47c & 28a + + sii 4294.43 & 35b & 50b & 37c & + oii 4294.79 & & & & + + oii 4303.84 & 30c & 62a & 54c & 26a + + cii 4313.30 & 17c & & & + + oii 4317.14 & 58a & 76a & 64b & 56a + cii 4317.26 & & & & + + oii 4319.63 & 48a & 78a & 63b & 52a + + oii 4325.76 & 40b & 23b & 29c & 20a + + oii 4345.56 & 30c & & & 24a + + oii 4347.42 & 26c & & & 21a + + arii 4348.11 & 32c & & & + + oii 4349.43 & 66b & & & 63a + + oii 4351.26 & 31c & & & 41a + + oii 4366.89 & 60a & 61a & 32c & 48a + + oii 4369.27 & & & & 14a + + cii 4372.49 & 52a & 91a & 67c@xmath70 & + + cii 4374.27 & 32b & & & + + rcccc species and line & ls5130 & ls4419 & ls4784 & @xmath48peg + feiii 4395.78 & 40b & 60a & 58b & + oii 4395.94 & & & & + + cii 4411.20 & 38b & 63b & & + cii 4411.52 & & & & + + oii 4414.90 & 84a & 95a & 52c & 75a + + oii 4416.97 & 64a & 76a & 28c & 61a + + feiii 4419.59 & 44b & 32c & 35c & 34a + + feiii 4430.95 & 33c & & 39c & 24a + + nii 4432.74 & 38c & & 51c & 20a + + hei 4437.55 & 126a & 103a & 135a & 118a + + nii 4442.02 & & & 38c & 9a + + nii 4447.03 & 60a & & 72b & 15a + oii 4448.19 & & & & + + oii 4452.37 & 31c & & 29c & 25a + + aliii 4479.89 & 55a & & & 38a + aliii 4479.89 & & & & + + mgii 4481.13 & 202a & 196a & 265a & 160a + mgii 4481.33 & & & & + + aliii 4512.54 & 57a & 65a & 63a & 39a + + aliii 4528.91 & 73a & 144a@xmath71 & 135a@xmath71 & 66a + aliii 4529.20 & & & & + + nii 4530.40 & 29c & & & 25a + + nii 4552.53 & 174a & 199a & 187a & 114a + siiii 4552.62 & & & & + + siiii 4567.82 & 150a & 152a & 151a & 102a + + siiii 4574.76 & 81a & 97a & 90a & 69a + + oii 4590.97 & 52a & 87a & 67a & + + oii 4596.18 & 59a & 74a & 52a & 47a + + nii 4601.48 & 62b & 107a & 90a & 37a + oii 4602.13 & & & & 15a + + rcccc species and line & ls5130 & ls4419 & ls4784 & @xmath48peg + nii 4607.16 & 56a & 64a & 68a & 35a + + oii 4609.44 & 41b & 63a & 50a & 30a + + nii 4613.87 & 55a & 79a & 56a & 27a + + cii 4618.40 & 137c & 82a & 213a@xmath72 & + cii 4619.23 & & & & + + nii 4621.29 & 41b & 73a & & 26a + + nii 4630.54 & 92a & 122a & 123a & + + oii 4638.86 & 61b & 81a & 71a & 54a + + oii 4641.82 & 74b & 171a@xmath73 & 172a@xmath73 & 76a + + nii 4643.09 & 58b & & & 42a + + ciii 4647.42 & & 230a @xmath74 & 192a@xmath74 & + + oii 4649.14 & 90a & & & 92a + + oii 4650.84 & 64b & & & 52a + + oii 4661.63 & 67a & 94a & 75a & 54a + + oii 4673.74 & 26c & & & 19a + + oii 4676.24 & 44c & 45c & & 46a + + nii 4678.10 & 24c & & & + + nii 4694.70 & 15c & & & 9a + + oii 4696.35 & 15c & & & 12a + + oii 4699.00 & 42a & 69c & 37a & + oii 4699.22 & & & & + oii 4703.16 & & 66a@xmath74 & 53a@xmath74 & + + oii 4705.35 & 47b & & & + + oii 4710.01 & 30b & 40c & & 20a + + hei 4713.14 & 289a & 283a & 250a & 271a + hei 4713.37 & & & & + + rcccc species and line & ls5130 & ls4419 & ls4784 & @xmath48peg + siiii 4716.65 & 41b & 52b & & 12a + + nii 4779.67 & 18c & 36c & 44b & 10a + + cii 4802.70 & 28c & & & 15a + + nii 4803.29 & 37c & 61a & 74a & 18a + + arii 4806.07 & 29c & 21c & & 18a + + siiii 4813.30 & 31b & 41a & 75b@xmath75 & 21a + + sii 4815.52 & 54b & 27b & & 27a + + siiii 4819.72 & 55b & 55a & 50a & 40a + + sii 4824.07 & 12c & & 18c & + + siiii 4828.96 & 44b & 64a & 59a & + + arii 4879.14 & 28b & & 19c & + + oii 4890.93 & 21c & 22b & & 11a + + oii 4906.83 & 38b & & 29b & + + oii 4941.07 & 25b & 38b & 14c & 15a + + oii 4943.00 & 39b & 52b & 45c & 28a + + sii 4991.94 & 16c & & 24c & 22a + + nii 4994.36 & 31b & 55a & 59b & + + sii 5009.62 & 32b & & & + + nii 5010.62 & 40b & 40c & & + + sii 5014.03 & 52b & & & + + hei 5015.68 & 322a & 319a & 354a & + + sii 5032.41 & 66b & 56a & 91a & + + siii 5041.03 & 24c & & 30c & 15a + + rcccc species and line & ls5130 & ls4419 & ls4784 & @xmath48peg + nii 5045.09 & 54b & 59a & 47c & + + hei 5047.74 & 207a & 163a & 179b & + + siii 5055.98 & 35c & & 56b & 22a + siii 5056.31 & & & & + + sii 5103.30 & & & 20c & + + cii 5132.96 & 97a & 74a & 99a & + cii 5133.29 & & & & + + oii 5159.94 & 23a & 38a & 37a & + notes : + 1 . blend of oii 3982.72 and siii 3983.77 . + 2 . includes the oii , nii and cii lines at 40754076aa . + 3 . blended with arii 4131.71 . + 4 . blended with nii 4643.09 . blended with oii 4649.14 & 4650.84 . + 6 . blended with oii 4705.35 . + 7 . also includes cii 3920.69 . blended with oii 4277.40 . + 9 . blended with cii 4374.27 . blended with nii 4530.40 . + 11 . blended with nii 4621.29 . + 12 . blended with sii 4815.52 . + rcc species and line & ls5127 & @xmath49her + oii 3911.96 & 50c & 12a + + cii 3918.98 & 340b & 72a + nii 3919.01 & & + oii 3919.28 & & 8a + cii 3920.69 & & 88a + + oii 3954.37 & 44c & 20a + + hei 3964.73 & 210c & 132a + + nii 3995.00 & 108a & 40a + + nii 4041.31 & 57b & 19a + + nii 4043.53 & 59b & 9a + + oii 4069.62 & 182a & 16a + oii 4069.89 & & 19a + oii 4072.16 & & 27a + + cii 4074.48 & 185a & 15a + cii 4074.85 & & 9a + oii 4075.86 & & 32a + cii 4076.53 & & + nii 4076.91 & & + + oii 4119.22 & 324a & + oii 4120.28 & & + hei 4120.81 & & + hei 4120.99 & & + oii 4121.46 & & + feiii 4122.06 & & + feiii 4122.98 & & + + hei 4168.97 & 70b & 81a + + nii 4241.78 & 67a & 16a + + siii 4253.59 & 69a & 13a + oii 4253.90 & & 4a + oii 4254.13 & & 2a + + cii 4267.02 & 338a & 184a + cii 4267.27 + oii 4267.71 & & + siiii 4267.80 & & 26a + rcc species and line & ls5127 & @xmath49her + sii 4294.43 & 43a & 24a + oii 4294.79 & & 3a + + oii 4303.84 & 39a & 9a + + oii 4325.76 & 55c & 6a + cii 4325.83 & & + cii 4326.16 & & + + siii 4361.53 & 39c & 4a + + oii 4366.89 & 42c & 16a + + cii 4372.49 & 48c & + + feiii 4395.78 & 43b & 15a + + oii 4414.90 & 101b & 28a + oii 4416.97 & & 23a + + feiii 4419.59 & 53a & 23a + + hei 4437.55 & 101a & 102a + + nii 4447.03 & 65b & 18a + + sii 4463.58 & 36b & 14a + + mgii 4481.13 & 423a & 205a + mgii 4481.33 & & + + aliii 4512.54 & 60a & 20a + + sii 4524.95 & 47b & 22a + + aliii 4528.91 & 100b & 6a + aliii 4529.20 & & 28a + nii 4530.40 & & 7a + + rcc species and line & ls5127 & @xmath49her + feii 4549.47 & 49b & + + sii 4552.38 & 191a & 71a + siiii 4552.62 & & + + siiii 4567.82 & 156a & 59a + + siiii 4574.76 & 93a & 27a + + feii 4583.83 & 39b & 12a + + oii 4590.97 & 39c & 17a + + oii 4596.18 & 32c & 12a + + nii 4601.48 & 49c & 15a + + nii 4607.16 & 47c & 14a + + cii 4618.40 & 114c & + cii 4619.23 & & + nii 4621.29 & & + siii 4621.42 & & + siii 4621.72 & & + + feii 4629.34 & 107a & + nii 4630.54 & & + + oii 4638.86 & 43a & 21a + + oii 4641.82 & 108a & 24a + nii 4643.09 & & 17a + + sii 4648.17 & 170a & 5a + oii 4649.14 & & 34a + oii 4650.84 & & 17a + + sii 4656.74 & 26b & 10a + + oii 4661.63 & 65b & 15a + aliii 4663.05 & & 13a + + hei 4713.14 & 306a & 239a + hei 4713.37 & & + + siiii 4813.30 & 81b & 8a + sii 4815.52 & & 40a + + siiii 4819.72 & 47c & + + sii 5014.03 & 522a & 38a + hei 5015.68 & & 233a + + sii 5032.41 & 128a & 47a + + siii 5041.03 & 85a & 42a + + nii 5045.09 & 190a & + sii 5047.28 & & + hei 5047.74 & & + + siii 5055.98 & 182a & 82a + siii 5056.31 & & + +

accurate color reproduction is important in many applications ofd printing , from design prototypes tod color copies or portraits . although full color is available via other technologies , multi - jet printers have greater potential for graphicald printing , in terms of reproducing complex appearance properties . however , to date these printers can not produce full color , and doing so poses substantial technical challenges , from the shear amount of data to the translucency of the available color materials . in this paper , we propose an error diffusion halftoning approach to achieve full color with multi - jet printers , which operates on multiple isosurfaces or layers within the object . we propose a novel traversal algorithm for voxel surfaces , which allows the transfer of existing error diffusion algorithms fromd printing . the resulting prints faithfully reproduce colors , color gradients and fine - scale details . contact : \{alan.brunton , can.ates.arikan , philipp.urban}@igd.fraunhofer.de + web : www.cuttlefish.de

1506.02400

in this paper , we present algorithms for producing 3d color prints , which are highly accurate and detailed . from a technical standpoint , we consider the problem of halftoning a signal defined on a surface for the purposes of accurate color reproduction in @xmath0d printing when using translucent materials . more specifically , we present algorithms for error - diffusion on voxel representations of surfaces , which are compatible with translucent printing materials . accurate color reproduction is important in many applications of @xmath0d printing , especially for design - prototypes or @xmath0d color copies , where a texture - mapped @xmath0d scan of an object is to be reproduced in a color - consistent way ; this includes @xmath0d portraits . while @xmath0d color printing has existed for some time using powder - binder @xcite and layer - laminated @xcite systems , multi - jet @xmath0d printers with sufficient materials for color printing have only recently come on the market @xcite . in comparison to powder - based systems , multi - jet printers produce smoother surfaces at higher resolutions , and allow more control over the internal structure of the print . in the longer term , such printers have much greater potential to reproduce complex appearance properties . in fact , multi - jet printers with fewer materials have been used to approximate desired single - tone subsurface scattering properties @xcite , and opaque objects inside transparent shells @xcite . unfortunately , software allowing full color printing with multi - jet printers is not yet available ; the factory software allows colors to be selected from a palette of @xmath2 and manually assigned to shells . without accurate color reproduction , more complex appearance characteristics are limited in their applicability and appeal . therefore , accurate color reproduction is a crucial step to fully exploiting the appearance capabilities of these printers . we provide precisely this critical building block for graphical @xmath0d printing with this paper . however , there are substantial technical challenges in color reproduction with multi - jet printers . first , is the amount of data that must be processed . for accurate color , the material assigned to each voxel must be controlled . modern @xmath0d printers have a resolution of more than @xmath3 million voxels per cubic centimeter . thus , prints such as those in figure [ fig : teaser ] require tens of billions of voxels ( see table [ tab : performance ] for details ) . both resolution and build volume of @xmath0d printers are expected to increase in the coming years . second , for multi - jet printers currently on the market , the colored materials are highly translucent , which means that the organization of the materials a significant distance beneath the surface , up to a few millimeters , can greatly affect the perceived color as discussed in section [ sec : prelim ] . this translucency leads to both blurring of fine - scale details in textures , and to artifacts caused by severe dot - gain if care is not taken in the material arrangement . it is therefore highly important to control material placement several layers of voxels beneath the surface , considering the transmission and scattering properties of the materials , greatly complicating the computational aspects of halftoning algorithms . while we expect less translucent materials to come on the market in the future , not only will our proposed technique also work for these materials , it will actually perform better producing equal or larger color gamuts with less computational effort . further , employing translucent materials may have its own advantages , when we consider materials with similar translucency as observed in skin . third , while powder - binder and layer - laminated systems , like @xmath1d color printing , may combine up to @xmath0 inks at a single surface location , multi - jet printing allows exactly one material per voxel . in this paper , we leverage the knowledge of decades of research in color imaging , color management and @xmath1d color printing , to maximize the quality and exploit the full capabilities of high - resolution multi - material 3d printers and push their limits towards realism . to this end , we propose a geometry - adaptive error diffusion halftoning algorithm , which includes the following technical contributions : * a traversal algorithm for voxel representations of surfaces , which maps 2d anisotropic error diffusion filters onto a surface in a consistently oriented way ( section [ sec : traversal ] ) and requires only local information to do so . * a layered halftoning algorithm , which combines the traversal algorithm with an arbitrary 2d error diffusion algorithm , and can adapt to the translucency of the materials or increase the color gamut by varying the number of layers ( section [ sec : layered ] ) . we do not attempt to approximate the full bssrdf of an object , but rather present halftoning algorithms capable of reproducing an object s albedo texture even when the printing materials are highly translucent , under non - pathological illuminaton conditions , such as a light source in contact with ( as in figure [ fig : material_translucency ] ) or inside the printed object . note that a high - quality halftone , such as those produced by our algorithms , is necessary , but not sufficient to accurately reproduce an object s color . additionally , one needs colorimetrically or spectrally accurate measurements of the object s color , which are typically not provided by common 3d scanners , and an accurate optical printer model . in section [ sec : colormanagement ] we describe a fully empirical model for our printer . a more accurate model is left for future work . * existing 3d color printing technologies . * powder - binder @xcite and layer - laminated @xcite systems provide full color solutions . in both cases , cmyk inks are applied to a white base material . although these technologies are effective for color printing , they do not provide the resolution or smooth surfaces of multi - jet systems and are limited to nearly opaque materials . they therefore have less potential for future combinations of color with complex appearance properties . * appearance reproduction and multi - material fabrication . * spec2fab @xcite is a general reducer - tuner framework for specification - driven digital fabrication , which allows textures to be replicated on a 3d print . they use an error diffusion optimization of material layerings , effectively a contone , with a uniform error filter ( error is pushed equally to all neighbors ) . although an important first step in texture - mapping for multi - material printers , it does not allow for anisotropic error diffusion filters , and the iterative optimization prohibits a streaming architecture . openfab @xcite is a programmable fabrication pipeline for @xmath0d printers , which uses in - slice @xmath1d error diffusion dithering @xcite . the authors observe that by dithering in @xmath0d they could avoid streaks . our approach treats the color signal where it is defined on the surface by mapping @xmath1d filters into the tangent space of the surface . as discussed below , this allows us to colorimetrically characterize the 3d printer in a geometry - independent way . multi - material @xmath0d printing has been used to reproduce specified subsurface scattering properties @xcite . fabrication of directional brdfs for planar or near - planar surfaces has been done using multi - material printing @xcite and photolithography @xcite . given the work that has already been done using multi - material @xmath0d printing to reproduce complex appearance properties , the introduction of full color opens up tremendous possibilities for future appearance fabrication . * tone reproduction in fdm prints . * some recent work has focused on the challenging task of improving tone reproduction in fused deposition modeling ( fdm ) printing . hergel and lefebvre @xcite optimize seam placement in multi - filament fdm prints to hide or reduce artifacts from changing filaments . reiner et al . @xcite perform a type of halftoning for fdm printers while maintain long filament paths . switching filament heads not only creates artifacts , but also increases print time . both of these methods are specific to fdm printers . * 3d halftoning . * @xmath0d halftoning has been applied to material composition using @xmath0d error diffusion filters @xcite and @xmath0d dispersed - dot dithering @xcite . for color and appearance reproduction , 3d error diffusion is not appropriate because material assignments closer to the surface will have a greater influence on the appearance of the object than material assignments deeper within the object . thus , a 3d error diffusion filter would have to adjust its orientation during traversal to account for this and maintain a consistent orientation with respect to the surface . an isotropic filter would produce similar artifacts as are observed with isotropic filters in 2d error diffusion . in contrast , our approach of halftoning on multiple offset surfaces within the object , in addition to the surface itself , results in a halftone that inherently accounts for the geometry of the surface . the relative influence of voxels at different depths from the surface is calibrated in an offline process and built into an international color consortium ( icc ) profile . such an offline color calibration process would be very challenging for a 3d filter , because it would require calibrating every possible surface orientation . * 2d halftoning . * generations of researchers in the field of 2d halftoning focused on finding methodologies to optimally arrange printed dots for maximizing print quality ( by preserving tone and structure and shifting quantization errors to the highest spatial frequencies possible - see section [ sec : perceptionprinting ] ) subject to technical limitations of printing systems ( e.g. the ability to accurately deposit isolated dots ) @xcite . one category of algorithms , called _ point processes _ , allow a very fast computation of the halftone screens by thresholding pixels using a precomputed threshold mask that is tiled over the 2d image . the traditional clustered - dot order dithering to create amplitude modulated screens and dispersed - dot ordered dithering @xcite fall into this category . the latter technique was adapted to 3d printing @xcite accounting particularly for dot placement limitations of binder - jetting systems . one drawback of dispersed - dot ordered dithering is that frequency components given by the screen period are visible resulting in cross - hatch pattern artifacts . avoiding such artifacts in point process techniques , requires large threshold masks e.g. blue - noise masks @xcite or green - noise masks @xcite which are heavily distorted if applied on surface manifolds with non - zero gaussian curvature . a second category of algorithms , called _ neighborhood processes _ , considers a local pixel neighborhood for thresholding . error - diffusion methods ( see appendix ) belong to this category and were pioneered by floyd - steinberg @xcite . since then , the first error diffusion techniques affected by visual artifacts from low - frequency structural patterns ) were improved drastically producing visually pleasing halftones without artifacts . multiple modifications have been proposed including edge enhancement @xcite , tone - dependent diffusion filters @xcite , threshold modulation @xcite structure preservation @xcite or memory efficient processing @xcite . a last category employs models of the human visual system to optimize halftone pattern iteratively @xcite . such methods can produce the highest halftone quality , but are in general computationally much more expensive than methods belonging to the other categories . furthermore , these methods iterate over the whole data and require substantial modifications for adapting them into a streaming architecture . since iterative methods aim to preserve local structure , their quality advantages over other methods diminish with increasing print resolution . due to the low computational effort and high quality of 2d error diffusion achieved with small diffusion filters , we decided to adapt it to 3d color printing . one prior work addresses error diffusion on a surface @xcite . this approach operates on meshes and traverses the vertices based on the availability of subsequent moves or neighbors to diffuse error to . while this approach could be applied to any graph structure , including voxels , it is not clear that it can be applied in a streaming architecture . to the best of our knowledge , we are the first to consider error diffusion halftoning in the context of both non - euclidean domains and highly translucent materials . moreover , we are the first to propose such a technique demonstrated to be applicable in practice to tens of billions of elements . given an arrangement of multiple non - fluorescent printing materials with similar refractive indexes within a shape @xmath4 , light propagation within this shape can be described by the steady - state radiative transfer equation @xcite ( see supplemental material for details ) , which shows that the intensity of light traveling through the material is attenuated by absorption ( dependent on wavelength , independent of direction ) and is redistributed by scattering . thus , a fraction of light entering the print at one location may be emitted from the surface at a different location due to scattering ( see figure [ fig : material_translucency ] ) . if light travels through different materials its spectral power distribution is modulated by each material s absorption coefficients and the path length within this material . this has multiple implications for arranging translucent ( high scattering , low absorption ) printing materials colored in cyan ( c ) , magenta ( m ) , yellow ( y ) and white ( w ) for full color 3d printing . _ highest reflectance ( white point ) : _ to maximize reflectance , not only the surface must be covered with white material , but also the space beneath . a significant fraction of light travels deep into the object due to the low absorption . each non - white material with higher absorption placed into the light path would absorb light that would otherwise contribute to the albedo . there exists a minimum thickness @xmath5 of a white layer to create almost maximum reflectance , i.e. for each @xmath6 : @xmath7 , where @xmath8 and @xmath9 are spectral reflectances of white layers with thicknesses @xmath5 and @xmath10 , and @xmath11 is the maximally acceptable reflectance difference . _ lowest reflectance ( black point ) : _ to minimize reflectance , the printing material arrangement must ensure that light in the whole visible wavelength range is maximally absorbed . this can only be achieved by appropriately arranging c , m , y materials on the surface and beneath . the minimal layer thickness @xmath12 of such an arrangement for creating almost minimal reflectance ( deviating only by @xmath11 from minimum ) satisfies @xmath13 because the color materials absorption coefficients are larger than that of white ( assuming similar scattering properties of all materials ) . therefore , not all voxels in the shape addressable by the printer need to be considered . we can drastically reduce computation by filling all voxels within the shape by default with white ( maximizing reflectance of the white point ) and compute the arrangement of remaining materials only within @xmath12 of the surface . although material placement with a distance @xmath14 $ ] from the surface may impact reflectance , it affects only very bright colors which are also reproducible by color halftones in upper layers . while theoretically one could derive @xmath12 based on measurements of the material parameters , section [ sec : layered : construct ] shows how selecting a maximum distance @xmath15 from the surface allows an empirical trade - off between the number of voxels that must be addressed and the achievable gamut volume . _ optical dot gain : _ voxels occupied by a colored material and surrounded by white material appear larger because the spectral power distribution of light traveling through both materials is modulated mostly by absorption of the colored material . this phenomenon is known in 2d printing as _ optical dot gain _ the more colored voxels are stacked on top of each other aligned with the surface normal the more laterally scattered light is modulated by the colored material . thus , the colored area on the surface appears not only darker , but also larger as shown in figure [ fig : dot_gain ] . algorithms that use surface halftones to assign materials to inner voxels must consider this phenomenon to avoid dot gain artifacts particularly visible in bright areas by isolated disturbing large dots . section [ sec : layered : assign ] describes how independently halftoning layers allows to decorrelate material assignments in different layers , which greatly reduces dot gain artifacts in translucent materials . each printing system is limited in reproducing reflectances due to the small number of available printing materials . still , accurate color reproduction is possible by material arrangements creating physical errors for which the human visual system ( hvs ) is insensitive . two perceptual aspects are particularly important in printing . _ contrast perception : _ contrast sensitivity functions ( csfs ) describe the hvs s sensitivity to achromatic and chromatic contrasts as a function of spatial frequency , orientation and luminance @xcite . the achromatic csf has bandpass and chromatic csfs have lowpass characteristics . each csf monotonically decreases for frequencies higher than 10 cycles / degree resulting in an apparent blurring of high frequency contrasts . this is used in printing by arranging printing materials in high frequency patterns ( halftones ) to create various color shades from only a few colored printing materials . _ color perception : _ retinal processing reduces the information content of a spectral stimulus to only three values the cone responses @xcite . therefore , different reflectance spectra viewed under the same illuminant match visually if corresponding spectral stimuli yield similar cone responses . by agreeing on the viewing illuminant , accurate color printing is possible by material arrangements that reproduce cone responses instead of reflectances . a drawback of this _ metameric color reproduction _ is a likely mismatch between prints and originals for other illuminants . cie colorimetry provides two standardized color spaces for metameric printing ( see e.g. @xcite ) : ciexyz , which is linearly related to cone responses ; and cielab , an opponent color space derived from ciexyz that allows simple access to color attribute predictors for lightness , chroma and hue and is perceptually more uniform . color control in metameric printing is done by _ separation_relating ciexyz or cielab values to material arrangements . since this is generally impossible for all colors due to gamut limitations of the printing system , a preceding _ gamut mapping _ method must transform all colors into the device gamut aiming to minimize perceptual errors between original and reproduction @xcite . given the resolution of current 3d printers , and expected increases in both resolution and build space , we need a streaming model for computation , wherein only localized data and computation are needed . figure [ fig : flowchart ] shows such a streaming framework , which takes as input shape and color information , and produces printer compatible data in which a single material is assigned to each voxel . the error diffusion and material assignment algorithms described in sections [ sec : traversal ] and [ sec : layered ] are our core contributions and take place in the parts colored reddish . each aspect of our framework is described briefly below . * shape @xmath16 color : * the input to our pipeline is a shape @xmath17 defined by a surface @xmath18 ( e.g. tessellated mesh ) with attached color information @xmath19 ( texture data or per - vertex colors ) , where @xmath20 is specified by an invertible transformation from ciexyz defined for an illuminant ( e.g. cied50 ) and a set of color matching functions ( cmfs ) ( e.g. cie 1931 cmfs ) . examples for @xmath20 are standard rgb color spaces @xcite . by default we interpret rgb data as srgb . * voxelization : * the axis - aligned bounding box of our shape @xmath21 is discretized with a fixed , regular grid of voxels @xmath22 at the resolution given by the printer specification . we slice the surface and identify interior , @xmath23 , and surface voxels , @xmath24 , where a _ slice _ refers to a 2d array of voxels with a constant @xmath25-coordinate . the full slice is denoted by @xmath26 for slice @xmath27 with center @xmath25-coordinate @xmath28 , and @xmath29 denotes the surface voxels in slice @xmath27 . our voxelizer generates slices in chunks of @xmath30 slices@xmath31 for our hardware setup depending on the number of layers and layer thickness as discussed in section [ sec : layered ] , until the object is completely sliced . each chunk proceeds through the pipeline until materials have been assigned , and printer - specific output has been generated . then the next chunk is generated . during the voxelization process , we assign colors to the surface voxels ; abusing notation slightly , we redefine @xmath32 as a function on the surface voxels . we sample the texture data using trilinear interpolation ( mipmapping ) with the level - of - detail computed per surface voxel . this is key to avoid aliasing while there are no viewpoint - dependent effects as in rendering , the texture map is in general non - uniform over the surface . * color management : * in this stage , which includes gamut mapping and color separation , we map color data to printer tonal values using icc color management @xcite ( see section [ sec : colormanagement ] ) : @xmath33 , where each element of the tonal value space @xmath34 corresponds to the amount of available printing materials required to best reproduce a desired color . by function composition we attach one tonal value vector to each surface voxel : @xmath35 . * layer construction : * to account for the translucency of the materials , see figures [ fig : material_translucency ] and [ fig : dot_gain ] , we transfer the tonal values stored in the surface voxels to the interior voxels within a distance @xmath36 of the surface . see figure [ fig : layer_illustrate ] for an example . these voxels have the greatest influence on the appearance of the printed object . from these voxels , we extract a set of layers as isosurfaces of distance to the surface of the object . the number and thickness of the layers determine @xmath36 , and these are chosen as described in section [ sec : layered : construct ] . * halftoning : * we then treat each layer as a distinct surface with tonal values , and halftone them independently using the surface traversal algorithm described in section [ sec : traversal ] and the error diffusion and material assignment algorithm described in section [ sec : layered ] . * material arrangement : * the output of the halftoning is a voxel - level material arrangement @xmath37 , where the set of materials @xmath38 is discrete and small@xmath39 in our case . we store @xmath40 per - slice , and convert it to a printer - specific format to send to the device . classical error diffusion algorithms for 2d images , e.g. @xcite , rely on the partial ordering property of subsets of euclidean domains , to define a traversal order , eg . raster scan or serpentine scan order , in which error is never pushed to pixels that have already been halftoned . on a surface , which has non - zero curvature and is in general not a topological disc , such a partial ordering property does not exist without introducing a seam or a singularity , a consequence of the hairy ball theorem . any traversal scheme will eventually encounter an element that has already been halftoned , requiring either a stop and restart , or a change in direction . excessive stopping and restarting can lead to poor distribution of error , and we design a traversal that avoids this . to map anisotropic error diffusion filters on to the surface requires a local coordinate system of the tangent plane at a given point , which is fixed with the surface normal and one additional direction . mapping a filter to the surface so that it is consistently oriented for adjacent points is equivalent to parameterizing the tangent space of a surface in a consistent way , or finding smooth vector fields on a surface , which is in general an ill - posed problem and finding an optimal solution requires considering the entire surface . instead , we propose a voxel - level traversal algorithm that maintains a consistent orientation of the filter , traverses the surface voxels in long runs allowing error to accumulate and diffuse evenly , and does so using only local information . our traversal avoids situations where voxels only distribute or obtain error during diffusion . we maintain a consistent orientation of the filter by traversing the voxels in a consistent direction , and using the traversal direction to fix a coordinate system in the tangent plane . while our traversal provides no theoretical guarantees , it performs well in practice . as a consistent direction , we always travel clockwise or counterclockwise about a line parallel with the positive @xmath25-axis . for brevity , we describe only the counterclockwise case . note that we want to traverse counterclockwise , not about a global axis , but rather about a line through the center of the local _ connected component _ of voxels . depending on the geometry of @xmath4 , when it intersects a slice plane @xmath41 and is discretized into a voxel representation , the resulting voxels in slice @xmath27 may belong to multiple connected components . an example of a slice with multiple connected components can be seen in figure [ fig : layer_illustrate ] . if the surface normal is close to vertical ( almost parallel with the positive or negative @xmath25-axis ) , then @xmath42 will be multiple voxels wide . these surface voxels @xmath42 may enclose interior voxels in slice @xmath27 , and will be themselves enclosed by exterior or empty voxels in slice @xmath27 . figure [ fig : slice_geom ] shows a 2d cross - section of this . if we consider a part of the shape , which forms a connected component in consecutive slices , and where the surface faces almost in the negative @xmath25 direction , but slopes or curves slightly upward , we have the following situation , which is shown in 2d in figure [ fig : slice_geom ] . when projected into the @xmath43-plane , the surface voxels of @xmath42 will be located inside the surface voxels of @xmath44 . if we consider two surface voxels within the same connected component in slice @xmath27 , the innermost should be traversed first , since they are next to @xmath45 and have already received error diffused from slice @xmath46 . the outermost should be traversed last , which are next to @xmath44 . the situation is reversed for upward facing surfaces . we construct an undirected graph for traversal as follows : for each surface voxel @xmath47 , we connect it with an edge to each other surface voxel within the @xmath48 voxel window @xmath49 centered on @xmath50 . we denote this set of one - ring neighbors @xmath51 . we traverse the voxel representation of the surface slice by slice , analogous to halftoning an image row by row , by selecting the next voxel from the set of neighbors that are within the same slice , @xmath52 . error is always diffused upwards to the next slice , and to other surface voxels within the slice , which have not yet been assigned materials , as shown for a very small example ( @xmath53 ) in figure [ fig : traversal ] . one might also consider standard graph traversals , such as breadth - first or depth - first , also constrained to neighbors within the same slice , but these do not maintain a consistent direction of travel and therefore only permit symmetric error diffusion filters ( e.g. uniform , gaussian , etc . ) . as described in section [ sec : traversal : geometry ] , the traversal winds up the surface on a slice - by - slice basis counterclockwise about a line through the center of each connected component . however , finding the center of the current connected component would require a search over all voxels in the slice to determine the connected components , find their centers , and to which connected component each surface voxel belongs . such a search would prove too expensive essentially involving a pre - traversal . we can determine whether a potential next voxel in proper traversal order winds about the connected component in the correct direction using only two pieces of local information . the first is the surface normal . the second is the in - slice distance to the exterior of @xmath4 , or distance - to - empty , denoted by @xmath54 for voxel @xmath55 @xmath56 which can be pre - computed for each slice see figure [ fig : slice_geom ] for an example . we use the @xmath57-norm for simplicity and efficiency , but another norm would also work . the gradient of @xmath54 always points to the interior of the connected component to which @xmath50 belongs . this is important for situations where the current slice may have multiple disconnected components , as in figure [ fig : layer_illustrate ] . this allows us to efficiently test if a potential next voxel @xmath58 continues in the same direction around the connected component by computing @xmath59 . for example , if we are traversing counterclockwise , we can rule out all neighbors of @xmath50 for which this cross - product has a negative @xmath25-component . we denote the subset of @xmath60 satisfying this direction criterion @xmath61 . referring to figure [ fig : slice_geom ] and recalling the observations of section [ sec : traversal : geometry ] we get the following additional traversal criterion . if the surface faces downward , we wish to stay as _ inside_away from exterior voxels as possible . if the surface faces upward , we wish to stay as _ outside _ as possible . hence we choose the neighbor @xmath62 for down - facing surfaces , and @xmath63 otherwise . because the surface orientation is determined locally using the surface normal , the traversal adjusts as it traverses a slice with both up- and down - facing segments . figure [ fig : traversal_order ] shows the traversal order for such a slice . * start point selection . * we also use the distance - to - empty , in combination with one other local quantity , to select the start point in each slice . first , potential starting voxels ( voxels in the slice , which have not yet been traversed ) are filtered based on their _ error diffusion count _ , or number of times they have had error diffused to them from previous slices . we select the subset of voxels with the largest error diffusion count , and subsequently select from this subset using the distance - to - empty . if the @xmath25-component of the surface normal is negative , the start point is the point , which maximizes the distance - to - empty . if the @xmath25-component of the surface normal is positive , the start point is the point , which minimizes the distance - to - empty ( i.e. one of outer - most points ) . once the start point and traversal direction are selected , the direction is tested to ensure it is not a dead end . if no voxel is available in that direction , the direction is reversed . this reduces the occurrence of single voxels with few neighbors to push error . * birth and death of components . * during the slicing process , new parts of the shape are encountered , creating new connected components of voxels in a given slice . when a new component is encountered in a given slice , the surface voxels will form a disc in that slice , meaning it can be traversed like an image . we handle these cases specifically as follows . rather than spiral out , as with other downward facing surfaces , we select a global axis , @xmath64 or @xmath65 , and perform a 2d serpentine scan in this direction . this avoids situations where a long strip of voxels mostly has error pushed away from it , creating a start - up artifact noticeable for low tonal values . we also use a 2d serpentine traversal for components , which are ending in the current slice , rather than spiraling in . we detect the birth of a component when all possible starting points have an error diffusion count of @xmath66 , and the death of a component when all possible starting points have no neighbors in slice @xmath67 . * serpentine surface scan . * when we can no longer traverse in the same direction locally ( all @xmath68 have been traversed ) , we reverse direction , as in the serpentine traversal scheme from 2d error diffusion , which is shown to provide significant improvement over raster - scan traversal in images @xcite . further , when selecting a new start point , we set the traversal direction opposite the traversal direction , in which error was last diffused to the new starting voxel . once we have chosen the next voxel to traverse , we can establish a local coordinate system at the current surface voxel , with which we can align the error diffusion filter and the neighbors of the current voxel . as in image error diffusion , one non - zero filter element is aligned with the next voxel to be traversed . this gives us a direction , which together with the surface normal , gives us the necessary coordinate system for the tangent plane at each surface voxel . depending on whether the direction is clockwise or counterclockwise , we use either a left - handed or right - handed coordinate system . figure [ fig : traversal ] shows how an error diffusion filter with @xmath0 non - zero elements is mapped to the neighbors in the tangent space of a voxel as it traverses the surface voxels . this is closely related to decal mapping methods such as discrete exponential maps @xcite , and to parallel transport on manifolds , and for large decals exponential maps provide a more general solution . however , the filters used for error diffusion are typically small enough that the neighbors can simply be projected into the tangent plane while preserving distances . we let the error diffusion filter live in the local tangent coordinate system , orthogonally project the neighbors onto the tangent space , and find the best alignment of neighbors and non - zero filter elements . we align neighbors with filter elements using symmetric closest points to avoid a single neighbor being assigned to multiple filter elements . that is , both the filter element and the neighbor must be closest to each other . the distance to empty can be computed in time linear in the number of voxels in each slice using a distance transform @xcite . while this means over the whole print job , every voxel in @xmath22 must be visited , the constant hidden in the @xmath69 notation is very small and the distance transform is highly efficient . selection of the starting point requires a search over all unprocessed surface voxels , but for non - pathological cases this will only be necessary a small number of times per slice . each traversal operation requires a constant number of operations , as the number of neighbors is constant . normals can be computed in @xmath70 per slice using a signed distance field as an implicit surface definition , as described in section [ sec : layered : construct ] . thus , the total time complexity to fully process @xmath4 is linear in the total number of voxels @xmath71 with a small constant , which is very well suited to a streaming architecture . applying a halftone only to the surface voxels will result in a very low contrast print , due to the translucency of the materials . we can see this from figure [ fig : dot_gain ] , which shows how increasing the depth of color material assignment increases the contrast . in figure [ fig : layer_gamut ] , we see how increasing the depth of color assignment increases the color gamut . for this reason , we introduce _ layered halftoning _ , in which color data is transferred from the surface to additional isosurfaces , or layers , within the object , which are then independently halftoned using the traversal scheme presented in section [ sec : traversal ] and an adaptive filter @xcite and threshold error diffusion algorithm @xcite . layered halftoning begins by transferring the surface tonal values @xmath72 from @xmath73 to a subset of @xmath74 within a distance @xmath36 of the surface . all interior voxels more distant from the surface are assigned a white material to maximize reflectance as explained in section [ sec : prelim ] . this requires both many distance computations and a transfer operator @xmath75 , which maps functions on @xmath73 to functions on @xmath74 . to compute the distance field @xmath76 of voxels to the surface , we leverage the fact that we are only interested in voxels within a given distance @xmath36 this allows us to construct a @xmath0d mask containing these distances at the printer resolution . the distance field @xmath77 is initialized with a large value @xmath78 . distance computation then proceeds by moving this mask over the surface voxels @xmath73 , and writing the mask value to the voxel at the corresponding offset , if that value is less than the value already there . while this approach is efficient only for relatively small distances , it has the advantage of being embarrassingly parallel and requiring only a less - than operation at each voxel . as a transfer operator , we simply take the tonal value @xmath79 of the nearest surface voxel @xmath80 , allowing the transfer operator to piggy - back on the distance computation at minimal extra cost , i.e. interior tonal values are assigned as @xmath81 for @xmath82 . aside from being easy to compute , this has the benefit of preserving high - frequencies in @xmath72 . figure [ fig : layer_illustrate ] shows the process of assigning srgb color values to surface voxels , converting them to tonal values , transferring the tonal values to interior voxels and applying error diffusion to get the final material assignment . note that because tonal values are pushed to the interior in 3d , interior tonal values may include surface tonal values from different slices . although there are inevitably discontinuities in @xmath83 at the cut loci of the distance field @xmath10 ( voxels @xmath50 where multiple surface voxels are equidistant ) , we found this not to be a problem as the tonal values mostly vary smoothly without significant correlation between high frequencies in @xmath72 and high curvature of the surface . in this case we simply assign tonals in a first - come fist - served manner . similarly , although areas with varying curvature result in slight tone shifts from @xmath72 to @xmath83 , in particular for areas of high positive mean curvature , we did not find this to be a problem , even for small prints . we extract layers of voxels with tonal values from @xmath74 by defining isosurfaces of @xmath10 with which to segment @xmath74 . as isosurfaces we choose @xmath84 , where @xmath85 is the layer thickness , which we choose to be the voxel size along the dimension of minimum resolution . we use @xmath86 , where @xmath87 is an integer constant , which defines @xmath88 . ideally @xmath89 ( see section [ sec : lightscatteringabsorption ] ) to minimize reflectance for maximizing contrast . however , we make a tradeoff between computational effort and gamut volume and set @xmath90 for all prints shown in this paper . figure [ fig : layer_gamut ] shows color gamuts computed using @xmath91 with the gamut volume as a fraction of the volume of srgb . we can see how the volume increases with the number of layers . however , we found that the black point of the @xmath3-layer profile , @xmath92 , is close to the minimum black point achievable with current cmyw materials@xmath93 for a @xmath94 cube of full cmy mixture . thus , little additional gamut is likely to be gained by more layers . voxel layer @xmath95 is then defined as the set of voxels between isosurfaces @xmath95 and @xmath96 , @xmath97 where the second condition is to ensure that the layers form thin voxel approximations of isosurfaces as much as possible , and @xmath98 is the voxel window defined in section [ sec : traversal : algorithm ] . due to different resolutions along each axis , @xmath99 may be multiple voxels thick depending on the orientation . in the case of @xmath100 , this condition is replaced by @xmath101 . note that we can very efficiently compute normals for all isosurfaces , required for traversal of the error diffusion filter ( see section [ sec : traversal : algorithm ] ) : @xmath102 , @xmath103 , where @xmath104 is a signed distance field , for @xmath105 . we use finite differences to compute @xmath106 . we treat each voxel layer as a set of surface voxels , traverse them as described in section [ sec : traversal ] , and halftone them as follows . each tonal channel is halftoned independently using a standard error diffusion filter , mapped onto @xmath107 as described in section [ sec : traversal : mapping ] and in the appendix . at a given voxel @xmath103 , this results in a halftone vector @xmath108 where @xmath109 . thus , multiple materials may have a tonal value of @xmath110 indicating they should be assigned to @xmath50 . since we can only assign a single material @xmath111 to each voxel , we apply the following tie - breaking scheme . for each tonal channel @xmath112 , we maintain a tie - breaker error , @xmath113 , which is initialized to @xmath114 for all @xmath115 at the start of each slice . if at a given voxel , one or more tonal values are halftoned to @xmath110 , the tonal channel @xmath116 with the largest tie - breaker error is declared the `` winner '' and we assign the corresponding material to the voxel . we then reset @xmath117 and increment @xmath113 for all other @xmath115 . if there is no @xmath115 such that @xmath118=1 $ ] , we assign white material to the voxel . note that this tie - breaking scheme is independent of the per - tonal error diffusion , and the error diffusion does not know when a tonal is quantized to @xmath110 , but the material is not assigned . this in general leads to a slight loss of tone , but is mitigated by the large number of voxels , and it is anyway accounted for by the color management , as described in section [ sec : colormanagement ] , because we characterize the printer using targets printed with the same algorithm . note that errors are only diffused within the same layer , and each layer is halftoned independently . this has the following benefits . first , it allows different layers to be halftoned in parallel . second , it further avoids the issue of resampling a halftoned signal , which would be necessary if the halftone was performed on the surface and then transfered to the interior . this is problematic because the halftoned signal is by design high - frequent and resampling it introduces artifacts . third , we can use the threshold modulation of the halftoning algorithm @xcite to avoid inter - layer correlations , which can create dot gain artifacts ( see section [ sec : lightscatteringabsorption ] ) . we do this by seeding the pseudo - random number generators used for the threshold modulation differently for each slice and layer . as discussed in section [ sec : layered : construct ] , due to differences in resolution along the different axes , some interior voxels will fall in between layers , and not be directly halftoned . for these voxels , we simply assign the material of the nearest voxel that does belong to a layer . we use an objet500 connex3 from stratasys @xcite , for which the vendor has provided an interface allowing us to specify material assignments at the printer resolution . as the connex3 allows only 3 model materials and 1 support material , we expanded the possible gamut by coloring the support material with a yellow acid dye . thus , we print with white , cyan , magenta and yellow materials . care must be taken that there are no large agglomerations of support within the object , which would weaken the structural integrity of the print . fortunately , as our halftone is by design high - frequency , this is avoided by scaling the yellow tonal value by a constant factor between @xmath66 and @xmath110 . we found @xmath119 to provide both a chromatic yellow and structurally stable prints . our layer - based approach also allows us to prevent support from being present in the outermost layer ( the surface ) , by applying a layer and tonal dependent soft - thresholding to the tonal values . for error diffusion we use a tone - adaptive filter @xcite and threshold modulation technique @xcite . specifying the color - to - tonal transformation @xmath120 requires at least two icc profiles : an input profile that specifies texture - data colorimetrically and a printer profile that includes gamut mapping and color separation . generating a printer profile involves a colorimetric characterization of the printer , i.e. we need to predict the printout s ciexyz color for each tonal value vector in @xmath121 . for this , the tonal value space is sampled and the color of corresponding printouts is measured and used to fit a colorimetric printer model . we used a fully empirical model , which interpolates between color - measured printouts of densely sampled tonal values for our cmyw printer we use a uniform @xmath122 sampling of the channel - wise linearized cmy tonal value cube . linearization is performed employing the broadband murray - davies model according to @xcite . this target is printed using the algorithm described in sections [ sec : traversal ] and [ sec : layered]exactly the same algorithm as used to generate the results shown in figures [ fig : visual : objects ] and [ fig : gallery ] , without color management as the input are direct tonal values . a picture of the resulting printout is shown in figure [ fig : color_manage ] ( left ) . measuring colors of such highly translucent printing materials is a challenging problem . it can be shown that measurements made by spectrophotometers used in the graphic arts community are systematically biased towards lower reflectance due to subsurface light transport away from the detection area . for our color measurements , we used a bidirectional @xmath123 measurement geometry , an almost colorimetric dslr camera ( vora value of 0.9455 @xcite ) , diffuse broadband illumination simulating cied50 , flat fielding ( to account for optical path variations between light source , printed color patches and camera ) , and a polynomial approach to map camera rgb values to ciexyz . color errors determined by spectroradiometric measurements are within the interinstrument - variability of spectrophotometers used in graphic arts for opaque materials . for more details , the reader is referred to @xcite . we evaluated the effectiveness of color management as follows . we took a standard color checker and measured each of the 24 patches with a spectrophotometer . we then mapped each of these colors into our printer gamut ( 12-layers , figure [ fig : layer_gamut ] , second from right ) , and printed planar patches . we then measured the printed patches with the almost colorimetric dslr camera , and computed the ciede2000 @xcite differences to the gamut mapped colors i.e . the colors predicted by our empirical printer model . we observed a median error of @xmath124 , a mean error of @xmath125 , and minimum and maximum errors of @xmath126 and @xmath127 . the predicted and printed colors are shown in figure [ fig : color_manage ] ( right ) . note that , even though the proposed half - toning algorithm was used to print the planar patches , this evaluates _ only the empirical printer model _ and not the halftone , which is evaluated in sections [ sec : experiments : eval ] and [ sec : experiments : visual ] . * performance . * our non - optimized implementation supports multi - threading for some computations ( for example the transfer operator and halftoning each layer in parallel ) , but does not exploit any gpu capabilities , although several aspects would lend themselves well to gpu implementation . table [ tab : performance ] shows performance data for several prints : the number of voxels , both those assigned a material ( @xmath128 ) and in the build volume ( @xmath129 ) , computation times include the time to deliver the first chunk of slices to the printer and the total , and the print times . the results were collected on a standard desktop pc with an intel core i7 - 4770 processor and @xmath130 gb of memory . see text for a description of how we compute the tonal errors . @xmath131print job terminated after @xmath132 of slices . [ tab : performance ] * tone preservation . * it is in general difficult to evaluate how well a 3d print reproduces the color of the input texture , due to the large change in appearance resulting from different illuminants and illumination distribution in combination with the curved surface . as discussed further in section [ sec : experiments : limits ] , the printing materials we use are highly color inconstant , meaning the perceived color changes under different illuminants . further , typically the textured model is not rendered using the reflectance properties of the printing materials . in the case of scanned objects , evaluating with respect to the original object is meaningless , since the capture systems are typically not color calibrated . therefore , we evaluate the quality of tone preservation , rather than the similarity of perceived colors , as follows . we compute the root mean squared error ( rmse ) between material assignments @xmath133 and tonal values @xmath83 , which are given for several models in the last column of table [ tab : performance ] . first , we compute material fractions for each slice the number of voxels assigned a given material in the slice divided by the total number of voxels @xmath134 such that @xmath135 . then we determine expected material fractions from the tonal values @xmath136 using the demichel equations @xcite . for example , we compute the expected material fractions for cyan and white from tonal values as @xmath137 and @xmath138 respectively , where @xmath139 , @xmath140 and @xmath141 are the average tonal values over the same set of voxels . the formulas are similar for magenta and yellow . finally , we compute the errors as the difference between the expected material fraction and the actual material fraction for each tonal channel . we see that our algorithm preserves tone very well , with most errors being less than @xmath142 of the tonal value range . visual experiments were conducted to assess the structural quality of the proposed halftone method , in particular the level of graininess and the visibility of structural artifacts . for this , we generated a test surface with the following function @xmath143 where @xmath144 , @xmath145 , @xmath146 with both @xmath64 and @xmath65 ranging from @xmath147 to @xmath148 . we then applied two textures to this surface , as shown in figure [ fig : visual : objects ] ; one with smooth color gradients and one with high - frequency patches ; the colors were selected to uniformly cover the a*b*-plane of lightness l*=70 and gamut - mapped using the absolute - colorimetric intent of our icc profile . the printed surfaces were placed on a colorimetrically characterized display next to their rendered versions ( gap of approx . 1.25 cm - the left / right position of rendering and print was randomly chosen for each subject ) . we used an _ eizo coloredge cg301w _ @xmath149-inch display with a resolution of @xmath150 pixels . the printed surfaces were @xmath151 , which covers @xmath152 pixels on the screen . the screen was placed horizontally and illuminated from directly above using a _ just normlicht led color viewing light m _ viewing booth at a distance of @xmath153 from the screen to provide a diffuse illumination . the illuminant used was led - simulated cied50 . a chin - rest was placed so to obtain @xmath154 pixels - per - degree ( ppd ) at the screen center ( @xmath155 distance ) and a @xmath156 viewing angle to the screen . the renderings were done to approximately match this viewing condition and the luminance level of the print . figure [ fig : visual : objects ] shows pictures taken with an almost colorimetric dslr camera of what the subjects viewed . a detailed image and diagram of the setup can be found in the supplemental material . + 30 naive subjects with normal visual acuity tested with the snellen test participated in the experiments ( 10 female and 20 male ; average age of 30 years ) . for each of the two textures , a total of 10 distorted textures were created by adding zero - mean gaussian pixel - noise with standard deviations of @xmath157 ciede2000 units , simulating different levels of graininess . in a first experiment , subjects were asked to select the rendering with the distorted texture matching the graininess of the print . in a second experiment , they had to judge the level of structural artifacts in the print showing the noise - free rendering as reference on a quality scale of 0 - 5 : 0 ( not visible ) , 1 ( visible but not disturbing ) , 2 ( slightly disturbing ) , 3 ( disturbing ) , 4 ( very disturbing ) , 5 ( extremely disturbing ) . [ tab : experiment ] results of the experiments are shown in table [ tab : experiment ] . for both textures , the average perceived graininess level corresponds to a noise standard deviation smaller or equal to 2 ciede2000 units . to put this value into perspective , we computed the pixel - wise ciede2000 differences between the noise - free and distorted texture ( noise std . of 2 ciede2000 units ) using s - cielab @xcite with a visual resolution of 60 ppd @xcite . the 99th percentile of ciede2000 errors is 1.644 ( smooth texture ) and 1.694 ( patch texture ) . since this is only slightly above the just noticeable distance , we can conclude that the perceived graininess of the halftone is very low . the second experiment revealed that also structural artifacts do not adversely affect the perceived quality of the halftone : for both textures , average quality scores are smaller than 1.4 , i.e. artifacts are visible but not even slightly disturbing . note that there are some drying related artifacts , which are likely considered in the quality scores but independent of the halftone . + + figure [ fig : gallery ] shows some qualitative results demonstrating the detail and realism produced by our software using our hardware and material setup . part ( a ) shows ( left - to - right ) a head scan @xcite with original texture rendered with diffuse shading , the same model with the texture gamut - mapped using our profile , and the resulting @xmath0d print ( @xmath158 tall ) . note how well both smooth tones and high - frequency details are reproduced . part ( b ) shows the same model printed @xmath148 tall to show the effects of scale on the prints note how similar the features of the two prints are despite the difference in scale . part ( c ) shows a realistic - looking printed apple @xcite . parts ( d ) and ( e ) demonstrate how well our halftoning algorithms reproduce high - frequency details in the textures . prints ( c)-(e ) were generated with an error in the icc profile , so the colors are off in absolute terms when compared to the textures . in part ( d ) @xcite , the back of the ear is rendered next to the printed ear , which is about @xmath159 high . the images of the prints in parts ( a ) and ( e ) were taken under simulated cied50 , to which we color - characterized the camera . we transformed the raw rgb images to ciexyz values as described in ( * ? ? ? 3.2 ) , and transformed these values to srgb for display . all other photographs of prints were taken using automatic camera settings . our setup is limited primarily from the hardware side . with four materials we are forced to print without a black . this means we have difficulty reproducing dark colors and a lack of color constancy a change of illuminant alters the perceived color . we measured the color inconstancy index ( cii ) between cied50 and cief11 of cyan , magenta , yellow and white , to be @xmath160 , @xmath161 , @xmath162 and @xmath163 , respectively , using cat02 @xcite for chromatic adaptation and ciede2000 for color difference @xcite . we also have no control over the translucency , as the materials have very similar scattering and transmission properties . we can only use their translucency to create realistic looking objects , not control it to a desired level . the translucency of the materials also results in blurring of some details ; an interesting approach has been proposed to address this in a way similar to unsharp masking @xcite . the introduction of printers with more materials , and new , more opaque materials would address these limitations , and open the possibility of combining color with desired scattering and reflective properties . a compelling open question is whether such complex appearance properties can be achieved using error diffusion - style techniques . a further limitation is due to the colored support material . during the uv - curing process , support material adjacent to model material mixes and binds with the model material . this results in a yellow tinge on some sloping and vertical surfaces . this could be overcome without additional materials by a multi - pass printing , in which the support material is cured separately from the other materials . algorithmically , our approach has the same limitations as @xmath1d error diffusion methods : for extremely low tonal values , we encounter mild start - up artifacts materials are not placed entirely uniformly . we have found this only in extreme cases . the curvature of the surface affects the transfer of tonal values to the interior layers when the mean absolute curvature @xmath164 approaches @xmath165 , where @xmath5 is the depth of highest reflectance as discussed in section [ sec : prelim ] ; equivalently as the radius of curvature approaches @xmath5 . high negative mean curvature ( convex ) regions may darken and high - frequency texture features that coincide with high positive mean curvature ( concave ) regions will bloom " . we have not found this to be a major problem , although such effects show up strongest in small prints , e.g. the nose of the small ruslan print in figure [ fig : gallery ] ( b ) . we plan to address curvature - dependent effects in future work , e.g. by adjusting the tonal values according to curvature , which could be estimated using the signed distance field @xmath166 in . the surfaces used in the visual experiment also demonstrate the extent to which curvature influences the perceived color . these surfaces , shown in figure [ fig : visual : objects ] , contain both concave and convex regions with high curvature , and relatively flat regions on the periphery . in this paper , we have presented algorithms for precise and efficient control of material placement in multi - jet @xmath0d printers for the purposes of halftoning a critical ingredient in accurate color reproduction . we have presented a layered halftone , which accounts for the high - translucency of currently available color materials , and a traversal algorithm for voxel representations of surfaces , which allows us to map @xmath1d error diffusion filters onto a surface in a consistently oriented way . this allows us to leverage decades of knowledge from @xmath1d error diffusion research . our algorithms fit within an efficient streaming architecture , preserve tone and do not produce artifacts . we have further shown how our algorithms seamlessly allow the printer s support material to be colored , thereby expanding the printer s color gamut while preserving structural stability . the @xmath0d color prints generated with our setup exhibit a high level of detail and realism . due to the ability to arbitrarily configure materials with different optical properties through an object , multi - jet printing has tremendous potential for graphical @xmath0d printing in terms of reproducing complex appearance properties . by providing the first full color capabilities for these printers , we have provided a critical building block for graphical @xmath0d printing . the introduction of more opaque color materials will allow our algorithms to print with larger color gamuts using fewer layers , resulting in a performance boost . better exploiting parallelism in our computations is a key direction for future work as printer resolutions and build volumes increase . as the number of printing materials increase , it will be important to consider printer models in characterizing the printers and computing profiles . we give a big shout - out to tejas tanksale ( tt ) for his help in generating icc profiles and conducting measurements . in addition , we thank : christoph godau for access to the equipment at the idd of tu darmstadt ; stratasys for providing the voxel - level printing interface to the printer ; pedro santos and the cultlab3d of fraunhofer igd for making the nefertiti buste available to us ; christian zeh from clariant produkte gmbh for coloring the support material ; ten24 for making the head model available , and scanlab and turbosquid for making the ruslan scan available . this work was funded by the fhg internal programs under grant no . attract 008 - 600075 . our halftoning algorithms we propose are based on error diffusion , in which after a pixel or voxel is quantized , the resulting error is distributed to nearby pixels or voxels using an error diffusion filter . thus , if a pixel with a value of @xmath126 is quantized to @xmath66 , the error is distributed to its neighbors , making it more likely that they are quantized to @xmath110 . formally , we are given an input tonal signal @xmath72 and an error signal @xmath167 , which is initialized @xmath168 . the error diffused signal is defined as @xmath169 which is then quantized using a threshold @xmath170 the resulting error is then diffused to the neighbors @xmath171 where @xmath172 is a neighbor of @xmath50 , @xmath173 is an element of the error diffusion filter diffusing error from @xmath50 to @xmath80 . . projet 860pro . http://www.3dsystems.com/3d-printers/professional/projet-860pro . model - based color halftoning using direct binary search . _ 14 , _ 12 , 19451959 . error diffusion on meshes . _ 46 _ , 336344 . , brunton , a. , tanksale , t. , and urban , p. 2015 . . spie / is&t electronic imaging conference ( accepted included in supplemental material)_. san francisco . an optimum method for two - level rendition of continuous - tone pictures . in _ ieee intl . conf . on comm . _ seattle , wa , 1115 . , kulikowski , j. , and levinson , j. 1966 . the effect of orientation on the visual resolution of gratings . _ 187 , _ 2 , 427436 . courier dover publications . , alain , b. , and ostromoukhov , v. 2009 . structure - aware error diffusion . _ 28 , _ 5 , 162:1162:8 . memory efficient error diffusion . _ 12 , _ 11 , 13521366 . , levin , d. , didyk , p. , sitthi - armorn , p. , and matusik , w. 2013 . spec2fab : a reducer - tuner model for translating specifications to 3d prints . _ 32 , _ 4 . , sachs , e. , patrikalakis , n. m. , and troxel , d. e. 2003 . a dithering algorithm for local composition control with three - dimensional printing . _ 35 , _ 9 , 851867 . . . tech . rep . , central bureau of the cie , vienna , austria . . vienna , austria . , gobbetti , e. , pintus , r. , and scopigno , r. 2008 . color enhancement for rapid prototyping . in _ vast_. \1924 . le procd . _ 26 , _ 3 , 1721 , 2627 . , wang , j. , pellacini , f. , tong , x. , and guo , b. 2010 . fabricating spatially - varying subsurface scattering . _ 29 , _ 4 . , tsai , e. , dikovsky , d. , geraedts , j. , herr , h. , and oxman , n. 2014 . voxel - based fabrication through material property mapping : a design method for bitmap printing . . error - diffusion algorithm with edge enhancement . _ 8 , _ 12 , 18441850 . \2013 . , 3 ed . john wiley & sons , inc . , west sussex , england . distance transforms of sampled functions . tech . rep . , cornell university . an adaptive algorithm for spatial grey scale . _ 17 , _ 1 . , fuchs , m. , matusik , w. , pfister , h. , and rusinkiewicz , s. 2010 . physical reproduction of materials with specified subsurface scattering . _ 29 , _ 3 . clean color : improving multi - filament 3d prints . _ 33 , _ 2 , 469478 . , 4.3.0.0 ed . http://www.color.org . \2003 . . _ 28 , _ 6 , 425435 . , dong , y. , pellacini , f. , and tong , x. 2013 . bi - scale appearance fabrication . _ 32 , _ 4 . , arce , g. , and gallagher , n. 1999 . digital halftoning by means of green - noise masks . _ 16 , _ 7 , 15751586 . crc press . , glasner , d. , xiong , y. , durand , f. , freeman , w. , matusik , w. , and zickler , t. 2013 . fabricating brdfs at high spatial resolution using wave optics . _ 32 , _ 4 . tone - dependent error diffusion . _ 13 , _ 2 , 201215 . \1998 . fundamentals of 3d halftoning . _ 1375 _ , 224239 . . 2014 . or iris . http://mcortechnologies.com/3d - printers / iris/. \1992 . digital halftoning technique using a blue - noise mask . _ 9 , _ 11 , 19201929 . john wiley & sons . _ 359 , _ 1 , 381 . \2001 . a simple and efficient error - diffusion algorithm . in _ proc . siggraph_. , qu , y. , wong , t .- t . , cohen - or , d. , and heng , p. 2008 . structure - aware halftoning . _ 27 , _ 3 , 89 . , carr , n. , mech , r. , stava , o. , dachsbacher , c. , and miller , g. 2014 . dual - color mixing for fused deposition modeling printers . _ 33 , _ 2 , 479486 . \1997 . . _ 41 _ , 643656 . . http://www.turbosquid.com/fullpreview/index.cfm/id/777450 . , grimm , c. , and wyvill , b. 2006 . interactive decal compositing with discrete exponential maps . _ 25 , _ 3 , 605613 . objet500 connex3 . http://www.stratasys.com/3d-printers/production-series/connex3-systems . , buckley , r. , and swen , s. 1999 . . is&t / sid , 7th cic_. scottsdale ariz . , 127134 . . http://www.ten24.info/?p=1164 . http://www.turbosquid.com/3d-models/free-max-model-apple/549455 . the mit press . spatial modulation transfer in the human eye . _ 57 , _ 3 , 401406 . , wang , s .- , ragan - kelley , j. , and matusik , w. 2013 . openfab : a programmable pipeline for multi - material fabrication . _ 32 , _ 4 . measure of goodness of a set of color - scanning filters . _ 10 _ , 14991508 . \2000 . . _ 25 , _ 1 , 419 . \2000 . , 2 ed . john wiley & sons , inc . \1996 . . _ 27 _ , 731734 . improving mid - tone quality of variable - coefficient error diffusion using threshold modulation . _ 22 , _ 3 , 437444 . given an arrangement of multiple non - fluorescent printing materials with similar refractive indexes within a shape @xmath4 , light propagation within this shape can be described by the steady - state radiative transfer equation @xcite @xmath174 where @xmath175 is the radiant intensity at location @xmath176 propagating in direction @xmath177 for wavelength @xmath178 $ ] nm ( visible wavelength range ) , @xmath179 is the spectral absorption coefficient , @xmath180 is the spectral scattering coefficient , @xmath181 is the sphere and @xmath182 is the scattering function satisfying @xmath183 . solving the radiative transfer equation with an appropriate boundary condition at the surface gives us the radiation that is diffusely emitted from the shape . adding the radiation directly reflected from the surface due to fresnel reflection , gives us the total radiation emitted from the object s surface . shows that the intensity of radiant energy traveling through the material is attenuated by absorption ( note that @xmath184 depends on the wavelength but is independent of the traveling direction ) and is redistributed by scattering ( scaled by @xmath185 ) . thus , a fraction of light entering the print at one location may be emitted from the surface at a different location due to scattering . if light travels through different materials its spectral power distribution is modulated by each material s absorption coefficients and the path length within this material . figure [ fig : experiment_setup ] shows the physical setup we used to conduct the visual experiment . subject placed their chin on the chinrest and viewed the printed object next to a rendering of it on a color - calibrated display . the color characterization of the display was performed from the chinrest position considering the viewing booth illumination . in this way , tristimulus values from objects placed on the display and illuminated by the viewing booth could be reproduced on the screen if viewed from the chinrest position .

we study the predictions of anomalous inflationary models on the abundance of structures in large scale structure observations . the anomalous features encoded in primordial curvature perturbation power spectrum are ( a ) : localized feature in momentum space , ( b ) : hemispherical asymmetry and ( c ) : statistical anisotropies . we present a model - independent expression relating the number density of structures to the changes in the matter density variance . models with localized feature can alleviate the tension between observations and numerical simulations of cold dark matter structures on galactic scales as a possible solution to the missing satellite problem . in models with hemispherical asymmetry we show that the abundance of structures becomes asymmetric depending on the direction of observation to sky . in addition , we study the effects of scale - dependent dipole amplitude on the abundance of structures . using the quasars data and adopting the power - law scaling for the amplitude of dipole we find the upper bound for the spectral index of the dipole asymmetry . in all cases there is a critical mass scale in which for the enhancement in variance induced from anomalous feature decreases ( increases ) the abundance of dark matter structures in universe .

1406.7277

the standard model of cosmology , @xmath4cdm , explains anisotropies on cosmic microwave background ( cmb ) radiation @xcite and the large scale structure ( lss ) observations with good accuracies @xcite . this model is based on the assumption of cold dark matter as the skeleton of the structures in the universe @xcite and the cosmological constant as the source of late time acceleration while the initial conditions are set by the inflationary paradigm @xcite . inflation generates nearly scale - invariant , nearly gaussian and nearly adiabatic perturbations which are imprinted on the temperature map of cmb and also provides the initial distribution for the seeds of structures in the early universe @xcite . accordingly , one of the triumphs of modern cosmology is the compatibility of the dark matter tracers distribution with the statistics of cmb temperature anisotropy @xcite . thanks to vast cosmological data available during last decade , models of early universe are well - constrained by cosmological observations . in particular , the high precision maps of cmb have been studied extensively to put constraints on inflationary models @xcite . any deviation from the gaussian , adiabatic , isotropic and scale - invariant initial conditions will open distinct fingerprints on cmb map as tools to study the physics of early universe such as inflation @xcite . on the other hand , the lss observables , by transferring the initial condition of perturbations to the late time observables , are potentially very useful tools to test models of early universe in sub - cmb scales @xcite . the abundance of the structures in the universe @xcite , the scale - dependence of bias parameter between dark matter halos as the host of galaxies and the dark matter perturbations @xcite and higher order statistics of dark matter tracers ( e.g. the bispectrum of galaxies ) @xcite are all used to probe the deviations from the standard initial conditions . there are indications for anomalous features in cmb map such as power asymmetry between the two cmb hemispheres @xcite , power deficit in low multipoles @xcite and the coincidence of the quadrupole and octupole directions @xcite . these observations trigger the motivation to study the predictions of inflationary models with anomalous features in lss observables . we note that the cmb maps cover the first few e - folds of the inflationary period , say the first 5 - 7 e - folds , while the remaining 50 or so e - folds are below the cmb scales . it is likely that the primordial power spectrum may have non - trivial features not only on cmb scales but also on scales which are beyond the probe of cmb but are accessible to lss observations . in this work we investigate the framework under which one can study the probable observational fingerprints of anomalous inflationary power spectrum on lss , specially in the abundance of the structure . having this said , there are open questions in lss , specially on galactic scale , such as the missing satellite " problem @xcite and the too big to fail " problem @xcite which may be considered as hints for anomalies / features in primordial curvature perturbation power spectrum on galactic scales ( for a recent work addressing this idea see @xcite ) . it is worth to mention that the theoretical predictions of standard model for matter power spectrum are in reasonable agreements with the observational data in the range of wavenumber ( scales ) @xmath5 . however , the amplitude of small scale dark matter power spectrum is not strongly constrained by lss observables . accordingly , there are freedoms on the amplitude of the power on galactic scales which in turn can be used as an explanation for the missing satellite problem . therefore the anomalous inflationary models , by modifying the power on small scales , can be considered as a possible solutions to the tension between observations and theory on galactic scales . the rest of the paper is organized as follows . in section [ sec - ii ] first we review the procedure for obtaining the number density of structures based on excursion set theory @xcite and then we study the effects of the primordial feature on the number density of structures in a model - independent way . in section [ theory ] we present the theoretical motivations for generating primordial features studied in this paper . in section [ sec - iii ] , we study the predictions of each anomalous inflationary model on the abundance of the structures . the reader who is not interested in theoretical realization of the features can skip section [ theory ] and go directly to section [ sec - iii ] . the conclusion and future prospects are presented in section [ sec - iv ] . the mass function of the dark matter halo structures can be found from the excursion set theory ( est ) as a barrier crossing problem of density perturbation in different mass scales ( window functions ) @xcite . the number density @xmath6 of virialized objects with the mass between @xmath7 and @xmath8 is @xcite [ eq : massfunc ] n(m , z)=-2f ( ) where @xmath9 is the mean matter density and @xmath10 is the universality function of halo distribution which is a function of height parameter @xmath11 . here @xmath12 is the critical density ( in spherical collapse ) @xcite and @xmath13 is the variance of the structures with mass @xmath7 at the specific redshift @xmath14 . the variance is defined as [ eq : sigma ] ^2(m , z)= p_l(k , z ) w^2(kr)=k^2p_l(k , z)w^2(kr ) where @xmath15 is the linear matter power spectrum and the second equality in eq . ( [ eq : sigma ] ) is obtained with the assumption of isotropic density perturbations . in addition , @xmath16 is the window function , which for the top - hat filter in real space , has the following form in momentum space w(x)=3 , in which modes with @xmath17 are smoothed out . the variance obtained for each mass scale is related to the radius of window function via the relation @xmath18 . an important point to mention is the universality of mass function , i.e. it is only a function of the height parameter @xmath19 . the universality means that on each mass range we have the same dependence on the height parameter . the height parameter @xmath19 encodes the information about the local threshold of collapse as well as the variance of matter perturbations which the latter depends on the cosmological model . the press - schechter mass function , as one of the first proposals for the universality function , is defined as @xcite f()= e^-^2/2 , assuming that the primordial seeds of the structures have a gaussian distribution . the n - body simulations have investigated the validity of the press - schechter conjecture and have found that it works with @xmath20 accuracy . the sheth - tormen mass function , incorporating the idea of ellipsoidal collapse @xcite , has the universality function @xmath10 given by @xcite f()= c e^-a^2/2 , where @xmath21 and @xmath22 are constants which are determined from n - body simulations of lss and @xmath23 is determined from the requirement that @xmath24 . in the limit @xmath25 , @xmath26 and @xmath27 , we recover the press - schechter proposal . the discrepancy between press - schechter and the results obtained from the n - body simulation indicates the complexity of non - linear structure formation . in addition , it may indicate that the assumptions such as the spherical collapse or the markovianity of the random walks of matter density perturbations in est framework are only approximation to the real process of collapse . there is a vast literature on the prediction of dark matter clustering in est framework with deviations from gaussianity , markovianity and spherical collapse @xcite . in this work , in order to simplify the analysis and also in order to get some intuitions on the effects of anomalous features on lss abundances , we consider only the press - schechter and sheth - tormen universality functions . in order to study the effects of feature in primordial curvature perturbations on the abundance of the structures we should relate the late - time linear matter power spectrum to the initial curvature perturbations power spectrum . the poisson equation relates the matter density contrast in fourier space , @xmath28 , to the gravitational potential ( bardeen potential ) @xmath29 via _ m ( k , z ) = , where @xmath30 is the matter density parameter at the present time ( hereafter we omit the superscript @xmath31 in density parameter ) and @xmath32 is the present value of hubble parameter . now we can relate the gravitational potential @xmath29 to the primordial curvature perturbation @xmath33 by the transfer function @xmath34 and the growth function @xmath35 ( which are separable functions in @xmath4cdm model ) via ( k , z)=t(k)d(z)(1+z)_k . note that we set the primordial ( initial ) bardeen potential in radiation dominated era as @xmath36 . consequently , the matter power - spectrum is related to the primordial curvature perturbation via p_l(k , z)=k_(k ) , where @xmath37 is the dimensionless power spectrum . usually the curvature perturbations power spectrum is parameterized as _ = a ( ) ^n_s-1 , where @xmath38 is the amplitude of the fluctuations at pivot wavenumber @xmath39 ( this value is from planck collaboration ) and @xmath40 is the spectral index which shows the scale - dependence of power spectrum . the constraints from the planck collaboration are @xmath41 and @xmath42 at @xmath43 @xcite . the question we are interested to answer is _ if we put a feature in primordial curvature perturbations power spectrum , how does this affect the abundance of the structures in lss ? _ with the inclusion of feature in power spectrum , we schematically have [ eq : power - mod ] _ ( * k * ) = ( * k * ) + _ ( ) , in which @xmath44 is the standard isotropic and nearly scale - invariant primordial curvature perturbation power spectrum and @xmath45 is a scale - dependent feature added to the curvature power . consequently , the variance of matter perturbation from eq.([eq : power - mod ] ) is modified as follows ( m , z ; a_*)=|(m , z)+(m , z;a _ * ) , where @xmath46 is the variance in @xmath47 model obtained from @xmath48 and @xmath49 is its modification induced from the correction in primordial power spectrum . in this notation , @xmath50 is a symbolic parameter which collectively represents the set of the parameters which define the feature ( e.g. the amplitude of the feature or the location of feature in momentum space ) . the leading order correction in the variance is given by [ eq : deltasigma0 ] ( m , z;a _ * ) = d_k w^2(kr ) t^2(k ) _ ( ; a _ * ) , where @xmath51 is the solid angle of the spherical coordinates in the fourier space . in the case in which the correction to the power spectrum is isotropic we have [ eq : variance - change ] ( m , z;a _ * ) = k^3 dk w^2(kr ) t^2(k ) _ ( k;a _ * ) . the next step is to translate the effects of the corrections in variance to the number density of structures . for this purpose , we use the fact that the variance @xmath52 in @xmath4cdm model has a power - law relation with the radius ( mass ) of the enclosed over - dense region as follows @xcite [ sigma8 ] |_r(z)_*(z)()^- where @xmath53 is the normalization factor and @xmath54 is the fitting parameter . for the mass range @xmath55 , @xmath56 , accordingly the normalization parameter is @xmath57 and the parameter @xmath54 is approximately given by @xmath58 $ ] in which to leading order @xmath59 @xcite . for the mass range @xmath60 , @xmath61 , accordingly the normalization parameter is @xmath62 and the parameter @xmath63)$ ] . neglecting the logarithmic dependence of the variance to the mass ( this approximation introduces an error of less than @xmath20 ) , in these two intervals the mass function of the structures becomes [ eq : fwn ] depends on the mass interval we are interested . in eq.([eq : fwn ] ) @xmath64 is the number density per mass of structures with the standard initial conditions , i.e. in the absence of the feature . now we define a new parameter , @xmath65 , as the fractional excess of the number density of the structures in the mass range @xmath7 , @xmath8 as follows [ eq : deltan ] _ n ( m , z ; a_*)= where @xmath66 is the mass function of the structures in the presence of the feature in primordial curvature power spectrum collectively represented by the parameter @xmath50 . we also consider the fact that the number density of the structures can be calculated at any redshift . therefore , the change in number density depends on the redshift through the growth function @xmath35 which appears in the variance of the matter perturbation . now , assuming the universality of @xmath67 which is encoded in @xmath10 , @xmath68 at any mass scale is obtained to be [ eq : deltan1 ] _ n ( m , z ; a_*)=(m , z;a _ * ) , where a prime represents the derivative with respect to the height parameter @xmath19 . note that eq . ( [ eq : deltan1 ] ) is obtained from the taylor expansion of eq . ( [ eq : deltan ] ) with respect to change in variance induced from @xmath50 at a fixed mass . using the definition of height parameter , ( [ eq : deltan1 ] ) reduces to [ deltan1 ] _ n(m , z ; a_*)= - . from this equation we see that the change in number density , @xmath68 , is proportional to the change in the variance @xmath69 induced by the primordial feature . we note that the proportionality factor is given by the logarithmic derivative of the universal mass function . consequently , an excess or deficit in variance does not have a uniform effect on all mass scales ; its effect is non - trivial which depends on the mass scale under consideration . for the press - schechter universality function we have [ nps ] _ n @xmath70 is the change in number density at a specific mass @xmath7 . this change is a function of the mass through the height parameter . now consider the critical height @xmath71 with the critical mass @xmath2 which depends on redshift . the interesting point is that for the structures with the mass range @xmath72 in which @xmath73 , the number density contrast has an excess for @xmath74 while opposite conclusion is reached for @xmath75 in which @xmath76 . note that the sign flip for @xmath68 at the critical height parameter @xmath77 is a result of the gaussian probability distribution function ( pdf ) which is assumed in the press - schechter formalism . for a gaussian pdf with a fixed value of the density contrast @xmath78 the probability is maximum if the variance satisfies the relation @xmath79 . if one fixes the density contrast to the critical value @xmath80 , which is the definition of critical height parameter @xmath81 , then the maximum probability occurs at @xmath82 , or equivalently at @xmath83 . that is why @xmath68 for the collapsed objects as a function of the variance flips the sign at @xmath83 . a crucial point to discuss is that for the mass scales @xmath84 where @xmath85 , we are in quasi - linear regime where the structures are on the verge of formation and the density contrast is almost near the critical density @xmath86 ( the uncertainty comes from the fact that the condition of spherical collapse is @xmath87 which is not exactly equal to @xmath88 ) . accordingly , our result is more reliable for masses and redshifts in which @xmath89 while it is not applicable for @xmath90 . similarly , we can find the number density contrast for the sheth - tormen mass function as [ eq : deltanst ] _ n sheth - tormen number density contrast reduces to that of the press - schechter obtained in eq.([nps ] ) . interestingly , the critical height in which the big bracket in eq.([eq : deltanst ] ) vanishes is more complicated than the press - schechter function in which one simply has @xmath81 . this indicates that in a more realistic model of collapse the change in the statistics of dark matter halos is a non - trivial function of the mass scale @xmath2 which in itself is a redshift dependent quantity . we remark that in this work we assume that introducing a feature in primordial power spectrum does not invalidate the press - schechter procedure of structure formation . this means that the processes of barrier crossing , the magnitude of @xmath86 obtained from spherical collapse and the environment - independence of standard est do not change dramatically in the models we study in next section and the corrections are sub - leading . this is because the mechanism of collapse occurs in late time universe and it is almost scale - free . the inflationary models with feature mainly change the initial conditions of perturbations which to first order modifies the variance as we have discussed here . with these model - independent analysis for @xmath91 and @xmath68 , in section [ sec - iii ] we calculate these quantities for various features in curvature perturbation power spectrum presented in section [ theory ] . in this section we present three different kinds of anomalous features in inflationary power spectrum which will be used for the abundance of structures in next section . the reader who is not interested in mechanism of generating these features in inflationary model building can directly jump to section [ sec - iii ] . our discussions here will be brief , we only provide the motivations and the overviews of mechanisms generating the corresponding features . for detail studies , we refer the reader to the extensive literature on each topic . first we consider models of inflation in which there is a sharp jump in curvature perturbations power spectrum in fourier space . mathematically , we parameterize this localized feature in power spectrum by [ eq - delta ] _ = _ in which @xmath92 is the dirac delta function , @xmath93 is the usual nearly scale - invariant power spectrum while the amplitude of feature is represented by @xmath50 in units of @xmath94 . in order for this localized feature to be relevant for lss observations , we require @xmath95 . we note that @xmath50 can be either positive or negative , corresponding respectively to an increase or a decrease in the primordial power spectrum . the idea of generating local features have been extensively studied in the literature . here we follow the mechanism employed in @xcite . the model contains two scalar fields @xmath96 and @xmath97 in which the light field @xmath96 is the inflaton field and the heavy field @xmath97 is the waterfall field . the potential is as in hybrid inflation @xcite given by [ potential ] v= ^2 + ( ^2 - ) ^2 + ^2 ^2 . however , unlike models of hybrid inflation , in this scenario inflation mainly proceeds as in chaotic model with potential @xmath98 with some effective mass @xmath99 which undergoes a small but abrupt change when @xmath96 reaches the critical value @xmath100 . the waterfall field @xmath97 is introduced to induce the change in mass . in this picture inflation has three stages as follows . as in chaotic models inflation starts with a large super - planckian value , @xmath101 in which @xmath102 is the reduced planck mass , so one can obtain about 60 @xmath103-folds of inflation to solve the horizon and the flatness problems . as in hybrid inflation the waterfall field is very heavy during the first stage of inflation so it is stuck to its instantaneous minimum @xmath104 . however , once the inflaton field has reached the critical value @xmath105 , the waterfall field becomes tachyonic and it rapidly rolls towards its global minimum @xmath106 . this represents the second stage of inflation which may be taken to be very short for our purpose . the final stage of inflation , corresponding to @xmath107 , follows as in chaotic inflation but with the effective mass of the inflaton , @xmath108 , given by m_+^2 = m^2 + g^2 ^2 = m^2 ( 1 + ) , in which the constant @xmath109 is defined via @xmath110 . we assume the change in inflaton mass is small , corresponding to @xmath111 , so the slow - roll conditions are satisfied throughout inflation . in order to bring this local feature into the lss window we assume that the short waterfall stage begins when the lss modes @xmath112 leaves the horizon . typically , this corresponds to around 50 @xmath103-folds or so before the end of inflation . the details of correction in power spectrum and bispectrum have been studied in @xcite . the correction in power spectrum from waterfall @xmath113 compared to the original power spectrum coming from the inflaton @xmath114 is obtained to be [ powers - ratio ] 10 ^ 3 ^2 ( ) ^4 ( ) ^-10/3 , in which @xmath115 is the slow - roll parameter and @xmath116 is a parameter defined in @xcite measuring the sharpness of the phase transition . for a very sharp waterfall phase transition , we require @xmath117 , say @xmath118 . with @xmath119 the above ratio can be as big as 10 . in principle by taking @xmath116 very big , one can make the duration of the waterfall phase transition ( the intermediate second stage of inflation ) as short as one desires . in this limit , we approach the localized feature presented in eq . ( [ eq - delta ] ) . there are indications of power asymmetry in cmb map as observed by planck satellite @xcite , for earlier reports of hemispherical asymmetry in wmap data see @xcite . the observations indicate that the power spectrum in northern hemisphere is different than the the power spectrum in southern hemisphere . using the following dipole modulation on the cmb temperature fluctuations @xcite [ dipole - power0 ] t ( ) = , in which @xmath120 is the statistically isotropic temperature fluctuations , @xmath121 is the amplitude of the dipole asymmetry , @xmath122 is the preferred direction in the sky , @xmath123 is the direction of the look to sky and @xmath124 is the angle between @xmath122 and @xmath125 , the planck collaboration has found @xmath126 and @xmath127 in galactic coordinates @xcite , see also @xcite . whether this detection is statistically significant is under debate and we shall wait for the upcoming planck data release to see if the significance of the detection will improve . the best known method to generate a hemispherical asymmetry is the idea of long mode modulation @xcite . in this picture , a long super - horizon mode with the wavelength @xmath128 modulates the curvature perturbations at the desired scales ( either the cmb or the lss scales ) . an observer probing a cosmological distance @xmath129 much smaller than the long mode ( @xmath130 can not probe the wave - like nature of the long mode . for this observer , the long mode only modifies the background quantities such as @xmath131 for the inflaton field or modulates the surface of end of inflation . this in turn induces a directional - dependence ( a gradient ) in power spectrum taking the form [ dipole - power ] _ = _ . note that the additional factor 2 in the above formula compared to eq . ( [ dipole - power0 ] ) is because @xmath132 . for this picture to work , there should exist a coupling between the long mode @xmath133 and the smaller modes i.e. the cmb - scale or the lss - scale modes . this rings the bell for the relation between the amplitude of dipole asymmetry and the `` squeezed - limit '' non - gaussianity @xcite , however see @xcite for a different view . imposing the quadrupole constraints one finds @xcite @xmath134 . with these discussions in mind , in order to get large enough dipole asymmetry one has to start with a model with large enough squeezed - limit non - gaussianity . as is well - known from maldacena s analysis @xcite , many single field models of inflation fail to do this job . indeed , maldacena s non - gaussianity consistency condition @xcite implies that @xmath135 so one can not generate observable dipole asymmetry from the long mode modulation in conventional single field models of inflation . this argument also implies that one has to look for non - trivial models of inflation in which the maldacena s consistency condition is violated such as in curvaton models or in multiple fields models of inflation . the observations on sub - cmb scales indicate that dipole asymmetry becomes less significant for @xmath136 @xcite . this indicates a non - trivial scale - dependent dipole asymmetry , i.e. @xmath137 which should be considered in any theoretical modeling of dipole asymmetry . we take into account the upper bound imposed from @xcite when consideration the effects of dipole asymmetry on lss in next section . homogeneity and isotropy are two pillars of cosmological principles yielding the standard frw cosmology . there were indications of statistical anisotropies in wmap data @xcite , but the follow up analysis demonstrated that the systematic errors can accommodate the apparent source of quadrupole asymmetry @xcite . having this said , the possibilities of violating statistical isotropy is intriguing both observationally and also from the theoretical point of view . indeed , it is likely that there is no statistical anisotropy at cmb scales but there may exist statistical anisotropy at lss scales . one interesting mechanism to generate statistical anisotropies are based on the dynamics of gauge fields . the presence of a background gauge field breaks the three - dimensional rotation invariance and the space - time takes the form of bianchi i metric . specifically , assuming that the gauge field @xmath138 is turned on along the @xmath129-direction the metric line element is given by ds^2 = -dt^2 + a(t)^2 dx^2 + b(t)^2 ( dy^2 + dz^2 ) , in which @xmath139 and @xmath140 are two scale factors with the corresponding hubble expansion rates @xmath141 and @xmath142 . a vector field in an expanding background suffers from the conformal invariance so its background value dilutes exponentially during inflation . in order to break the conformal invariance and prevent the background gauge field to decay exponentially , one can couple the gauge field non - trivially to inflaton field . an interesting realization of anisotropic inflation is proposed in @xcite based on the following lagrangian [ action ] s= d^4 x , where @xmath102 is the reduced planck mass , @xmath96 is the inflaton field and @xmath143 is the gauge field strength . it is shown in @xcite that with a suitably chosen gauge kinetic coupling @xmath144 the inflationary system reaches an attractor solution in which the gauge field energy density approaches a small but observationally detectable fraction of the total energy density . in is shown that the anisotropies produced are at the order of slow - roll parameters . the anisotropic power spectrum in this model has the quadrupole asymmetry given by [ quad - p ] _ ( k ) = ( k ) ( 1 + g _ * ^2 _ * p * ) in which @xmath145 is the anisotropic direction ( the @xmath129-direction in our example ) and @xmath146 is the momentum direction in fourier space so @xmath147 represent the angles between these two directions . with this definition , @xmath148 is the amplitude of quadrupole asymmetry . for the model of anisotropic inflation introduced in @xcite one can show that @xmath149 in which @xmath150 is the number of e - folds of inflation @xcite . imposing the recent planck data @xcite ( see also @xcite ) one obtains @xmath151 . comparing the two power spectra eqs . ( [ quad - p ] ) and ( [ dipole - power ] ) we stress that they have distinct structural forms . first , the quadrupole asymmetry eq . ( [ quad - p ] ) is in fourier space while the dipole asymmetry eq . ( [ dipole - power ] ) is in real space . second , eq . ( [ quad - p ] ) represents anisotropy at each point on cmb sphere so it does not distinguish between two different cmb hemispheres . however , eq . ( [ dipole - power ] ) represents asymmetry between two cmb hemispheres , i.e. the asymmetry depends on the direction of the look to sky given by the unit vector @xmath152 . motivated by the quadrupole asymmetry obtained for the specific model of anisotropic inflation @xcite ( see also @xcite for a different realization of quadrupole anisotropy ) we can consider the following general presentation of statistical anisotropies in power spectrum @xcite @xmath153,\ ] ] where @xmath154 ( with @xmath155 ) are spherical harmonics and the parameter @xmath156 quantifies the departure from statistical isotropy as a function of wavenumber @xmath157 . imposing the reality of the power spectrum one obtains @xmath158 . for our specific anisotropic inflation given by the action eq . ( [ action ] ) we have @xmath159 while the rest of @xmath160 are zero . cdm @xmath161 , induced from localized feature . the curves are for localized feature introduced in eq . ( [ eq - delta ] ) with @xmath162 , @xmath163 ( red - solid curve ) ; @xmath164 , @xmath163 ( green dashed curve ) and @xmath162 , @xmath165 ( blue dotted curve ) . ] having studied the effects of the primordial features on the number density of structures in section [ sec - ii ] and after presenting three different types of features in inflationary power spectrum in section [ theory ] , in this section we study the effects of these features on the variance of linear order perturbations and consequently on the number density of the structures . we study the predictions of each feature in turn . the matter power spectrum of structures are nearly well constrained for the scales in the range @xmath166 by large scale structure surveys like sloan digital sky survey ( sdss)-iii baryon oscillation spectroscopic survey ( boss ) project@xcite . however , beyond the mentioned scales , the data are not strong enough to put constraints on the matter power spectrum . for @xmath167 the volume of the sky to find the statistics is much bigger than the coverage of the ongoing surveys . on the other hand , for scales @xmath168 , there is a lack of statistics for the dwarf satellite galaxies to get a significant limit measurement for the power . accordingly , constraining observationally the matter power spectrum on small scales is a challenging task . there are many proposals to detect the dark matter sub - halos gravitationally which can be used to put constraints on the matter power spectrum on small scales @xcite . also the detection / null detection of gamma rays that are anticipated from the annihilation of dark matter particles can constrain the matter power spectrum and the initial curvature perturbation power respectively @xcite . from the theoretical point of view , as discussed in sec.([sec - i ] ) , the primordial power spectrum of curvature perturbations can be translated to the number density of the structures in the universe . consequently , constraining the mass profile of the structures or statistics of dark matter tracers can be used as a probe of primordial curvature perturbation . now consider the local feature in power spectrum given by eq . ( [ eq - delta ] ) localized at the scales @xmath169 with the amplitude @xmath50 in units of @xmath170 . as mentioned before , @xmath50 can take either signs . the change in variance from the general formula eq . ( [ eq : variance - change ] ) is obtained to be [ eq : deltasigma ] = _ ( k_*)w^2(k_*r ) in fig . [ fig : variance - ratio ] , we have plotted the ratio of the change in variance to variance in @xmath4cdm model , ( @xmath69 ) , for this feature . the red solid curve , the green dashed curve and the blue dotted curve respectively are for @xmath171 at @xmath163 , @xmath172 at @xmath163 and @xmath172 with @xmath165 . it is obvious that @xmath173 linearly depends on the amplitude of @xmath174 . it is worth to mention that the oscillatory features in fig . [ fig : variance - ratio ] for high masses are not physical . these oscillatory features are caused by the choice of the window function in eq.([eq : deltasigma ] ) . accordingly the physical feature is for the mass range @xmath175 , where @xmath176^{1/3}$ ] . on the other hand , the position of feature affects the amplitude of @xmath173 too . the window function smoothes all the modes smaller from the specific mode @xmath112 in the gaussian window function in fourier space . for each specific mode @xmath112 , we can find the corresponding critical mass @xmath2 as follows m_c=(k_*)^-3_m(z ) where @xmath32 is the hubble constant and @xmath177 is the density parameter of matter in redshift @xmath14 given by m_c=1.1 * 10 ^ 12 h^2 ( ) ^-3 _ m(z ) . for an inflationary model with @xmath163 , the effect of feature on the number count of the structures would be valid for @xmath178 , where for the redshift @xmath179 the critical mass is less than the milky way mass . accordingly , the number density of milky way type galaxies can not be used as a viable observational constraint on the amplitude of the localized feature for the wavenumber @xmath180 . if the feature is placed in smaller wavenumbers ( larger scales ) it affects @xmath173 more significantly . this is because the matter transfer function decays like @xmath181 so it obtains less contributions from power spectrum modifications on small scales . another important point to notice is that @xmath182 is independent of redshift . as a result the redshift - dependence of the modifications in abundance comes from the universality function and not from the change in variance . in the next step , we calculate the change in number density , @xmath68 , for both press - schechter and the sheth - tormen mass functions . for the press - schechter mass function we obtain _ n sheth - tormen universality function the result is _ n |_st(m , z;a_*,k _ * ) = _ ( k_*,z)w^2(k_*r ) the change in number density , @xmath68 , from the localized feature is plotted in fig . [ fig - n - ratio ] . in this figure we plot @xmath68 for press - schechter and sheth - tormen universality functions versus the mass scale . the feature is located at @xmath163 with the amplitude of @xmath162 . the black solid ( red long dashed ) curve is for press - schechter ( sheth - tormen ) universality function at @xmath183 . these two curves show that the change in abundance is smaller for the more realistic @xmath10 function in sheth - tormen proposal . therefore , in the follow up figures we present the sheth - tormen prediction for @xmath68 . the green - dashed ( blue dotted ) curve in fig . [ fig - n - ratio ] is the prediction for @xmath184 ( @xmath185 ) . [ fig - n - ratio ] shows that in the mass range of @xmath175 where the variance change is physical and not the artifact of window function , @xmath68 decrease with redshifts . this means that we find the most noticeable deviation from the standard case at smaller redshifts ( present time ) . this can be explained by the fact that the variance becomes larger at small redshifts so the height parameter decreases . accordingly , as discussed in sec.([sec - ii ] ) , for the range @xmath175 ( @xmath186 ) we have more deviation from standard case . note that , as in the case of @xmath173 , the oscillatory patterns for higher mass ranges are the artifact of the window function . as a result of eq . ( 32 ) the change in the number density of the structures , @xmath68 , is proportional to @xmath187 . this means that for the mass range with @xmath188 the positive feature in the power - spectrum increases the number of structures in comparison to the standard case , while for the mass range @xmath88 we have the opposite effect . the main point here is that the regime @xmath88 corresponds to the collapse regime where @xmath189 and the actual structures are formed . this is the regime where the positive bump in primordial power - spectrum reduces the number of structures . the intuition for this comes from the probability distribution function of primordial density seeds ( like in press - schechter case @xmath190 ) , where the probability is maximum for a fixed density contrast when @xmath191 . in other words the first crossing rate in ps formalism depends on @xmath192 and accordingly the change in variance has an extremum at @xmath83 as we anticipated from the above discussions . with the amplitude @xmath162 for press - schechter mass function at redshift @xmath183 ( black solid line ) and for sheth - tormen mass function at @xmath183 , @xmath184 and @xmath193 shown respectively by the red long dashed curve , the green dashed curve and the blue dotted curve . ] for sheth - tormen universality mass function . the features have the amplitude @xmath162 with @xmath163 ( solid black curve ) ; @xmath165 ( red long dashed curve ) ; @xmath194 ( green dashed curve ) . ] in fig . [ fig - n - ratioa ] we have plotted the density contrast versus mass at redshift @xmath184 for features localized at @xmath163 , @xmath165 and @xmath195 . it is obvious that by changing the position of feature we get different mass scales @xmath2 for the deficiency in the number of structures for a given value of @xmath50 . also the amplitude of @xmath68 decreases by increasing @xmath112 . an important point to mention is that the matter power spectrum is nearly well constrained for the scales in the range @xmath196 by the lyman - alpha forest @xcite . accordingly , there are flexibilities in turning on features on smaller scales with @xmath197 . in figs . [ fig - n - ratio ] and [ fig - n - ratioa ] , we observe that a feature with the positive amplitude , @xmath198 , reduces the number density of structures for the mass smaller than @xmath199 . as a result , the number of structures decreases by the amount n(m_i < m < m_f)= f_sky_0^z_obs dz _ m_i^m_f |n(m , z)_n ( m , z;a_*,k _ * ) where @xmath200 is the change in the number of the structures , @xmath201 is the fraction of the sky in which a specific survey collects the data and @xmath202 is the initial ( final ) mass range which we are interested in . the volume integration is substituted with integral over redshift as the cosmological volumes are redshift - dependent ( @xmath203 ) . this mechanism provides a chance to resolve the missing satellite problem , the tension which is raised between the simulations and observations on the number density of dwarf galaxies @xcite . in order to address this discrepancy , different proposals were put forward . one proposal is to look for astrophysical solution , asserting that there is a minimum dark matter halo mass below which halos can not host a galaxy . this is because the gravitational potential of the dark matter halo is not sufficient to encapsulate the baryons in it . astrophysical processes like snie explosion and agn feedbacks wash the baryons out of the gravitational potential of matter @xcite . the second proposal is assuming that dark matter halos themselves do not exist in these scales . this can be explained if dark matter is not cold or not completely dark such as in warm dark matter and self - interacting dark matter models @xcite . it is worth to mention that there are proposals to detect the dark matter substructures which are not hosting baryonic matter with gravitational lensing technique @xcite . our model based on local feature in inflationary power spectrum can be considered as the third proposal . in our proposal , by changing the variance of dark matter perturbation , we can change the statistics of the structures in the universe . this proposal requires further cosmological observations to be tested . in this subsection we study the effects of dipole asymmetry in primordial power spectrum on the statistics of the structures . as discussed in subsection [ dipole - sec ] there are indications of dipole asymmetry in cmb map as observed by planck collaboration @xcite . the primordial power spectrum with dipole modulation is given in eq . ( [ dipole - power ] ) . first we consider the simple case in which the amplitude of dipole asymmetry , @xmath121 , does not depend on scales . in this approximation the change in the variance from the general formula eq . ( [ eq : variance - change ] ) is obtained to be ( m;a_d,)=a_d(_.)|(m ) . this indicates that @xmath49 depends on both the amplitude of the dipole and the direction of the observation @xmath204 . the maximum ( minimum ) change in variance occurs along the direction of dipole ( perpendicular to direction of dipole ) . now we can find the change in the statistics of the structures . for the press - schechter mass function the change in number density is obtained to be _ n |_ps(m ; a_d , ) = ( ^2(m)-1 ) a_d ( _ . ) , while for the sheth - tormen case we have _ n |_st(m ; a_d , ) = ( a^2(m , z)-1 + ) a_d ( _ . ) . below we present the predictions for @xmath68 with the dipole amplitude @xmath126 and @xmath127 in galactic coordinate as reported by the planck observation @xcite . vs. mass for the model with dipole asymmetry . the black solid line and the red long dashed curve are for @xmath205 and @xmath206 at the redshift @xmath184 respectively . the predictions for @xmath207 at redshifts @xmath184 , @xmath185 and @xmath208 are presented by green dashed , blue dotted and magenta dash - dotted curves respectively . ] with the same caption . ] in fig . [ fig - n - ratio - dipole ] we plot @xmath68 versus mass . the horizontal black solid line represents the prediction at redshift @xmath184 with the observation direction angle @xmath205 . this curve shows that for the direction of sight perpendicular to dipole direction the number density of the structures does not change compared to the standard case . the red long dashed and the green dashed curves respectively are for @xmath206 and @xmath207 at the redshift @xmath184 . the blue dotted and the magenta dash - dotted curves respectively are for redshifts @xmath193 and @xmath209 at angle @xmath207 . the essential point here is that @xmath68 increases with redshift while it also depends on @xmath210 , the angle between the direction of the observation and the dipole direction in sky . in fig . [ fig - n - ratio - dipole1 ] we zoom in mass scale where one can see the @xmath77 barrier crossing at different redshifts . the critical mass , where @xmath173 changes sign , increases with redshift . this is understandable intuitively because of the hierarchy in structure formation . when one is in the region @xmath186 at higher redshifts one gets less deviation from standard case . this is because at higher redshifts the growth function is smaller and consequently the power spectrum and the variance are smaller . this means that the height parameter approaches the critical value and the effects of dipole corrections become small . on the other hand , when one crosses the @xmath77 barrier again in higher redshift then the height parameter becomes larger but it deviates from @xmath77 and consequently one obtains more structures . as discussed in subsection [ dipole - sec ] the cmb dipole asymmetry shows non - trivial scale - dependence such that for scales @xmath211 there is the upper bound on the amplitude of dipole @xmath212 from the quasar data @xcite . therefore , it is desirable to consider the predictions of a scale - dependent dipole asymmetry on the abundance of structures . the simplest proposal is to assume a power - law behavior for @xmath213 as follows [ eq : powerdiprun ] _ = , where @xmath214 is the amplitude of the dipole asymmetry at the pivot wavenumber @xmath215 and @xmath216 shows the spectral index of the dipole amplitude . for @xmath217 we obtain the scale - invariant dipole asymmetry studied above . now substituting the primordial power spectrum eq . ( [ eq : powerdiprun ] ) in eq . ( [ eq : variance - change ] ) , @xmath218 is obtained to be ( m ; a_d,)=a_dp(_.)/(|(m))k^n_a+2 _ ( k)w^2(k ) . correspondingly , the change in number density is given by _ n(m ; a_d , ) = ( ^2(m)-1 ) a_dp(_.)k^n_a+2 _ ( k)w^2(k ) . vs. mass for the model with dipole asymmetry at @xmath219 . the red long dashed curve and the green dashed curve show the prediction for @xmath217 and @xmath220 with @xmath207 respectively . the black solid curve shows the constraints from quasars on dipole asymmetry . the vertical solid black lines indicate the mass ranges of dark matter halos which are the hosts of quasars . the mass - range of quasar host halos are @xmath221 . ] hirata @xcite has used the quasar data to check the amplitude of hemispherical asymmetry in lss in which the constraint @xmath222 at redshift @xmath219 is interpreted as the bound @xmath223 with 1-sigma accuracy marginalized over directions . now we can use this observation to constrain the spectral index of the dipole asymmetry . in fig . [ fig : running ] we have plotted @xmath68 versus mass for sheth - tormen universality function at redshift @xmath219 . the solid black curve represents the limit obtained from quasar analysis . in order to constrain the spectral index of dipole anisotropy , we should consider the mass between @xmath7 and @xmath8 of the host halos of quasars . the minimum mass of the host dark matter halo typically are @xmath224 @xcite . as a result we obtain the bound @xmath1 to satisfy the lss constraints on the dipole asymmetry . it is important to mention that we assume the quasars are a fair sample of galaxies with a host halo of masses @xmath225 . accordingly counting the number of quasar in a specific redshift bin gives us information on the distribution of dark matter halos . then we assume that the amplitude of dipole asymmetry seen in quasar data , @xmath226 , can be assigned as an upper limit for @xmath227 obtained from the number count of dark matter halos . we change the spectral index of dipole asymmetry in the primordial power spectrum in order to have exactly the same amount of structures predicted by the observed upper limit dipole asymmetry in quasar data . the value of @xmath228 is the upper bound for dipole spectral index in that sense . the mass range that we can trust our bound is @xmath178 . the quasar sample is in redshift @xmath229 ( @xmath230 ) , accordingly the mass range of the host dark matter halos are in valid range of our calculations . in this subsection we consider models of inflation which generate statistical anisotropies in curvature perturbation power spectrum as given in eq . ( [ eq - psaniso0 ] ) . the case with @xmath231 is the known quadratic anisotropy @xcite which is generated in models of anisotropic inflation @xcite . as in previous examples , to calculate the change in the abundance of structures , first we calculate the change in the variance . the important point in this type of feature is that the variance is obtained from the integration in @xmath157-space so we can use the orthogonality of spherical harmonics to obtain @xcite = |(m ) . this means that in anisotropic models only the amplitude of the variance is modified . for press - schechter mass function @xmath68 becomes _ n have [ eq : deltanqu ] _ n(m ; g_lm ) |_st = ( a^2(m , z)-1 + ) . eq.([eq : deltanqu ] ) shows that although the variance has changed by a constant amplitude @xmath232 but the abundance of structures does not change just by a constant amplitude in which it depends non - trivially on the mass of structures which appears implicitly in the height parameter in eq . ( [ eq : deltanqu ] ) . as in previous examples , define @xmath77 the point where @xmath173 changes sign with the mass @xmath2 . for the mass scales @xmath175 with @xmath186 a positive ( negative ) change in the amplitude of the power - spectrum @xmath233 causes a decrease ( increase ) in the abundance of structures . interestingly , we see that quadrupole asymmetry with @xmath234 does not affect the abundance of structures . although the anisotropic power spectrum introduced in eq . ( [ eq - psaniso0 ] ) does not change the statistics of the structures but it has a secondary effect through the bias parameter @xcite . the bias parameter relates the abundance of dark matter halos , which are the host of galaxies ( observable quantities ) , to the underlying dark matter density perturbations . therefore , models with anisotropy in the form of eq . ( [ eq - psaniso0 ] ) can have non - trivial predictions for galaxy number density due to bias and not from the statistics of the structure such as in press - schechter formalism . in this work we studied the effects of inflationary models with anomalous feature in power spectrum on the number density of large scale structure and dark matter halos which are the host environment of galaxies and group of galaxies . the mass function of structures in universe is a function of local collapsed parameter ( i.e. @xmath86 ) and the variance of mass distribution . the latter itself is the fourier transform of matter density contrast power spectrum which is a function of the mass scale . therefore , the variance is related to the underlying cosmological model and the initial conditions . we have shown that any feature in primordial power spectrum induces a change in the variance of structures in large scales . accordingly , in the light of press - schechter idea ( to obtain the number density of structures ) , a change in variance results in a change in number density of structures . we also assumed that to leading order the features in primordial power spectrum only modify the variance and the processes of excursion set theory do not change compared to standard case as we discussed at the end of in sec . [ sec - ii ] . first we have shown that the change in number density , @xmath68 , is proportional to the change in variance , @xmath182 . however the sign of this relation depends on the mass scale and the redshift which one observes the structure . at any specific redshift , there is a mass scale @xmath2 in which for @xmath235 corresponding to @xmath236 we have @xmath237 @xmath238 . it is important to note that the regime @xmath239 corresponds to a quasi linear regime , where we should interpret the results in this regime with cautious and the results in this work does not apply for @xmath240 as the structures are not yet formed at these scales . three distinct anomalous features are considered in sec . [ sec - iii ] . first we considered models with localized feature in power spectrum . we have shown that localized features with positive amplitude ( excess in power spectrum ) can reduce the number of the structures at corresponding scales . while the power spectrum of matter distribution is well constrained at scales @xmath241 but the missing satellite problem at galactic scales opens up the possibility that the inflationary models with features can be a promising new candidate to solve the cold dark matter challenges on galactic scales . the other interesting point is that the statistics of structures decreases with redshift . consequently , we have the maximum effect from the localized feature in local universe . next we have investigated the effects of primordial curvature perturbation with dipole asymmetry on the abundance of structures . the change in the abundance of structures is proportional to the amplitude of the dipole asymmetry . there is no change for the abundance of structures in the direction perpendicular to the dipole direction while this change is maximum along the direction of dipole . in the limit @xmath186 the change in abundance is larger at low redshifts while for @xmath85 this behavior is reversed . in order to be consistent with observations on sub - cmb scales we have considered dipole asymmetry in the form @xmath242 . imposing the constraints from quasar data we have found @xmath1 . finally we studied the imprints of primordial curvature perturbations with statistical anisotropies , such as in models of anisotropic inflation , on the abundance of structures . we have shown that the change in the amplitude of variance is controlled by @xmath233 while the change in abundance of structures depends only on the universal mass function . the large scale structure observations are promising arena for observational cosmology to probe the physics of early universe . any specific deviation from the standard nearly adiabatic , nearly scale - invariant , nearly gaussian and isotropic initial conditions can be probed in lss . in addition the scale - dependence of these features ( i.e. non - gaussianity , dipole asymmetry etc ) can be probed by the tracers of dark matter . however , non - linear structure formation , the mechanism of collapse , dark matter and baryon biases and redshift space distortions all are the complications which need to be understood in order to investigate the physics of early universe . a. mana , t. giannantonio , j. weller , b. hoyle , g. huetsi and b. sartoris , mon . not . roy . astron . soc . * 434 * , 684 ( 2013 ) [ arxiv:1303.0287 [ astro-ph.co ] ] . d. k. hazra , jcap * 1303 * , 003 ( 2013 ) [ arxiv:1210.7170 [ astro-ph.co ] ] . m. boylan - kolchin , j. s. bullock and m. kaplinghat , mon . not . * 415 * , l40 ( 2011 ) [ arxiv:1103.0007 [ astro-ph.co ] ] . s. garrison - kimmel , s. horiuchi , k. n. abazajian , j. s. bullock and m. kaplinghat , arxiv:1405.3985 [ astro-ph.co ] . a. r. zentner , int . j. mod . d * 16 * , 763 ( 2007 ) [ astro - ph/0611454 ] . h. j. mo and s. d. m. white , mon . not . soc . * 282 * , 347 ( 1996 ) [ astro - ph/9512127 ] . j. e. gunn and j. r. gott , iii , astrophys . j. * 176 * , 1 ( 1972 ) . w. h. press and p. schechter , astrophys . j. * 187 * , 425 ( 1974 ) . m. maggiore and a. riotto , astrophys . j. * 717 * , 526 ( 2010 ) [ arxiv:0903.1251 [ astro-ph.co ] ] . a. de simone , m. maggiore and a. riotto , mon . not . * 412 * , 2587 ( 2011 ) [ arxiv:1007.1903 [ astro-ph.co ] ] . g. damico , m. musso , j. norena and a. paranjape , jcap * 1102 * , 001 ( 2011 ) [ arxiv:1005.1203 [ astro-ph.co ] ] . p. t. p. viana and a. r. liddle , mon . not . soc . * 281 * , 323 ( 1996 ) [ astro - ph/9511007 ] . m. j. white , p. t. p. viana , a. r. liddle and d. scott , mon . not . soc . * 283 * , 107 ( 1996 ) [ astro - ph/9605057 ] . a. liddle and d. lyth , _ cosmological inflation and large - scale structure _ , cambridge university press , 2000 . a. a. abolhasani , h. firouzjahi , s. khosravi and m. sasaki , jcap * 1211 * , 012 ( 2012 ) [ arxiv:1204.3722 [ astro-ph.co ] ] . a. d. linde , phys . d * 49 * , 748 ( 1994 ) [ arxiv : astro - ph/9307002 ] . a. r. liddle , d. h. lyth , e. d. stewart and d. wands , phys . d * 49 * , 6410 ( 1994 ) [ arxiv : astro - ph/9401011 ] . h. k. eriksen , a. j. banday , k. m. gorski , f. k. hansen and p. b. lilje , astrophys . j. * 660 * , l81 ( 2007 ) [ astro - ph/0701089 ] . j. hoftuft , h. k. eriksen , a. j. banday , k. m. gorski , f. k. hansen and p. b. lilje , astrophys . j. * 699 * , 985 ( 2009 ) [ arxiv:0903.1229 [ astro-ph.co ] ] . f. k. hansen , a. j. banday , k. m. gorski , h. k. eriksen and p. b. lilje , astrophys . j. * 704 * , 1448 ( 2009 ) [ arxiv:0812.3795 [ astro - ph ] ] . c. gordon , astrophys . j. * 656 * , 636 ( 2007 ) [ astro - ph/0607423 ] ; y. akrami , y. fantaye , a. shafieloo , h. k. eriksen , f. k. hansen , a. j. banday and k. m. gorski , astrophys . j. * 784 * , l42 ( 2014 ) [ arxiv:1402.0870 [ astro-ph.co ] ] . g. aslanyan , l. c. price , k. n. abazajian and r. easther , arxiv:1403.5849 [ astro-ph.co ] . a. notari , m. quartin and r. catena , jcap * 1403 * , 019 ( 2014 ) [ arxiv:1304.3506 [ astro-ph.co ] ] . l. dai , d. jeong , m. kamionkowski and j. chluba , phys . d * 87 * , 123005 ( 2013 ) [ arxiv:1303.6949 [ astro-ph.co ] ] . m. h. namjoo , s. baghram and h. firouzjahi , phys . d * 88 * , 083527 ( 2013 ) [ arxiv:1305.0813 [ astro-ph.co ] ] . a. a. abolhasani , s. baghram , h. firouzjahi and m. h. namjoo , arxiv:1306.6932 [ astro-ph.co ] . h. firouzjahi , j. -o . gong and m. h. namjoo , arxiv:1405.0159 [ astro-ph.co ] . m. h. namjoo , a. a. abolhasani , s. baghram and h. firouzjahi , arxiv:1405.7317 [ astro-ph.co ] . d. h. lyth , arxiv:1405.3562 [ astro-ph.co ] . d. h. lyth , jcap * 1308 * , 007 ( 2013 ) [ arxiv:1304.1270 [ astro-ph.co ] ] . j. m. maldacena , jhep * 0305 * , 013 ( 2003 ) [ astro - ph/0210603 ] . c. m. hirata , jcap * 0909 * , 011 ( 2009 ) [ arxiv:0907.0703 [ astro-ph.co ] ] . s. flender and s. hotchkiss , jcap * 1309 * , 033 ( 2013 ) [ arxiv:1307.6069 [ astro-ph.co ] ] . c. l. bennett , r. s. hill , g. hinshaw , d. larson , k. m. smith , j. dunkley , b. gold and m. halpern _ et al . _ , astrophys . j. suppl . * 192 * , 17 ( 2011 ) [ arxiv:1001.4758 [ astro-ph.co ] ] . n. e. groeneboom and h. k. eriksen , astrophys . j. * 690 * , 1807 ( 2009 ) [ arxiv:0807.2242 [ astro - ph ] ] ; d. hanson and a. lewis , phys . d * 80 * , 063004 ( 2009 ) [ arxiv:0908.0963 [ astro-ph.co ] ] ; n. e. groeneboom , l. ackerman , i. k. wehus and h. k. eriksen , astrophys . j. * 722 * , 452 ( 2010 ) [ arxiv:0911.0150 [ astro-ph.co ] ] ; d. hanson , a. lewis and a. challinor , phys . d * 81 * , 103003 ( 2010 ) [ arxiv:1003.0198 [ astro-ph.co ] ] . m. a. watanabe , s. kanno and j. soda , phys . lett . * 102 * , 191302 ( 2009 ) [ arxiv:0902.2833 [ hep - th ] ] . j. kim and e. komatsu , phys . d * 88 * , 101301 ( 2013 ) [ arxiv:1310.1605 [ astro-ph.co ] ] . s. r. ramazanov and g. rubtsov , phys . d * 89 * , 043517 ( 2014 ) [ arxiv:1311.3272 [ astro-ph.co ] ] . l. ackerman , s. m. carroll and m. b. wise , phys . d * 75 * , 083502 ( 2007 ) [ erratum - ibid . d * 80 * , 069901 ( 2009 ) ] [ astro - ph/0701357 ] . a. r. pullen and m. kamionkowski , phys . d * 76 * , 103529 ( 2007 ) [ arxiv:0709.1144 [ astro - ph ] ] . l. anderson _ et al . _ [ boss collaboration ] , arxiv:1312.4877 [ astro-ph.co ] . s. baghram , n. afshordi and k. m. zurek , phys . rev . d * 84 * , 043511 ( 2011 ) [ arxiv:1101.5487 [ astro-ph.co ] ] . s. rahvar , s. baghram and n. afshordi , phys . d * 89 * , 063001 ( 2014 ) [ arxiv:1310.5412 [ astro-ph.co ] ] . p. scott , t. bringmann and y. akrami , j. phys . ser . * 375 * , 032012 ( 2012 ) [ arxiv:1205.1432 [ astro-ph.co ] ] . r. a. c. croft , d. h. weinberg , m. bolte , s. burles , l. hernquist , n. katz , d. kirkman and d. tytler , astrophys . j. * 581 * , 20 ( 2002 ) [ astro - ph/0012324 ] . m. tegmark and m. zaldarriaga , phys . d * 66 * , 103508 ( 2002 ) [ astro - ph/0207047 ] . n. y. gnedin and a. j. s. hamilton , mon . not . soc . * 334 * , 107 ( 2002 ) [ astro - ph/0111194 ] . u. seljak _ et al . _ [ sdss collaboration ] , phys . d * 71 * , 103515 ( 2005 ) [ astro - ph/0407372 ] . a. a. klypin , a. v. kravtsov , o. valenzuela and f. prada , astrophys . j. * 522 * , 82 ( 1999 ) [ astro - ph/9901240 ] . a. v. kravtsov , a. a. berlind , r. h. wechsler , a. a. klypin , s. gottloeber , b. allgood and j. r. primack , astrophys . j. * 609 * , 35 ( 2004 ) [ astro - ph/0308519 ] . a. m. brooks , m. kuhlen , a. zolotov and d. hooper , astrophys . j. * 765 * , 22 ( 2013 ) [ arxiv:1209.5394 [ astro-ph.co ] ] . m. goetz and j. sommer - larsen , astrophys . space sci . * 284 * , 341 ( 2003 ) [ astro - ph/0210599 ] . m. vogelsberger , j. zavala , c. simpson and a. jenkins , arxiv:1405.5216 [ astro-ph.ga ] .

orbits of 55 visual binary stars are computed using recent speckle interferometry data from the soar telescope : 33 first - time orbits and 22 revisions of previous orbit calculations . the orbital periods range from 1.4 to 370 years , the quality of orbits ranges from definitive to preliminary and tentative . most binaries consist of low - mass dwarfs and have short periods ( median period 31 years ) . the dynamical parallaxes and masses are evaluated and compared to the _ hipparcos _ parallaxes . using differential speckle photometry , binary components are placed on the color - magnitude diagram .

1608.01242

this paper presents new or updated orbits for 55 binary systems or subsystems . it is based on speckle interferometric measurements made at the 4.1-m southern astrophyisical research ( soar ) telescope @xcite combined with archival data collected in the washington double star catalog , wds @xcite . it continues previous work on binary orbits resulting from the soar speckle program . the sixth catalog of visual binary orbits , vb6 @xcite , presently contains orbital elements of more than 2600 pairs . knowledge of binary orbits is needed in various areas such as : ( i ) measurement of stellar masses ( especially for stars of very low or high mass , unusual chemical composition or at various evolutionary stages ) ; ( ii ) statistics of orbital elements in relation to mechanisms of binary formation and evolution ; ( iii ) accurate models of stellar motion for astrometry ; ( iv ) binary systems of special interest , for example exohosts or young binaries with circumstellar material . the latter class is poorly defined , given that a common `` uninteresting '' binary may suddenly become important in the light of new discoveries , e.g. the visual triple star hd 131399 hosting an unusual planet @xcite . accurate parallaxes soon to be available from _ gaia _ will greatly enhance the value of binary orbits for calibration of stellar masses . during five years of this mission , only portions of long binary periods will be covered , making the ground - based monitoring an essential complement . these arguments indicate that the work on binary orbits is part of the astronomical infrastructure and has its own value . most binaries studied here consist of late - type dwarfs in the solar neighborhood , including close pairs first resolved by _ these pairs have orbital periods measured in years and decades , while wider classical visual binaries have periods of a few centuries or millenia . efforts to monitor the motion of these fast binaries by speckle interferometry have been made by @xcite , @xcite , and others mostly on the northern sky . the soar data on their southern counterparts reveal a fast motion and the lack of coverage after the _ hipparcos _ discovery in 1991 . first , still preliminary orbits of several such pairs are presented here . another class of preliminary orbits have long periods and incomplete orbit coverage , so typical for visual orbits . unlike spectroscopists who accumulate data for one or several orbital revolutions before publishing the orbit , the visual binary community has a tradition of sharing the observations , because the time needed for an orbit calculation may exceed the human life span . the downside of this tradition is publication of premature orbits and frequent , sometimes poorly justified orbit revisions . one of the reasons for publishing here uncertain first - time orbits is the need to understand the orbital motion and to plan further observations for their improvement . compared to visual micrometer measures , speckle interferometry has a much improved accuracy ( the soar data have random and systematic errors of a few mas ) . some classical binaries discovered and measured at the limit of visual techniques have now reliable and accurate orbits based on speckle interferometry . large resolving power of the 4.1-m soar telescope allows to map previously inaccessible portions of their orbits near periastron . this is particularly important for difficult orbits with high eccentricity or oriented edge - on . such orbits are under - represented in the vb6 and hence distort the statistics . traditionally , visual binary orbits were mostly computed by the observers themselves who had a deep understanding of the underlying data and its reliability . speckle interferometry is more reliable than micrometer measures ; however , measures of close or very unequal pairs near the limit of the technique can also be unreliable or distorted by instrumental artifacts . access to the original soar data and the possibility to reprocess doubtful measures distinguishes this work from orbit calculation made by others . orbital elements and their errors are determined by least - squares fitting using the idl code orbit . ] when no prior orbit is available , the initial approximation is chosen interactively , considering the observed motion of each pair . this step is non - trivial because position angles are determined by speckle interferometry only modulo 180 , allowing quadrant `` flips '' , the orbit coverage often has large gaps , and some measurements are simply erroneous . sometimes the same data can be represented by several very different orbits . the least - squares method implies that the measurement errors are known and normally distributed ; in such case the optimum weights are inversely proportional to the squares of the errors . visual measurements do not match this model , having random and/or systematic errors that are difficult to quantify . the errors of speckle interferometry behave better , but depend on the instrument , telescope aperture , magnitude difference , wavelength , data quality , etc . , so there are no simple prescriptions for their evaluation . here the errors are assigned subjectively based on these considerations and are corrected iteratively by assigning larger errors ( hence smaller weights ) to the outliers . the result is checked by computing the @xmath0 metric , where @xmath1 is the number of degrees of freedom . in the ideal case , @xmath2 . the same error in the radial and tangential directions is adopted for each observation . this is true for speckle interferometry , while the known tendency of visual observers to measure angles more accurately than separations is ignored here , considering low weight of the visual data . for many pairs observed at soar , the rms residuals from the orbits are on the order of 1mas . the weighting scheme adopted in the vb6 does not fully account for the high accuracy of soar data ; as a result , some recent orbits ( e.g. * ? ? ? * ) are `` pulled away '' by inaccurate old measures and leave systematic residuals to the soar data . as mentioned above , for some binaries the scarce available observations do not fully constrain their orbital elements and are compatible with a wide range of orbits . in such cases , the mass sum calculated using the hip2 parallax @xcite provides a guidance on selecting the most plausible orbit . some orbital elements can be fixed while fitting the remaining elements , so that the resulting mass sum takes a reasonable value . this approach does not work for binaries with unknown or small parallax . table [ tab : orb ] lists the orbital elements and their errors in common notation ( @xmath3 orbital period , @xmath4 epoch of periastron , @xmath5 eccentricity , @xmath6 semimajor axis , @xmath7 position angle of the node for the equinox j2000.0 , @xmath8 argument of periastron , @xmath9 inclination ) . the first column gives the wds code of the binary and , in the following line , its _ hipparcos _ number when available . the system identifier adopted in the wds ( `` discoverer code '' ) is given in the second column . for each pair , the first line contains orbital elements , while the following line gives their formal errors . the errors are omitted for some preliminary orbits of grade 5 where they have no sense . provisional grades are assigned based on the principles described in the vb6 , where grades 4 and 5 mean preliminary orbits and grade 1 are definitive and accurate orbits . the last column contains references to previously computed orbits , when available . figure [ fig : mosaic ] gives samples of orbital plots , illustrating their different quality . its top row shows three first - time , but well defined orbits . the second row shows new preliminary orbits that risk substantial revisions in the future , mostly because the coverage is still insufficient . drastic revisions of previous orbits are shown in the bottom row . individual observations and residuals are listed in table [ tab : obs ] , available in full electronically . it contains still unpublished measures made at soar in 2016 , while some published soar measures were reprocessed . its first column identifies the binary by its wds code . then follow the time of observation @xmath10 in besselian years , position angle @xmath11 for the time of observation ( correction for precession is done internally during orbit calculation ) , separation @xmath12 in arcseconds , and measurement error @xmath13 . unrealistically large errors are assigned to the discarded observations , so that they have no influence on the fitted elements but are still kept in the table . the last two columns of table [ tab : obs ] contain the residuals o@xmath14c in angle and separation . table [ tab : ptm ] provides additional information , namely the the spectral type as given in _ hipparcos _ or simbad and the hip2 parallax @xmath15 , to be compared to the dynamical parallax @xmath16 in the next column . the latter is evaluated by the baize - romani method , applicable when binary components follow the standard main - sequence relation between mass and absolute magnitude . taking the initial mass sum @xmath17 , i compute the dynamical parallax @xmath18 , use it to evaluate the absolute @xmath19 magnitudes of both components , and estimate their masses @xmath20 and @xmath21 by means of the standard relation by @xcite , valid for @xmath22 and extrapolated to larger masses in a few cases . the new mass sum leads to the new value of @xmath23 , and the iterations converge rapidly . the components masses @xmath20 and @xmath21 obtained in this procedure are listed in the columns ( 6 ) and ( 7 ) of table [ tab : ptm ] . asterisks mark the dynamical parallaxes derived from reliable orbits of grade 3 or better . figure [ fig : dynpar ] compares the dynamical and _ hipparcos _ parallaxes . the combined @xmath19 magnitude and @xmath24 color index in columns ( 8) and ( 9 ) are taken mostly from the _ hipparcos _ catalog . for hip 83716b and 87914b , the combined magnitudes refer to the secondary subsystems , while @xmath25 mag is assumed to match the estimated spectral type k0v of both pairs . the last four columns of table [ tab : ptm ] provide differential photometry resulting from the soar speckle interferometry , where the filters @xmath26 and @xmath27 have central wavelengths and bandwidths of 543/22 and 788/132 nm , respectively . the @xmath28 and @xmath29 are average values , while @xmath30 and @xmath31 stand for the rms scatter of magnitude difference in each filter if measured several times , indicating internal consistency of the differential photometry . for some pairs this scatter is as good as 0.1 mag , but it can be up to @xmath320.5 mag . speckle differential photometry is unreliable for very close pairs at or below the diffraction limit . for faint or wide binaries , @xmath33 can be systematically overestimated . despite these caveats , speckle interferometry at soar is the only source of differential photometry for some close binaries . when both @xmath28 and @xmath29 are measured , the empirical linear relation @xmath34 holds , allowing to estimate differential magnitudes in both filters even when measurements in only one filter are available . it is also reasonable to assume that @xmath35 and @xmath36 . therefore , magnitudes and colors of each binary component can be estimated using combined and differential photometry . figure [ fig : cmd ] shows the @xmath37 color - magnitude diagram ( cmd ) where only binaries with @xmath38mas are plotted . for reference , the 1-gyr isochrones from @xcite and @xcite are shown . brief comments on some binaries are provided in this section . the acronym rv stands for radial velocity . _ _ hds 546 is a typical case of a _ hipparcos _ binary that has not been observed after its discovery in 1991 . it has made one full revolution since , and the short arc observed at soar shows a rapid motion . this is a pair of k2v chromospherically active dwarfs . _ 05229 - 4219 . _ tok 93 aa , ab was first resolved in 2011 at gemini - s @xcite . it has a variable rv and astrometric acceleration . some observations have been re - processed . the faint star b at 14 is physical . _ 05542 - 2909 . _ the orbit of fin 382 with @xmath39yr given in @xcite is wrong ( figure [ fig : mosaic ] ) , the true period is 10yr . the revised orbit has a very high quality , with the rms residuals of 3.6 and 1.8 mas in two coordinates . _ . _ koh 79 is a young variable star eg cha in the @xmath40 cha association . two preliminary orbits proposed by @xcite are now replaced by one non - ambiguous , though still preliminary , orbit . the dynamical parallax of 11.5mas matches the known distance to this association and may eventually help to measure it better . _ 09149 + 0427 . _ hei 350 is a nearby red dwarf pair gj 390 ( the spectral type of the primary is k4v ) . despite the large @xmath41 , the two first micromemeter measures by w. heintz in 1987 and 1988 have contradictory quadrants , and both do not agree well with the accurate measures available since 2003 . the pair has not been resolved by _ hiparcos _ , presumably because of the large @xmath33 . the orbit is only loosely constrained . the binary might be young and belong to the castor moving group . its secondary with @xmath42 mag is located above the main sequence in figure [ fig : cmd ] ; the large magnitude difference @xmath43 mag has been confirmed in 2016.4 . _ 09275 - 5806 . _ the orbit of chr 240 published in @xcite is substantially revised , with a new period of 1.42yr instead of 2.62yr . this very close pair is often under the diffraction limit of soar . _ 09466@xmath144966 . _ our relative photometry implies that @xmath44 , i.e. the secondary is bluer than the primary . the g5iv primary is evolved ; it is located in the subgiant branch of the cmd . the _ hipparcos _ photometry @xmath45 mag is suspect because in 1991 the pair was very close , 015 . _ 09510 - 8004 . _ hds 1421 ( b8iv ) belongs to the sco - cen association ; the dynamical parallax is 6.4mas . the orbit is still preliminary , but its curvature is well defined . _ 11102 - 1122 . _ hds 1590 ( hd 97038 ) is a double - lined spectroscopic binary ( d. latham , 2012 , private communication ) . its corrected orbit is now well defined after coverage of the periastron . _ 13072@xmath145420 . _ despite the drastic revision of the previous orbit , the elements of fin 54 remain tentative , given that only a small arc is covered ( figure [ fig : mosaic ] ) . one of the components is an eclipsing binary v949 cen . the dynamical parallax is 4.8mas , substantially larger than 1.3mas measured by _ hipparcos_. _ 13377@xmath142337 . _ two orbits of rst 2856 ( bd@xmath1422 3633 ) by @xcite with periods of 110 yr and 51.5 yr are now replaced by the new , but still tentative orbit with @xmath46yr . it is based on two hrcam measures and a handful of visual micrometer measures . the @xmath47 mag is derived from the very noisy data and is likely wrong ; the pair has not been resolved by _ hipparcos _ despite its 03 separation . the dynamical parallax is 12.3mas , the matching spectral type of the primary is g5v . _ this star hr 5278 ( f6v ) possibly contains a debris disk . a slight revision of the previous orbit by @xcite leaves very well defined elements . both components are at about 1.5 mag above the main sequence , while the mass sum is slightly larger than expected . the stars could have evolved off the main sequence or are still contracting . _ 14383@xmath144954 . _ a drastic revision of the orbit by @xcite changes the period of fin 371 from 27.6yr to 50.7yr . _ 14453@xmath143609 . _ the first tentative orbit for i 528 , with a poor coverage and doubtful visual measures , is suggested . the formal errors of the elements are misleadingly small . the component c is at 45 . the dynamic parallax is 5.3mas , and the binary may belong to the sco - cen association . _ 14589 + 0636 . _ wsi 81 ( hip 73314 ) is on the california planet search program @xcite . the new orbit is well defined and leads to the mass sum of 2.2 @xmath48 , while the dynamical parallax corresponds to 1.7 @xmath48 . both components are located on the main sequence . _ 15035@xmath144035 . _ the revision of the orbit by @xcite for i 1262 nearly doubles its period , from 134 to 251 yr , but the orbit still remains preliminary and poorly constrained . the binary belongs to the sco - cen association , its dynamic parallax is 5.7mas . _ 15042@xmath141530 . _ the new edge - on ( @xmath49 ) orbit of rst 3906 is very different from the previous one by @xcite with @xmath50 . the new orbit fits the speckle data well but implies that many micrometer measures of this pair are totally wrong ( possibly observers confused it with another pair ? ) . the pair was unresolved at soar on 2016.14 in agreement with the new orbit . _ 15282@xmath140921 . _ this nearby k2v star gj 586 is a spectroscopic binary with a record - high eccentricity of @xmath51 . precise rvs from @xcite constrain all spectroscopic elements very tightly , leaving for adjustment only the `` visual '' elements @xmath6 , @xmath7 , and @xmath9 . the orbital parallax is @xmath52mas , the orbital masses are @xmath53 and @xmath54 @xmath55 . _ 15290@xmath142852 . _ bu 1114 is a nearby dwarf with a physical tertiary component c at 96 . some published non - resolutions of ab at soar refer in fact to this component c , pointed erroneously instead of ab . the edge - on orbit is preliminary and yields the mass sum of 2.0 @xmath48 , less than 2.9 @xmath56 estimated from the luminosity . _ 15317 + 0053 . _ tok 48 ( hip 76031 , hd 138369 ) is a single - lined spectroscopic binary with a period of 619.3 days @xcite . resolved measures from soar are combined with the rvs of the primary component in the orbital fit . however , the solution converges to @xmath57 , contradicting the observed rv variation . therefore , in the combined speckle - spectroscopic orbit the inclination was fixed at 150 . the differential photometry @xmath58 mag and @xmath59 mag leaves no doubt that the spectrum should contain double lines . however , griffin could no find them , despite dedicated effort . his hypothesis that the secondary component is a close binary is not supported by the speckle photometry , but the secondary could have a fast axial rotation , reducing the depth of its lines in the blended spectrum . both components are located on the main sequence . _ _ chr 232 aa , ab ( hr 5765 ) is an a7v chemically peculiar binary @xcite with a new well - constrained 16.5-yr orbit ( rms residuals 1mas ) . a very similar orbit has just been published by @xcite . it has made more than one full revolution since its discovery in 1996.18 . the visual companion b at 92 is also a resolved close binary see 238 bc with a 60.4-yr orbit ; its observation with hrcam in 2008.54 in fact refers to the aa , ab pair . the two subsystems aa , ab and bc in this 2 + 2 quadruple are not co - planar . the dynamical parallaxes of aa , ab and bc are 9.4 and 10.3 mas , respectively , while the hip2 parallax is 10.4mas . the component ab is located above the main sequence in the cmd . _ 15348@xmath142808 . _ the first preliminary 26-yr orbit of tok 49 aa , ab is computed . the outer binary rst 1847 ab has a separation of 1 . _ 15548@xmath146554 . _ the first orbit of the nearby red dwarf binary nzo 65 ( g2v ) is well constrained . its components are on the main sequence . _ 16544@xmath143806 . _ this is a dramatic revision of the previous orbit of hds 2392 by @xcite . _ 17018@xmath145108 . _ the orbit of i 1306 by @xcite with @xmath60 yr is revised to @xmath61 yr and still remains preliminary ; only a half of it is covered in almost a century of observation . this is a bright star hr 6312 , a9iii . the mass sum of 1.0 @xmath48 does not match the spectral type , throwing suspicion on the hip2 parallax of 9.6mas ( the dynamical parallax is 5.8mas ) . _ 17066 + 0039 . _ after a slight correction , the orbit of the subsystem ba , bb with @xmath62yr is now well constrained . the outer pair bu 823 ab has a poorly defined 532-yr orbit . coplanarity between the inner and outer orbits is not excluded . _ _ a drastic revision of the orbit of fin 373 ( hr 6411 , @xmath9 aps , b8/9v ) . compared to the orbit by @xcite , even the direction of the rotation has changed from prograde to retrograde , see figure [ fig : mosaic ] . the published soar measure in 2009.26 was wrong and corresponded to the image doubling caused by telescope vibration ; it has been reprocessed . the rms residuals to the new orbit are 1.8mas , and it is now well defined . _ 17575@xmath145740 . _ the first 9-yr orbit of the subsystem ba , bb is well constrained . the outer pair ab has a separation of 25 and no orbit so far . _ 17584 + 0428 . _ kui 84 ( gj 9609 , k8 ) is a nearby close triple system where the secondary component of the 14.7-yr visual binary contains a 34.5-day spectroscopic subsystem @xcite . the secondary component is located above the main sequence in the cmd , as expected . deviations of the speckle measures from the latest visual orbit by @xcite prompted its slight revision here . rvs were used together with the speckle measures , leading to a very well - defined orbit . the hip2 parallax of 22.9mas corresponds to the mass sum of 2.0 @xmath56 , while the orbital parallax of 26.3mas and the mass sum of 1.3 @xmath56 are a better match to the masses estimated by @xcite . _ 18150@xmath145018 . _ the first orbit of i 429 ( hd 166839 , a0v ) looks reasonable , but then the discovery measure by innes in 1902.5 at a separation of 05 must be totally wrong . this and other measures made before 1929 were ignored in the orbit calculation . presently the pair is approaching periastron . _ 18368@xmath142617 . _ the new orbit of rst 2187 ab looks good , but corresponds to the mass sum of 1.4 @xmath56 . a revision of the hip2 parallax from 21.5@xmath631.7mas to 19mas is needed to increase the mass sum of this pair of nearly equal g3v stars to 2 @xmath55 . the tertiary component c at 126 is physical . _ _ hds 2665 is a nearby k2v dwarf . the primary component is located on the main sequence , while the secondary is above it . the mass sum of 2.0 @xmath56 is somewhat larger than expected , so there may be inner subsystems . _ 19581@xmath144808 . _ some observations of the nearby g0v dwarf binary hds 2842 were reprocessed for a minor revision of its previous 32-yr orbit , now definitive . _ 20217@xmath143637 . _ hds 2908 ( hd 193464 , f8v ) is another nearby dwarf binary . double lines were noticed , but there is no spectroscopic orbit in the literature . the @xmath64 mag is based on two well - resolved observations and is more reliable than @xmath65 mag . _ _ the 66-yr first orbit of chr 230 aa , ab ( k1/k2iii ) is well defined . the outer binary ab has a separation of 09 . the large eccentricity @xmath66 of aa , ab could be caused by the kozai - lidov cycles . both components are located on the giant branch in the cmd ( figure [ fig : cmd ] ) . new and updated orbits presented here make an incremental improvement of the vb6 content and thus contribute , albeit in a small way , to the observational foundations of astronomy . some of these binaries are interesting because they are members of multiple systems with three or more components , are young , lead to the useful measurements of masses , or for other reasons . this work used the simbad service operated by centre des donnes stellaires ( strasbourg , france ) , bibliographic references from the astrophysics data system maintained by sao / nasa , and the washington double star catalog maintained at usno . 04180@xmath143826 & hds 546 & 22.674 & 2014.439 & 0.422 & 0.2333 & 53.2 & 211.1 & 110.6 & 4 & new + 20048 & & @xmath630.218 & @xmath630.087 & @xmath630.028 & @xmath630.0052 & @xmath632.2 & @xmath635.2 & fixed & & + 05229@xmath144219 & tok 93 aa , ab & 5.680 & 2013.511 & 0.481 & 0.0594 & 225.6 & 92.6 & 59.6 & 3 & new + 25148 & & @xmath630.270 & @xmath630.088 & @xmath630.041 & @xmath630.0033 & @xmath634.1 & @xmath633.4 & @xmath633.0 & & + 05542@xmath142909 & fin 382 & 10.070 & 2015.397 & 0.537 & 0.1117 & 194.8 & 200.1 & 127.7 & 1 & tok2015c + 27901 & & @xmath630.036 & @xmath630.018 & @xmath630.004 & @xmath630.0014 & @xmath631.2 & @xmath631.6 & @xmath631.0 & & + 06018@xmath147158 & hei 676 & 102.0 & 2006.39 & 0.657 & 0.3988 & 301.9 & 340.0 & 169.0 & 5 & new + 28573 & & @xmath6312.5 & @xmath630.18 & @xmath630.035 & @xmath630.0237 & @xmath631.2 & fixed & fixed & & + 06032@xmath671922 & hds 823 aa , ab & 100 & 2044.6 & 0.93 & 0.449 & 51.5 & 80.2 & 95.9 & 5 & new + 28671 & & & & & & & & & & + 08369@xmath147857 & koh 79 ab & 77.1 & 1996.4 & 0.291 & 0.2441 & 182.0 & 320.7 & 106.2 & 4 & koh2002 + & & @xmath638.9 & @xmath631.5 & @xmath630.052 & @xmath630.0207 & @xmath631.3 & @xmath6311.3 & @xmath631.5 & & + 08398@xmath146604 & hds 1245 & 33.84 & 2013.175 & 0.781 & 0.2118 & 299.6 & 97.3 & 129.5 & 2 & new + 42496 & & @xmath630.90 & @xmath630.055 & @xmath630.013 & @xmath630.0059 & @xmath631.8 & @xmath631.7 & @xmath632.6 & & + 08494@xmath670341 & a 2899 & 300 & 2123 & 0.231 & 1.53 & 4.0 & 250.4 & 107.8 & 5 & new + 43304 & & fixed & @xmath6351 & @xmath630.059 & @xmath630.19 & @xmath635.1 & @xmath6339.9 & @xmath633.6 & & + 09149@xmath670427 & hei 350 & 62.68 & 1984.83 & 0.20 & 0.810 & 102.3 & 263.6 & 136.0 & 5 & tok2015c + 45383 & & @xmath632.88 & @xmath631.60 & fixed & @xmath630.037 & @xmath632.1 & @xmath6310.6 & @xmath633.0 & & + 09186@xmath672049 & ho 43 & 358.6 & 1941.2 & 0.40 & 0.610 & 80.0 & 214.9 & 131.6 & 4 & baz1989a + 45671 & & @xmath639.4 & @xmath632.5 & fixed & @xmath630.013 & @xmath633.6 & @xmath636.7 & @xmath632.3 & & + 09275@xmath145806 & chr 240 & 1.421 & 2015.376 & 0.391 & 0.0324 & 259.2 & 141.3 & 130.0 & 2 & tok2015c + 46388 & & @xmath630.007 & @xmath630.027 & @xmath630.038 & @xmath630.0009 & @xmath635.3 & @xmath638.6 & fixed & & + 09320@xmath140111 & chr 174 & 6.09 & 2015.06 & 0.471 & 0.0342 & 29.6 & 351.1 & 159.4 & 2 & new + 46776 & & @xmath630.14 & @xmath630.04 & @xmath630.019 & @xmath630.0022 & @xmath6332.5 & @xmath6330.6 & @xmath6312.1 & & + 09466@xmath144955 & hds 1414 & 60 & 2018.67 & 0.80 & 0.144 & 291.3 & 243.1 & 62.8 & 5 & new + 47969 & & fixed & @xmath632.76 & @xmath630.35 & @xmath630.128 & @xmath6310.6 & @xmath6330.9 & @xmath6334.1 & & + 09510@xmath148004 & hds 1421 & 63.52 & 2015.24 & 0.50 & 0.1583 & 142.3 & 135.2 & 34.5 & 4 & new + 48320 & & @xmath635.68 & @xmath630.37 & fixed & @xmath630.0057 & @xmath6315.8 & @xmath6320.9 & @xmath637.0 & & + 10112@xmath143245 & hds 1469 & 30.54 & 2015.323 & 0.636 & 0.1482 & 206.9 & 156.0 & 134.8 & 4 & new + 49883 & & @xmath633.92 & @xmath630.474 & fixed & @xmath630.0235 & @xmath6311.2 & @xmath6317.6 & @xmath6315.6 & & + 11102@xmath141122 & hds 1590 & 20.61 & 1995.616 & 0.783 & 0.1613 & 254.7 & 74.6 & 120.7 & 3 & tok2015c + 54580 & & @xmath630.43 & @xmath630.424 & @xmath630.009 & @xmath630.0044 & @xmath631.6 & @xmath630.8 & @xmath631.0 & & + 11238@xmath143829 & chr 241 & 3.559 & 2013.264 & 0.322 & 0.0576 & 154.2 & 257.8 & 114.7 & 3 & new + 55628 & & @xmath630.022 & @xmath630.078 & @xmath630.029 & @xmath630.0014 & @xmath632.1 & @xmath633.7 & @xmath631.8 & & + 11425@xmath672355 & cou 390 & 92.75 & 2007.86 & 0.786 & 0.5498 & 231.4 & 100.9 & 43.6 & 3 & new + 57112 & & @xmath635.57 & @xmath630.28 & @xmath630.014 & @xmath630.0168 & @xmath633.2 & @xmath632.7 & @xmath633.2 & & + 11525@xmath141408 & hds 1676 & 16.71 & 1999.60 & 0.681 & 0.1474 & 324.6 & 307.5 & 48.2 & 3 & tok2014a + 57894 & & @xmath630.42 & @xmath630.42 & @xmath630.010 & @xmath630.0030 & @xmath632.0 & @xmath632.9 & @xmath631.7 & & + 13072@xmath145420 & fin 54 & 373.7 & 2189.08 & 0.01 & 0.3516 & 113.3 & 79.7 & 135.3 & 5 & hrt2011d + 64025 & & & & & & & & & & + 13081@xmath147719 & hds 1839 & 17.44 & 2014.11 & 0.168 & 0.2007 & 349.2 & 320.5 & 120.5 & 4 & new + 64091 & & @xmath631.03 & @xmath630.84 & @xmath630.104 & @xmath630.0124 & @xmath635.4 & @xmath6319.0 & @xmath632.5 & & + 13306@xmath144238 & hds 1891 & 12.51 & 2014.76 & 0.192 & 0.185 & 99.9 & 273.2 & 162.6 & 4 & new + 65906 & & @xmath630.05 & @xmath631.38 & @xmath630.046 & @xmath630.027 & @xmath63103.7 & @xmath63157.7 & @xmath6327.1 & & + 13377@xmath142337 & rst 2856 & 122 & 2038.6 & 0.55 & 0.352 & 108.7 & 290.2 & 81.9 & 5 & hei1997 + & & & & & & & & & & + 13598@xmath140333 & hds 1962 & 9.496 & 2008.139 & 0.374 & 0.0734 & 205.6 & 49.5 & 49.9 & 3 & hor2011b + 68380 & & @xmath630.071 & @xmath630.060 & @xmath630.032 & @xmath630.0029 & @xmath633.7 & @xmath632.7 & @xmath633.2 & & + 14020@xmath142108 & wsi 79 & 22.41 & 2015.258 & 0.735 & 0.1967 & 137.2 & 146.6 & 140.5 & 3 & tok2015c + 68552 & & @xmath632.00 & @xmath630.022 & @xmath630.015 & @xmath630.0067 & @xmath633.1 & @xmath635.0 & @xmath632.8 & & + 14383@xmath144954 & fin 371 & 51.13 & 2015.39 & 0.205 & 0.0998 & 234.4 & 38.9 & 109.1 & 3 & zir2014a + 71577 & & @xmath633.79 & @xmath631.50 & @xmath630.028 & @xmath630.0041 & @xmath631.7 & @xmath6313.6 & @xmath631.5 & & + 14453@xmath143609 & i 528 ab & 28.92 & 2008.87 & 0.45 & 0.0748 & 88.8 & 276.4 & 56.2 & 5 & new + 72140 & & @xmath630.93 & @xmath631.11 & fixed & @xmath630.0065 & @xmath637.2 & @xmath635.6 & @xmath635.1 & & + 14589@xmath670636 & wsi 81 & 9.364 & 2006.981 & 0.145 & 0.1344 & 65.0 & 279.1 & 130.5 & 3 & new + 73314 & & @xmath630.049 & @xmath630.086 & @xmath630.008 & @xmath630.0015 & @xmath631.7 & @xmath634.5 & @xmath631.1 & & + 15035@xmath144035 & i 1262 & 235 & 1920.6 & 0.257 & 0.3122 & 33.6 & 189.6 & 52.6 & 5 & hei1996a + 73667 & & & & & & & & & & + 15042@xmath141530 & rst 3906 & 53.897 & 2018.396 & 0.740 & 0.1570 & 154.5 & 3.7 & 89.4 & 4 & hei1981a + 73724 & & @xmath6315.772 & @xmath633.868 & @xmath630.337 & @xmath630.0278 & @xmath631.3 & @xmath6332.1 & @xmath631.5 & & + 15251@xmath142340 & rst 2957 & 53.52 & 2020.795 & 0.850 & 0.2762 & 96.4 & 302.0 & 86.0 & 4 & new + 75478 & & @xmath636.57 & @xmath630.722 & @xmath630.086 & @xmath630.0076 & @xmath631.3 & fixed & @xmath630.6 & & + 15282@xmath140921 & bag 25 aa , ab & 2.43623 & 2007.2603 & 0.976 & 0.1074 & 272.8 & 255.6 & 55.4 & 2 & new + 75722 & & fixed & fixed & fixed & @xmath630.0081 & @xmath632.1 & fixed & @xmath633.6 & & + 15290@xmath142852 & bu 1114 ab & 205 & 2047.16 & 0.498 & 0.7219 & 138.3 & 359.5 & 90.6 & 4 & new + 75790 & & fixed & @xmath634.45 & @xmath630.045 & @xmath630.0272 & @xmath630.5 & @xmath635.4 & @xmath630.3 & & + 15317@xmath670053 & tok 48 & 1.713 & 2016.470 & 0.435 & 0.0427 & 161.9 & 45.8 & 150.0 & 3 & new + 76031 & & @xmath630.002 & @xmath630.010 & @xmath630.017 & @xmath630.0010 & @xmath632.1 & @xmath632.3 & fixed & & + 15332@xmath142429 & chr 232 aa , ab & 16.468 & 2011.536 & 0.419 & 0.0933 & 10.4 & 321.6 & 123.9 & 2 & new + 76143 & & @xmath630.194 & @xmath630.064 & @xmath630.009 & @xmath630.0019 & @xmath631.9 & @xmath632.6 & @xmath630.9 & & + 15348@xmath142808 & tok 49 aa , ab & 26.27 & 2006.28 & 0.342 & 0.1814 & 344.1 & 131.9 & 56.0 & 5 & new + 76275 & & & & & & & & & & + 15548@xmath146554 & nzo 65 & 120.60 & 2009.88 & 0.464 & 0.6831 & 122.2 & 219.0 & 85.7 & 3 & new + 77921 & & @xmath632.18 & @xmath630.11 & @xmath630.010 & @xmath630.0085 & @xmath630.4 & fixed & @xmath630.2 & & + 16544@xmath143806 & hds 2392 & 26.48 & 2009.459 & 0.50 & 0.1858 & 101.3 & 154.8 & 134.7 & 4 & tok2015c + 82709 & & @xmath630.67 & @xmath630.086 & fixed & @xmath630.0045 & @xmath633.2 & @xmath632.8 & @xmath632.0 & & + 17018@xmath145108 & i 1306 & 193.7 & 1989.59 & 0.187 & 0.317 & 12.8 & 149.0 & 84.6 & 5 & ole2004b + 83321 & & & & & & & & & & + 17066@xmath670039 & tok 52 ba , bb & 6.339 & 2012.979 & 0.386 & 0.0725 & 15.9 & 179.2 & 34.0 & 2 & tok2014a + 83716 & & @xmath630.146 & @xmath630.076 & @xmath630.023 & @xmath630.0027 & @xmath6310.7 & @xmath6313.6 & @xmath635.5 & & + 17221@xmath147007 & fin 373 & 56.90 & 2011.48 & 0.839 & 0.0779 & 323.0 & 51.2 & 126.3 & 3 & doc2013d + 84979 & & @xmath637.67 & @xmath630.39 & @xmath630.023 & @xmath630.0106 & @xmath634.2 & @xmath635.0 & @xmath638.1 & & + 17575@xmath145740 & tok 55 ba , bb & 8.92 & 2015.16 & 0.387 & 0.1225 & 345.7 & 342.3 & 104.1 & 2 & new + 87914 & & @xmath630.36 & @xmath630.13 & @xmath630.020 & @xmath630.0026 & @xmath631.3 & @xmath638.7 & @xmath630.8 & & + 17584@xmath670428 & kui 84 & 14.715 & 2001.496 & 0.496 & 0.1746 & 351.5 & 190.0 & 62.8 & 1 & doc2005f + 87991 & & @xmath630.040 & @xmath630.051 & @xmath630.008 & @xmath630.0014 & @xmath630.8 & @xmath631.8 & @xmath631.2 & & + 18092@xmath142211 & rst 3157 & 10.685 & 2015.031 & 0.409 & 0.1575 & 232.4 & 55.5 & 50.2 & 2 & tok2015c + 88932 & & @xmath630.045 & @xmath630.020 & @xmath630.008 & @xmath630.0019 & @xmath631.3 & @xmath631.3 & @xmath631.2 & & + 18150@xmath145018 & i 429 & 63.10 & 2021.4 & 0.84 & 0.150 & 132.6 & 228.2 & 84.0 & 4 & new + & & @xmath635.75 & @xmath633.7 & @xmath630.16 & @xmath630.031 & @xmath632.9 & @xmath6314.0 & @xmath632.1 & & + 18368@xmath142617 & rst 3187 ab & 18.37 & 2017.30 & 0.364 & 0.1661 & 40.6 & 7.2 & 70.1 & 3 & new + 91253 & & @xmath630.18 & @xmath630.16 & @xmath630.022 & @xmath630.0020 & @xmath630.9 & @xmath632.9 & @xmath630.8 & & + 18480@xmath141009 & hds 2665 & 31.212 & 2020.007 & 0.705 & 0.4535 & 192.8 & 44.2 & 49.1 & 3 & new + 92250 & & @xmath630.003 & @xmath630.194 & @xmath630.014 & @xmath630.0178 & @xmath631.3 & @xmath633.4 & @xmath633.4 & & + 19581@xmath144808 & hds 2842 & 35.28 & 2016.036 & 0.738 & 0.2364 & 254.2 & 60.6 & 70.2 & 3 & tok2015c + 98274 & & @xmath632.48 & @xmath630.065 & @xmath630.037 & @xmath630.0140 & @xmath632.4 & @xmath634.1 & @xmath632.2 & & + 20057@xmath143743 & hds 2865 & 50.0 & 2018.71 & 0.328 & 0.1778 & 4.2 & 131.7 & 131.5 & 4 & new + 98979 & & fixed & @xmath630.56 & @xmath630.040 & @xmath630.0043 & @xmath633.7 & @xmath636.6 & @xmath634.2 & & + 20104@xmath141923 & hds 2873 & 90.0 & 2007.46 & 0.484 & 0.3843 & 35.2 & -0.3 & 94.0 & 5 & new + 99391 & & fixed & @xmath630.67 & @xmath630.006 & @xmath630.0087 & @xmath630.9 & @xmath634.3 & @xmath630.4 & & + 20202@xmath143435 & i 1416 & 20.16 & 2000.19 & 0.90 & 0.163 & 140.1 & 37.2 & 113.9 & 3 & b__1961d + 100266 & & @xmath630.31 & @xmath631.24 & fixed & @xmath630.035 & @xmath633.2 & @xmath6319.7 & @xmath6311.6 & & + 20217@xmath143637 & hds 2908 & 12.74 & 2007.44 & 0.518 & 0.1120 & 107.8 & 338.9 & 80.8 & 3 & new + 100417 & & @xmath630.24 & @xmath630.52 & @xmath630.046 & @xmath630.0065 & @xmath630.6 & @xmath6310.7 & @xmath632.7 & & + 20248@xmath141943 & hds 2919 & 65.0 & 2020.90 & 0.474 & 0.3243 & 269.7 & 311.6 & 60.9 & 5 & new + 100685 & & fixed & @xmath630.75 & @xmath630.062 & @xmath630.0086 & @xmath632.9 & @xmath633.6 & @xmath632.5 & & + 21073@xmath145702 & hds 3009 & 54.08 & 2013.30 & 0.311 & 0.3482 & 350.1 & 143.3 & 56.8 & 3 & new + 104256 & & @xmath634.21 & @xmath630.75 & @xmath630.030 & @xmath630.0050 & @xmath631.0 & @xmath638.3 & @xmath631.9 & & + 22116@xmath143428 & chr 230 aa , ab & 66.5 & 2010.37 & 0.929 & 0.1156 & 318.7 & 305.0 & 70.5 & 3 & new + 109561 & & @xmath6311.0 & @xmath630.39 & @xmath630.024 & @xmath630.0168 & @xmath633.5 & @xmath637.5 & @xmath634.1 & & c r rrr rr 04180 - 3826 & 1991.2500 & 230.0 & 0.1360 & 0.0100 & 0.1 & @xmath140.0002 + 04180 - 3826 & 2014.0430 & 228.6 & 0.1332 & 0.0015 & 0.3 & 0.0003 + 04180 - 3826 & 2014.8536 & 207.0 & 0.1022 & 9.0047 & @xmath144.5 & 0.0051 + 04180 - 3826 & 2015.1053 & 201.1 & 0.0826 & 0.0020 & @xmath141.8 & @xmath140.0006 + 04180 - 3826 & 2015.7385 & 165.3 & 0.0566 & 0.0020 & 1.9 & 0.0001 + 04180 - 3826 & 2015.9080 & 148.8 & 0.0559 & 0.0020 & 0.6 & 0.0008 + 04180 - 3826 & 2016.1373 & 127.1 & 0.0578 & 0.0020 & @xmath141.3 & @xmath140.0010 ccc cc cc rr rr rr 04180@xmath143826 & 20048 & k2v & 26.8 & 25.0 & 0.8 & 0.7 & 8.89 & 1.11 & & & 0.50 & 0.14 + 05229@xmath144219 & 25148 & g5v & 15.1 & 15.1 * & 1.1 & 0.8 & 8.70 & 0.76 & & & 1.22 & 0.38 + 05542@xmath142909 & 27901 & f3v & 18.5 & 17.2 * & 1.5 & 1.2 & 6.34 & 0.43 & 1.49 & 0.18 & 1.06 & + 06018@xmath147158 & 28573 & k2/k3iv & 15.0 & 15.6 & 0.9 & 0.8 & 9.77 & 1.09 & & & 0.60 & 0.12 + 06032@xmath671922 & 28671 & g0v & 16.8 & 18.0 & 0.9 & 0.6 & 9.28 & 0.71 & & & 1.90 & 0.48 + 08369@xmath147857 & & k4ve & & 11.5 & 0.8 & 0.8 & 10.46 & 1.45 & & & 0.39 & 0.21 + 08398@xmath146604 & 42496 & k1v : & 16.6 & 17.7 * & 0.9 & 0.7 & 9.73 & 0.97 & 1.93 & 0.73 & 1.51 & 0.34 + 08494@xmath670341 & 43304 & g5 & 28.2 & 29.7 & 1.0 & 0.5 & 7.96 & 0.79 & & & 2.94 & 0.52 + 09149@xmath670427 & 45383 & k0 & 55.7 & 46.4 & 0.8 & 0.5 & 7.91 & 1.19 & 3.98 & & 2.13 & 0.46 + 09186@xmath672049 & 45671 & f5 & 9.2 & 9.1 & 1.3 & 1.1 & 8.64 & 0.51 & & & 0.77 & + 09275@xmath145806 & 46388 & g2v & 20.0 & 19.5 * & 1.2 & 1.1 & 7.20 & 0.67 & 0.73 & 0.34 & & + 09320@xmath140111 & 46776 & a3v & 6.3 & 5.4 * & 3.6 & 3.3 & 4.54 & 0.16 & 0.44 & 0.16 & & + 09466@xmath144955 & 47969 & g5iv & 6.8 & 6.9 & 1.4 & 1.2 & 8.73 & 0.83 & 0.99 & 0.10 & 1.23 & 0.17 + 09510@xmath148004 & 48320 & b8iv & 3.9 & 6.3 & 2.4 & 1.6 & 6.47 & 0.12 & 2.05 & 0.14 & 1.69 & 0.18 + 10112@xmath143245 & 49883 & g1v & 11.6 & 11.9 & 1.3 & 0.8 & 8.30 & 0.68 & 3.04 & 0.87 & 2.14 & 0.48 + 11102@xmath141122 & 54580 & g0 & 16.2 & 16.4 * & 1.1 & 1.1 & 7.70 & 0.72 & 0.30 & 0.14 & 0.22 & 0.19 + 11238@xmath143829 & 55628 & k1v & 20.7 & 20.3 * & 0.9 & 0.9 & 8.51 & 0.83 & 0.37 & 0.17 & 0.39 & 0.46 + 11425@xmath672355 & 57112 & g5 & 21.9 & 23.3 * & 0.9 & 0.6 & 9.02 & 0.83 & & & 1.51 & 0.07 + 11525@xmath141408 & 57894 & g0v & 15.6 & 17.5 * & 1.2 & 1.0 & 7.66 & 0.64 & 1.30 & 0.19 & 1.15 & 0.01 + 13072@xmath145420 & 64025 & g1/g2v & 1.4 & 4.8 & 1.5 & 1.2 & 9.08 & 0.73 & & & 0.90 & 0.06 + 13081@xmath147719 & 64091 & k3v & 23.3 & 26.1 & 0.9 & 0.6 & 8.76 & 1.01 & & & 1.79 & 0.07 + 13306@xmath144238 & 65906 & k4v & 30.6 & 31.2 & 0.7 & 0.6 & 9.56 & 1.47 & & & 0.97 & 0.11 + 13377@xmath142337 & & g5 & & 12.3 & 0.9 & 0.7 & 10.17 & 0.80 & 2.10 & & 0.25 & + 13598@xmath140333 & 68380 & f8v & 12.9 & 11.0 * & 1.8 & 1.5 & 6.36 & 0.56 & 1.13 & 0.12 & 0.93 & + 14020@xmath142108 & 68552 & k0/k1v & 20.0 & 22.1 * & 0.8 & 0.6 & 9.51 & 0.97 & 2.77 & 0.06 & 1.64 & 0.18 + 14383@xmath144954 & 71577 & a0iv & 4.2 & 4.3 * & 2.7 & 2.2 & 6.51 & 0.05 & 0.98 & 0.20 & & + 14453@xmath143609 & 72140 & a1iv / v & 4.8 & 5.1 & 2.1 & 1.6 & 7.37 & 0.12 & 1.44 & 0.23 & & + 14589@xmath670636 & 73314 & k0 ? & 23.2 & 25.5 * & 0.9 & 0.8 & 8.45 & 0.88 & 0.69 & 0.12 & 0.56 & 0.11 + 15035@xmath144035 & 73667 & f3v & 6.1 & 5.7 & 1.4 & 1.3 & 8.82 & 0.54 & 0.61 & & 0.44 & + 15042@xmath141530 & 73724 & f5/f6v & 8.4 & 7.9 & 1.5 & 1.2 & 8.03 & 0.60 & 1.15 & 0.51 & & + 15251@xmath142340 & 75478 & g2v & 15.3 & 15.5 & 1.0 & 1.0 & 8.48 & 0.65 & 0.52 & 0.36 & 0.28 & 0.14 + 15282@xmath140921 & 75718 & k2v & 48.6 & 51.4 * & 0.9 & 0.6 & 6.89 & 0.86 & 3.40 & 0.04 & 2.21 & 0.24 + 15290@xmath142852 & 75790 & g0v & 17.1 & 14.6 & 1.6 & 1.3 & 6.43 & 0.68 & 0.93 & & 0.90 & 0.02 + 15317@xmath670053 & 76031 & g0 & 19.7 & 23.8 * & 1.1 & 0.9 & 7.42 & 0.71 & 1.56 & 0.43 & 1.25 & 0.15 + 15332@xmath142429 & 76143 & a7+k ... & 10.3 & 9.5 * & 2.0 & 1.4 & 6.26 & 0.38 & 1.75 & 0.17 & 1.18 & 0.02 + 15348@xmath142808 & 76275 & g8v & 17.4 & 17.4 & 1.0 & 0.7 & 9.08 & 0.77 & 2.46 & 0.16 & 1.84 & 0.16 + 15548@xmath146554 & 77921 & g2v & 21.0 & 22.1 * & 1.1 & 1.0 & 7.55 & 0.69 & 0.52 & 0.61 & 0.35 & + 16544@xmath143806 & 82709 & g0v & 15.6 & 17.3 & 1.1 & 0.7 & 8.40 & 0.69 & 3.27 & 0.15 & 2.32 & + 17018@xmath145108 & 83321 & a7v & 9.6 & 5.8 & 2.3 & 2.1 & 6.43 & 0.32 & 0.47 & 0.35 & & + 17066@xmath670039 & 83716 & k0v & 17.3 & 18.4 * & 0.8 & 0.8 & 9.81 & 1.00 & 0.05 & 0.07 & 0.07 & 0.06 + 17221@xmath147007 & 84979 & b8/b9vn ... & 2.5 & 2.6 * & 4.2 & 4.1 & 5.39 & & 0.15 & 0.22 & & + 17575@xmath145740 & 87914 & k0v & 25.1 & 24.9 * & 0.8 & 0.7 & 9.31 & 1.00 & 0.42 & 0.22 & 0.28 & 0.14 + 17584@xmath670428 & 87991 & k8 & 22.9 & 26.3 * & 0.7 & 0.6 & 9.80 & 1.40 & 1.16 & & 0.48 & + 18092@xmath142211 & 88932 & k3/k4v & 30.5 & 28.2 * & 0.8 & 0.7 & 8.88 & 1.04 & 0.79 & 0.39 & 0.14 & 0.17 + 18150@xmath145018 & & a0v & & 6.2 & 1.9 & 1.7 & 7.22 & & 0.79 & 0.51 & 1.36 & + 18368@xmath142617 & 91253 & g3v & 21.5 & 18.7 * & 1.1 & 1.0 & 7.80 & 0.70 & 0.78 & 0.25 & 0.43 & 0.20 + 18480@xmath141009 & 92250 & k0 & 36.0 & 41.9 * & 0.8 & 0.5 & 8.45 & 0.93 & 3.65 & & 2.18 & 0.02 + 19581@xmath144808 & 98274 & g0v & 18.6 & 18.0 * & 1.1 & 0.8 & 8.35 & 0.68 & 2.28 & 0.62 & 1.68 & 0.18 + 20057@xmath143743 & 98979 & f3v & 6.7 & 9.6 & 1.4 & 1.1 & 7.94 & 0.49 & 1.20 & & 0.41 & 0.06 + 20104@xmath141923 & 99391 & g0v & 16.4 & 14.6 & 1.4 & 0.8 & 7.28 & 0.65 & 3.09 & 0.38 & 2.29 & 0.17 + 20202@xmath143435 & 100266 & f8v & 16.1 & 15.9 * & 1.5 & 1.2 & 6.62 & 0.66 & 1.24 & 0.09 & 1.10 & 0.02 + 20217@xmath143637 & 100417 & g0/g1v & 16.8 & 15.4 * & 1.2 & 1.1 & 7.49 & 0.66 & 0.49 & 0.15 & 0.20 & 0.05 + 20248@xmath141943 & 100685 & k3v & 18.3 & 17.0 & 0.9 & 0.8 & 9.40 & 0.98 & & & 0.68 & 0.11 + 21073@xmath145702 & 104256 & k1v & 18.6 & 20.8 * & 0.9 & 0.7 & 8.82 & 0.84 & 2.33 & 0.13 & 1.58 & 0.05 + 22116@xmath143428 & 109561 & k1/k2iii+ .. & 4.2 & 4.2 * & 2.7 & 1.9 & 6.68 & 0.64 & 1.67 & 0.63 & &

we discuss the properties of lorentzian wormholes in dilatonic einstein - gauss - bonnet theory in four spacetime dimensions . these wormholes do not need any form of exotic matter for their existence . a subset of these wormholes is shown to be linearly stable with respect to radial perturbations . we perform a comprehensive study of their domain of existence , and derive a generalised smarr relation for these wormholes . we also investigate their geodesics determining all possible particle trajectories , and perform a study of the acceleration and tidal forces that a traveler crossing the wormhole would feel .

1111.4049

black holes are by far the most celebrated class of solutions derived from einstein s field equations . being among the first types of solutions to be found almost a century ago , they have undergone an extensive investigation over the years . their existence conditions , forms of solutions and set of properties have been studied in the context of both traditional general relativity and generalized gravitational theories admitting either four or more dimensions . the most well - known example of such a generalized gravitational theory in four dimensions is provided by the low - energy heterotic string effective theory @xcite . in this theory , the scalar curvature term of einstein s theory is only one part of a more complex action functional where higher - curvature gravitational terms as well as kinetic and interaction terms of a variety of additional fields ( axions , fermions and gauge fields ) make their appearance . the dilatonic einstein - gauss - bonnet theory is a minimal version of the aforementioned theory and contains the scalar curvature term @xmath0 , a scalar field called the dilaton , and a quadratic curvature term , the gauss - bonnet ( gb ) term , given by @xmath1 . the gb term in four dimensions can be expressed as a total derivative term and , normally , makes no contribution to the field equations ; however , the exponential coupling to the dilaton field , that emerges in the context of the dilatonic einstein - gauss - bonnet theory , ensures that the gb term is kept in the theory . in the framework of the dilatonic einstein - gauss - bonnet theory , new types of black hole solutions emerged that are endowed with non - trivial dilaton hair ( for an indicative list of works on this topic , see @xcite ) . the presence of the gb term in the theory caused the circumvention of the traditional no - hair theorems as it bypassed the conditions for their validity . in reality , the existence of the dilatonic black holes violated only the ` letter ' of the no - hair theorems , and not the ` essence ' of them since the dilaton charge was of a ` secondary ' type . however , it became evident that in the context of this type of generalized gravitational theories solutions with a much richer structure and a modified set of properties , compared to the ones in general relativity , can emerge . another class of gravitational solutions whose properties are strictly set by the general theory of relativity are wormholes . they were first discovered in 1935 as a feature of schwarzschild geometry @xcite and named the ` einstein - rosen bridge ' as they connect two different universes . their importance was realised in full when wheeler @xcite showed that such a bridge , or wormhole , can connect also two distant regions of our own universe thus opening the way for fast interstellar travel . however , it was soon demonstrated that this is not possible for the following reasons : ( i ) the schwarzschild wormhole is hidden inside the event horizon of the corresponding black hole , therefore , is not static but evolves with time ; as a result , the circumference of its ` throat ' is not constant but opens and closes so quickly that not even a light signal can pass through @xcite , ( ii ) even if a traveler could somehow pass the throat , she would be bound to exit the wormhole through the past horizon of the schwarzschild geometry ; this horizon was shown to be unstable against small perturbations and that it would change to a proper , and thus non - traversable , horizon at the mere approaching of the traveler @xcite . in @xcite a new class of wormhole solutions was found that possess no horizon and could in principle be traversable . however , some form of exotic matter whose energy - momentum tensor had to violate all ( null , weak and strong ) energy conditions was necessary in order to keep the throat of the wormhole open . several studies have considered a phantom field , i.e. a scalar field with a reversed sign in front of its kinetic term , as a candidate for such a form of matter @xcite . it was demonstrated in @xcite that the gb term leads to an effective energy - momentum tensor that also violates the energy conditions . it is in fact this violation that causes the circumvention of the no - hair theorem forbidding the existence of regular black holes with non - trivial scalar hair . it was noted recently in @xcite , too , that the presence of the gb term in the context of a scalar - tensor theory has the property to evade the various no - go theorems of general relativity . indeed , various wormhole solutions were found in the context of gravitational theories with higher curvature terms @xcite . in the presence of the gauss - bonnet term in particular , wormhole solutions were found in the context of higher - dimensional gravitational theories @xcite . in this work , we will investigate the properties of wormhole solutions that arise in the context of the four - dimensional dilatonic einstein - gauss - bonnet theory , first reported in @xcite . the presence of the higher - curvature gb term , that follows naturally from the compactification of the 10-dimensional heterotic superstring theory down to four dimensions , suffices to support these types of solutions without the need for phantom scalar fields or other forms of exotic matter . the outline of our paper is as follows : in section ii , we present the theoretical context of our model and discuss the asymptotic forms of the sought - for wormhole solutions at the regions of radial infinity and the regular throat . based on the latter , we derive the embedding diagram and study the violation of the energy conditions . we also present a smarr relation for the wormhole solutions . in section iii , we present the results of our numerical analysis that reveal the existence of wormhole solutions in the dilatonic einstein - gauss - bonnet theory , and discuss their properties . we demonstrate the stability with respect to radial perturbations of a subset of these solutions in section iv . in section v we discuss the junction conditions . the geodesics in these wormhole spacetimes are presented in section vi . we calculate the magnitude of the acceleration and tidal forces that a traveler traversing the wormhole would feel in section vii , and conclude in section viii . we consider the following effective action @xcite motivated by the low - energy heterotic string theory @xcite @xmath2 , \label{act}\end{aligned}\ ] ] where @xmath3 is the dilaton field with coupling constant @xmath4 , @xmath5 is a positive numerical coefficient given in terms of the regge slope parameter , and @xmath6 is the gb correction . the dilaton and einstein equations are given by @xmath7 \nonumber\\ & & -\alpha e^{-\gamma \phi } \left [ h_{\mu\nu } + 4\left(\gamma^2\nabla^\rho \phi \nabla^\sigma \phi -\gamma \nabla^\rho\nabla^\sigma \phi\right ) p_{\mu\rho\nu\sigma } \right ] \label{einsteq}\end{aligned}\ ] ] with @xmath8 -\frac{1}{2}g_{\mu\nu}r^2_{\rm gb } \ , \\ p_{\mu\nu\rho\sigma } & = & r_{\mu\nu\rho\sigma } + 2 g_{\mu [ \sigma } r_{\rho ] \nu } + 2 g_{\nu [ \rho } r_{\sigma ] \mu } + r g_{\mu [ \rho } g_{\sigma ] \nu } \ . \label{hp}\end{aligned}\ ] ] throughout this paper we consider only static , spherically - symmetric solutions of the above set of equations . thus we can write the spacetime line - element in the form @xcite @xmath9 as was demonstrated in @xcite , the dilatonic - einstein - gauss - bonnet ( egbd ) theory admits black hole solutions whose gravitational background has the line - element of eq . ( [ metrics ] ) . it was also shown that further classes of solutions emerge in the context of the same theory . one of them , in particular , possesses no curvature singularity and no proper horizon ( with @xmath10 being regular over the whole radial regime ) . however , the @xmath11 metric component as well as the dilaton field showed some pathological behavior at a finite radius @xmath12 , as seen from the expansion near @xmath13 @xcite @xmath14 the absence of any singular behaviour of the curvature invariants at @xmath13 signifies that the aforementioned pathological behaviour is merely due to the particular choice of the coordinate system . in @xcite we have argued that this class of asymptotically flat solutions can be brought to a regular form by employing the coordinate transformation @xmath15 . then , the metric becomes @xmath16 the above form is regular and describes a wormhole solution , where @xmath13 is the radius of the throat . indeed , in terms of the new coordinate @xmath17 , the expansion at @xmath18 assumes the form @xmath19 where @xmath20 , @xmath21 and @xmath22 are constant coefficients , and shows no pathology . both metric functions and the dilaton field remain finite in this asymptotic regime . thus these solutions possess no horizon . in addition , all curvature invariant quantities including the gb term turn out to be finite at @xmath18 , a result that demonstrates the absence of any singularity . substitution of the metric eq . ( [ metricl ] ) into the dilaton equation ( [ dileq ] ) and einstein equations ( [ einsteq ] ) yields a coupled system of ordinary differential equations ( ode s ) for the metric functions and the dilaton field @xmath23\right\ } , \hspace*{0.3cm}\,\,\ , \label{ode1 } \\[1 mm ] & & \nonumber\\ \nu ' - \frac{r^2 f -l^2}{2 l r^2 } & = & \frac{\phi'^2 r^2}{8 l } + 2\alpha \gamma\frac{e^{-\gamma\phi}}{l r^2 f } \,\nu'\phi ' ( r^2 f- 3 l^2)\,,\label{ode2 } \\[1 mm ] & & \nonumber \\ \nu''+\nu'^2 + \frac{\nu ' ( 2 l f - r^2 f')}{2 r^2 f } & + & \frac { 2 r_0 ^ 2 f - l r^2 f'}{2 r^4 f } \label{ode3 } \\ & = & -\frac{\phi'^2}{4 } + 2\alpha \gamma\frac{e^{-\gamma\phi}}{r^2 f } \left\ { 2 l \left[\nu ' ( \gamma\phi'^2-\phi '' ) -\phi ' ( \nu'^2+\nu'')\right ] + \nu ' \phi ' \left(\frac{3lf'}{f}-\frac{2r_0 ^ 2}{r^2}\right)\right\ } , \nonumber\\[1 mm ] \phi '' + \nu'\phi ' + \frac{\phi ' ( 4 l f - r^2 f')}{2 r^2 f } & = & 4\alpha \gamma\frac{e^{-\gamma\phi}}{r^4 f } \left\ { -2 ( r^2 f - l^2)(\nu'^2+\nu '' ) + \nu'\left[\frac{f'}{f}\,(r^2 f-3 l^2 ) + \frac{4 l r_0 ^ 2}{r^2}\right]\right\}. \label{ode4}\end{aligned}\ ] ] here @xmath24 and the prime denotes the derivative with respect to @xmath17 . equations ( [ ode1 ] ) , ( [ ode2 ] ) and ( [ ode3 ] ) follow from the @xmath25 , @xmath26 and @xmath27 components of the einstein equations , respectively , whereas the last equation ( [ ode4 ] ) follows from the dilaton equation . for the numerical computation we ` diagonalize ' eqs . ( [ ode1 ] ) , ( [ ode2 ] ) and ( [ ode4 ] ) with respect to @xmath28 , @xmath29 and @xmath30 . the remaining equation , eq . ( [ ode3 ] ) , involving also second derivatives of @xmath31 and @xmath3 , is satisfied if the other three equations are fulfilled . thus a system of ode s must be solved that consists of two first - order equations for the metric functions @xmath32 and @xmath31 and a second - order equation for the dilaton field . the expansion near the throat , eqs . ( [ expansion - throat1])-([expansion - throat3 ] ) , once substituted into the set of equations leads to a number of recursive constraints that determine the higher - order coefficients in terms of the lower ones . looking at the lowest order , we observe that @xmath33 , @xmath34 and @xmath35 are free parameters . also , the set of parameters of the theory includes the radius of the throat @xmath13 and the value of @xmath36 . the value of the constant @xmath4 will be later fixed to 1 for simplicity . however , not all of the above parameters are actually independent . to start with , we observe that the field equations remain invariant under the simultaneous changes @xmath37 in addition , the following transformation @xmath38 is also a symmetry of the equations . in the light of the above , we conclude that only one parameter out of the set @xmath39 is independent . we may therefore choose to have a zero asymptotic value of the dilaton field at infinity which entails fixing the value of the dilaton field at the throat @xmath35 . the remaining two parameters can be combined to give a dimensionless parameter @xmath40 that will be used throughout our analysis . finally , among the two parameters associated with the metric functions , ( @xmath34 , @xmath33 ) , again only the latter is independent it is only the derivatives of the metric function @xmath31 that appear in the field equations of motion , therefore , we may use this freedom to fix the value of @xmath34 in order to ensure asymptotic flatness at radial infinity . thus , our class of wormholes is a two - parameter family of solutions that , in the context of our analysis , have been chosen to be @xmath33 and @xmath40 . when the expansion of the metric functions and dilaton field near the throat ( [ expansion - throat1])-([expansion - throat3 ] ) are substituted into the field equations , we obtain a set of constraints on the higher - order coefficients . the constraint on the value of the first derivative of the dilaton field at the throat is particularly interesting and takes the form @xmath41}\ . \label{constraint - phi1}\ ] ] as the left - hand - side of the above equation is positive - definite , the same must hold for the right - hand - side . we may easily see that the expression inside the square brackets in the denominator remains positive and has no roots if @xmath42 this inequality is automatically satisfied for the set of wormhole solutions presented in the next section , and gives a lower limit on the size of the throat of the wormhole @xmath13 . then , the positivity of the right - hand - side of eq . ( [ constraint - phi1 ] ) demands that @xmath43 . the solutions satisfying @xmath44 comprise a boundary in the phase space of the wormhole solutions and their physical significance will be discussed in the next section . as already discussed above , at the asymptotic regime of radial infinity , i.e. as @xmath45 , we demand asymptotic flatness for the two metric functions and a vanishing value of the dilaton field . then , the asymptotic expansion at infinity takes the form @xmath46 in the above , @xmath47 and @xmath48 are identified with the mass and dilaton charge of the wormhole , respectively . we remind the reader that in the case of the black hole solutions @xcite , the parameters @xmath47 and @xmath48 were related , and that rendered the dilatonic hair as `` secondary '' . however , in the case of the wormhole solutions these parameters are not related ; this result , together with the fact that the number of independent parameters near the throat is also two , confirms the classification of this group of solutions as a two - parameter class of solutions . a general property of a wormhole is the existence of a throat , i. e. a surface of minimal area ( or minimal radius for spherically symmetric spacetimes ) . indeed , this property is implied by the form of the line element ( [ metricl ] ) above , with @xmath49 and @xmath50 finite . to cast this condition in a coordinate independent way , we define the proper distance from the throat in the following way @xmath51 if we impose the condition for a minimal radius @xmath52 at @xmath18 , this translates to @xmath53 it is easily seen that the first condition is indeed satisfied . for the second condition we find @xmath54 this gives a coordinate independent meaning to the parameter @xmath33 , @xmath55^{-1 } \ .\ ] ] a particularly useful concept with which we may examine the geometry of a given manifold is the construction of the corresponding embedding diagram . in the present case , we consider the isometric embedding of a plane passing through the wormhole . due to the spherical symmetry of the solutions , we may simplify the analysis and choose @xmath56 . then , we set @xmath57 where @xmath58 are a set of cylindrical coordinates in the three - dimensional euclidean space @xmath59 . regarding @xmath60 and @xmath61 as functions of @xmath17 , we find @xmath62 from the last equation it follows that @xmath63 hence @xmath64 is a parametric representation of a slice of the embedded @xmath56-plane for a fixed value of @xmath65 . we observe that the curvature radius of the curve @xmath64 at @xmath18 is given by @xmath66 . thus @xmath33 has a geometric meaning : @xmath67 is the ratio of the curvature radius to the radius of the throat . as was already mentioned in the introduction , the existence of the wormhole solution relies on the violation of the null energy condition . the null energy condition holds if @xmath68 for any null vector field @xmath69 . in the particular case of spherically symmetric solutions , and if we employ the einstein equations , this condition can be expressed as @xmath70 if one or both of the above conditions do not hold in some region of spacetime , then the null energy condition is violated . by making use of the expansion of the fields near the throat ( [ expansion - throat1])-([expansion - throat3 ] ) , we find that , close to @xmath13 , @xmath71_{l=0 } = -\frac{2}{f_0 r_0 ^ 2 } < 0 \ , \label{eng_pert_cond}\ ] ] provided @xmath72 , i.e. in the absence of a horizon . thus for the wormhole solutions there is always a region close to the throat where the null energy condition is violated . on the other hand , from the asymptotic expansion of the solutions at infinity , we find @xmath73 we observe that , for solutions with a positive dilaton charge @xmath48 ( and a positive mass @xmath47 ) the null energy condition is satisfied in the asymptotic region . however , if the dilaton charge is negative , the null energy condition is violated also in that asymptotic region . note that @xmath74 for the black hole solutions , where it was given by a positive - definite combination of @xmath75 @xcite . however , @xmath74 does not necessarily hold for all of the wormhole solutions . to derive the smarr - like mass formula for the wormhole solutions we start with the definition of the komar mass @xmath76 where @xmath77 is the timelike killing vector field , @xmath78 is a spacelike hypersurface , @xmath79 is a normal vector on @xmath78 and @xmath80 is the natural volume element on @xmath78 . here @xmath81 denotes the contribution of the throat , @xmath82 where @xmath83 is the area of the throat , and @xmath84 is the surface gravity at the throat @xmath85 to obtain the mass formula we express @xmath86 in terms of the effective stress energy tensor @xmath87\ , , \label{r00_th}\ ] ] where we have added a zero in the form of the dilaton equation . multiplication by @xmath88 yields @xmath89 substitution of the metric yields a total derivative for the right - hand - side of the above equation . thus , after integration , we find @xmath90 now substitution into eq . ( [ m_adm ] ) gives the smarr - like formula @xmath91 with @xmath92 here @xmath93 is the induced spatial metric on the throat , @xmath94 is the scalar curvature of @xmath93 , and the integral is evaluated at @xmath18 . defining the normal vector @xmath95 on the surface @xmath18 the mass formula becomes @xmath96 with @xmath97 according to the above , the smarr - like mass formula for wormholes is obtained by replacing the horizon properties by the corresponding throat properties in the known egbd mass formula for black holes @xcite . in addition , an extra contribution appears that may be interpreted as a modified throat dilaton charge , while the gb modification is of the same type as the gb modification of the area ( or entropy in the case of black holes ) . for the numerical calculations we use the line - element ( [ metricl ] ) , since the functions are well behaved at @xmath13 . for the representation of the numerical results we employ the metric ( [ metrics ] ) as well . to solve the odes numerically , we introduce the compactified coordinate @xmath98 , thus mapping the semi - infinite range of @xmath17 to the finite range of @xmath99 , i.e. @xmath100 . we cover the parameter range @xmath101 and @xmath102 , keeping @xmath103 fixed . also we set @xmath104 . let us start the discussion of the solutions by recalling the boundary conditions for the system of equations , consisting of two first order and one second order odes . at the throat @xmath18 regularity requires @xmath105 } \right\}_{l=0 } = 0\ , . \label{bc_phi}\ ] ] this boundary condition has to be supplemented by @xmath106 or @xmath107 in order to obtain a specific solution . we note that the asymptotic condition @xmath108 for @xmath109 is always satisfied . this can be seen from the asymptotic form of eq . ( 14 ) , @xmath110 , which has the general solution @xmath111 . this leaves the asymptotic boundary conditions at radial infinity @xmath112 wormhole solutions can be found for every value of @xmath40 below @xmath113 . this upper bound on @xmath40 translates into a lower bound on the radius of the throat @xmath13 ( for a given @xmath36 ) . thus the radius of the throat @xmath13 can be arbitrarily large . in figs . 1a , b and c , we show the metric functions @xmath114 and @xmath115 and the dilaton function @xmath116 , respectively , for an indicative set of wormhole solutions . we also exhibit the scaled gb term @xmath117 in fig . close to the throat the functions show a distinct dependence on both parameters , @xmath33 and @xmath40 . however , for intermediate values of @xmath17 , the functions @xmath114 , and likewise the functions @xmath115 , corresponding to the same value of @xmath40 , approach each other and form clusters . for larger values of @xmath17 , solutions obtained for different values of @xmath40 also merge together . the same behaviour is observed for the gb term whereas for the dilaton function @xmath116 the same tendency also exists but becomes pronounced at slightly larger values of @xmath17 . [ f-1 ] + + + in fig . [ f-2 ] we visualize the geometry of the wormhole solutions . in particular , we present as a typical example in fig . [ f-2]a the isometric embedding of the solution with @xmath118 and @xmath119 @xcite . here also the curvature radius at the throat , @xmath66 , is shown . also for @xmath118 , the curves @xmath60 versus @xmath61 near the throat are shown in fig . [ f-2]b for several values of @xmath33 . [ f-2 ] + we may also define the profile function @xmath120 via the equation @xmath121 we note that , at the throat , the profile function @xmath122 goes to one and , thus , @xmath123 vanishes . for small values of @xmath33 , the value of the metric function @xmath31 at the throat keeps decreasing as @xmath124 . therefore , in the limit @xmath124 , the metric function @xmath10 tends to zero and a horizon emerges , thus recovering the class of dilatonic black hole solutions @xcite . let us now explore the domain of existence of these wormhole solutions . to that end we construct a set of families of wormhole solutions , where for each family the value of the parameter @xmath40 is fixed , while the second parameter @xmath33 varies within a maximal range of @xmath125 . the values of the parameter @xmath40 cover the range @xmath126 . for larger values of @xmath40 no wormhole solutions are found . in fig . [ f-3 ] we present the domain of existence of the wormhole solutions . [ f-3]a shows the scaled area of the throat @xmath127 versus the scaled dilaton charge @xmath128 . the domain of existence is mapped by the families of solutions obtained for a representative set of fixed values of @xmath40 . we observe that the domain of existence of the wormhole solutions is bounded by three curves indicated by asterisks , crosses and dots . the boundary indicated by asterisks coincides with the egbd black hole curve @xcite and corresponds to the limit @xmath124 . the limit @xmath129 on the other hand is indicated by crosses . at the third boundary , marked by dots , solutions are encountered that are characterized by a curvature singularity . [ f-3 ] + + before discussing the three limiting cases in more detail , let us consider further the boundary conditions at the throat . we illustrate the metric function @xmath33 at the throat for the same set of families of wormhole solutions in fig . [ f-3]b . for fixed values of @xmath40 below the limiting value for egbd black holes , @xmath130 @xcite , @xmath33 covers the full range @xmath125 . beyond this critical value , however , the minimal value of @xmath33 increases with increasing @xmath40 . we note that , for a certain intermediate range of @xmath40 , @xmath33 is not monotonic . in fig . [ f-3]c we show the dilaton field @xmath35 at the throat versus the scaled dilaton charge @xmath128 for the same set of solutions . clearly , the domain of @xmath35 is bounded . again we observe that in the limit @xmath124 the black hole values are obtained . inspection of the redshift function @xmath34 at the throat , exhibited in fig . [ f-3]d , reveals that @xmath131 tends to zero as @xmath124 . thus a horizon emerges in this limit . [ f-5 ] + to study the black hole limit in more detail let us consider a sequence of wormhole solutions approaching the black hole solution . we demonstrate this limiting behaviour for solutions with @xmath132 in fig . [ f-5 ] . here , we exhibit a sequence of solutions with values of @xmath33 tending to one . clearly , the profile function @xmath120 ( fig . [ f-5]a ) and the dilaton function @xmath133 ( fig . [ f-5]b ) tend fast to the limiting black hole solutions . for @xmath134 the wormhole functions and their black hole counterparts are already very close . the redshift functions @xmath31 are not distinguishable for this set of solutions except very close to the throat , where @xmath34 diverges in the limit . [ f-6 ] + next we consider the limit of large @xmath33 , indicated by crosses in figs . [ f-3 ] . as @xmath129 , the mass @xmath47 and the dilaton charge @xmath48 assume finite values . the same holds for the redshift function @xmath34 and the dilaton field @xmath35 at the throat , although the derivative of the dilaton field @xmath135 with respect to the coordinate @xmath17 diverges like @xmath136 , as seen from the boundary condition eq . ( [ bc_phi ] ) . as an example we exhibit the function @xmath114 in fig . [ f-6]a for two values of @xmath40 and increasing values of @xmath33 . we observe that for a given value of @xmath40 , the functions @xmath114 deviate from each other only for small values of @xmath17 close to the throat , but coincide for larger @xmath17 . moreover , the region where the functions coincide increases with increasing values of @xmath33 . thus the solutions approach a limiting solution for @xmath129 , which depends on @xmath40 . this limiting behaviour is even more pronounced in the scaled gb term @xmath137 , a curvature invariant which is exhibited in fig . [ f-6]b for the same set of parameters . extrapolating the value of @xmath137 at the throat for large @xmath33 indicates that @xmath137 remains finite for @xmath129 . [ f-7 ] finally we turn to the third boundary curve , indicated by dots in figs . this boundary emerges when branches of solutions with fixed @xmath40 terminate at singular configurations . we demonstrate that these singular configurations possess a curvature singularity at a critical value of @xmath17 , @xmath138 , in fig . [ f-7]a . here the scaled gb term @xmath137 is shown versus @xmath17 for @xmath139 for a sequence of solutions approaching the singular configuration . we observe a sharp peak in the vicinity of @xmath138 , which increases in size and diverges as the singular configuration is approached . to elucidate this emergence of a curvature singularity at some finite critical value of @xmath17 , we show in fig . [ f-7]b the metric function @xmath114 , in fig . [ f-7]c its derivative @xmath140 , and in fig . [ f-7]d the derivative of the dilaton function @xmath141 for a fixed value of @xmath139 and increasing values of @xmath33 approaching the singular configuration . we note that all functions are continuous , but as the singular configuration is approached , their derivatives develop a discontinuity at some point @xmath138 outside the throat . in fact the derivative of the function @xmath32 has a pole at @xmath138 . we now turn to the energy conditions , eqs . ( [ nulleng ] ) . we demonstrate the violation of the null energy condition in figs . here we show ` normalized ' quantities in order to emphasize the change of sign . the normalization factor is @xmath142 this normalization is responsible for the steep rise observed in the figures for @xmath119 , which occurs when all quantities are close to zero . the null energy condition is violated close to the throat for all wormhole solutions , as seen explicitly from eq . ( [ eng_pert_cond ] ) and demonstrated for the set of solutions of fig . however , for large values of @xmath33 and at the same time small values of @xmath40 the dilaton charge @xmath48 can become negative . in this case the null energy condition is violated also in the asymptotic region , as seen from eq . ( [ term2_as ] ) . in fact we observe that for the larger values of @xmath33 in fig . [ f-9]b the combination @xmath143 is negative everywhere . thus in this case the null energy condition is violated in the whole spacetime . [ f-9 ] + + [ smarrfig ] let us finally turn to the smarr - like mass relation , eq . ( [ smarr ] ) . this mass relation is a perfect check for the numerical accuracy and thus for the quality of the solutions . we demonstrate the smarr relation in figs . [ smarrfig ] by comparing the mass @xmath47 obtained from the asymptotic fall - off of the metric functions with the mass @xmath144 obtained by evaluating the various terms in the smarr relation eq . ( [ smarr ] ) . the smarr - like formula is well satisfied for all sets of solutions . in fig . [ smarrfig]a , we keep @xmath145 as in the main body of the paper and exhibit the masses @xmath47 and @xmath144 versus @xmath146 for @xmath139 . the relative error in this case is below @xmath147 ( for @xmath148 it is below @xmath149 , and for @xmath118 it is below @xmath150 ) . in fig . [ smarrfig]b we have fixed @xmath118 , @xmath151 and varied @xmath4 to address the @xmath4-dependence of the mass formula for a set of solutions . also for these solutions the relative error is small and , in particular , it is below @xmath152 . [ f - smarr ] + + to gain a better understanding of the quantities contributing in the smarr formula , we now consider the various throat properties . first , we exhibit in fig . [ f - smarr]a the scaled throat quantity @xmath153 versus the scaled dilaton charge @xmath128 for the full domain of existence of the wormhole solutions . @xmath153 resembles in its structure the entropy of black holes , since it contains the same correction term to the throat area @xmath83 as in the black hole case to the horizon area . in the black hole limit , @xmath154 assumes the meaning of the black hole entropy . the scaled entropy - analogue throat quantity @xmath153 is always greater than one . only in the schwarzschild black hole limit , where the dilaton vanishes , it assumes precisely the value one . a comparison with fig . [ f-3]a shows , that the quantity @xmath155 is always larger than the scaled throat area @xmath127 . this was observed before for black holes @xcite . the color coding of fig . [ f - smarr]a is related to the stability of the solution , discussed in the next section . the scaled surface gravity @xmath156 at the throat of the wormhole solutions is exhibited in fig . [ f - smarr]b . it is bounded , with its lower boundary given by the black hole values , which remain close to the schwarzschild value of 1/2 . the product of the quantities @xmath154 and @xmath84 represents the first term of the smarr formula , up to numerical factors . we exhibit this first term , divided by the mass , in fig . [ f - smarr]c . since the first term is always greater than one ( except for the schwarzschild case ) , the remaining pieces of the mass formula must contribute negatively to cancel this excess . the second term in the mass formula , @xmath157 , is negative , except for the small region with large @xmath33 and small @xmath40 . its contribution to the scaled mass formula can be read off the horizontal axis of fig . [ f - smarr]c . we exhibit the third term of the scaled mass formula in fig . [ f - smarr]d . it contains the scaled throat dilaton charge , @xmath158 , which is modified by the gb term , analogous to the modification of the area by the gb term . as expected , this contribution is negative for the wormhole solutions . it vanishes for the black holes , which represent the upper limit for @xmath158 . our starting point for the study of the stability is the line element for spherically symmetric solutions @xmath159 here we consider only the pulsation modes . thus we allow the metric and dilaton functions to depend on the radial coordinate @xmath17 and the time coordinate @xmath160 , @xmath161 for the study of the stability behaviour of our solutions , we also need the time - dependent einstein and dilaton equations . these were presented in @xcite in the context of the stability analysis of the dilatonic black hole solutions , and thus we refrain from repeating them here . ) was used for the derivation of the time - dependent field equations . the new set of equations with respect to the new coordinate @xmath17 may be derived under the redefinitions @xmath162 , @xmath163 $ ] and the change of variable @xmath164 . also , the following changes should be made due to the different conventions used : @xmath165 and @xmath166 . ] next , we decompose the metric and dilaton functions into the unperturbed functions and the perturbations @xmath167 where we assume a harmonic time dependence of the perturbations and @xmath168 is considered as small . now we substitute the perturbed functions into the einstein and dilaton equations and linearize in @xmath168 . this yields a system of linear odes for the functions @xmath169 , @xmath170 and @xmath171 , where the coefficients depend on the unperturbed functions and their derivatives . we use the unperturbed equations to eliminate @xmath29 , @xmath172 , @xmath28 and @xmath30 . the @xmath173 part of the einstein equations yields @xmath174 in terms of @xmath175 and @xmath176 . thus @xmath174 and @xmath177 can be eliminated from the rest of the odes . the dilaton equation and the @xmath27 part of the einstein equations can be diagonalized with respect to @xmath178 and @xmath179 . finally @xmath180 can be eliminated from the dilaton equation by adding the @xmath181 part of the einstein equations with a suitable factor . thus we end up with a single second order equation for @xmath182 , @xmath183 where the coefficients @xmath184 , @xmath185 and @xmath186 depend on the unperturbed solution . the coefficient @xmath186 tends to one asymptotically and is bounded at @xmath18 . the coefficients @xmath184 and @xmath185 tend to zero asymptotically , however , they diverge at @xmath18 as @xmath187 . thus , to obtain solutions which are regular at @xmath18 suitable boundary conditions are needed . since the perturbations are assumed to be normalizable , @xmath175 has to vanish asymptotically . in order to obtain a unique solution we also fix @xmath182 at @xmath18 . thus we are left with three boundary conditions for a second order ode , which can be satisfied only for certain values of the eigenvalue @xmath188 . we can avoid the singularity of @xmath184 and @xmath185 at @xmath18 , by employing the transformation @xmath189 , where @xmath190 satisfies @xmath191 . this yields @xmath192 where the coefficient @xmath193 is bounded at @xmath18 . we note that @xmath190 diverges like @xmath187 at @xmath18 . thus for acceptable solutions we have to impose the boundary condition @xmath194 at @xmath18 . in addition @xmath195 has to vanish asymptotically to ensure normalized solutions . however , these boundary conditions do not lead to unique solutions , since the ode ( [ lineq2 ] ) is homogeneous . therefore , we supplement eq . ( [ lineq2 ] ) with the auxiliary ode @xmath196 and impose the boundary conditions @xmath197 and @xmath198 asymptotically . these conditions give exactly the normalization @xmath199 . again , solutions exist only for certain values of the eigenvalue @xmath188 . the ode ( [ lineq2 ] ) can be written as a schrdinger equation by introducing the new coordinate @xmath200 via @xmath201 , and eliminating the first derivative @xmath202 as above . this yields @xmath203 where @xmath204 ^ 2 + \frac{q_0}{q_\sigma } \ , \ ] ] and @xmath205 is related to @xmath195 . the new potential @xmath206 is bounded for all @xmath200 . note that the derivation of eq . ( [ lineq3 ] ) holds only if the function @xmath186 is positive for all @xmath17 . however , if @xmath207 on some interval , one can not rely on the rule that the smallest eigenvalue corresponds to the eigenfunction without knots . indeed we found in some cases two different eigenfunctions @xmath208 , @xmath209 without knots which satisfy the orthogonality condition @xmath210 . in order to determine the change of stability we consider families of wormhole solutions with fixed @xmath40 the ode ( [ lineq1 ] ) is then solved together with the normalization constraint for varying values of @xmath33 . in fig . [ f - pert](a ) we show the eigenvalue @xmath188 versus @xmath33 for several values of @xmath40 . we observe that solutions for negative values of @xmath188 exist for large values of @xmath33 , corresponding to an instability of the wormhole solutions . for these solutions the function @xmath175 , respectively @xmath195 , decays exponentially . thus these solutions correspond to bound states of the equivalent schrdinger equation . as @xmath33 decreases , the eigenvalue increases and tends to zero for some critical value @xmath211 , which depends on @xmath40 . no solutions were found for @xmath212 . solutions of eq . ( [ lineq1 ] ) with positive @xmath188 would oscillate in the asymptotic region , and are not normalizable . [ f - pert ] + the interpretation is that the instability mode vanishes at some critical value @xmath213 . hence we conclude that the wormhole solutions are stable for @xmath212 . however , this holds only if @xmath214 for all @xmath17 . otherwise one can not be sure that @xmath188 is the smallest possible eigenvalue . therefore , eigenfunctions with negative eigenvalues can not be excluded even if @xmath215 . consequently , if @xmath186 is not positive for all @xmath17 the question of stability can not be decided by considering the standard equivalent schrdinger eigenvalue problem . numerically , we found that @xmath186 is positive for all wormhole solutions with @xmath216 ; but wormhole solutions where @xmath186 is negative on some interval exist if @xmath217 . in fig . [ f - pert](b ) we show the stability region in the domain of existence @xcite . also shown is the line where the eigenvalue @xmath188 changes sign . the instability region ( red or upper ) is characterized by negative @xmath188 , whereas the stability region ( lilac or lower ) is characterized by the absence of negative eigenvalues and positive @xmath186 . in the remaining ( white ) region the question of stability is not yet decided . up to this point , we have discussed the behaviour of the metric functions and of the dilaton field in only half of the wormhole spacetime , i.e. the part with @xmath218 . our solutions should naturally be extended to the second asymptotically flat part of the manifold ( @xmath219 ) . if this is performed by demanding that the derivatives of the metric and dilaton functions are continuous , we observe a singular behaviour corresponding to curvature singularities . this is demonstrated in fig . [ nonsymmext-1 ] for the wormhole solution with parameters @xmath118 , @xmath119 and @xmath104 . [ nonsymmext-1 ] + however , wormhole solutions without curvature singularities can be constructed when we extend the wormhole solutions to the second asymptotically flat part of the manifold in a symmetric way . in this case jumps appear in the derivatives of the metric and dilaton functions at @xmath18 . the jumps can be attributed to the presence of matter located at the throat of the wormhole . the corresponding junction conditions are of the form @xmath220 where @xmath221 is the stress - energy tensor of the matter at the throat , and @xmath222 the corresponding source term of the dilaton field . the lhs of the junction conditions can be derived in a standard way by integrating the einstein and dilaton equations across the boundary @xmath18 , i. e. @xmath223 . this yields @xmath224 where the subscript @xmath225 indicates evaluation at @xmath18 . next we assume that the matter at the throat takes the form of a perfect fluid with energy density @xmath226 and pressure @xmath227 and a dilaton charge @xmath228 . we also introduce the action @xmath229 at the throat , where @xmath230 denotes the ( 2 + 1)-dimensional induced metric on the throat , @xmath231 the corresponding ricci scalar , and @xmath232 , @xmath233 are constants . inserting the metric this brings the non - trivial junction conditions eqs . ( [ junc00])-([juncdil ] ) to the form @xmath234 using these equations @xmath226 , @xmath227 and @xmath228 can be expressed in terms of the metric and dilaton functions and the constants @xmath235 , @xmath232 . in figs . [ junction-1 ] we give an example for @xmath236 and @xmath145 . we note that the stable wormhole solutions possess positive energy density @xmath226 . [ junction-1 ] + + + + let us also consider the special case @xmath237 ( i.e. dust ) and choose the dilaton charge density @xmath228 at the throat to be twice the dilaton charge density of the wormhole in the ` bulk ' , @xmath238 . this yields @xmath239 interestingly , @xmath240 and @xmath226 are positive for all wormhole solutions , as shown in fig [ junction-3 ] . [ junction-3 ] + next we consider perturbations of the wormhole solutions in the form @xmath241 , etc . for symmetric wormholes the derivatives of the perturbations also have a jump at @xmath78 . introducing perturbations of the energy density @xmath242 , pressure @xmath243 and dilaton charge @xmath244 we find @xmath245 thus these equations give the perturbations of the energy density , the pressure and the dilaton charge on the throat . thus for any perturbation in the ` bulk ' , we can find the corresponding perturbation on the throat . consequently , the stability of the solutions is not affected by the matter on the throat . the study of the orbits of test particles and light is essential to fully understand the properties of a spacetime . the motion of test particles in a egbd wormhole spacetime with metric @xmath246 , eq . ( [ metricl ] ) , and dilaton field @xmath3 is governed by the lagrangian @xmath247 where the dot denotes the derivative with respect to the affine parameter @xmath248 , and @xmath249 is a constant ( @xmath250 for heterotic string theory ) . the conjugate momenta @xmath251 are found to be @xmath252 first integrals of motion are given by @xmath253 we refer to @xmath254 as the energy of the test particle and to @xmath255 as its angular momentum . the affine parameter can be chosen such that @xmath256 , with @xmath257 for time - like geodesics and @xmath258 for null geodesics . since we are considering spherically symmetric spacetimes we may choose @xmath259 , which implies @xmath260 . employing eqs . ( [ first_int ] ) the lagrangian then reduces to @xmath261 = \hat{\kappa } \ . \label{lageff}\ ] ] let us first consider time - like geodesics . for @xmath262 we obtain @xmath263 \ , \label{dotl2}\ ] ] where we introduced the effective potential @xmath264 the effective potential @xmath206 is a suitable quantity to discuss qualitatively the trajectories of test particles . we note that the effective potential approaches the value of one asymptotically . thus it follows from eq . ( [ dotl2 ] ) that unbound trajectories are only possible if @xmath265 . the turning points @xmath266 of trajectories for a given energy @xmath254 and angular momentum @xmath255 are determined by the condition @xmath267 we note that , in contrast to black hole spacetimes , the wormhole spacetimes do not possess an event horizon . consequently , particles can not disappear behind an event horizon . instead they can travel through the throat of the wormhole from one asymptotically flat part of the manifold to the other . [ ff-1 ] to discuss the various types of trajectories in these wormhole spacetimes we consider as an example the wormhole solution with parameters @xmath268 and @xmath119 . we exhibit the corresponding effective potential @xmath269 in fig . [ ff-1 ] for several values of the angular momentum @xmath255 . in particular , we first discuss the dependence of the shape of the effective potential on the angular momentum @xmath255 of the test particle , and next consider the different types of trajectories with their dependence on the energy @xmath254 of the test particle . in the case @xmath270 the effective potential is a monotonically increasing function of @xmath271 , assuming its minimum at the throat @xmath12 and tending to one asymptotically . when the angular momentum is increased , the effective potential remains monotonic until it develops a saddle point for some critical value @xmath272 . for @xmath273 the effective potential is no longer monotonic , since a local maximum @xmath274 and a local minimum @xmath275 occur . if @xmath255 is larger than a certain value , @xmath276 ( say ) , the maximum @xmath274 exceeds the asymptotic value of the effective potential , and thus it becomes the global maximum . we note that the essential features of the effective potential @xmath269 are the same for all the wormhole solutions considered . concerning the types of orbits of test particles we note that there are two kinds of trajectories . for trajectories of the first kind the particles remain on a single asymptotically flat manifold , whereas for trajectories of the second kind the particles travel from one asymptotically flat part of the manifold to the other , passing through the throat . trajectories of the first kind exist only if the effective potential is non - monotonic , i.e. for @xmath273 . bound orbits exist for @xmath277 , e.g. @xmath278 in fig . [ ff-1 ] , whereas unbound orbits exist for @xmath279 , e.g. @xmath280 in fig . for unbound orbits we need @xmath281 , thus @xmath274 must be the absolute maximum . trajectories of the second kind exist for all values of @xmath254 above the minimum of the effective potential @xmath282 . the particles move on bound orbits if @xmath283 , e.g. @xmath284 , @xmath278 , @xmath280 in fig . thus for @xmath285 the particles move on circles at the throat . radial trajectories are obtained for @xmath270 . in this case the particles oscillate about the throat . in the limit @xmath286 they possess the minimal possible energy and are at rest at the throat . on the other hand , if @xmath287 , e.g. @xmath288 in fig . [ ff-1 ] , the particles travel along unbound orbits , starting at infinity on one asymptotically flat part of the manifold , passing through the throat and reaching infinity on the other asymptotically flat part of the manifold . in fig . [ ff-2 ] we exhibit the minimal possible energy @xmath286 versus @xmath128 for several values of @xmath40 . note that @xmath289 approaches zero in the black hole limit . the black line on the left , highlighted in the inset , indicates the stability change of the wormhole solutions . [ ff-2 ] in order to calculate the trajectories we consider @xmath290 from which we find @xmath291 where @xmath17 is restricted to intervals where @xmath292 and @xmath293 is an integration constant . the trajectories are then displayed in the @xmath294-plane , where @xmath295 in fig . [ ff-3 ] we show examples of bound orbits of test particles with angular momentum @xmath296 and several values of the energy @xmath254 for the wormhole solution with @xmath148 and @xmath119 . for an energy of @xmath297 ( fig . [ ff-3]a ) there exists only a bound orbit of the second kind . this is highlighted in the figure by using different colors ( line styles ) for the two asymptotically flat manifolds , when projecting into the @xmath294-plane . in contrast , for @xmath298 ( fig . [ ff-3]b ) there exists in addition a bound orbit of the first kind . examples for unbound orbits are shown in fig . [ ff-4]a for energies @xmath299 and @xmath300 . [ ff-3 ] \a ) + b ) [ ff-4 ] + let us now consider the null geodesics in these wormhole spacetimes . setting @xmath258 in eq . ( [ lageff ] ) we find @xmath301 = e^2 \frac{e^{4\beta\phi}e^{-2\nu}}{f } \left[1 -\left(\frac{l}{e}\right)^2 \frac{e^{2\nu}}{r_0 ^ 2+l^2}\right]\ . \label{dotl2n}\ ] ] we define an effective potential @xmath302 by @xmath303 we note that for the discussion of the null geodesics it is more convenient to consider the `` normalized '' effective potential @xmath304 , since it is independent of @xmath305 . as an example , we show the `` normalized '' effective potential @xmath306 for the parameter values @xmath307 and @xmath119 in fig . the `` normalized '' effective potential possesses a local minimum @xmath308 at the throat , a maximum @xmath309 at some distance from the throat and tends to zero asymptotically . as for the timelike geodesics there are two kinds of trajectories : either the massless test particle remains on one of the asymptotically flat parts of the manifold , or it passes through the throat from one asymptotically flat part to the other . the first kind of ( unbound ) trajectories exists only if the ratio @xmath310 is smaller than the maximum of the `` normalized '' effective potential . the second kind of trajectories includes unbound geodesics if @xmath311 and bound geodesics if @xmath312 . circular orbits exist for @xmath313 ( and unstable ones for @xmath314 ) . following the formalism of @xcite , we will now calculate the magnitude of the acceleration and tidal forces that a traveler traversing the wormhole would feel . for this purpose , it is particularly convenient to make two changes of reference frames : first , starting from the standard reference frame with basis vectors @xmath315 , in which the line - element takes the form of eq . ( [ metricl ] ) , we change to an orthonormal reference frame with @xmath316 in terms of which the metric tensor assumes the form : @xmath317 . alternatively , we may write the set of equations in ( [ ortho1 ] ) as @xmath318 , with @xmath319 . \label{trans1}\ ] ] the above allows us to write the transformation law of the components of the riemann tensor as @xmath320 with @xmath321 . next , we introduce the orthonormal reference frame of the traveler which is related to the previous one by a lorentz transformation @xmath322 with @xmath323^{-1/2}$ ] and @xmath324 the radial velocity of the traveler at radius @xmath17 as measured by a static observer there . as before , we may write @xmath325 , with @xmath326 , \label{trans2}\ ] ] so that @xmath327 we note that @xmath328 , and thus @xmath329 can be naturally considered as the traveler s normalized vector of four - velocity @xmath330 . for the magnitude of @xmath330 to be fixed in the rest frame of the traveler , the four - acceleration @xmath331 should be orthogonal to the four - velocity , i.e. @xmath332 , and therefore @xmath333 if we further assume that the traveler moves radially , then @xmath334 . for a traveler moving along their world - line in which the tangent vector is @xmath335 , the acceleration is given by the formula @xmath336 or @xmath337 . in order to compute the magnitude of the acceleration that the traveler feels , we work in the following way : we calculate the @xmath338 component in the @xmath339 coordinate frame first according to the formula @xmath340 where we have assumed that the four - velocity is a function of the radial variable @xmath17 and rewritten its expression as @xmath341 however , it also holds that @xmath342 combining the above two results for @xmath338 , we find that the magnitude of the acceleration is @xmath343 for the wormhole to be traversable , the above quantity must remain always finite and , if possible , take a small value . in @xcite , it was also demanded that the gravitational acceleration should be small at the location of the stations where the trip starts and ends . demanding that the time needed to complete the whole trip should not be too large , the stations should be fairly close to the throat . both of these conditions are satisfied in our case , since an asymptotically flat regime is reached fairly quickly . the above also means that the traveler , starting from the stations , does not need to travel with a relativistic velocity in order to approach the wormhole in a reasonable time . then , if @xmath344 , @xmath345 , and the acceleration ( [ con - acc ] ) reduces to @xmath346 in fig . [ fig_acc ] , we depict the dimensionless acceleration @xmath347 at the throat as a function of @xmath128 . we observe that @xmath348 ranges roughly between 10 and 100 for a set of wormhole solutions with parameter @xmath40 ranging between 0.128 and 0.01 . in the context of superstring theory , @xmath349 , therefore , for the above solutions @xmath350 and the magnitude of the acceleration turns out to be @xmath351 , where @xmath352 is the acceleration of gravity at the surface of the earth . since there is no upper limit for @xmath13 in our solutions , one may be tempted to increase the size of the throat of the wormhole so that @xmath353 is of the order of @xmath352 ; this would demand a throat radius of the order of at least @xmath354 light - years . the aforementioned result could perhaps have been anticipated since our analysis is performed within the context of superstring theory , a theory whose fundamental scale is tied to the planck scale . [ fig_acc ] next we turn to the tidal acceleration that a traveler feels between two parts of her body as the wormhole is crossed . let @xmath355 be the vector separation between these two parts , which in the reference frame of the traveler is purely spatial , @xmath356 . then , the tidal acceleration is given by the expression @xcite @xmath357 where @xmath358 takes on only spatial values and where we have used that @xmath359 . but since the metric is diagonal and the riemann tensor is antisymmetric in the first two indices , the superscript @xmath360 should also take only spatial values . then , the non - vanishing components of the tidal acceleration are : @xmath361 the transformation laws ( [ trans1 ] ) and ( [ trans2 ] ) allow us to compute the components of the riemann tensor in the orthonormal frame of the moving observer in terms of the ones in the @xmath339 coordinates . we then obtain @xmath362 similarly , we find : @xmath363\,.\end{aligned}\ ] ] as in the case of the acceleration , we should demand that the magnitude of the above components should be small , i.e. of the order of the acceleration of gravity at the surface of the earth . using the fact that @xmath364 , the above constraints may be written as @xmath365 \frac{\gamma^2}{r_0 ^ 2+l^2 } \left|\frac{l\nu'}{f}+\frac{v^2}{c^2 f}\,\left(\frac{l f'}{2f}- \frac{r_0 ^ 2}{r_0 ^ 2+l^2}\right)\right| & \leq & \frac{g_\oplus}{c^2 ( 2\,{\rm m})}=\frac{1}{(10^{10 } { \rm cm})^2}\,.\end{aligned}\ ] ] the second inequality involves the velocity with which the traveler moves and thus may be considered as a constraint on this quantity . the first inequality restricts again the profile of the metric functions - a similar analysis to the one above leads to the same results regarding the magnitude of the tidal forces and the necessary size of the throat in order to bring these down to a reasonable value . the existence of traversable wormholes in the context of general relativity relies on the presence of some form of exotic matter . however , in the framework of a string - inspired generalized theory of gravity , the situation may be completely different . here we have investigated wormhole solutions in egbd theory , which corresponds to a simplified action that is motivated by the low - energy heterotic string theory . indeed , as we have demonstrated , egbd theory allows for stable , traversable wormhole solutions , without the need of introducing any form of exotic matter . the violation of the energy conditions , that is essential for the existence of the wormhole solutions , gravitational theories , where @xmath366 is the torsion , without the energy conditions being violated . ] is realized via the presence of an effective energy - momentum tensor generated by the quadratic - in - curvature gauss - bonnet term . we have determined the domain of existence of these wormhole solutions and shown that it is bounded by three sets of limiting solutions . the first boundary consists of the egbd black hole solutions of @xcite . the second boundary is approached asymptotically , when the curvature radius at the throat of the wormhole diverges ( the @xmath129 limit ) . finally , at the third boundary , solutions with a curvature singularity at a finite distance from the throat are encountered . we have investigated the properties of these egbd wormholes and derived a smarr - like mass relation for them . in this , the horizon properties in the black hole case are replaced by the corresponding throat properties of the wormholes , thus the area and surface gravity here refer to the ones at the throat . moreover , as is well - known for black holes in egb theories , their entropy does not correspond merely to their horizon area but it receives a gb correction term . similarly , we find that the mass formula for the wormhole solutions includes an analogous gb correction term ; in addition , another term , that vanishes in the black hole case , appears that represents the gb corrected dilaton charge at the throat . we have demonstrated that the smarr relation is satisfied very well by the numerical solutions . we have also investigated the stability of the solutions . we have shown that a subset of our wormhole solutions , the one that lies close to the border with the linearly stable dilatonic black holes in the domain of existence , is also linearly stable with respect to radial perturbations . while we have also shown that another subset is unstable , we have concluded that the study of the standard equivalent schrdinger equation can not determine the stability for the full domain of existence we hope to resume the question of the existence of an alternative method for the study of the stability of all of our wormhole solutions at some future work . when the wormhole solutions are extended to the second asymptotically flat region , this extension must be made in a symmetric way , since otherwise a singularity is encountered . as a consequence of this symmetric extension , the derivatives of the metric and dilaton functions become discontinuous at the throat . this discontinuity demands the introduction of some matter distribution at the throat . we have shown that this may be realized by the introduction of a perfect fluid at the throat whose energy density is positive for the subset of stable wormhole solutions . next , we have studied and classified the geodesics of massive and massless test particles in these wormhole spacetimes . depending on the respective effective potential , there are two general kinds of trajectories for these particles . the particles may remain on bound or escape orbits within a single asymptotically flat part of the spacetime , or they may travel from one asymptotically flat part to the other on escape orbits , and travel back and forth on bound orbits . in addition , we have calculated the acceleration and tidal forces which travelers traversing the wormhole would feel . we find that their magnitude may be small for fairly large values of the size of the throat of the wormhole . according to our findings , the radius of the wormhole throat is bounded from below only , therefore , the wormholes can be indeed arbitrarily large . astrophysical consequences will be addressed in a forthcoming paper as well as the existence of stationary rotating wormhole solutions in the egbd theory . we gratefully acknowledge discussions with eugen radu . b.k . acknowledges support by the dfg . p. kanti , n. e. mavromatos , j. rizos , k. tamvakis , e. winstanley , phys . d * 54 * ( 1996 ) 5049 [ arxiv : hep - th/9511071 ] . z. -k . guo , n. ohta and t. torii , prog . phys . * 120 * , 581 ( 2008 ) [ arxiv:0806.2481 [ gr - qc ] ] ; prog . phys . * 121 * , 253 ( 2009 ) [ arxiv:0811.3068 [ gr - qc ] ] . n. ohta and t. torii , prog . phys . * 121 * , 959 ( 2009 ) [ arxiv:0902.4072 [ hep - th ] ] ; prog . phys . * 122 * , 1477 ( 2009 ) [ arxiv:0908.3918 [ hep - th ] ] ; prog . theor . * 124 * , 207 ( 2010 ) [ arxiv:1004.2779 [ hep - th ] ] . k. -i . maeda , n. ohta and y. sasagawa , phys . d * 80 * , 104032 ( 2009 ) [ arxiv:0908.4151 [ hep - th ] ] ; phys . d * 83 * , 044051 ( 2011 ) [ arxiv:1012.0568 [ hep - th ] ] . b. kleihaus , j. kunz , e. radu , phys . lett . * 106 * , 151104 ( 2011 ) [ arxiv:1101.2868 [ gr - qc ] ] .

we investigate the linear growth and vertical structure of the magnetorotational instability ( mri ) in weakly ionised , stratified protoplanetary discs . the magnetic field is initially vertical and dust grains are assumed to be well mixed with the gas over the entire vertical dimension of the disc . for simplicity , all the grains are assumed to have the same radius ( , or m ) and constitute a constant fraction ( 1 % ) of the total mass of the gas . solutions are obtained at representative radial locations ( and au ) from the central protostar for a minimum - mass solar nebula model and different choices of the initial magnetic field strength , configuration of the diffusivity tensor and grain sizes . we find that when no grain are present , or they are m in radius , the midplane of the disc remains magnetically coupled for field strengths up to a few gauss at both radii . in contrast , when a population of small grains ( m ) is mixed with the gas , the section of the disc within two tidal scaleheights from the midplane is magnetically inactive and only magnetic fields weaker than mg can effectively couple to the fluid . at 5 au , ohmic diffusion dominates for when the field is relatively weak ( a few milligauss ) , irrespective of the properties of the grain population . conversely , at 10 au this diffusion term is unimportant in all the scenarios studied here . high above the midplane ( ) , ambipolar diffusion is severe and prevents the field from coupling to the gas for all . hall diffusion is dominant for a wide range of field strengths at both radii when dust grains are present . the growth rate , wavenumber and range of magnetic field strengths for which mri - unstable modes exist are all drastically diminished when dust grains are present , particularly when they are small ( m ) . in fact , mri perturbations grow at au ( 10 au ) for mg ( 130 mg ) when m grains are mixed with the gas . this upper limit on the field strength is reduced to only mg ( 10 mg ) when the grain size is reduced to m . in contrast , when the grains are assumed to have settled , mri unstable modes are found for mg at 5 au and 250 mg at 10 au ( salmeron & wardle 2005 ) . similarly , as the typical size of the dust grains diminishes , the vertical extent of the dead zone increases , as expected . for m grains , the disk is magnetically inactive within two scaleheights of the midplane at both radii , but perturbations grow over the entire section of the disk for grain sizes of m or larger . when dust grains are mixed with the gas , perturbations that incorporate hall diffusion grow faster , and are active over a more extended cross section of the disc , than those obtained under the ambipolar diffusion approximation . we conclude that in protoplanetary discs , the magnetic field is able to couple to the gas and shear over a wide range of fluid conditions even when small dust grains are well mixed with the gas . despite the low magnetic coupling , mri modes grow for an extended range of magnetic field strengths and hall diffusion largely determines the properties of the perturbations in the inner regions of the disc . [ firstpage ] accretion , accretion discs instabilities magnetohydrodynamics stars : formation .

0801.2428

magnetic fields may regulate the ` disc accretion phase ' of star formation by providing means of transporting away the excess angular momentum of the disc , enabling matter to accrete . the more generally relevant mechanisms associated with this are mhd turbulence induced by the magnetorotational instability ( mri ; balbus & hawley 1991 , 1998 ) and outflows driven centrifugally from the disc surfaces ( blandford & payne 1982 , wardle & knigl 1993 ; see also the review by knigl & pudritz 2000 ) . these processes are , in turn , thought to play key roles in the dynamics and evolution of astrophysical accretion discs . magnetically driven turbulence is likely to impact disc chemistry ( e.g. semenov , wiebe & henning 2006 , ilgner & nelson 2006 ) as well as the properties and evolution of dust grains mixed with the gas ( e.g. turner et al . magnetic resonances have been shown to modify the net tidal torque exerted by the disc on a forming planet and thus , alter the speed and direction of the planet s migration through the disc ( terquem 2003 ; johnson , goodman & menou 2006 ) . finally , the magnetically inactive ( dead ) zones ( gammie 1996 ; wardle 1997 ) are not only the regions where planets are thought to form , but they have also been invoked as a possible mechanism to stop their inward migration ( e.g. matsumura & pudritz 2005 ) . in protoplanetary discs , however , the magnetic diffusivity can be high enough to limit or even suppress these processes . the specific role magnetic fields are able to play in these environments is , therefore , largely determined by the degree of coupling between the field and the neutral gas . a critical parameter for this analysis is the ionisation fraction of the fluid , which reflects the equilibrium between ionisation and recombination processes taking place in the disc . ionisation processes outside the disc innermost sections ( @xmath20 au ) are non - thermal , driven by interstellar cosmic rays , x - rays emitted by the magnetically active protostar and radioactive decay ( hayashi 1981 ; glassgold , najita & igea 1997 ; igea & glassgold 1999 ; fromang , terquem & balbus 2002 ) . on the other hand , free electrons are lost through recombination processes which , in general , take place both in the gas phase and on grain surfaces ( e.g. nishi , nakano & umebayashi 1991 ) . dust grains affect the level of magnetic coupling in protoplanetary discs when they are well mixed with the gas ( e.g. in relatively early stages of accretion and/or when turbulence prevents them from settling towards the midplane ) . they do so in two ways . first , they reduce the ionisation fraction by providing additional pathways for electrons and ions to recombine on their surfaces . second , charged dust particles can become important species in high density regions ( umebayashi & nakano 1990 ; nishi , nakano & umebayashi 1991 ) . for example , at @xmath1 au in a disc where @xmath17 m grains are present , positively charged particles are the most abundant ionised species within two scaleheights from the midplane ( see fig . 6 of wardle 2007 ; hereafter w07 ) . as these particles generally have large cross sections , collisions with the neutrals are important and they become decoupled ( or partially decoupled ) to the magnetic field at densities for which smaller species , typically ions and electrons , would still be well tied to it . both of these mechanisms act to lower the conductivity of the fluid , especially near the midplane where the density is high and ionisation processes are inefficient . in the minimum - mass solar nebula disc ( hayashi 1981 , hayashi et el . 1985 ) , for example , x - rays are completely attenuated below @xmath21 ( see fig . @xmath1 of salmeron & wardle 2005 ; hereafter sw05 ) . as a result , in the disc inner sections the magnetic coupling may be insufficient to provide adequate link between the neutral particles and the field . moreover , recent calculations for a minimum - mass solar nebula disk exposed to x - ray and cosmic ray ionising fluxes ( w07 ) indicate that near the surface ( above 3 - 4 tidal scaleheights ) the magnetic diffusivity can also be severe , even though in these regions the ionising flux is strongest and the electron fraction is significantly larger than it is in the disc interior . this effect results from the strong decline in the number density of charged particles high above the midplane . these considerations suggest that magnetic activity in the inner regions of weakly ionised discs may well be confined to intermediate heights above the midplane ( @xmath22 ) . it is clear that the level of magnetic diffusion is strongly dependent on the presence , and size distribution , of dust particles suspended in the gas phase . in fact , once dust grains have settled , the ionisation fraction may be enough to produce adequate magnetic coupling over the entire vertical extension of the disk even at @xmath23 au ( w07 ) . a realistic study of the properties of the mri in these discs must , therefore , incorporate a consistent treatment of dust dynamics and evolution ( unless they are assumed to have settled , a good approximation to model relatively late accretion stages , as was the case in sw03 and sw05 ) . this analysis is further complicated because dust grains have complex spatial and size distributions ( e.g. mathis , rumpl & nordsieck 1977 , umebayashi & nakano 1990 , dalessio et al . 2006 ) , determined by the competing action of processes involving sticking , shattering , coagulation , growth ( and/or sublimation ) of ice mantles and settling to the midplane ( e.g. weidenschilling & cuzzi 1993 ) . previous results ( sw03 , sw05 ) highlight also the importance of incorporating in these studies all three diffusion mechanisms between the magnetic field and the neutral fluid ( namely , the ohmic , hall and ambipolar diffusivities ) , as hall diffusion largely determines the growth and structure of mri perturbations , particularly in the disc inner regions ( e.g. within distances of the order of a few au from the central protostar ; sw05 ) . this is implemented here via a diffusivity tensor formalism ( cowling 1957 , norman & heyvaertz 1985 , nakano & umebayashi 1986 , wardle & ng 1999 , w07 ) . dust grains can affect the structure and dynamics of accretion discs via two additional mechanisms : dust opacity can modify the radiative transfer within the disc which , in turn , can dramatically alter its structure and dust particles may become dynamically important if their abundance is sufficiently high . in this study both effects are small because the disc is vertically isothermal and grains constitute only a small fraction of the mass of the gas ( see below ) . in this paper we study the vertical structure and linear growth of the mri in a disc where dust grains are well mixed with the gas over its entire vertical dimension . results are presented for two representative distances ( @xmath3 and @xmath24 au ) from the central protostar . for simplicity , we assume that all particles have the same radius @xmath0 , @xmath1 or @xmath15 m and constitute @xmath1% of the total mass of the gas , a typical assumption in studies of molecular clouds ( umebayashi & nakano 1990 ) . this fraction is constant with height , which means that we have also assumed that no sedimentation has occurred . although this is a very simplified picture , the results illustrate the importance of dust particles in the delicate ionisation equilibrium of discs , and consequently , on their magnetic activity . the paper is organised as follows . the adopted disc model is described in section [ sec4:cond ] . section [ sec4:formulation ] briefly summarises the formulation and methodology , which are based on sw05 . we refer the reader to that study and references therein for further details . this section also includes a discussion of the typical dependence of the components of the diffusivity tensor and magnetic coupling with height with ( and without ) grains for the two radial positions of interest . section [ sec4:mri ] then presents the vertical structure and linear growth of unstable mri modes at these radii and compares solutions incorporating different diffusion mechanisms and assumptions regarding the presence , and size , of dust grains . these results , and possible implications for the dynamics and evolution of low conductivity discs , are discussed in section [ sec4:discussion ] . our main conclusions are summarised in section [ sec : summary ] . our fiducial disc , assumed to be geometrically thin and vertically isothermal , is based on the minimum - mass solar nebula model ( hayashi 1981 , hayashi et al 1985 ) . our formulation incorporates the disc vertical stratification but following common practice it neglects radial gradients . this is appropriate , as these gradients typically occur over a much larger length scale than those in the vertical direction . under these assumptions , the equilibrium structure of the disc is the result of the balance between the vertical component of the gravitational force exerted by the central object and the pressure gradient within the disc . the vertical profile of the density is then given by @xmath25 \ , , \label{eq : rhoinitial}\ ] ] where @xmath26 is the midplane density and @xmath27 is the tidal scaleheight of the gas . the neutral gas is assumed to be composed of molecular hydrogen and helium , such that @xmath28 , which results in @xmath29 . as already mentioned above , a key feature of protoplanetary accretion discs is that they are weakly ionised . this is because ionisation processes are generally ineffective ( except possibly in the vicinity of the star and in the surface regions ) , while the recombination rate is accelerated by the high density of the fluid and the removal of charges by dust grains ( if present ) . in fact , outside the innermost @xmath30 - @xmath31 au , where thermal effects are relevant ( e.g. hayashi 1981 ) , the main ionising sources are x - rays and uv radiation emanating from the central object ( e.g. glassgold , feigelson & montmerle 2000 ) and to a much lesser extent the decay of radioactive materials , particularly @xmath32 ( consolmagno & jokipii 1978 , sano et al . interstellar cosmic rays may also be important as they can potentially reach deeper into the disc than x - ray and uv fluxes do . in fact , cosmic rays are the dominant ionising source at 1 au for @xmath33 ( they even reach the midplane at this radius , albeit significantly attenuated ; e.g. sw05 ) . however , their actual contribution is unclear because the low - energy particles responsible for ionisation may be scattered by outflows launched from the protostar - disc system ( e.g. fromang et al . 2002 ) . on the other hand , recombination processes in the disc generally occur both in the gas phase ( through the dissociative recombination of electrons with molecular ions and the radiative recombination with metal ions ) and on grain surfaces ( e.g. oppenheimer & dalgarno 1974 , spitzer 1978 , umebayashi & nakano 1980 , nishi et al . 1991 , sano et al . 2000 ) . for a typical abundance of metal atoms in the gas phase of @xmath34 is the fraction of heavy metal atoms in the gas phase , estimated from interstellar absorption lines in diffuse clouds ( morton 1974 ) . ] ( umebayashi & nakano 1990 ) , the radiative recombination rate of metal ions is dominant for all vertical locations of interest , with the exception of the uppermost sections of the disc ( see fig . 2 of sw05 ) . the resulting ionisation fraction of the fluid largely determines the ability of the magnetic field to couple to the gas and shear and thus , regulates the magnetic activity in these astrophysical systems . in protoplanetary discs , in particular , the field has been envisioned to be dynamically important near the surface , whereas magnetic activity may be suppressed in their inner sections ( the ` layered accretion ' scenario ; gammie 1996 , wardle 1997 ) . however , the existence and configuration of a magnetically inactive dead zone in the disk interior has been shown to be critically dependent on the presence and properties of dust grains mixed with the gas ( w07 ) . as this study shows , in a minimum - mass solar nebula model at 1 au the entire cross section of the disc is magnetically coupled when dust grains are assumed to have settled to the midplane . it is , in particular , the presence of small grains what most severely affects the magnetic coupling . for example , the presence of a standard interstellar population of @xmath17 m grains at 1 au reduces the total active layer of the disc from @xmath35 to @xmath36 . this column density increases to @xmath37 once the grains aggregate to @xmath15 m . at 5 au , in contrast , the entire cross section of the disc is coupled once the grains have grown to @xmath19 m ( we refer the reader to w07 for further details of these models ) . the solutions presented in this paper are based on the formulation detailed in sw03 and sw05 . only a brief summary is given here . we write the equations of non - ideal mhd about a local keplerian frame that corotates with the disc at the keplerian frequency @xmath38 and express the velocity field as a departure from exact keplerian motion . we further assume that the abundances of charged species are sufficiently low to be able to neglect their inertia , thermal pressure and the effect of ionisation and recombination processes on the neutrals . under these conditions , only the equations of motion for the neutral gas are required : @xmath39 @xmath40 @xmath41 in the equation of motion ( [ eq : momentum ] ) , @xmath42 is the gravitational potential for a non self - gravitating disc and @xmath43 is the centripetal term generated by keplerian motion . coriolis terms @xmath44 and @xmath45 are associated with the use of a local keplerian frame and @xmath46 is the isothermal sound speed . in the induction equation ( [ eq : induction ] ) , @xmath47 is the electric field in the frame comoving with the neutrals and @xmath48 accounts for the generation of toroidal field by the disc differential rotation . finally , the magnetic field must also satisfy the constraint @xmath49 and the current density must satisfy ampere s law , @xmath50 and ohm s law @xmath51 following wardle & ng ( 1999 ) and wardle ( 1999 ; hereafter w99 ) , the current density is expressed as , @xmath52 where subscripts @xmath53 and @xmath54 denote vector components parallel and perpendicular to @xmath55 . in this expression , @xmath56 , @xmath57 and @xmath58 are the ohmic , hall and pedersen conductivity terms given by , @xmath59 @xmath60 and @xmath61 subscript @xmath62 is used here to label the charged species . they are characterised by their number density @xmath63 , particle mass @xmath64 , charge @xmath65 and hall parameter @xmath66 ( the ratio of the gyrofrequency and the collision frequency with the neutrals ) , which measures the relative importance of the lorentz and drag forces in balancing the electric force on the particle . equation ( [ eq : j - e ] ) can be inverted to find an expression for @xmath47 . this leads to the following form of the induction equation ( w07 ) @xmath67 \ , , \label{eq1:induction_drift}\end{aligned}\ ] ] where @xmath68 @xmath69 and @xmath70 are the ohmic , hall and ambipolar diffusivities ; and @xmath71 is the total conductivity perpendicular to the magnetic field . when ions and electrons are the only charged species , it can be shown that ( w07 ) @xmath72 and @xmath73 note that the ohmic ( @xmath74 ) and ambipolar ( @xmath75 ) diffusivity terms are always positive , as the former does not depend on the magnetic field strength and the second scale quadratically with it . as a result , they are both invariant under a reversal of the magnetic field polarity . on the contrary , the hall term ( @xmath76 ) scales linearly with @xmath11 and thus , can become negative . the change in sign of @xmath76 corresponds , in turn , to a change in the direction of the magnetic field at the height where particular species become decoupled to it by collisions with the neutrals . it corresponds , therefore , to changes in the contribution of different charged species to this component of the diffusivity tensor . the relative importance of the diffusion terms in ( [ eq : etao ] ) to ( [ eq : etaa ] ) differentiate three _ diffusivity regimes _ : 1 . in the _ ambipolar diffusion _ regime , @xmath77 for most charged species and @xmath78 . in this limit , which is typically dominant in low density regions ( e.g. in molecular clouds and near the surface of protoplanetary discs ) , the magnetic field is effectively frozen into the ionized component of the fluid and drifts with it through the neutrals . 2 . _ ohmic ( resistive ) _ limit . in this case @xmath79 for most charged species , resulting in @xmath80 . the magnetic field can not be regarded as being frozen into any fluid component and the diffusivity is a scalar , the well - known ohmic diffusivity . this regime dominates close to the midplane in the inner regions ( @xmath81 au ) of protoplanetary discs when the magnetic field is relatively weak ( w07 ) . _ hall diffusion _ limit , which occurs when @xmath77 for charged species of one sign ( typically electrons ) and @xmath82 for those of the other sign ( e.g. ions ) . in this case , @xmath83 and @xmath74 . this regime is important at intermediate densities ( between those at which ambipolar and ohmic diffusion regimes dominate ) . it has been shown to prevail under fluid conditions satisfied over vast regions in protoplanetary discs ( e.g. sano & stone 2002a , w07 ) . the diffusivities used in this study were calculated using the procedure described in w07 , to which we refer the reader for details . essentially , the adopted chemical reaction scheme is based on that of nishi , nakano & umebayashi ( 1991 ) , but it allows for higher charge states on dust grains . this is necessary because of the higher temperature and density of protoplanetary discs in relation to those typically associated with molecular clouds . finally , in the following sections we will also use the magnetic coupling parameter ( w99 ) @xmath84 the ratio of the critical frequency ( @xmath85 ) _ above _ which flux - freezing conditions break down and the dynamical ( keplerian ) frequency of the disc . for the sake of clarity we now sketch the arguments leading to this expression ( see also wardle & ng 1999 ) . first , recall that non - ideal mhd effects become important when the inductive and diffusive terms in the induction equation ( [ eq : induction ] ) are comparable ; or @xmath86 where ( see equation [ eq1:ohm ] ) @xmath87 in expression ( [ eq : ohm10 ] ) , @xmath88 is taken to be a characteristic measure of the conductivity of the gas . next , we adopt the following typical values for the various terms above @xmath89 where @xmath90 is the wavenumber of the mri perturbations in flux - freezing conditions and @xmath91 is the local alfvn speed . substituting these relations in ( [ eq : ind1 ] ) yields the desired expression for @xmath85 . this parameter is useful because in ideal - mhd conditions , the instability grows at @xmath92 ( balbus & hawley 1991 ) . consequently , if @xmath93 ( or @xmath94 ) the field is poorly coupled to the disc at the frequencies of interest for the analysis of the mri and non - ideal mhd effects are expected to be important . equations ( [ eq : continuity ] ) to ( [ eq1:j_curlb ] ) and ( [ eq : j - e ] ) are linearized about an initial ( labeled by a subscript ` 0 ' ) , steady state where the magnetic field is vertical and the current density , fluid velocity and electric field in the frame comoving with the neutrals all vanish . as a result of the last condition , the changes in the components of the diffusivity tensor are not relevant in this linear formulation ( e.g. @xmath95 ) and only the initial , unperturbed values are required . taking perturbations of the form @xmath96 and assuming @xmath97 we obtain a system of ordinary differential equations ( ode ) in @xmath98 ( the perturbations of the electric field in the laboratory frame ) , @xmath99 , and the perturbations growth rate @xmath100 . three parameters are found to control the dynamics and evolution of the fluid : ( @xmath1 ) @xmath101 , the local ratio of the alfvn speed and the isothermal sound speed of the gas , which measures the strength of the magnetic field . ( @xmath102 ) @xmath103 , the coupling between the ionised and neutral components of the fluid . ( @xmath104 ) @xmath105 , the ratio of the diffusivity terms perpendicular to @xmath55 , which characterises the diffusivity regime of the fluid . these parameters are evaluated at different locations ( @xmath106 , @xmath107 ) of the disc taking the magnetic field strength ( @xmath108 mg ) as a free parameter . the system of equations is integrated vertically as a two - point boundary value problem for coupled ode with boundary conditions @xmath109 and @xmath110 at @xmath111 and @xmath109 at @xmath112 . given that magnetic diffusion can have such a dramatic effect on the properties of magnetically - driven turbulence in protoplanetary discs , we now present calculations of the magnetic diffusivity at 5 and 10 au . we explore which disc regions are expected to be magnetically coupled and which diffusion mechanism is dominant at different positions ( @xmath106 , @xmath107 ) as a function of @xmath11 . this discussion is relevant for the analysis of our mri results at these locations ( section [ sec4:mri ] ) . fig . [ fig:10au_eta_1 ] shows the components of the diffusivity tensor ( @xmath74 , @xmath113 and @xmath75 ) as a function of height for @xmath114 au and a representative field strength ( @xmath115 mg ) . the solutions in the top panel have been obtained assuming that dust grains have settled into a thin layer about the midplane , so the charges are carried by ions and electrons only . the results in the middle and bottom panels incorporate a population of 1 and 0.1 @xmath116m - sized grains , respectively . note that when dust grains are present , all diffusivity terms increase drastically in relation to their values in the no - grains scenario . this effect becomes more accentuated as the grain size diminishes , given the efficiency of small grains in removing free electrons from the gas . for example , in the case that incorporates @xmath17m - sized particles ( bottom panel ) , the diffusivity components at the midplane are larger , by 4 7 orders of magnitude , than their corresponding values when the grains have settled . when dust grains have either settled or aggregated to at least 1 @xmath116 m in size ( see top and middle panels ) , @xmath113 is dominant for @xmath117 and the fluid in this region is in the hall diffusivity regime . also , @xmath118 there which implies that , if ions and electrons are the sole charge carriers , @xmath119 ( see equation [ eq : etaa1 ] ) . for higher @xmath107 , @xmath120 and ambipolar diffusion dominates . note that the hall diffusivity term increases less sharply than the ambipolar diffusivity in response to the fall in fluid density . this is a general feature , and an expected one , given that the former scales with @xmath121 and the latter with @xmath122 . as a result , @xmath75 is typically several orders of magnitude larger than @xmath113 near the surface regions of the disc . finally , for the @xmath17m - sized grain population ( bottom panel ) , @xmath123 for all @xmath107 . this can be traced back to the fact that hall diffusion is suppressed here because of the nearly equal abundances of ( negatively charged ) grains and ions in this region . note also that the hall diffusivity component shows characteristic ` spikes ' , at the heights where it changes sign . this effect is also particularly evident for the @xmath17m - sized grains . in this scenario , in particular , @xmath76 is positive when @xmath124 and @xmath125 . it is negative at all other vertical locations . as mentioned above , this corresponds to different charged species contributing to this diffusion term . in order to explore the contributions to @xmath76 in each of the vertical sections in which it has a different sign , we now describe the behaviour of the charged species at four representative heights , namely @xmath126 and @xmath104 ( for which @xmath127 ) , and @xmath128 and @xmath13 ( where @xmath129 ) . at @xmath126 , the density is sufficiently high for electrons to be able to stick to the grains . as a result , they reside mainly on dust particles , as do about a third of the ions . the contribution of ( negatively charged ) grains and ions to the hall diffusivity term are very similar , with a small positive excess , which determines the sign of @xmath76 in this region . on the contrary , at @xmath128 , ions are the dominant positively charged species , while the negative charges are still contained in dust grains and drift together with the neutrals . consequently , @xmath76 is negative in this section of the disc . at @xmath130 , ions and electrons are the main charge carriers and ions dominate the contribution to the hall term , which makes @xmath76 positive . finally , for @xmath131 , the dominant contribution to the hall diffusivity term comes from the small percentage of remaining negatively charged grains . as a result , @xmath76 is negative ( and very small ) . we now turn our attention to @xmath3 au ( fig . [ fig:5au_eta_1 ] ) . at this radius , when the dust particles are at least @xmath132 m in size ( or are absent ) , hall diffusion dominates for @xmath133 . in contrast with the solutions at 10 au , however , ohmic diffusion dominates over ambipolar diffusion ( @xmath134 ) for most of this region ( which implies that @xmath135 in this section of the disc when no grains are present ) . this is expected , given the larger fluid density at this radius in comparison with the 10 au case discussed above . when the grains are small ( @xmath136 m , see bottom panel ) , ambipolar diffusion is dominant ( @xmath120 ) for @xmath137 and @xmath138 . at all other heights , @xmath139 . [ fig : contour_10au ] and [ fig : contour_5au ] generalise the analysis of the previous paragraphs to other field strengths for @xmath114 and @xmath13 au , respectively . as before , the top panels refer to discs where no grains are present . the middle and bottom panels incorporate 1 and 0.1 @xmath116m - sized grains , respectively . in these plots , the contours show the values of @xmath140 and the background shading ( from dark to light ) denotes the dominant diffusion mechanism as ohmic , hall or ambipolar . the solid line is the critical value of the diffusivity @xmath141 ( w07 ) , above which the diffusion term in the induction equation @xmath142 is larger than the inductive term @xmath143 ( see equation [ eq : induction ] ) , a situation that effectively limits the ability of the field to couple to the keplerian shear . the dashed lines correspond to increases ( decreases ) in @xmath144 by factors of 10 in the direction of a stronger ( weaker ) magnetic field . note that at 10 au , in the no - grains case ( top panel of fig . [ fig : contour_10au ] ) , field strengths @xmath145 a few gauss are able to couple to the gas at the midplane . hall diffusion dominates at this location for @xmath146 g but the range of field strengths over which this occurs decreases gradually with height , so that when @xmath147 ambipolar diffusion is dominant for all @xmath11 . the coupled region potentially extends to @xmath148 , depending on the field strength . when the particles have aggregated to 1 @xmath116 m in size ( middle panel of fig . [ fig : contour_10au ] ) , the situation is qualitatively similar to the one just discussed , the only significant difference being that in this case hall diffusion is dominant at the midplane for all @xmath11 that can couple to the gas . the previous results , however , are significantly modified when small dust grains are present ( @xmath136 m , bottom panel ) . in this case , the section of the disc below two scaleheights is magnetically inactive and the magnetic diffusivity is severe enough to prevent coupling over the entire disc thickness when @xmath149 mg . ambipolar diffusion dominates for all field strengths above @xmath150 four scaleheights and also for @xmath151 mg @xmath152 mg below this height ( the actual upper limit of this range varies with @xmath107 ) . hall diffusion is inhibited close to the midplane for this range of field strengths because the number density of positively and negatively charged grains are very similar . note , finally , that ohmic diffusion is not dominant in any of the depicted scenarios at this radius , as expected , given the relatively low density of the fluid . the results at 5 au are qualitatively similar ( fig . [ fig : contour_5au ] ) to the ones just discussed . one key difference , however , is that ohmic diffusion dominates in all cases for @xmath153 provided that the field is sufficiently weak ( @xmath154 a few milligauss ) . ambipolar diffusion , on the other hand , is dominant for relatively strong fields ( e.g. @xmath155 g at @xmath156 in the no - grains case ) as well as for all @xmath11 near the disk surface ( @xmath157 ) . hall diffusivity is the most important diffusion mechanism for fluid conditions in - between those specified above . note that the coupled region extends up to @xmath158 in all cases . when no grains are present ( top panel ) the midplane is coupled for fields up to a few gauss . this upper limit drops to @xmath159 mg when @xmath19 m grains are mixed with the gas . when the grains are @xmath17 m in size , however , only the region above @xmath160 is coupled , and only for @xmath161 mg . this ` dead ' zone is slightly more extended at this radius than at 10 au , but stronger fields are able to couple to the gas in this case . the steep contours close to the midplane in this scenario result from the pronounced increase in the diffusivity in response to the increase in fluid density and the decline in the ionisation fraction of the gas as @xmath107 diminishes . , @xmath162 and @xmath163 ) as a function of height for @xmath114 au and @xmath115 mg . in the top panel , dust grains have settled to the midplane . the middle and bottom panels show the solutions when dust grains of radius @xmath164 and 0.1 @xmath116 m , respectively , are well mixed with the gas over the entire disc thickness . when the grains are @xmath165 m in size ( or are absent ) , hall diffusion dominates for @xmath166 . in contrast , for the 0.1 @xmath116m - sized grains , @xmath123 for all @xmath107 and @xmath167 above @xmath168 . note that @xmath113 shows ` spikes ' at the @xmath107 where it changes sign in response to different charged species becoming decoupled to the magnetic field by collisions with the neutrals . , scaledwidth=45.0% ] for r = 5 au . note that when the grains are @xmath169 m in size or are absent ( top and middle panels ) , hall diffusion dominates for @xmath133 . at this radius , however , @xmath134 for @xmath170.,scaledwidth=45.0% ] in a log(b ) - @xmath171 plane for @xmath114 au . in the top panel , dust particles are assumed to have settled to the midplane . the effect of dust grains is included in the middle ( @xmath172 m ) and bottom ( @xmath136 m ) panels . the solid line is the critical value of the diffusivity @xmath141 ( w07 ) above which the magnetic field does not couple to the gas and shear . the background shading ( from dark to light ) denotes the dominant diffusion mechanism as ohmic , hall or ambipolar . note that when the grain size is 1 @xmath116 m or larger ( or they have settled ) , the disc midplane is magnetically coupled for a range of @xmath11 . in contrast , when the grains are small ( bottom panel ) , magnetic diffusion prevents the field to couple to the gas for @xmath170 . , scaledwidth=40.0% ] for @xmath3 au . note that in all cases , ohmic diffusion is dominant close to the midplane ( @xmath153 ) when the field is weak ( @xmath173 a few mg ) . the midplane can be magnetically coupled when the grains are absent or have aggregated to @xmath174 m . however , a dead zone develops below @xmath175 scaleheights when they are small ( @xmath136 m ) . in this scenario , only @xmath161 mg can couple to the gas ( and only for @xmath176 ) . note that ambipolar diffusion dominates for all @xmath11 when @xmath157 ( all panels ) and for 1 mg @xmath152 mg below this height ( bottom panel only).,scaledwidth=40.0% ] in this section we address the following question : which magnetic diffusion mechanism determines the properties of the mri in different regions in protoplanetary discs ? in this connection , it is useful to recall ( see w99 , sw03 and references therein for details ) that in the disc regions where the magnetic coupling @xmath177 , ideal - mhd conditions hold and the particular configuration of the diffusivity tensor has little effect on the behaviour of the mri . when @xmath103 is weaker than this but @xmath178 , ambipolar diffusion dominates . finally , in the regions where @xmath179 , hall diffusion modifies the structure and growth of mri unstable modes , provided that this degree of coupling is sufficient for unstable modes to grow . figs . [ fig : chi_10au ] and [ fig : chi_5au ] compare @xmath103 with the ratio @xmath180 at @xmath114 and @xmath13 au , respectively , as a function of the magnetic field strength and for different assumptions regarding the presence , and radius , of dust grains mixed with the gas . in each figure , the top panels depict the results obtained assuming that no grains are present whereas the effect of 1 @xmath116 m ( 0.1 @xmath116m)-sized grains is shown in the middle ( bottom ) panels . note the dramatic impact dust grains have in the level of magnetic coupling at both radii . for example , introducing a population of @xmath17 m grains causes @xmath103 to drop by 6 - 8 orders of magnitude at the midplane . bottom ( solid ) and leftmost ( dashed ) lines correspond to @xmath181 mg in all cases . the field strength increases by a factor of ten towards larger @xmath103 , except that at both radii the top ( and rightmost ) curves for @xmath172 m show the maximum field strength for which the mri grows . this is also the case for the solutions with no grains at 5 au ( the maximum values of @xmath11 for these cases are noted in the captions of figs . [ fig : chi_10au ] and [ fig : chi_5au ] ) . note that at @xmath24 au ( fig . [ fig : chi_10au ] ) , hall diffusion is not expected to play an important role in the local properties of the mri once the grains have settled ( @xmath182 for all @xmath107 and @xmath11 ) . in this scenario , ambipolar diffusion dominates in the inner sections of the disc when the field is weak ( @xmath170 and @xmath183 mg ) while ideal - mhd holds at all heights for stronger @xmath11 . on the contrary , when either 1 or 0.1 @xmath116 m grains are mixed with the gas ( middle and bottom panels ) , hall diffusion has an impact on the mri within @xmath150 three scaleheights of the midplane . at higher @xmath107 , the ionisation fraction is such that @xmath184 and ambipolar diffusion determines the local properties of mri unstable modes . the corresponding solutions at @xmath13 au are shown in fig . [ fig : chi_5au ] . in this case , hall diffusion dominates within two scaleheights from the midplane if the magnetic field is weak ( @xmath185 mg ) and the grains have settled ( top panel ) . this is consistent with the higher column density at this radius in comparison to the 10 au case discussed above ( for which hall diffusion was unimportant ) . ambipolar diffusion is dominant in this scenario for @xmath24 mg @xmath186 mg but for stronger fields , @xmath177 for all @xmath107 and the fluid is in ideal - mhd conditions over the entire section of the disc . on the other hand , when dust grains are present ( middle and bottom panels ) , hall diffusion determines the properties of mri - unstable modes in the inner sections of the disc ( @xmath133 ) for all magnetic field strengths for which they grow . ambipolar diffusion is locally dominant at higher @xmath107 . finally , note that when dust grains are present , the extent of the disc section where hall diffusion has an impact on the mri at both radii is quite insensitive to the strength of the field . evidently , dust particles can efficiently reduce the degree of magnetic coupling in the disc inner sections for a wide range of magnetic field strengths . in light of the concepts presented so far , we now analyse the properties of the mri at the two radii of interest . ( solid lines ) and critical coupling @xmath187 ( dashed lines ) for @xmath188 au as a function of the magnetic field strength and for different assumptions regarding the presence ( and size ) of dust grains . hall diffusion modifies the structure and growth rate of mri unstable modes in the regions where @xmath189 provided that the coupling is sufficient for the instability to operate ( sw03 ) . ambipolar diffusion is important if @xmath190 . for stronger @xmath103 , ideal - mhd describes the fluid adequately ( w99 ) . the bottom ( and leftmost ) lines correspond to @xmath191 mg in all panels . @xmath11 increases by a factor of 10 between curves ( towards larger @xmath192 ) , except that the top curve for @xmath172 m grains corresponds to @xmath193 mg , the strongest field for which perturbations grow . , scaledwidth=40.0% ] for @xmath3 au . note , however , that the top ( and rightmost ) lines correspond to @xmath194 mg ( top panel ) and @xmath195 mg ( middle panel ) , the maximum field strength for which unstable modes grow in each scenario.,scaledwidth=40.0% ] figs . [ fig : grains1 ] and [ fig : grains1_1 ] compare the vertical structure and growth rate of the most unstable mri modes at @xmath114 and 5 au , respectively , for different choices of the magnetic field strength . the left column of each figure displays solutions obtained assuming that grains have settled out of the gas phase . the remaining columns from left to right show results that incorporate the effect of a different population of single- sized dust particles of radius @xmath196 3 , 1 and @xmath197 m , respectively . note how the growth rate , wavenumber and range of magnetic field strengths for which unstable modes exist are all drastically diminished when dust grains are present . this is expected , given the reduction in the coupling between the neutral and ionised components of the fluid when dust grains ( particularly if they are small ) are well mixed with the gas . note also that the range of field strengths for which unstable modes are found matches quite well with the range for which the magnetic field is expected to couple to the fluid , as discussed in section [ subsec : diffusivity ] ( see figs . [ fig : contour_10au ] and [ fig : contour_5au ] and compare the maximum magnetic field strength for which the mri grows and the maximum @xmath11 that can couple to the gas at the @xmath107 where the modes peak ) . for the discussion that follows , it is useful to keep in mind that both the growth rate and the envelope of these perturbations are shaped by the interplay of different diffusion mechanisms , whose relative importance vary strongly with height . in particular , the ambipolar diffusion component of the diffusivity tensor drives the local growth rate ( and thus the amplitude of global perturbations ) to increase with @xmath107 ( sw03 ) . this is because , in this diffusion regime , the maximum growth rate increases with the local @xmath103 ( w99 ) and is , therefore , a strong function of the vertical location . accordingly , the envelopes of the perturbations driven by this term typically peak at an intermediate height above the midplane . on the contrary , the maximum growth rate of hall perturbations is insensitive to @xmath103 ( w99 ) . because they are not driven from any particular vertical location , their envelope is fairly flat ( sw03 ) . we now analyse the structure and growth of the perturbations at the two radii of interest . the properties of these modes without dust grains were discussed in detail in sw05 . for the sake of clarity , these results are briefly summarised as part of this discussion . at @xmath24 au , ambipolar diffusion drives the mri when the field is weak ( @xmath183 mg ) and ions and electrons are the sole charge carriers ( see top panel of fig . [ fig : chi_10au ] ) . as a result , unstable modes peak above the midplane in this scenario ( left column of fig . [ fig : grains1 ] ; e.g. the mode computed with @xmath181 mg peaks at @xmath198 ) . for stronger fields , ideal - mhd holds at all @xmath107 . this explains the flat envelope , and fast growth , of the perturbations obtained with @xmath115 and 100 mg . unstable modes are found in this case for @xmath199 mg and they grow at about the ideal - mhd rate for 2 mg @xmath200 mg . when @xmath11 is even stronger than 250 mg , the wavelength of the most unstable mode is @xmath201 , the disc tidal scaleheight , and the perturbations are strongly damped ( balbus & hawley 1991 ) . finally , note that no dead zone develops in this scenario , given that the magnetic field is coupled to the gas even at the midplane ( see top panel of fig . [ fig : contour_10au ] ) . we now turn our attention to the solutions obtained under the assumption that dust grains are present at this radius . note that when the dust particles are relatively large ( @xmath202 and @xmath19 m , central two columns of fig . [ fig : grains1 ] ) , the perturbations exhibit the flat envelope that is typical when hall diffusion controls their behaviour ( for these fluid conditions hall diffusion is expected to modify mri modes for @xmath203 , fig . [ fig : chi_10au ] ) . these modes grow even at the midplane , a result consistent with the level of magnetic coupling associated with this disc model ( e.g. see middle panel of fig . [ fig : contour_10au ] ) . on the other hand , when the dust grains are small ( rightmost column of fig . [ fig : grains1 ] ) the low magnetic coupling , especially within two scaleheights of the midplane ( fig . [ fig : contour_10au ] , bottom panel ) , causes the amplitude of all perturbations in this section of the disc to be severely reduced . unstable modes were found here only for @xmath183 mg , a much reduced range compared with the @xmath204 mg for which they exist when ions and electrons are the only charge carriers . [ fig : grains1_1 ] shows the solutions obtained for @xmath205 au . at this radius , mri unstable modes grow without grains for @xmath206 mg . moreover , they grow at essentially the ideal - mhd rate for a significant subset of this range ( @xmath207 mg @xmath208 mg ) . hall diffusion modifies the structure and growth of these modes for @xmath183 mg . when the field is within this limit , a small magnetically dead region develops but it disappears for stronger @xmath11 . note that all the solutions that incorporate dust grains peak at a height above the midplane where ambipolar diffusion is locally dominant ( e.g. @xmath209 for @xmath115 mg and @xmath172 m ; see also the middle panel of fig . [ fig : chi_5au ] ) . this signals that this diffusivity term shapes the structure of these perturbations and explains why their amplitude increase with height . the solutions computed with the small grain population ( @xmath136 m ) exhibit , as in the @xmath114 au case , an extended dead zone encompassing the region where the magnetic coupling is insufficient to sustain the mri ( @xmath203 ; see bottom panel of fig . [ fig : contour_5au ] ) . solutions are found in this case for @xmath210 mg . the solutions described so far in this section incorporate all diffusion mechanisms ( represented by @xmath75 , @xmath76 and @xmath74 ) . for comparison , fig [ fig : grains3_1 ] displays how the full @xmath211 modes at 10 au and including @xmath17m - sized grains ( left column ) , are modified if only ambipolar diffusion ( middle column ) and hall diffusion ( right column ) are considered . note that when the field is weak ( e.g. the solutions for @xmath181 and 4 mg ) , full @xmath211 perturbations grow faster , and are active closer to the midplane , than modes obtained in the ambipolar diffusion limit . this is the result of the contribution of the hall diffusivity term ( e.g. note that the solution in the hall limit for @xmath181 mg grows at the ideal - mhd rate ) . for @xmath115 mg , the structure of the modes computed with and without the hall term are fairly similar , as this diffusion mechanism is no longer important at the @xmath107 where they peak ( see bottom panel of fig . [ fig : contour_10au ] ) . note also that modes in the hall limit do not grow for @xmath212 mg , a result consistent with ambipolar diffusion being dominant at these field strengths ( see bottom panel of fig . [ fig : contour_10au ] ) . a similar comparison is shown in fig . [ fig : grains5 ] for @xmath3 au and incorporating the effect of @xmath15m - sized dust particles . the hall term is important here also , as evidenced by the faster growth and more extended unstable zone of the perturbations that include this diffusion mechanism . ambipolar diffusion strongly influences the structure of the modes for intermediate field strengths ( e.g. compare the solutions with @xmath115 mg and different configurations of the diffusivity tensor ) and the perturbations grow in this limit for a reduced range of @xmath11 in relation to that associated with full @xmath211 ( or hall limit ) modes . the dependence of the growth rate of the most unstable mode ( @xmath213 ) with the strength of the magnetic field is summarised in fig . [ fig : grains6 ] for different assumptions regarding the presence , and radius , of dust grains mixed with the gas . results are shown for @xmath188 au ( top panel ) and 5 au ( bottom panel ) . note that @xmath213 drops sharply at a characteristic field strength ( @xmath214 ) , which is a function of both the properties of the grain population and the radial position . the maximum field strength for which perturbations grow in a particular scenario is weaker at larger radii , an effect that is particularly noticeable when no grains are present . this behaviour results from the instability being ( generally ) damped when the midplane ratio of the alfvn to sound speed approaches unity ( e.g. @xmath215 ; balbus & hawley 1991 ) ( sw03 ) when the magnetic field is counteraligned with the disc angular velocity vector ( @xmath216 ) . ] and the wavelength of the most unstable mode becomes @xmath201 . as the midplane density and temperature decrease with radius , the ratio @xmath217 , associated with a particular field strength , increases at larger radii and the perturbations are damped at a weaker field as @xmath106 increases . note also that for each radii , the range of field strengths over which unstable modes exist is smaller as the grain size diminishes . this is consistent with the drop in magnetic coupling for a particular field strength as the dust particles are smaller ( see figs . [ fig : contour_10au ] and [ fig : contour_5au ] ) . finally , note that at 5 au and for weak fields ( @xmath218 mg , bottom panel of fig . [ fig : grains6 ] ) the mri grows faster when the small grains are considered than it does in the other scenarios . this is because in this case the modes grow high above the midplane , where the magnetic coupling is more favourable , and are completely suppressed at lower @xmath107 . in the other cases , on the contrary , the perturbations grow over a more extended section of the disc and the comparatively low magnetic coupling closer to the midplane reduces their global growth rate . this effect is not so noticeable at 10 au by the relatively high magnetic coupling even at @xmath156 at this radius . au ( top panel ) and @xmath3 au ( bottom panel).,scaledwidth=45.0% ] in this paper we have examined illustrative examples of the impact of dust grains in the magnetic activity of protoplanetary discs and , in particular , in the linear growth and vertical structure of mri perturbations . solutions were computed for @xmath114 and @xmath13 au assuming a single size grain population of radius @xmath0 , 1 or 3 @xmath116 m , and constituting 1 % of the total mass of the gas , is well mixed with the gas phase over the entire vertical extent of the disc . this fraction is independent of height , so we have also assumed that the grains have not sedimented towards the midplane . our results indicate that the perturbations wavenumber and growth rate are significantly reduced when grains are present . furthermore , the magnetically inactive dead zone , which was practically non - existent when grains were settled , extends to @xmath219 at either radii when 0.1 @xmath116m - sized grains are considered . at 10 au ( 5 au ) , unstable perturbations were found in this case for @xmath183 mg ( 16 mg ) , a much reduced range compared with the strengths for which they exist when no grains are involved ( 250 and 795 mg , respectively ) . this maximum field strength corresponds well to the equipartition field at the height at which the perturbations peak , as expected ( e.g. @xmath220 at 10 au ; see lower right panel of fig . [ fig : grains1 ] ) . these results illustrate the impact of dust particles on the dynamics and evolution of low conductivity discs . they can also be used to estimate the maximum magnetic field strength to support magnetic activity at @xmath1 au . the magnetic coupling at this radius , in a disc including @xmath221 m grains , is too weak below @xmath222 to allow the field to sufficiently couple with the gas ( w07 ) . assuming that the maximum field strength for the mri to grow is also of the order of the equipartition field at about this height , we can roughly estimate that the mri should be active for @xmath223 mg at @xmath1 au . this is a smaller range than the several gauss for which unstable modes exist when dust grains are not present ( sw05 ) . however , mri modes still grow in this case for a wide range of field strengths . the results just presented were obtained assuming that all grains have the same size and are well mixed with the gas at all @xmath107 . more realistic spatial , and size , distributions must incorporate the effects of dust dynamics and evolution within the disc . observations of the mid and far infrared spectra of discs have provided credible indications that dust properties in discs are indeed different from those of particles in diffuse clouds ( e.g. van dishoeck 2004 , dalessio et al . 2006 and references therein ) . two aspects of this dust evolution have been clearly identified . first , dust grains coagulate from @xmath17 m to @xmath151 mm particles . second , ( silicate ) material becomes crystallised . it is believed that this crystallisation occurs in the disc , given that crystalline silicates are absent from the interstellar medium . furthermore , the presence of this material at radial locations where the temperature is too low to produce them , suggests significant radial mixing takes place as well ( e.g. van dishoeck 2004 and references therein ) . finally , simulations of dust dynamics and evolution also suggest that in quiescent environments , the grains tend to settle and agglomerate into bigger particles ( e.g. weidenschilling & cuzzi 1993 ) and efficiently coagulate and grow icy mantles ( ossenkopf 1993 ) . all these processes modify the surface area of dust grains and impact the recombination rate on their surfaces and the way they drift in response to magnetic stresses . it is also expected that a residual population of small grains would remain ( e.g. nomura et al . 2006 ) even when the mean grain size maybe relatively large ( @xmath224 m ) . this is an important consideration because these small grains tend to carry a significant fraction of the grain charge ( s07 ) . the effect of this settling in the spectral energy distribution ( and optical appearance ) of protostellar discs has been investigated in recent studies ( e.g. dullemond & dominik 2004 ) . how quickly , and to what height , dust particles are able to settle is an important , and largely unanswered , question . according to nakagawa , nakazawa & hayashi ( 1981 ) , the mass fraction of @xmath151 - @xmath225 m grains well mixed with the gas , diminishes from @xmath226 to @xmath227 in a timescale of about @xmath228 to @xmath229 years . moreover , although the timescale for dust grains to sediment all the way to the midplane may exceed the lifetime of the disc , they may be able to settle within a few scaleheights from the midplane in a shorter timescale ( dullemond & dominik 2004 ) . this is complicated even more by the expectation that the transition between sections were dust grains are well mixed with the gas , and those completely depleted of them , occurs gradually ( dullemond & dominik 2004 ) . mhd turbulence may itself be an important factor for the settling of dust particles . it may , in particular , produce sufficient vertical stirring to prevent settling below a certain height ( dullemond & dominik 2004 , carballido et al . 2005 , turner et al . 2006 ) . however , this is contingent on the disc being able to generate and sustain mhd turbulence in the vertical sections where the dust is present . this is not guaranteed , even if turbulence exists in other regions , as dust grains efficiently reduce the ionisation fraction ( and magnetic coupling ) of the gas . as a result , the efficiency and even the viability of mhd turbulence in the presence of dust grains , is an important topic that merits careful investigation . we have explored in this paper the linear growth and vertical structure of mri unstable modes at two representative radii ( @xmath3 and 10 au ) in protoplanetary discs . dust grains are assumed to be well mixed with the fluid over the entire section of the disc and , for simplicity , are taken to have the same radius ( @xmath202 , 1 or 0.1 @xmath116 m ) . they constitute a constant fraction ( 1% ) of the total mass of the gas . these solutions are compared with those arrived at assuming that the grains have settled to the midplane of the disc ( sw05 ) . we have also explored which disc sections are expected to be magnetically coupled and the dominant diffusion mechanism as a function of height and the strength of the magnetic field , which is initially vertical . our models use a minimum - mass solar nebula disc ( hayashi 1981 , hayashi et al . 1985 ) and incorporate all three diffusion mechanisms between the magnetic field and the neutral gas : ohmic , hall and ambipolar . the diffusivity components are a function of height ( as is the density ) and are obtained using the method described in w07 , to which we refer the reader for details . essentially , this formalism uses a chemical reaction scheme similar to that of nishi , nakano & umebayashi ( 1991 ) , but it incorporates higher dust - grain charge states that are likely to occur in discs on account of their larger gas density and temperature in relation to those of molecular clouds . our calculations also include a realistic ionisation profile , with the main ionising sources being cosmic rays , x - rays and ( to a lesser extent ) radioactive decay . solutions were obtained at the two radii of interest for different grain sizes and configurations of the diffusivity tensor as a function of the magnetic field strength . we refer the reader to sw03 and sw05 for further details of the integration procedure . the main findings of this study are detailed below . 1 . when no grain are present , or they are @xmath5 m in radius , the midplane of the disc remains magnetically coupled for field strengths up to a few gauss at both radii 2 . in contrast , when a population of small grains ( @xmath6 m ) is mixed with the gas , the section of the disc below @xmath230 ( @xmath160 ) is magnetically inactive at @xmath114 au ( 5 au ) . only magnetic fields weaker than 25 mg ( 50 mg ) can couple to the gas . au , ohmic diffusion dominates for @xmath153 when the field is relatively weak ( @xmath9 a few milligauss ) , irrespective of the properties of the grain population . conversely , at 10 au this diffusion term is unimportant in all the scenarios studied here . 4 . high above the midplane ( @xmath231 , depending on the specific model ) , ambipolar diffusion is severe and prevents the field from coupling to the gas for all @xmath11 . this is consistent with previous results by w07 . 5 . hall diffusion is dominant for a wide range of field strengths and grain sizes at both radii ( see figs . [ fig : contour_10au ] and [ fig : contour_5au ] ) . 1 . the growth rate , wavenumber and range of magnetic field strengths for which unstable modes exist are all drastically diminished when dust grains are present , particularly when they are small ( @xmath232 m ; see figs . [ fig : grains1 ] and [ fig : grains1_1 ] ) . 2 . in all cases that involve dust grains , perturbations that incorporate hall diffusion grow faster than those obtained under the ambipolar diffusion approximation . 3 . at 10 au , unstable mri modes grow for @xmath233 mg ( 10 mg ) when the grain size is @xmath19 m ( @xmath17 m ) , a much reduced range compared with the @xmath234 mg for which they exist when ions and electrons are the only charge carriers ( sw05 ) . when the grains are relatively large ( @xmath164 and @xmath15 m ) , hall diffusion controls the structure the modes , which grow even at the midplane . in contrast , when the grains are small ( @xmath136 m ) , the perturbations grow only for @xmath235 and are shaped mainly by ambipolar diffusion . au , mri perturbations exist for @xmath233 mg ( 16 mg ) when the grains are @xmath19 m ( @xmath17 m ) in size . for comparison , the upper limit when no grains are present is @xmath236 mg ( sw05 ) . these modes are shaped largely by ambipolar diffusion , the dominant mechanism at the height where they peak . we conclude that in protoplanetary discs , the magnetic field is able to couple to the gas and shear over a wide range of fluid conditions even when small dust grains are well mixed with the gas . despite the low magnetic coupling , mri modes grow for an extended range of magnetic field strengths and hall diffusion largely determines the properties of the perturbations in the inner regions of the disc . this research has been supported by the australian research council . rs acknowledges partial support from nasa theoretical astrophysics program grant nng04g178 g . balbus s. a. , hawley j. f. , 1991 , apj , 376 , 214 balbus s. a. , hawley j. f. , 1998 , rev . phys . , 70 , 1 blandford r.d . , payne d.g . , 1982 , mnras , 199 , 883 ( bp82 ) carballido a. , stone j. m. , pringle j. e. , 2005 , mnras , 358 , 1055 consolmagno g. j. , jokipii j. r. , 1978 , moon planets , 19 , 253 cowling , t. g. 1957 , magnetohydrodynamics ( new york : interscience ) dalessio p. , calvet n. , hartmann l. , franco hernandez r. , servin h. , 2006 , apj , 638 , 314 dullemond c. p. , dominik c. , 2004 , a&a , 421 , 1075 fromang s. , terquem c. , balbus s. a. , 2002 , mnras , 329 , 18 gammie c. f. , 1996 , apj , 457 , 355 glassgold a. e. , feigelson e. d. , montmerle t. , 2000 , in protostars & planets iv , ed . v. g. mannings , a. p. boss , s. russell ( tucson : univ . arizona press ) , p. 429 glassgold a. e. , najita j. , igea j. , 1997 , apj , 480 , 344 hayashi c. , 1981 , prog theor phys supp , 70 , 35 hayashi c. , nakasawa k. , nakagawa y. , 1985 , in protostars & planets ii , ed . black , m. s. mathews ( tucson : univ . arizona press ) , p. 1100 igea j. , glassgold a. e. , 1999 , apj , 518 , 848 ilgner m. , nelson r. p. , 2006 , , 445 , 223 johnson e. t. , goodman j. , menou k. , 2006 , apj , 647 , 1413 knigl a. , pudritz r. e. , 2000 , in mannings , v. g. , boss , a. p. , russell , s. eds , protostars & planets iv . arizona press , tucson , p. 759 mathis j. s. , rumpl w. , nordsieck k. h. , 1977 , apj , 217 , 425 matsumura s. , pudritz r. e. , 2005 , apj , 618 , l137 morton d. c. , 1974 , astrophys . j. , 193 , l35 nakagawa y. , nakazawa k. , hayashi c. , 1981 , icarus , 45 , 517 nakano t. , umebayashi t. 1986 , mnras , 218 , 663 nishi r. , nakano t. , umebayashi t. , 1991 , apj , 368 , 181 nomura h. , nakagawa y. , 2006 , apj , 640 , 1099 norman c. , heyvaerts j. , 1985 , aa , 147 , 247 oppenheimer m. , dalgarno a. , 1974 , apj , 192 , 29 ossenkopf v. , 1993 , aa , 280 , 617 salmeron r. , wardle m. , 2003 , mnras , 345 , 992 ( sw03 ) salmeron r. , wardle m. , 2005 , mnras , 361 , 45 ( sw05 ) sano t. , stone j. m. , 2002a , apj , 570 , 314 sano t. , miyama s. , umebayashi j. , nakano t. , 2000 , apj , 543 , 486 semenov d. , wiebe d. , henning th . , 2006 , apj , 647 , l57 spitzer , l. 1978 , physical processes in the interstellar medium ( new york : wiley ) terquem c. , 2003 , mnras , 341 , 1157 turner n. j. , willacy k. , bryden g. , yorke h. w. , 2006 , apj , 639 , 1218 umebayashi t. , nakano t. , 1980 , pasj , 32 , 405 umebayashi t. , nakano t. , 1990 , mnras , 243 , 103 van dishoeck e. f. , 2004 , , 42 , 119 wardle , m. , knigl , a. 1993 , apj , 410 , 218 wardle m. , 1997 , in proc . iau colloq . 163 , accretion phenomena and related outflows , ed . d. wickramasinghe , l. ferrario , g. bicknell ( san francisco : asp ) , p. 561 wardle m. , 1999 , mnras , 307 , 849 ( w99 ) wardle m. , 2007 , astro - ph/0704.0970 wardle m. , ng c. , 1999 , mnras , 303 , 239 weidenschilling s. j. , cuzzi j. n. , 1993 , in protostars & planets iii , ed . e. h. levy , j. i. lunine ( tucson : univ . arizona press ) , p. 1031

for classification problems with significant class imbalance , subsampling can reduce computational costs at the price of inflated variance in estimating model parameters . we propose a method for subsampling efficiently for logistic regression by adjusting the class balance locally in feature space via an accept reject scheme . our method generalizes standard case - control sampling , using a pilot estimate to preferentially select examples whose responses are conditionally rare given their features . the biased subsampling is corrected by a post - hoc analytic adjustment to the parameters . the method is simple and requires one parallelizable scan over the full data set . standard case - control sampling is inconsistent under model misspecification for the population risk - minimizing coefficients . by contrast , our estimator is consistent for provided that the pilot estimate is . moreover , under correct specification and with a consistent , independent pilot estimate , our estimator has exactly twice the asymptotic variance of the full - sample even if the selected subsample comprises a miniscule fraction of the full data set , as happens when the original data are severely imbalanced . the factor of two improves to if we multiply the baseline acceptance probabilities by ( and weight points with acceptance probability greater than 1 ) , taking roughly times as many data points into the subsample . experiments on simulated and real data show that our method can substantially outperform standard case - control subsampling .

1306.3706

in recent years , statisticians , scientists and engineers are increasingly analyzing enormous data sets . when data sets grow sufficiently large , computational costs may play a major role in the analysis , potentially constraining our choice of methodology or the number of data points we can afford to process . computational savings can translate directly to statistical gains if they : enable us to experiment with and prototype a variety of models , instead of trying only one or two , allow us to refit our models more often to adapt to changing conditions , allow for cross - validation , bagging , boosting , bootstrapping or other computationally intensive statistical procedures or open the door to using more sophisticated statistical techniques on a compressed data set . @xcite discuss the tradeoffs arising when we adopt this point of view . one simple manifestation of these tradeoffs is that we may run out of computing resources before we run out of data , in effect making the sample size @xmath4 a function of the efficiency of our fitting method . class imbalance is pervasive in modern classification problems and has received a great deal of attention in the machine learning literature [ @xcite ] . it can come in two forms : : : _ marginal imbalance_. one of the classes is quite rare ; for instance , @xmath5 . such imbalance typically occurs in data sets for predicting click - through rates in online advertising , detecting fraud or diagnosing rare diseases . : : _ conditional imbalance_. for most values of the features @xmath6 , the response @xmath7 is very easy to predict ; for instance , @xmath8 but @xmath9 . for example , such imbalance might arise in the context of email spam filtering , where well - trained classifiers typically make very few mistakes . both or neither of the above may occur in any given data set . the machine learning literature on class imbalance usually focuses on the first type , but the second type is also common . if , for example , our data set contains one thousand or one million negative examples for each positive example , then many of the negative data points are in some sense redundant . typically in such problems , the statistical noise is primarily driven by the number of representatives of the rare class , whereas the total size of the sample determines the computational cost . if so , we might hope to finesse our computational constraints by subsampling the original data set in a way that enriches for the rare class . such a strategy must be implemented with care if our ultimate inferences are to be valid for the full data set . this article proposes one such data reduction scheme , local case - control sampling , for use in fitting logistic regression models . the method requires one parallelizable scan over the full data set and yields a potentially much smaller subsample containing roughly half of the information found in the original data set . the simplest way to reduce the computational cost of a procedure is to subsample the data before doing anything else . however , uniform subsampling from an imbalanced data set is inefficient since it fails to exploit the unequal importance of the data points . case - control sampling sampling uniformly from each class but adjusting the mixture of the classes is a more promising approach . this procedure originated in epidemiology , where the positive examples ( cases ) are typically diseased patients and negative examples ( controls ) are disease - free [ @xcite ] . often , an equal number of cases and controls are sampled , resulting in a subsample with no marginal imbalance , and costly measurements of predictor variables are only made for selected patients [ @xcite ] . this method is useful in our context as well , since a logistic regression model fitted on the subsample can be converted to a valid model for the original population via a simple adjustment to the intercept [ @xcite ] . however , standard case - control sampling still may not make most efficient use of the data . for instance , it does nothing to exploit conditional imbalance in a data set that is marginally balanced . even with some marginal imbalance , a control that looks similar to the cases is often more useful for discrimination purposes than one that is obviously not a case . we propose a method , local case - control sampling , which attempts to remedy imbalance _ locally _ throughout the feature space . given a pilot estimate @xmath10 of the logistic regression parameters , local case - control sampling preferentially keeps data points for which @xmath7 is surprising given @xmath6 . specifically , if @xmath11 , we accept @xmath12 with probability @xmath13 , the @xmath14 residual of the pilot model . in the presence of extreme marginal or conditional imbalance , these errors will generally be quite small and the subsample can be many orders of magnitude smaller than the full data set . just as with case - control sampling , we can fit our model to the subsample and make an equally simple correction to obtain an estimate for the original data set . when the logistic regression model is correctly specified and the pilot is consistent and independent of the data , the asymptotic variance of the local case - control estimate is exactly twice the variance of a logistic regression fit on the ( potentially much larger ) full data set . this factor of two improves to @xmath1 if we accept with probability @xmath15 and weight accepted points by a factor of @xmath16 . for example , if @xmath17 then the variance of the subsampled estimate is only 20% greater than the variance of the full - sample mle . the subsample we take with @xmath2 is no more than @xmath18 times larger than the subsample for @xmath19 , and for data sets with large imbalance is roughly @xmath3 times as large . local case - control sampling also improves on the bias of standard case - control sampling . when the logistic regression model is misspecified , case - control sampling is in general inconsistent for the risk minimizer in the original population . by contrast , local case - control sampling is always consistent given a consistent pilot , and is also asymptotically unbiased when the pilot is . sections [ secsimulations ] and [ secspam ] present empirical results demonstrating the advantages of our approach in simulations and on the yahoo ! webspam data set . our setting is that of predictive classification : we are given @xmath4 independent and identically distributed observations , each consisting of predictors @xmath20 and a binary response @xmath21 , with joint probability measure @xmath22 . for our purposes , we assume the predictors are mapped into some real covariate vector space , so that @xmath23 our aim is to learn the function @xmath24 or equivalently to learn @xmath25 which could be infinite for some @xmath26 . a linear logistic regression model assumes @xmath27 is linear in @xmath26 ; that is , @xmath28 where @xmath29 . this is less of a restriction than it might seem , since @xmath26 may represent a very large basis expansion of some smaller set of `` raw '' features . nevertheless , in the real world , @xmath27 is unlikely to satisfy our parametric model for any given basis @xmath26 . when the model is misspecified , we can still view logistic regression as an m - estimator with convex loss equal to the negative log - likelihood for a single data set : @xmath30 as an m - estimator , under general conditions logistic regression in large samples will converge to the minimizer of the population risk @xmath31 [ @xcite ] . that is , @xmath32 converges to the population maximizer of the expected log - likelihood @xmath33.\end{aligned}\ ] ] if @xmath34 for some @xmath35 , then @xmath36 ; otherwise @xmath37 is the best linear approximation to @xmath27 in the sense of ( [ eqnpoprisk ] ) . for a misspecified model , @xmath38 can not possibly converge to @xmath27 no matter what sampling scheme or estimation procedure we use , or how much data we obtain . consistency , then , will mean that @xmath39 . model misspecification is ubiquitous in real - world applications of regression methods . for reasons of exposition , the misspecification always takes a simple form in our simulations , for example , in example [ ex1 ] there are two binary predictors , and we would have correct specification if only we added one interaction but in the real world it usually is neither possible nor even desirable to expand the feature basis until the model is correctly specified . for instance , if @xmath40 , then there are @xmath41500,500 quadratic terms . even if we included all those terms as features , we would still be missing cubic terms , quartic terms , and so on . some kinds of misspecification are milder than others , and some are easier to find and fix than others . seeking better - specified models ( without adding too much model complexity ) is a worthy goal , but realistically perfect specification is unattainable . our goal , then , is to speed up computation while still obtaining a good estimate of @xmath0 , the population logistic regression parameters . as we will see , standard case - control sampling achieves the first goal , but may fail at the second . recent years have seen substantial work on classification in imbalanced data sets . see @xcite and @xcite for surveys of machine learning efforts on this problem . many of the methods proposed involve some form of undersampling the majority class , oversampling the minority class , or both . @xcite examined the limit of marginally imbalanced logistic regression and proved it is equivalent to fitting an exponential family model to the minority class . one recurring theme is to preferentially sample negative examples that lie near positive examples in feature space . for example , @xcite propose selecting majority - class examples whose average distance to its three nearest minority examples is smallest . our method has a similar flavor since the probability of sampling a negative example @xmath42 is @xmath43 , which is large when the features @xmath26 are similar to those characteristic of positive examples . our proposal lies more in the tradition of the epidemiological case - control sampling literature . in particular , case - control sampling within several categorical strata has been studied by @xcite . typically , the strata are based on easy - to - measure screening variables available for a wide population , with more laborious - to - collect variables being measured on the sampled subjects . @xcite discuss survey calibration methods for efficient regression in two - stage sampling schemes , which are interesting but too computationally intensive for our purposes here . case - control sampling is commonly carried out by taking all the cases and exactly @xmath18 times as many controls for some fixed @xmath18 ( e.g. , @xmath44 ) . however , for our purposes it will be simpler to consider a nearly equivalent procedure based on accept reject sampling . define some acceptance probability function @xmath45 and let @xmath46 , the log - selection bias . consider the following algorithm : generate independent @xmath47 . fit a logistic regression to the subsample @xmath48 , obtaining unadjusted estimates @xmath49 . assign @xmath50 and @xmath51 . specifically , we could generate the @xmath52 by first generating @xmath53 mutually independent of the pilot , the data , and each other , then taking @xmath54 . note that steps ( 2)(3 ) are equivalent to logistic regression with offset @xmath55 for each data point . this variant is convenient to analyze because the subsample thus obtained is an i.i.d . sample from a new population : @xmath56 with @xmath57 , the marginal probability of @xmath58 . the estimate @xmath59 is motivated by a simple application of bayes rule relating the odds of @xmath60 in @xmath22 and @xmath61 . if @xmath62 is the true conditional log - odds function for @xmath63 , we have @xmath64 that is , the log - odds @xmath62 in our biased population is simply a vertical shift by @xmath55 of the log - odds @xmath65 in the original population , so given an estimate of @xmath66 we can subtract @xmath55 to estimate @xmath27 . if the model is correctly specified , logistic regression on the subsample yields a consistent estimate for the function @xmath62 , so the estimate for @xmath65 is also consistent . note that the derivation ( [ eqnbayesccbegin])([eqnbayesccend ] ) is equally valid if the sampling bias @xmath55 depends on @xmath26 , in which case we have @xmath67 . local case - control sampling exploits this more general identity . if @xmath6 is integrable , then upon differentiating the population risk ( [ eqnpoprisk ] ) with respect to @xmath32 we obtain the population score criterion : @xmath68 = \int { \bigl(p(x ) - p_\theta(x ) \bigr ) \pmatrix{1 \cr x } \,d\mathbb{p}(x)}.\end{aligned}\ ] ] informally , the best linear predictor is the one that gets the conditional probabilities right on average . note this is not the same as a predictor that gets the conditional log - odds right on average . to illustrate the difference between approximating probabilities and approximating logits , suppose that @xmath69 and @xmath70 . the left panel of figure [ figjumpplots ] shows @xmath65 as a solid line and its best linear approximation as a dashed line . on the logit scale , the dashed line appears to be a very poor fit to the black curve . it fits reasonably well for large @xmath26 , but it appears more or less to ignore the smaller values of @xmath26 . the right panel of figure [ figjumpplots ] shows why . when we transform both curves to the probability scale , the fit looks much more reasonable . @xmath71 need not approximate @xmath65 particularly well for small @xmath26 , because in that range even a large change in the log - odds produces a negligible change in the conditional probability @xmath72 . by contrast , @xmath71 needs to approximate @xmath65 well for larger @xmath26 , where @xmath72 changes more rapidly . approximates the true log - odds @xmath65 in the sense of matching its implied conditional probabilities , not logits . ] in general , logistic regression places highest priority on fitting @xmath27 where @xmath73 is largest : where @xmath74 and @xmath75 . in this example , with its strong marginal imbalance , the regions that matter most are those where @xmath72 is largest . this often makes sense in applications such as medical screening or advertising click - through rate prediction , where accuracy is most important when the probability of disease or click - through is nonnegligible . in section [ secdiscussion ] , we consider how to modify the method to obtain classifiers that prioritize correctness near some other , user - defined level curve of @xmath72 . finally , note that figure [ figjumpplots ] suggests the case - control sampling estimate is unlikely to be consistent for @xmath0 in general . the nature of our linear approximation in the left panel is intimately related to the fact that @xmath76 everywhere in the sample space . if @xmath65 were shifted upward by some constant , the response of the dashed curve would be more complicated than a simple constant shift by @xmath55 , since the relative importance of the two segments would change . therefore , estimating @xmath77 and then subtracting @xmath55 may not be a successful strategy . if the linear model is misspecified , the case - control estimate is generically not consistent for the best linear predictor @xmath0 as @xmath78 [ @xcite ] . the unadjusted estimate will instead converge to the best linear predictor of @xmath66 for the distribution @xmath63 , which solves the score criterion @xmath79 let @xmath80 be the large - sample limit of the _ adjusted _ case - control sampling estimate with bias @xmath55 . then @xmath80 solves the population score criterion @xmath81 which differs from ( [ eqnpopscore ] ) in two ways . first , the integral is taken over a different distribution for @xmath6 . second , and more importantly , the integrand is different . we are now approximating @xmath65 in a different sense than we were . in general under misspecification , @xmath80 is different for every @xmath55 . if we sample cases and controls equally , in the limit we will get a different answer than if we sample twice as many controls ; and in either case we will get a different answer than if we use the entire data set or subsample uniformly . these differences can be quite consequential for our inferences about @xmath82 or the predictive performance of our model , as we see next . [ ex1 ] in this fictitious example , we consider estimating the effect of exposure to oatmeal on a person s risk of developing some rare disease . suppose that @xmath83 of the population has a family history of the disease , half the population eats oatmeal ( independently of family history ) , and that both exposure and family history are binary predictors . suppose further that the true conditional log - odds function @xmath65 is given by the top - left panel of table [ taboatmeal ] . [ taboatmeal ] @lcc@ + + & * history * @xmath84 & * history * @xmath85 + oatmeal @xmath86 & @xmath865 & @xmath864 + oatmeal @xmath87 & @xmath8610 & @xmath861 + + 6pt @lcc@ + & * history * @xmath84 & * history * @xmath85 + oatmeal @xmath86 & 0.007 & 0.02 + oatmeal @xmath87 & 5e@xmath865 & 0.37 + @lcc@ + + & * history * @xmath84 & * history * @xmath85 + oatmeal @xmath86 & @xmath861.2 & @xmath860.2 + oatmeal @xmath87 & @xmath866.2 & 2.8 + + 6pt @lcc@ + & * history * @xmath84 & * history * @xmath85 + oatmeal @xmath86 & 0.24 & 0.46 + oatmeal @xmath87 & 0.002 & 0.94 + the corresponding conditional probabilities @xmath72 are given in the lower - left panel of table [ taboatmeal ] . notice that oatmeal increases the risk for people who are already at risk by virtue of their family history , but has a protective effect for everyone else . this interaction means that the additive logistic regression model is misspecified . because only the probabilities in the `` history @xmath87 '' column are large enough to matter , the fitted model for @xmath65 pays more attention to the at - risk population , for whom oatmeal elevates the risk of disease . a logistic regression on a large sample from this population estimates the coefficient for oatmeal as @xmath88 , implying an odds ratio of about 4.0 . this is close to the marginal odds ratio of roughly 4.3 that we would obtain if we did not control for family history . suppose , however , that we sampled an equal number of cases and controls . then the conditional log - odds of disease in our sample would reflect the top - right panel of table [ taboatmeal ] , with all cells increased by the same amount . for large samples , the case - control estimate is @xmath89 , implying an odds ratio of about 0.44 . using case - control sampling has reversed our inference about the effect of oatmeal exposure , because after shifting the log - odds the left column becomes much more important . [ ex2 ] suppose that @xmath90 , and that @xmath91 . let @xmath92 data simulated from this model are shown in the left panel of figure [ fig2dmix ] . in this example , the true log - odds @xmath65 is an additive quadratic function of the two coordinates @xmath93 and @xmath94 . in this example as in the previous one , the population - optimal case - control parameters differ substantially from the optimal parameters in the original population , with dramatic effects for the predictive performance of the model . the boundaries for the two estimates are overlayed on the left panel of figure [ fig2dmix ] . in the right panel , we plot the precision recall curves resulting from each set of parameters on a large test set . a simple alternative to standard case - control sampling is to weight the subsampled data points by the inverse of their probability of being sampled . we include weighted case - control sampling as a competitor in our simulation studies in section [ secsimulations ] . because it is a horvitz thompson estimator with positive sampling probabilities for any @xmath95 pair , this method is @xmath96-consistent , and asymptotically normal and unbiased under general conditions [ @xcite ] . although weighting succeeds in removing the bias induced by the case - control sampling , this consistency comes at a cost of increasing the variance , since the effective sample size is reduced [ @xcite ( @xcite ) ] . and @xmath97 . ] despite its inefficiency , the weighted case - control method can be an attractive means of obtaining a consistent pilot if another good pilot is not immediately available , and we later will use it to that end in our experiments . in this section , we describe local case - control subsampling , a generalization of standard case - control sampling that both improves on its efficiency and resolves its problem of inconsistency . to achieve these benefits , we require a pilot estimate , that is , a good guess @xmath98 for the population - optimal @xmath0 . local case - control sampling differs from case - control sampling only in that the acceptance probability @xmath99 is allowed to depend on @xmath26 as well as @xmath100 . our criterion for selection will be the degree of `` surprise '' we experience upon observing @xmath101 given @xmath102 : @xmath103 where @xmath104 is the pilot estimate of @xmath105 . the algorithm is : generate independent @xmath106 . fit a logistic regression to the sample @xmath107 to obtain unadjusted estimates @xmath108 . assign @xmath109 and @xmath110 . as before , steps ( 2)(3 ) are equivalent to fitting a logistic regression in the subsample with offsets @xmath111 . the @xmath52 are generated as in section [ seccc ] , and the adjustment is again justified by ( [ eqnbayesccbegin])([eqnbayesccend ] ) , only now with the constant @xmath55 replaced by @xmath112 in other words , the subsample is drawn from a measure with @xmath113 if @xmath65 is well approximated by the pilot estimate , then @xmath114 throughout feature space . that is , conditional on selection into @xmath115 , @xmath101 given @xmath102 is nearly a fair coin toss . to motivate this choice heuristically , recall that the fisher information for the log - odds of a bernoulli random variable is maximized when the probability is @xmath116 : fair coin tosses are more informative than heavily biased ones . in effect , local case - control sampling tilts the conditional distribution of @xmath7 given @xmath117 to make each @xmath101 in the subsample more informative . we then fit a logistic regression in the more favorable sampling measure , and `` tilt back '' to obtain a valid estimate for the original population . in marginally imbalanced data sets where @xmath105 is small everywhere in the predictor space , a good pilot has @xmath118 for all @xmath26 , and the number of cases discarded by this algorithm will be quite small . if we wish to avoid discarding any cases , we can always modify the algorithm so that instead of keeping @xmath119 with probability @xmath120 , we keep it with probability 1 and assign weight @xmath120 . in many applications , there may be a natural choice of pilot fit @xmath121 ; for instance , if we are refitting a classification model every day to adapt to a changing world , then yesterday s fit is a natural choice for today s pilot . if no pilot fit is available from such a source , we recommend an initial pass of weighted case - control sampling ( described in section [ subsecwcc ] ) to obtain the pilot . weighted case - control sampling using a fixed fraction of the original sample is itself @xmath96-consistent and asymptotically unbiased for the true parameters . consequently , if the pilot were fit using an independent data set the second - stage estimate would enjoy consistency and asymptotic unbiasedness per the results in section [ sectheory ] . our experiments suggest that mild inaccuracy in the pilot estimate , and using a data - dependent pilot , do not unduly degrade the performance of the local case - control algorithm . for example , is simulation 2 of section [ subsecsim2 ] , the pilot is fifty times less efficient than the final local case - control estimate . the main role of the pilot fit is to guide us in discarding most of the data points for which @xmath101 is obvious given @xmath102 while keeping those for which @xmath101 is conditionally surprising . in our example and simulations , we use a pilot sample about the same size as the local case - control subsample , on the principle that we can afford to spend about as much time computing the pilot as computing the second - stage estimate . when @xmath122 is small throughout @xmath123 , this rule amounts roughly to weighted case - control sampling using all the cases and one control per case . although the above rule has worked reasonably well for us , at this time we can offer no finite - sample guarantees that any given pilot sample size is large enough . because standard case - control sampling amounts to local case - control sampling with a constant - only pilot fit , we might expect that the pilot fit need not be perfect to improve upon case - control sampling . our experiments in sections [ secsimulations ] and [ secspam ] support this intuition . as we will see in section [ subsecasymptotics ] , under correct model specification , and with an independent and consistent pilot , the baseline procedure described above has exactly twice the asymptotic variance as a logistic regression estimated with the full sample , despite using a potentially very small subset of the data . we can improve upon this factor of two by increasing the size of the subsample . one simple way to achieve this is to multiply all acceptance probabilities by some constant @xmath18 , for example , @xmath17 . when deciding whether to sample the point @xmath12 , we would then generate @xmath124 and assign weight @xmath125 to each sampled point . this amounts to a larger , weighted subsample from @xmath63 , and we can make the same correction to the estimates from the subsample . we see in section [ subsecasylargersample ] that for @xmath2 the factor of two is replaced by a factor of @xmath1 . in the case of large imbalance , most of the @xmath126 are near 0 or 1 . for @xmath2 , the marginal acceptance probability at @xmath102 becomes @xmath127 where the approximation holds for @xmath128 or 1 . for @xmath19 , the marginal acceptance probability is @xmath129 , so for @xmath2 we take roughly @xmath3 times as many data points as for @xmath19 . for example , if @xmath17 , the subsample accepted is roughly 3 times as large , and the relative efficiency improves from @xmath130 to @xmath131 . alternatively , if @xmath4 is extremely large , even a small fraction of the full data set may still be more than we want . in that case , we can proceed as above with @xmath132 , or simply sample any desired number @xmath133 of data points uniformly from the local case - control subsample . we now turn to examining the asymptotic behavior of the local case - control estimate . we first establish consistency , assuming a consistent pilot estimate @xmath121 . we expressly do not assume that the pilot estimate is independent of the data , since in some cases we may recycle into the subsample some of the data we used to calculate the pilot . by assuming independence of @xmath121 and the data , we can obtain finer results about the asymptotic distribution of @xmath134 . we show it is asymptotically unbiased when @xmath121 is , and derive the asymptotic variance of the estimate . when the logistic regression model is correctly specified , the local case - control estimate has exactly twice the asymptotic variance of the mle for the full data set . for better clarity of notation in this section , we will use the letter @xmath135 in place of @xmath121 to denote pilot estimates . additionally , we drop the notation @xmath136 and absorb the constant term into @xmath26 , so that @xmath137 . to avoid trivialities , assume without loss of generality that there is no @xmath138 for which @xmath139 ( if not , we can discard redundant features ) . for @xmath140 $ ] define the `` soft hinge '' function @xmath141 and note that @xmath142 = h\bigl(\theta'x ; p(x)\bigr).\ ] ] as a function of @xmath143 , @xmath144 is positive and strictly convex , its magnitude is bounded by @xmath145 , and it has lipschitz constant @xmath146 . if @xmath147 , @xmath144 diverges as @xmath148 , and if @xmath149 @xmath144 diverges as @xmath150 . as a function of @xmath135 , @xmath151 has lipschitz constant @xmath152 . hence , @xmath153 with lipschitz constant @xmath154 . the marginal acceptance probability given @xmath26 is @xmath155 given pilot @xmath135 , the local case - control subsampling scheme effectively samples from the probability measure @xmath156 , where @xmath157 and @xmath158 is the marginal probability of acceptance . under this measure , @xmath159 because @xmath160 , if @xmath161 is integrable under @xmath22 it is also integrable under any @xmath162 . conditioning on @xmath6 , we can write the population risk of the logistic regression parameters @xmath32 with respect to sampling measure @xmath162 as @xmath163 by cauchy schwarz , the integrand in ( [ eqnpoprisklam2 ] ) is bounded by @xmath164 . if @xmath165 , then , we may appeal to dominated convergence and take limits with respect to @xmath32 and @xmath135 inside the integral . @xmath166 is strictly convex because the integrand is , and always has a unique population minimizer if there is no separating hyperplane in the population . [ lemma1 ] assume there is no @xmath167 for which @xmath168 henceforth , we refer to this assumption as _ nonseparability_. then @xmath166 attains a unique minimum for every @xmath169 . denote by @xmath170 the empirical risk on a local case - control subsample taken using the pilot estimate @xmath135 . then @xmath171.\ ] ] it will be somewhat simpler to replace the random subsample size @xmath172 with its expectation @xmath173 . define @xmath174.\ ] ] since minimizing ( [ eqnemprisk0 ] ) with respect to @xmath32 is equivalent to minimizing ( [ eqnemprisk1 ] ) , the two are equivalent for our purposes . if the unadjusted parameters @xmath175 minimize @xmath176 , the local case - control estimate @xmath177 is an @xmath178-estimator minimizing @xmath179 . we use analogous notation for the population version : @xmath180 for any given pilot estimate @xmath135 and large @xmath4 , we expect @xmath181 define the right - hand side of ( [ eqnargminqlam ] ) to be @xmath182 , the large - sample limit of local case - control sampling with pilot estimate fixed at @xmath135 . the best linear predictor for the original population corresponds to the case @xmath183 ( uniform subsampling ) , that is , @xmath184 . consistency means that for large @xmath4 , @xmath185 . recall that if the model is correctly specified with true parameters @xmath35 , then @xmath186 for _ any _ fixed pilot estimate @xmath135 . minimizing @xmath187 therefore yields a estimate . unfortunately , in the misspecified case @xmath188 . in this sense , local case - control sampling with the pilot @xmath135 held fixed is in general _ not _ consistent for @xmath0 . however , we see below that it is consistent if @xmath189 . [ propthetastarthetastar ] assume @xmath165 , that the classes are nonseparable , and that @xmath184 is the best linear predictor for the original measure @xmath22 . then @xmath190 in other words , if we could only choose our pilot perfectly , then the local case - control estimate would converge to @xmath0 as @xmath78 . proof of proposition [ propthetastarthetastar ] write @xmath191 . the population optimality criterion for lcc with pilot @xmath135 is @xmath192.\end{aligned}\ ] ] noting that @xmath193 , if we evaluate the above at @xmath194 , we obtain @xmath195 & = & \tfrac{1}{2}\mathbb{e } \bigl [ x \bigl(p(x ) \bigl(1-p^*(x)\bigr ) - \bigl(1-p(x)\bigr)p^*(x ) \bigr ) \bigr]\hspace*{-25pt } \\ & = & \tfrac{1}{2}\mathbb{e } \bigl [ x \bigl(y - p^*(x ) \bigr ) \bigr]\end{aligned}\ ] ] which is exactly half the population score ( [ eqnpopscore ] ) for the original population . since @xmath0 optimizes the risk for the original population , this value is 0 . there is an intuitive explanation of this result : in @xmath196 , the _ acceptance probabilities _ are @xmath197 if @xmath198 and @xmath199 if @xmath60 ; hence they play the same role as the _ pseudoresiduals _ @xmath200 did in the original measure @xmath22 . for example , the point @xmath42 would contribute @xmath201 to the gradient if we evaluated the full - sample score at @xmath0 . evaluating the subsample score at 0 , the same point now contributes @xmath202 to the score but only if it is accepted , which occurs with probability exactly @xmath203 . so , in essence , the subsampling stands in for the reweighting that we otherwise would have done when fitting our logistic regression to the full sample . of course , in practice we never have a perfect pilot if we did we would not need to estimate @xmath0but proposition [ propthetastarthetastar ] suggests that if @xmath135 is near @xmath0 , minimizing @xmath187 yields a good estimate . in fact , we will see that if @xmath204 then @xmath39 as well . for our asymptotic results , assume we have an infinite reservoir @xmath205 of i.i.d . pairs , a sequence of i.i.d . @xmath206 variables @xmath207 for making accept reject decisions , and a sequence of pilot estimates @xmath208 the @xmath209 are possibly dependent upon the data , but the @xmath210 are assumed to be independent of everything else . @xmath211 is the local case - control estimate , computed using pilot @xmath209 , data@xmath212 , and accept reject decisions @xmath213 . the main result of this section is that if the pilot estimate @xmath209 is consistent for @xmath0 , then so is @xmath211 . the details are somewhat technical , especially the proof of proposition [ prop_pointwise ] , but the main idea is that if @xmath214 , then for large @xmath4 @xmath215 in the appropriate sense . @xmath216 is what the local case - control estimate actually minimizes , whereas the last function is minimized by @xmath0 , our ultimate target . first , we establish pointwise convergence . [ prop_pointwise ] if @xmath217 and @xmath218 , then for each @xmath219 , @xmath220 because we avoid assuming independence between the pilot @xmath209 and the data @xmath12 , the proof is technical and is deferred to the . the proof relies on the coupling of the acceptance decisions @xmath52 for different pilot estimates through @xmath210 . with this coupling , two nearby pilot estimates will differ on very few accept reject decisions . because neither @xmath221 nor @xmath222 changes very fast , pointwise convergence also implies uniform convergence on compacts . [ propuniform ] if @xmath217 and @xmath223 , then for compact @xmath224 , @xmath225 define @xmath226 by proposition [ prop_pointwise ] , @xmath227 pointwise . next , we show it is lipschitz . the integrand in ( [ eqnpopoptqlam ] ) is @xmath26 times two factors each bounded by @xmath228 , hence @xmath229 similarly for @xmath216 , we have @xmath230 so that @xmath231 it follows that , with probability tending to 1 , @xmath232 has lipschitz constant less than @xmath233 . now , for any @xmath234 , we can cover @xmath235 with finitely many euclidean balls of radius @xmath236 , centered at @xmath237 . let @xmath238 be the event that @xmath239 has lipschitz constant less than @xmath18 and @xmath240 on @xmath238 , we have @xmath241 , and @xmath242 as @xmath243 . finally , we come to the main result of the section , in which we prove that the local case - control estimate is consistent when the pilot is . because the functions are strictly convex , we can ignore everything but a neighborhood of @xmath0 . [ thmconsistency ] assume @xmath244 and the classes are nonseparable . if @xmath214 then the local case - control estimate @xmath245 as well . let @xmath224 be any compact set with @xmath246 in its interior , and let @xmath247 where the strict inequality follows from strict convexity . uniform convergence implies that with probability tending to 1 , @xmath248 which implies in turn that @xmath249 whenever this is the case , the strictly convex function @xmath216 has a unique minimizer in the interior of @xmath235 . since @xmath235 was arbitrary , we can take its diameter to be less than any @xmath250 hence , @xmath251 . in this section , we derive the asymptotic distribution of the local case - control logistic regression estimate , in the same asymptotic regime as the previous section . to prove our results here , we assume the pilot estimate @xmath209 is independent of our data set . this would not be the case if our pilot were based on a subsample of the data ( the procedure we use for all our simulations ) , but it could hold if the pilot came from a model fitted to data from an earlier time period . the main result of this section is that if the logistic regression model is correctly specified and the pilot is consistent , the asymptotic covariance matrix of the local case - control estimate for @xmath32 is exactly twice the asymptotic covariance matrix of a logistic regression performed on the entire data set . for the results in this section , we will need @xmath252 . it will be convenient to give names to some recurring quantities . first , we have seen that if @xmath165 we can differentiate @xmath253 inside the integral to obtain the gradient of the population risk : @xmath254 whereas @xmath255 is the expectation of the logistic regression score with respect to @xmath162 , we can also define its covariance matrix : @xmath256.\ ] ] when @xmath257 , @xmath258 , and is continuous in @xmath32 and @xmath135 by dominated convergence . since the derivatives of the integrand in ( [ eqnintegralg ] ) are uniformly bounded by @xmath259 , dominated convergence implies we can again differentiate inside the integral . differentiating with respect to @xmath32 we obtain @xmath260 here , the integrand is dominated by @xmath261 , so dominated convergence again applies and thus we see that @xmath262 is continuous in @xmath32 and @xmath135 . @xmath263 for any @xmath264 since we have assumed there is no nonzero @xmath167 for which @xmath139 . finally , define the matrix of crossed partials : @xmath265 to be concrete , @xmath266 . continuity of @xmath267 again follows from noting the derivative of the integrand in ( [ eqnintegralg ] ) with respect to @xmath135 is dominated by @xmath268 . to begin , we consider the behavior of @xmath182 for @xmath135 near @xmath0 . by proposition [ propthetastarthetastar ] , we have @xmath269 . since @xmath263 , we can apply the implicit function theorem to the relation @xmath270 to obtain @xmath271 by standard m - estimator theory , if we fix @xmath135 and send @xmath78 the coefficients of a logistic regression performed on a sample of size @xmath272 from @xmath162 would be asymptotically normal with covariance matrix @xmath273 in light of this and the fact that @xmath274 , we might predict the following . [ thm_condl_var ] assume @xmath257 . if @xmath275 independently of the data , then @xmath276 with @xmath277 . again , we defer the proof to the . we can combine ( [ eqncondlvar ] ) with ( [ eqncondlbias ] ) to immediately obtain the following reassuring facts . [ corbiasvar ] assume @xmath257 and @xmath209 is a sequence of pilot estimators given independently of the data . then : if @xmath209 is @xmath96-consistent , so is @xmath211 . if @xmath209 is asymptotically unbiased , so is @xmath211 . if @xmath278 then @xmath279 with @xmath280 in ( [ eqnfullvar ] ) , we have suppressed the arguments of @xmath0 in @xmath281 and @xmath282 . the first term in ( [ eqnfullvar ] ) characterizes the contribution of conditional bias ( given @xmath209 ) to the overall variance , and the second is the contribution of conditional variance . in the special case where logistic regression model is correctly specified , we have the following . [ thmasyvar ] assume the logistic regression model is correct and let @xmath283 be the asymptotic variance of the mle for the full sample . then if @xmath257 and @xmath284 independently of the data , we have @xmath285 hence , although the size of a local case - control subsample is roughly @xmath286 , the variance of @xmath134 is the same as if we took a simple random sample of size @xmath287 from the full data set . in other words , each point sampled is worth about @xmath288 points sampled uniformly . proof of theorem [ thmasyvar ] if logistic regression is correctly specified for @xmath22 , it is also for @xmath162 , regardless of @xmath135 , so @xmath289 . furthermore , by standard maximum likelihood theory @xmath290 for each @xmath135 . therefore , ( [ eqncondlvar ] ) specializes to @xmath291 but @xmath292 \biggl[\frac{e^{\lambda'x}+e^{f(x)}}{(1+e^{\lambda ' x})(1+e^{f(x ) } ) } \biggr ] xx ' \,d\mathbb{p}(x)}.\ ] ] if @xmath293 and @xmath294 , then ( [ eqngeneralhessian ] ) simplifies to @xmath295 this result is surprisingly simple . no characterization like theorem [ thmasyvar ] is available for the case - control and weighted case - control estimates , whose variances are not simple scalar multiples of @xmath296 . we can offer a simple heuristic argument for theorem [ thmasyvar ] , similar to that of proposition [ propthetastarthetastar ] . in @xmath297 , the acceptance probability @xmath298 for an observation at @xmath26 is @xmath299 , and given that it is accepted it contributes @xmath300 to the observed information . in the full sample , a point at @xmath26 is always accepted but contributes less , @xmath301 , to the observed information . again , the sampling probability stands in for the reweighting we would have done in the full sample . if @xmath302 is very small , we are discarding most of the data instead of keeping all of it and assigning it a tiny weight in the fit . the practical meaning of theorem [ thmasyvar ] is that local case - control sampling is most advantageous when @xmath303 is small , that is , when @xmath7 is easy to predict throughout much of the covariate space . this can happen as a result of marginal or conditional imbalance , or both . standard case - control sampling can also improve our efficiency in the presence of marginal imbalance , but unlike local case - control sampling , it does not exploit conditional imbalance . hence , we would expect local case - control to outperform standard case - control most dramatically when the marginal imbalance is very high , as in the simulation of section [ subsecsim2 ] . for data - dependent pilots , the efficiency picture is somewhat more complicated . for example , @xmath182 is approximately a linear function of @xmath304 . thus , if @xmath135 is unbiased but correlated with the noise in the data , we might get more or less variance relative to ( [ eqnasyvar2 ] ) , depending on how this correlation interacts with @xmath267 . if the model is correctly specified , it less clear whether an adversarially chosen pilot can affect the efficiency . either way , we do not anticipate serious problems from nonindependence . to stress - test our results against violations of independence , we expressly use a data - dependent pilot for all of our experiments : namely , a weighted case - control sample with sample points allowed to be recycled for the second - stage fit . in section [ subseclargersample ] , we proposed increasing the size of the local case - control subsample by multiplying all the acceptance probabilities @xmath305 by a constant @xmath2 and assigning weight @xmath306 when @xmath307 . we analyze the asymptotic variance here as a function of @xmath18 . to simplify matters , suppose the model is correctly specified and @xmath135 is fixed at @xmath35 . the weighted log - likelihood for the subsample and its derivatives are then @xmath308 conditionally on @xmath26 , there is a @xmath309 chance @xmath310 and @xmath311 , where @xmath312 . similarly , there is a @xmath313 chance @xmath314 , and @xmath315 . we immediately obtain @xmath316 the expectation and variance of the score evaluated at 0 are @xmath317 and the expected hessian is @xmath318 we have derived @xmath319 for @xmath19 , we recover the factor of two from ( [ eqnasyvar2 ] ) , but , for example , @xmath17 we only pay 20% increased variance relative to the full sample . here , we compare our method to standard weighted and unweighted case - control sampling for two - class gaussian models like the one considered in section [ subsecccbias ] . the standard case - control estimates use a 5050 split between the two classes . we begin with a five - dimensional two - class gaussian simulation where the classes have different covariance matrices . if @xmath320 , then @xmath321\\[-8pt ] & & { } + \frac{1}{2}(x-\mu_0)'\sigma_0^{-1}(x-\mu_0 ) + \operatorname{const.}\nonumber\end{aligned}\ ] ] equation ( [ eqnqda ] ) is linear if @xmath322 , and quadratic otherwise , so if the two covariance matrices were the same the linear logistic model would be correctly specified . in this case the model is incorrectly specified , letting us compare the behavior of the different methods under model misspecification . take @xmath90 , @xmath323 , and @xmath324 . the covariance matrices are @xmath325 and @xmath326 . hence @xmath65 is additive , but with a nonzero quadratic term in @xmath327 . for our simulation , we first generate a large ( @xmath328 ) sample from the population described above . second , we obtain a pilot model using the weighted case - control method on @xmath329 data points . next , we take a local case - control of size 1000 using that pilot model . [ tabmse ] @@l c d1.4d2.6 c d1.3d2.6@@xmath330 + & & & & & & + lcc & & 0.0049 & ( 0.00031 ) & & 0.025 & ( 0.00059 ) + wcc & & 0.023 & ( 0.0022 ) & & 0.16 & ( 0.0038 ) + cc & & 0.15 & ( 0.0016 ) & & 0.043 & ( 0.00096 ) + + & & & & & & + lcc & & 0.0037 & ( 0.0083 ) & & 0.039 & ( 0.00045 ) + wcc & & 0.59 & ( 0.064 ) & & 1.7 & ( 0.017 ) + cc & & 0.06 & ( 0.042 ) & & 0.87 & ( 0.0086 ) + for comparison , we obtain standard case - control ( cc ) and weighted case - control ( wcc ) estimates . for the comparison estimators we do not use a sample of size 1000 again but rather use the total number of observations seen by the lcc model or the pilot model , roughly 2000 , so the lcc estimate must pay for its pilot sample . we repeat this entire procedure 1000 times . table [ tabmse ] shows the squared bias and variance of @xmath331 over the 1000 realizations for each of the three methods . as expected , we face a bias - variance tradeoff in choosing between the wcc and cc methods , whereas the lcc method improves substantially on the bias of cc and the variance of wcc . standard errors for both bias and variance are computed via bootstrapping the 1000 realizations . more surprising is the fact that lcc enjoys smaller bias than wcc and smaller variance than cc , dominating the other two methods on both measures . the improvement in variance over the cc estimate is likely due to the conditional imbalance present in the sample , while the improvement in bias over the wcc estimate may come from the fact that the methods are only unbiased asymptotically and the lcc estimate is closer to its asymptotic limiting behavior . next , we simulate a two - class gaussian model with each class having the same variance , so that the true log - odds function @xmath27 is linear . we also increase the dimension to 50 for this simulation . since the model is now correctly specified , all three methods are asymptotically unbiased . however , in this case we introduce more substantial conditional imbalance , to demonstrate the variance - reduction advantages of local case - control sampling in that setting . for this example , @xmath332 , @xmath333 , @xmath334 , and @xmath335 . we repeat the procedure from section [ subsecsim1 ] , now with @xmath336 . instead of generating a full sample , the full data set is implicit and we sample directly from @xmath63 . in this example , the difference between the methods is more dramatic . table [ tabmse ] shows the squared bias and variance of the three methods . here , local case - control enjoys substantially better bias than the other two methods , improving on cc more than twenty - fold . for the correct pilot model , @xmath337 is roughly 0.005 , so the local case - control subsample size is around @xmath338 . since the model is correctly specified , the variance is roughly twice that of logistic regression on the full sample of size @xmath4 . in other words , local case - control subsampling is roughly 100 times more efficient than uniform subsampling . asymptotically , all three methods are unbiased but it appears that lcc again enjoys a smaller bias in finite samples . relative to standard case - control sampling , local case - control sampling is especially well - suited for data sets with significant conditional imbalance , that is , data sets in which @xmath101 is easy to predict for most @xmath102 . one such application is spam filtering . to demonstrate the advantages of local case - control sampling and compare asymptotic predictions to actual performance , we test our method on the web spam data available on the libsvm website . ] and originally from @xcite . the data set contains 350,000 web pages , of which about 60% are labeled as `` web spam , '' that is , web pages designed to manipulate search engines rather than display legitimate content . this data set is marginally balanced , though as we will see the conditional imbalance is considerable . as features , we use frequency of the 99 unigrams that appeared in at least 200 documents , log - transformed with an offset so as to reduce skew in the features . in this data set , the downsampling ratio @xmath339 is around 10% , that is , when using a good pilot we will retain about 10% of the observations . since we only have a single data set , we use subsampling as a method to assess the sampling distribution of our estimators . in each of 100 replications , we begin by taking a uniform subsample of size @xmath340100,000 from the population of 350,000 documents . after obtaining 100 data sets of size @xmath340100,000 , we use the same procedure as we used in our two simulations with @xmath34110,000 . our asymptotic theory predicts that the variance of the local case - control sampling estimate of @xmath32 should be a little more than twice the variance using the full sample ( more because the model is misspecified and our pilot has some variance ) . because the full sample is close to marginally balanced , the standard case - control sampling methods should do about as well as a uniform subsample of size 20,000that is , they should have variance roughly 5 times that of the full sample . note that 20,000 is roughly twice the size of the local case - control sample , since we are counting the pilot sample against the local case - control method . if we had a readily available pilot model , as we would in many applications , it would be more relevant to give the cc and wcc methods access to only 10,000 data points , doubling their variance relative to the observed variance in this experiment . the theoretical predictions come reasonably close in this experiment , as shown in figure [ figwebspamcoefs ] . the horizontal axis indexes each of the 100 coefficients to be fit ( there are 99 covariates and an intercept ) , and the vertical axis gives the variance of each estimated coefficient , relative to the variance of the same coefficient in a model fitted to the full sample . variance for local case - control , @xmath342 variance for standard ) are reasonably close to the mark , though a bit optimistic . ] the magnitude of our improvement over standard case - control sampling is substantial here , but could be much larger in a data set with an even stronger signal . the key point is that standard case - control methods have no way to exploit conditional imbalance , so the more there is , the more local case - control dominates the other methods . we have shown that in imbalanced logistic regression , we can speed up computation by subsampling the data in a biased fashion and making a post - hoc correction to the coefficients estimated in the subsample . standard case - control sampling is one such scheme , but it has two main flaws : it has no way to exploit conditional imbalance , and when the model is misspecified it is inconsistent for the population risk minimizer . local case - control sampling generalizes standard case - control sampling to address both flaws , subsampling with a bias that is allowed to depend on both @xmath26 and @xmath100 . when the pilot is consistent , our estimate is consistent even under misspecification , and if the model is correct then local case - control sampling has exactly twice the asymptotic variance of logistic regression on the full data set . our simulations suggest that local case - control performs favorably in practice . in the , we motivated our inquiry by identifying four ways that computational gains can translate to statistical ones . specifically , we suggested that computational savings can : enable us to experiment with and prototype a variety of models , instead of trying only one or two , allow us to refit our models more often to adapt to changing conditions , allow for cross - validation , bagging , boosting , bootstrapping , or other computationally intensive statistical procedures or open the door to using more sophisticated statistical techniques on a compressed data set . it is relatively clear how our proposed method can help with points ( 1 ) and ( 2 ) . as for point ( 3 ) , faster fitting procedures can directly speed up straightforward resampling techniques like bootstrapping or cross - validation , possibly making them feasible at scales where they previously were not . we discuss in section [ subsecextens ] how it can also help with boosting . the basic method as we have described it above does not deliver on point ( 4 ) , because the pilot model and second - stage model are the same . however , an extension of our method can help , which we discuss below . there is no reason in principle why the pilot model must be linear , or belong to the same model class as the model we fit to the local case - control sample . we can use any pilot predictions @xmath343 in the sampling algorithm , and then model the log - odds in the subsample quite flexibly by a gam , kernel logistic regression , random forests or any other method so long as we can use offsets @xmath344 in the second - stage procedure . for example , we could use as our pilot fit a simple model with a few important variables explaining most of the response , and in the second stage estimate more complex models refining the first . formally , our theoretical results may not cover this use . suppose the second - stage model can be written as a logistic regression after some basis expansion . then consistency of the second - stage estimate requires either that the pilot be consistent ( the new variables contribute nothing to the population fit ) or that the second - stage model be correctly specified . if neither of these assumptions holds approximately , then our estimate could be biased though perhaps not as biased as case - control sampling , which is a special case of local case - control with an intercept - only pilot . if we are prototyping , guarantees of consistency may not be a high priority . if they are , then as with case - control sampling , we can repair the inconsistency of the local case - control estimate by using a horvitz thompson estimator with weights @xmath345 . this may come at a cost of some added variance . it would be interesting to examine the bias of local case - control and the variance of weighted local case - control in this more general problem setting . this work suggests extensions in several directions , described below . in some applications ( e.g. , diagnostic medical screening ) , a false negative may be more costly than a false positive , or . one of the implications of the discussion in section [ subsecccbias ] is that the bernoulli log - likelihood implicitly places most emphasis on approximating the log - odds well near the 0 ( 50% probability ) level curve , which may not be appropriate if the decision boundary relevant to our application is at 10% . in general , we would expect to obtain a better model in the large-@xmath4 limit if we target the decision boundary we care most about . in a sense , the reason that standard case - control sampling performed so badly in example [ ex2 ] of section [ subsecccbias ] is that it targeted a level curve of @xmath346 other than 50% . specifically , it targeted the level curve corresponding to 50% in the subsampling population for equal - sampled case - control sampling , which corresponds to the marginal @xmath347 level curve in the original population . what happened by accident in example [ ex2 ] need not always be one , and it would be interesting to generalize our procedure so as to target any chosen decision threshold . more generally still , our indifference point could depend on our features @xmath26in online advertising , for instance , some advertisers may be willing to pay more per click than others . in section [ subsecadvantages ] we suggested using offsets to obtain a complex second - stage fit . alternatively , we can obtain any fitted log - odds function @xmath348 for the sample and simply add it to the pilot @xmath349 to obtain an estimate for @xmath65 . this observation suggests the possibility of iteratively fitting a `` base model '' to the subsample , then adding it to @xmath349 to obtain a new pilot for the next iteration . indeed , that iterative algorithm is closely related to the adaboost algorithm of @xcite . even more similarly to adaboost , we could weight each point by @xmath13 instead of sampling it with that probability . @xcite show that the adaboost algorithm can be thought of as fitting a logistic regression model additive in base learners . in adaboost , the function @xmath350 simply records the number of classifiers @xmath351 classifying @xmath26 as belonging to class @xmath352 minus the number classifying it as class @xmath353 , and friedman et al . show that @xmath354 can be thought of as approximating the log - odds of @xmath355 given @xmath117 . the difference is that while adaboost weights the point @xmath12 by@xmath356 , the local case - control version would use weights @xmath357 operationally , this alternative weighting scheme limits the influence of `` outliers , '' that is , hard - to - classify points that can unduly drive the adaboost fit . in high - dimensional settings , lasso- or ridge - penalized logistic regressions are often preferable to standard logistic regression , the model considered here . one could use local case - control sampling with a regularized version of logistic regression , but our asymptotic results might need revisiting in such a case especially in a high - dimensional asymptotic regime [ @xmath358 or @xmath359 . since the high - dimensional setting is important in modern statistics and machine learning , this bears further investigation . one way of viewing the method is as a way of `` tilting '' the conditional distribution of @xmath7 by a linear function of @xmath6 in the natural parameter space so as to enrich our subsample for more informative observations . we could use similar tricks on other glms . for instance , suppose we are given a poisson variable with natural parameter @xmath360 . by sampling with acceptance probability proportional to @xmath361 , we obtain ( conditional on acceptance ) a poisson with natural parameter @xmath362 . since poisson variables with larger means carry more information , this could yield a substantial improvement over uniform subsampling.=1 if our data arise from a poisson glm with @xmath363 , we could generalize the local case - control scheme by sampling @xmath12 with probability proportional to @xmath364 , where the extra parameter @xmath365 guarantees that we always tilt the conditional mean of @xmath101 upward . similar generalizations may apply for multinomial logit and survival models . because @xmath166 is strictly convex , it is sufficient to show that @xmath366 as @xmath367 in any direction . assume w.l.o.g . there is some neighborhood @xmath368 for which @xmath369 @xmath370 is linear in its second argument , and is increasing for sufficiently large @xmath143 . thus , for large enough @xmath371 , the population risk for @xmath22 is @xmath372 @xmath373 for any @xmath135 , so ( [ eqncond ] ) holds for @xmath162 with the same @xmath374 ( but a different @xmath375 ) . thus , we can repeat the same argument with @xmath22 replaced by @xmath376 . fix @xmath32 and begin by writing @xmath377 let @xmath378 be the bernoulli selection decisions , generated by comparing mutually independent @xmath53 to the threshold @xmath379 . the @xmath378 are independent conditional on @xmath135 and the data . also , write @xmath380 , so that @xmath381 . by the cauchy schwarz inequality , we have @xmath382 now , for @xmath250 define @xmath383 . for @xmath384 , we have @xmath385 which is integrable by assumption . finally let @xmath386 denote an average taken over indices @xmath387 , that is , @xmath388 . then @xmath389 by continuity , @xmath390 . therefore , it suffices to show @xmath391 . because @xmath392 by the law of large numbers , it suffices equally well to show that @xmath393 . now fix @xmath234 and take @xmath394 large enough that @xmath395 . for @xmath396 we have @xmath397 with probability one the second term is eventually less than @xmath398 . further , for @xmath399 , we have @xmath400 now , write @xmath401 @xmath402 iff @xmath210 lies between @xmath403 and @xmath404 . hence , conditionally on @xmath209 and the data , the @xmath405 are mutually independent nonnegative random variables bounded by @xmath394 with means @xmath406 since @xmath407 . continuing , we have @xmath408 conditioning on @xmath135 and @xmath409 , the first term is a sum of independent zero - mean random variables that are bounded in absolute value by @xmath394 . by hoeffding s inequality , @xmath410.\ ] ] since this bound is deterministic , the same applies to the unconditional probability that @xmath411 is large . take @xmath412 . with probability tending to 1 , @xmath399 and the event in ( [ eqnhoeffding ] ) holds , in which case @xmath413 since @xmath414 was arbitrary , the proof is complete . by the mean value theorem , we have for each @xmath4 @xmath415 where @xmath416 is some convex combination of @xmath211 and @xmath417 . noting that the lhs is by definition 0 and rearranging , we obtain @xmath418 if we can show the first factor tends in probability to @xmath419 and the second tends in distribution to @xmath420 , then by slutsky s theorem we have the desired result . using the skorokhod construction define a joint probability space for @xmath209 such that @xmath421 . we will condition on the sequence @xmath209 and use a triangular array central limit theorem for the random variables @xmath422 because @xmath209 is independent of the data , @xmath423 for any @xmath27 . the triangular array clt applies since @xmath424 \\ & = & \mathbb{p}(z_{ni}=1{|}\lambda_n)\bar a ( \lambda_n)^{-2 } { \operatorname{var}}_{\lambda_n}\bigl ( \nabla_\theta\ell\bigl(\bar\theta(\lambda_n)- \lambda_n;x_{ni},y_{ni}\bigr)\bigr ) \\ & = & \bar a(\lambda_n)^{-1 } j\bigl(\bar\theta ( \lambda_n),\lambda_n\bigr ) \\ & \stackrel{\mathrm{a.s . } } { \rightarrow } & \bar a\bigl(\theta^ * \bigr)^{-1}j\bigl(\theta^*,\theta^*\bigr).\end{aligned}\ ] ] therefore , defining @xmath425 and @xmath426 , the clt tells us @xmath427 whenever @xmath428 , which we assumed occurs with probability 1 . by dominated convergence , we also have @xmath429 . next we turn to the hessian . we have @xmath245 by theorem [ thmconsistency ] , so @xmath430 as well . writing @xmath431 we need to show that @xmath432 note that @xmath433 , which is integrable ; hence @xmath434 . since @xmath339 is continuous and strictly positive , and @xmath435 , it suffices to show that @xmath436 note that @xmath437 , and define @xmath438 . following the structure of the proof of proposition [ prop_pointwise ] , take @xmath394 large enough that @xmath439 and truncate the @xmath440 : @xmath441 the second term is eventually less than @xmath398 . now , @xmath442 is small , because @xmath443 hence , by cauchy schwarz @xmath444 so on the event @xmath445 , we have @xmath446 finally , we can bound the first term exactly as we did in the proof of proposition [ prop_pointwise ] , defining @xmath447 . the same argument implies @xmath448 $ ] , so as @xmath78 we have with probability approaching 1 , @xmath449 so taking @xmath450 , the right - hand side is less than @xmath451 . the authors are grateful to jerome friedman for suggesting that we investigate the bias of case - control sampling , and to nike sun for helpful comments and suggestions .

we present a sample of observations of thermonuclear ( type - i ) x - ray bursts , selected for comparison with numerical models . provided are examples of four distinct cases of thermonuclear ignition : he - ignition in mixed h / he fuel ( case 1 of * ? ? ? * ) ; he - ignition in pure he fuel , following exhaustion of accreted h by steady burning ( case 2 ) ; ignition in ( almost ) pure he accumulated from an evolved donor in an ultracompact system ; and an example of a superburst , thought to arise from ignition of a layer of carbon fuel produced as a by - product of more frequent bursts . for regular bursts , we measured the recurrence time and calculated averaged burst profiles from _ rxte _ observations . we have also estimated the recurrence time for pairs of bursts , including those observed during a transient outburst modelled using a numerical ignition code . for each pair of bursts we list the burst properties including recurrence time , fluence and peak flux , the persistent flux level ( and inferred accretion rate ) as well as the ratio of persistent flux to fluence . in the accompanying material we provide a bolometric lightcurve for each burst , determined from time - resolved spectral analysis . along with the inferred or adopted parameters for each burst system , including distance , surface gravity , and redshift , these data are suggested as a suitable test cases for ignition models . stars : neutron x - rays : bursts astronomical databases : miscellaneous methods : numerical nuclear reactions

1703.07485

thermonuclear ( type - i ) bursts arise from unstable ignition of accreted fuel ( typically mixed hydrogen and helium ) on the surface of accreting neutron stars in low - mass binary systems . such events are of high priority for observers , as they provide information about the fuel composition , accretion rate , and even neutron star spin , mass and radius ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? a key component contributing to our understanding of the burning physics is numerical modelling of the complex series of nuclear reactions which trigger and power the bursts ( e.g. * ? ? ? * ) . to date , the degree to which predictions of numerical models have been compared in detail to observations is limited ; this paper is part of a wider effort to address this situation . the physical conditions and processes which broadly influence the burst behaviour are relatively well understood . the primary determinant is the local accretion rate @xmath0 , frequently expressed as a fraction of the eddington rate @xmath1 , at which the outwards force due to radiation pressure ( assuming spherical symmetry ) equals the surface gravity . the eddington rate is also the predicted threshold at which the burning is expected to stabilise ( e.g. * ? ? ? * ) , although observationally bursting behaviour appears to cease for many sources at a substantially lower level @xcite . for systems accreting a mix of hydrogen and helium at typical accretion rates ( @xmath2 ) , the fuel layer is hot enough that h will burn stably via the hot - cno cycle . this process is limited by @xmath3-decay , and will exhaust the h at the base of the layer in a time @xmath4 ( e.g. * ? ? ? * ) which depends only on the accreted h - fraction @xmath5 and the metallicity @xmath6 . thus , the composition of the fuel layer at ignition may vary , and a number of categories have been identified by numerical studies . if the recurrence time is short , the burst will ignite in a mixed h / he environment , leading to a `` case 1 '' @xcite or `` prompt mixed '' @xcite burst . at lower accretion rates , the recurrence time may exceed @xmath7 , leading to ignition in a pure - he layer , i.e. `` case 2 '' or `` delayed helium '' bursts . at even lower rates , it is thought that h - burning too will become unstable , leading to ignition via that mechanism ; but no unambiguous examples of such bursts have been observed . an additional ignition case is provided by neutron stars in ultracompact systems , which likely accrete ( almost ? ) pure he . here the ignition will always be in a h - poor environment , largely independently of the accretion rate . the nuclear energy generation during the burst is related to the average h - fraction in the fuel @xmath8 , with @xmath9 . the composition of the fuel may be inferred from the @xmath10 parameter , the ratio of burst energy to accretion energy over the burst recurrence time . for a neutron star of mass @xmath11 , radius @xmath12 , @xmath13 where @xmath14 is the gravitational redshift at the neutron star surface , given by @xmath15 . provided the accretion rate is proportional to the persistent flux @xmath16 , @xmath10 may be measured ( up to a factor corresponding to the ratio of the anisotropy of burst and persistent emission ) as @xmath17 where @xmath18 is the bolometric correction to the ( band - limited ) flux @xmath16 , @xmath19 is the ( regular ) burst recurrence time , and @xmath20 the burst bolometric fluence ( total energy ) . the composition of the fuel may also be deduced from the shape of the lightcurve . he burns during the burst via the triple-@xmath10 reaction , which proceeds on a much faster timescale than the hot - cno , rp- and @xmath21 reactions which burn h. the higher the he mass fraction , the faster the burning will proceed during the burst , reflecting in the burst rise and duration , as well as the timescale @xmath22 ( the ratio of the fluence to the peak flux ) . the numerical codes that have been used to compare to observations fall into three broad classes . the first class determines the ignition conditions given the accretion rate and fuel composition , and predicts the burst recurrence time and energetics . while this model has been developed to compare with the inferred atmospheric expansion during a burst as suggested by measurements of burst oscillations @xcite , it has also been used to compare with observations of bursts at low accretion rates @xcite . the primary disadvantage of such models is that they do not follow the time - dependent compositional structure of the atmosphere , which is significantly modified by the thermonuclear burning . the next class of models are one - zone time - dependent codes that simulate some fraction of the nuclear reaction network . such codes have been used to demonstrate the extent of the rp - process that powers mixed h / he bursts @xcite , as well as serving to probe the sensitivity of burst lightcurves to individual reaction rates @xcite . the current ( practical ) state of the art for burst modelling is in 1-d multi - zone models that track the full extent of the nuclear reaction networks . codes in this class include kepler @xcite ; the general relativistic hydrodynamics code agile , coupled with a nuclear reaction network @xcite , and mesa @xcite . 1-d multi - zone models have been used to make the most detailed comparisons of observations to models , most notably for gs 1826@xmath2324 @xcite . however , these comparisons are generally limited in the extent of the sources and the observational data utilised . these limitations , despite the extensive modeling capabilities ( and accumulated observational data ) motivate the present study . here we present a sample of observations of thermonuclear bursts , intended for comparison with numerical models . the objectives of the sample assembly are twofold : first , as test cases to quantify variations between different codes ( and hence intrinsic model uncertainties ) ; and second , as examples that may be used to refine the system parameters by direct comparison with individual codes . we deliberately take a pragmatic approach which incorporates constraints on system parameters obtained by comparison with particular numerical models ( of varying degrees of fidelity ) . while this approach introduces dependence of the system parameters on the specific models chosen for these analyses , this dependence is unimportant for either of the two objectives described above . for the code comparisons , the specific choice of input parameters does not matter , provided they are plausible ; and the parameters we have adopted are the best current parameter estimates for the systems studied . as for improving the system parameters via more detailed comparisons with individual models , the adopted system parameters provided here should be viewed as a suggested starting point . this paper is presented as follows . in section [ sources ] we describe the sources selected to make up the sample . in [ observations ] we describe the observational data from which the sample lightcurves and other parameters are drawn . in section [ results ] we briefly describe the inferred properties of the bursting sources , and compare the lightcurves in the sample . finally , in [ discussion ] we suggest how the assembled data may be applied and used to test numerical models . we selected four well - studied burst sources , which span the range of theoretical ignition cases identified observationally ( table [ tab : sources ] ) . we also selected a fourth source to serve as an example of a superburst , 4u 1636@xmath23536 , which is notable as one of only two such events observed at high sensitivity with _ rxte _ ( e.g. * ? ? ? we list in table [ tab : sources ] those quantities required to simulate thermonuclear bursts and compare to the observations , including the distance , mass fraction of hydrogen @xmath5 and of cno nuclei , @xmath24 , and the adopted redshift @xmath25 and surface gravity @xmath26 . we caution that in most cases these parameters are not precisely known , however for the purposes of deciding upon suitable values for simulation tests , we adopt values which are generally consistent with the observed burst properties and/or the neutron star population . the surface gravity and fuel composition parameters are key input to the models , while the redshift is used to transform the burst lightcurves ( and recurrence times ) from the neutron - star frame to the observer s frame ( see [ discussion ] ) . the source distance @xmath27 is required to convert model predictions to observed quantities , and is subject to specific uncertainties depending upon the method by which it is estimated , as described further in appendix [ distance ] . [ cols="<,^,^,^,^,^,^ , < " , ] + values in italics indicate there are no constraints specific for that source . + references : 1 . @xcite ; 2 . this work ; 3 . @xcite ; 4 . @xcite ; 5 . @xcite below we describe the basic observational properties of each source . this source , discovered as a new transient by _ ginga _ in 1988 september @xcite , has been x - ray bright and exhibiting bursts ever since . it is known as the `` clocked '' or `` textbook '' burster , due to its unusually regular and consistent bursts @xcite . the inferred accretion rate varies little on short timescales , and is within the range 513% @xmath28 @xcite . this pattern of regular bursting behaviour has been broken only recently , with the transition to weaker , irregular bursts accompanied by a soft spectral state , in 2014 june @xcite . by virtue of the typically consistent bursting behaviour , gs 1826@xmath2324 is also one of the few sources for which detailed efforts have been made to deduce the composition of the accreted fuel . however , these efforts have not yet resulted in unambiguous results . the measured @xmath10-values for the source , at @xmath29 , strongly suggest a high fraction of h in the burst fuel , and so we adopt the solar value , i.e. @xmath30 , for the accreted h fraction . the lack of variation in burst fluence ( i.e. total burst energy ) over a range in accretion rate was taken as evidence by @xcite that the fuel must be deficient in cno nuclei . otherwise , steady ( hot - cno ) h - burning between the bursts would signficantly change the fuel composition at ignition , as the burst recurrence time decreased from around 6 hr to 4 hr . a subsequent comparison of burst properties , including the detailed shape of the lightcurve , with time - dependent 1-d model predictions by kepler @xcite led instead to the conclusion that the accreted fuel must be approximately solar in composition , i.e. @xmath31 , at odds with the earlier analysis @xcite . since kepler includes substantially more physics than the code used in the earlier analysis , we adopt solar composition following the most recent study . despite the uncertainty in the fuel composition , the properties of the bursts from gs 1826@xmath2324 strongly suggest burning of mixed h / he fuel by unstable he ignition , i.e. `` case 1 '' of @xcite ( or alternatively , `` prompt mixed '' bursts in regime 3 of @xcite ) . this source was discovered as a new transient in 1996 with _ bepposax _ @xcite , and has been active every 23 years ever since . observations made during the 1998 outburst by _ rxte _ revealed 401 hz x - ray pulsations , making it the first accretion - powered millisecond pulsar @xcite . bright , radius - expansion bursts have been observed in almost every outburst , and the detection of burst oscillations also at 401 hz in bursts during the outburst of 2002 october confirmed the link between oscillation frequency and neutron - star spin @xcite . the transient outbursts typically reach a maximum inferred accretion rate of only @xmath32 , and the bursts have fast ( @xmath33 s ) rises indicative of he - rich fuel . the wide ( 2.01 hr ) orbit can accommodate a ( h - rich ) companion @xcite , and it is thought that the accreted h is burned away by @xmath3-limited hot cno burning prior to ignition . the @xmath3430 hr recurrence times , given the low accretion rates , are sufficient provided the metallicity , @xmath6 is not far below the solar value . this type of burst corresponds to `` case 2 '' of @xcite , or `` delayed helium '' ( regime 4 ) of @xcite . comparisons of the burst properties with the predictions of numerical ignition models indicate a range of possible compositions , with @xmath35 @xcite . the lowest plausible h - fraction in the accreted fuel was @xmath36 . located in the globular cluster ngc 6624 , 4u 1820@xmath23303 ( also known as sgr x-4 ) was first detected as a bursting source with the _ astronomical netherlands satellite _ ( ans ; * ? ? ? when in the bursting state , the source exhibits quasi - regular radius - expansion bursts with recurrence times in the range 24 hr ( e.g. * ? ? ? the source intensity is modulated on a @xmath37-d periodicity @xcite , and it only exhibits bursts in the ( relatively short ) low - intensity phase of the cycle , when the inferred accretion rate is @xmath380.95 @xmath39 ( e.g. * ? ? ? the orbital period of 685 s is known from periodic modulation of both persistent x - rays and the optical counterpart @xcite , and indicates an `` ultracompact '' binary which is too close for a roche - lobe filling main - sequence star . the mass donor is thus assumed to be a very low - mass he white dwarf , and thus accreting pure he ( i.e. @xmath40 ) . some evolutionary models predict a small ( @xmath41 ) amount of h in the surface layers , and this can not be ruled out from comparing the observed bursts with ignition models @xcite . the mass and radius of the neutron star in this system have been estimated by equating the blackbody normalisation in the cooling tail and the peak flux of pre bursts , and adopting the inferred cluster distance @xcite . although unresolved systematic errors may yet affect such measurements ( e.g. * ? ? ? * ) , these values represent the best current estimates for the source , and are adopted in table [ tab : sources ] . 4u 1636@xmath23536 is a persistently - accreting burst source in a 3.8 hr orbit @xcite . the system is one of the most prolific type - i ( thermonuclear ) bursters known , and has been studied extensively with most major observatories ( e.g. * ? ? ? * ; * ? ? ? the neutron star spin has been measured from burst oscillations and transient pulsations during a superburst , at 579.3 hz @xcite . the relatively wide orbit , spectral features from hydrogen @xcite , and the ( at times ) long profiles of the bursts strongly suggest the mass donor is a main - sequence star which accretes hydrogen - rich fuel . the system has exhibited long - term variations in its persistent intensity . between 1996 and 2001 , the x - ray flux was in the range 3@xmath42 , indicating an accretion rate of 1121% @xmath28 ( cf . with * ? ? ? * ) . since then , the flux has been consistently approximately 50% lower . at these accretion rates , it would be expected that bursts would consistently exhibit long profiles arising from ignition of mixed h / he fuel , although this is not always the case . 4u 1636@xmath23536 is also well - known for short - recurrence time bursts occuring in trains of up to 4 @xcite . the first superbursts were detected from _ rxte_/asm data @xcite , and two additional events were subsequently reported , one also detected by the pca @xcite . all the superbursts were detected while the system was in it s higher flux range , prior to 2001 . the 2001 event , on february 22 ( mjd 51962.7 ) , was observed with the pca instrument , and has offered the highest signal - to - noise data during such an event , comparable only to the event seen from 4u 1820@xmath23303 @xcite . preliminary spectral fitting of the 4u 1636@xmath23536 burst indicated that the spectrum could not be adequately fitted with blackbody models @xcite , and subsequently it was shown that the burst spectrum included substantial contributions arising from reflection of the burst emission from the accretion disk @xcite . we used preliminary results from the multi - instrument burst archive ( minbar ) , which includes bursts observed by the _ rossi x - ray timing explorer _ proportional counter array ( _ rxte_/pca ) , _ bepposax _ wide - field camera ( wfc ; * ? ? ? * ; * ? ? ? * ) and _ integral _ joint european x - ray monitor ( jem - x ; * ? ? ? we calculated burst lightcurves from time - resolved spectroscopic analysis of selected bursts observed by _ rxte_/pca @xcite . the minbar analysis follows the approach of @xcite , incorporating the latest pca responses and the effects of deadtime , as for the measured peak pre burst fluxes ( see appendix [ distance ] ) . earlier analyses of samples including these data have demonstrated evidence for an increased contribution of persistent flux during bursts , possibly arising from the effects of poynting - robertson drag on the inner accretion disk @xcite . while the fractional increase in the persistent flux can be @xmath43 , this contribution is a small fraction of the burst flux , and so any correction required to the burst flux is also small ( of order a few per cent , similar to the intrinsic uncertainty of the time - resolved flux measurements ) . thus , we neglect this effect for the purposes of determining our mean profiles . for gs 1826@xmath2324 and 4u 1820@xmath23303 , we identified burst sequences in minbar for which we could reliably infer the burst recurrence time , even when the observations were interrupted . such interruptions arise for instruments in low - earth orbit , due to earth occultations and passages through the south atlantic anomaly . an example burst train is shown in figure [ example ] . we then selected and averaged bursts observed by the _ rxte_/pca within each burst train . minbar also includes bursts observed by _ bepposax_/wfc and _ integral_/jem - x , which are useful for establishing the regularity of the bursts , but the sensitivity of these instruments is much lower than _ rxte _ , and so are unsuitable for producing high - quality burst lightcurves . for gs 1826@xmath2324 , we augmented our burst sample with optical bursts observed during _ rxte _ observations in 1998 june , as also analysed by @xcite . we measured the average recurrence time of each train by assigning a trial integer value to each burst , and then performing a linear least - squares fit to the times . the trial values were determined by dividing the time since the first burst , by the minimum burst separation of the train ; we subsequently adjusted the sequence numbers to minimise the rms value of the fit . we typically obtained an rms value of @xmath34 min or better . from the set of bursts in each train @xmath44 , we then selected the _ rxte _ bursts ( @xmath45 , table [ tab : bursts ] ) , and created average lightcurves from each train . we also calculated averaged burst fluences , peak fluxes and @xmath46 , where @xmath16 is the persistent ( accretion ) flux , @xmath18 the bolometric correction ( see [ bolcorr ] ) , and @xmath20 the burst fluence . we estimated the accretion rate @xmath0 as a fraction of the eddington rate @xmath28 , as @xmath47 where @xmath16 is the average persistent flux over the burst train , and @xmath27 the estimated source distance from table [ tab : sources ] . this estimate neglects the possible anisotropy of the persistent emission , so must be taken as a guide only . the identification of suitable burst trains for 4u 1820@xmath23303 was made difficult by the relatively small number of bursts available in the minbar sample . just 67 bursts were detected by _ bepposax_/wfc , _ integral_/jem - x or _ rxte_/pca , with only 16 events observed with _ rxte_. we ultimately identified eight burst trains with three or more bursts and evidence for regular bursting behaviour ; the range of inferred recurrence time was 2.74.4 hr . only two of these sequences included bursts detected with _ rxte _ , each with just three bursts . for each sequence , we observed a pair of bursts within a few hours of each other , and a third burst some days earlier or later . we ultimately rejected one of these sequences , in 2009 may ( mjd 54980 ; see also * ? ? ? * ) because there were significant variations in the persistent flux ( and spectral shape ) over the interval , and it was not possible to be confident about the recurrence time for the close pair of bursts , due to data gaps between them . this was not the case for the remaining train , in 1997 may ( mjd 50572 ) , and so we retained this set of bursts . with only one burst observed by _ rxte _ in this sequence , we simply provide the time - resolved spectroscopy results from that burst , as was done for sax j1808.4@xmath233658 ( see [ outburst ] ) due to the sparse data for 4u 1820@xmath23303 , we augmented the sample with a pair of bursts observed by _ rxte_/pca on 2009 june 12 , separated by only 1.89 hr . the nearest observations in time preceeding ( following ) that observation were 13 d before ( 3 d after ) , so any possible test for periodicity over that time range would likely be uninformative given the likely phase drift of the burst sequence . assuming that the separation of 1.89 hr does reflect a regular recurrence time , this pair would be the most frequent bursts ever observed from this system ( cf . with * ? ? ? the constraints on recurrence times for the superbursts from 4u 1636@xmath23536 is poor due to the low duty cycle of observations between the detected events . the superburst in our sample was preceded by another @xmath48 years prior . however , the inferred accretion column is approximately twice the ignition column , suggesting that the actual recurrence time is closer to @xmath49 years @xcite . in addition to the properties of the burst itself reported in table [ tab : bursts ] , we constrained the `` quench time '' corresponding to the period following the superburst during which normal thermonuclear burning is suppressed . the first thermonuclear ( h / he ) event detected from 4u 1636@xmath23536 following the superburst was on mjd 51985.5 ( _ bepposax _ observation 10898 ) , giving @xmath50 d. the observation during which that event was detected commenced 17.5 hr before the burst was detected , and had a duty cycle of @xmath51% , with interruptions due to the low - earth orbit . the coverages suggests a @xmath52% chance that an earlier burst could have been missed , so that the quench time could have been shorter by up to 17.5 hr . however , no intervening observations were made to rule out additional bursts missed earlier than that observation . we measured the persistent flux in the 325 kev band from _ rxte _ and _ bepposax _ observations by fitting commonly - used phenomenological models , and converted the integrated flux to a bolometric value by applying a bolometric correction for each observation . the correction was estimated by fitting to _ rxte _ pca and hexte data a broad - band model typically comprising an absorbed comptonisation component ( comptt in xspec ; * ? ? ? * ) , usually with a gaussian component representing fe k@xmath10 emission in the range 6.46.7 kev . we measured the 325 kev flux from each fit , and then created an ideal response ( using dummyrsp in xspec ) covering the energy range 0.11000 kev , and integrated the flux over this range , excluding the effects of neutral absorption . we estimated the bolometric correction as the ratio of the unabsorbed 0.11000 kev flux to the absorbed 325 kev flux ; for the observations here , the correction was typically in the range 1.42 . this approach is reliable provided that the contribution to the bolometric flux outside the range to which pca and hexte are sensitive ( typically 3100 kev ) is small ( or at least consistent ) . however , it is known that for gs 1826@xmath2324 , that additional low - energy contributions may arise at times , so that the persistent flux measured by _ rxte _ in the 325 kev band is not always a reliable estimator of the accretion rate ( e.g. * ? ? ? * ) . in some observations , there is evidence for unusually low @xmath10-values , suggesting that a significant fraction of the x - ray flux is emitted outside the _ rxte _ band ; contemporaneous _ chandra _ or _ xmm - newton _ observations suggest that this flux may be emitted as a low - temperature ( @xmath53 kev ) thermal component . thus , we avoided from our data selection bursts from such epochs identified by @xcite , including those from 2003 april , and also the observations from 1998 june and 1997 november . an additional epoch of unusually low @xmath54 is from observations in 2006 august , and was also excluded . the remaining set of three burst trains have broadly consistent @xmath10-values , but we caution that there may yet be additional systematic errors contributing to the determination of the accretion rate in this ( and possibly other ) systems . here we describe the approach by which the values of the neutron star radius , redshift , and surface gravity for gs 1826@xmath2324 were chosen . we replicated the analysis of @xcite with updated analysis results and an improved method to determine the best - fit system parameters . we selected the train of bursts observed in 2000 september , with a recurrence time of 4.177 hr , similar to that used previously . the 7 bursts observed with _ rxte _ are augmented by the observation of 25 additional bursts with _ bepposax _ over the same interval , with the complete train spanning 12 d. however , the entire sequence of bursts is not consistent with a steady recurrence time , likely due to small variations in accretion rate . thus , we restricted the sequence to a shorter interval covering the _ rxte _ bursts , beginning on mjd 51810 and spanning almost 4 d ( as plotted in fig . [ example ] ) . we then compared the averaged lightcurve ( calculated from the _ rxte _ bursts only ) with examples predicted by the kepler code , tabulated in @xcite . @xcite found excellent agreement between the observed lightcurve and a model referred to as `` a3 '' , with solar composition , an accretion rate of @xmath55 . in the sample of @xcite , the accretion rate is quantified in terms of the eddington value , defined as @xmath56 . thus , for model a3 , the accretion rate was @xmath57 , and this model corresponds to `` a05d '' in the @xcite sample . in our revised comparison , we simultaneously matched the lightcurve and the recurrence time , using a markov - chain monte carlo ( mcmc ) code emcee @xcite to marginalise over the three parameters of interest : the redshift @xmath25 ( by which the lightcurve will be stretched when transforming into the observer s frame ) , the relative intensity scale , including the distance and redshift : @xmath58 where @xmath59 is the burst luminosity predicted by kepler , @xmath27 is the distance , @xmath60 takes into account the possible anisotropy of the burst emission ( see also appendix [ distance ] , and @xmath61 the observed burst flux . a third parameter is related to the offset in time between the model and observation , and has no physical meaning . our initial runs with this code did not match the recurrence time , and to ensure that the model - predicted burst interval ( when scaled by @xmath25 ) matched that observed , we weighted this parameter in the likelihood function by a factor of 50 . this ensured in the subsequent comparisons that the redshifted model recurrence time matched that observed to within a few percent . the comparison with model a05d , as used by @xcite , gave good agreement with the observed lightcurves , although requiring a redshift of @xmath62 . following the approach of @xcite , we transformed the model results ( calculated in a newtonian frame ) to the observer s frame assuming that the gravity , and the newtonian and the general - relativistic neutron star masses are identical . this implies that the radius required for the newtonian and general - relativistic gravity will be different , and the corresponding radius ( for a neutron star mass of @xmath63 ) implied by our comparison of model a05d was 8.5 km , well below the lower limit of 10.4 km inferred for a wide range of neutron stars ( e.g. * ? ? ? * ) . thus , we made a limited exploration of alternative models , focussing on those with solar composition . the comparison with model a028 gave a best - fit redshift of @xmath64 , implying a radius of 12.1 km , well within the expected range . the accretion rate ( in units of the eddington rate ) for this model is 0.082 , which is somewhat higher than that inferred for the 2000 sep epoch , at 0.0692 ; but may be consistent taking into account the persistent emission anistropy . @xcite estimated the ratio of persistent to burst anisotropies at @xmath65 , implying @xmath66 ( e.g. * ? ? ? * ) ; in that case , the luminosity ( and hence accretion rate ) inferred from the persistent flux would be higher , bringing the observation estimate roughly in line with the value assumed for run a028 . the corresponding distance is @xmath67 kpc . we do not quote errors on these parameters , as the lightcurve comparison ( fig . [ fig : gslcompare ] ) is not of sufficient quality to have confidence in the posterior distributions ( the estimated probability densities for the model parameters ) . however , the comparison is of comparable quality as in @xcite , and so for the present purposes , we adopt the values for the redshift and distance . 24 , with kepler model a028 . the symbols with errors show the average lightcurve observed by _ rxte _ in 2000 september , at which time the recurrence time was 4.177 hr the model curve , rescaled on both axis based on the best - fit distance and redshift , is the red solid curve . the best - fitting distance and redshift , determined from an mcmc analysis , were @xmath68 kpc and @xmath69 , respectively . ] the approach adopted for sax j1808.4@xmath233658 was necessarily different , as this source has not exhibited regular bursts in previous observations . the persistent flux ( and hence accretion rate ) varied significantly over the time at which the bursts were observed in 2002 october , and so the burst separations ( and hence , presumably , the recurrence times ) vary significantly . furthermore , the burst separations are all longer than 12 hr , so that many data gaps are present between each burst , introduced by the interruptions in the visibility of the source due to the observing satellite s @xmath70-min low - earth orbit . @xcite compared the observed burst times with the predictions of a numerical ignition model , to constrain the system parameters . for the purposes of this paper , we have revised the analysis of those authors , to reconcile the changes in the _ rxte_/pca calibration since that work was completed , to correctly treat the persistent and burst anisotropy factors , and to use a more rigorous parameter space exploration to determine the best - fit system parameters and uncertainties . we matched the burst times , fluences and @xmath10-measurements to the settle code developed by @xcite using mcmc and marginalising over the accreted hydrogen fraction @xmath5 , cno metallicity @xmath24 , base flux @xmath71 , and scaling factors incorporating the source distance and emission anisotropy . the settle model predicts burst parameters in the neutron star frame , given an inferred accretion rate . necessarily , we adopt the average inferred accretion rate ( based on linear interpolation between each observation ) for each burst interval , as settle does not simulate varying @xmath0 . the redshift @xmath25 and ( newtonian ) surface gravity @xmath26 are fixed in the settle code , at the values listed in table [ tab : sources ] so the results are implicitly for those values . the best - fit parameters we report are dependent on the fidelity of the model , which does not take into account all of the underlying physics , such as the effects of thermal and compositional inertia . more detailed models , such as kepler @xcite can provide more rigorous modelling of the underlying physics , however are computationally expensive and so impractical for the monte carlo approach adopted here . however , kepler simulations to reproduce the bursts observed from sax j1808.4@xmath233658 during the 2002 outburst have been performed recently , demonstrating that the assumption of a constant accretion rate between bursts results in the recurrence time being underestimated , when compared to the scenario in which the accretion rate decreases between bursts ( johnston et al . 2017 , in preparation ) . to account for this bias , we corrected the recurrence time predictions of settle using correction factors determined by kepler . the correction factor depends on the @xmath0 gradient , and thus is different for each burst . we adopt correction factors for each of the seven predicted bursts of sax j1808.4@xmath233658 , which vary from @xmath72 1 for the first burst to @xmath72 1.2 for the final burst . we infer the persistent and burst anisotropy factors ( @xmath73 and @xmath60 respectively ) from the scaling factors and find @xmath73 = 1.05@xmath740.01 and @xmath60 = 0.91@xmath75 . this implies an inclination of 62@xmath741@xmath76 . note that the error on the inclination is statistical only , and does not include the contribution from uncertainties in the mass and radius , due to the fixed values adopted in the settle model . as with the analysis of @xcite , our maximum - likelihood solution predicts an additional pair of bursts between the first pair of observed bursts , on mjd 52562.41363 and 52564.30515 ( # 1 and 2 from * ? ? ? these events would have fallen in data gaps , which explains why they were not observed . we thus report the burst separations for the last two bursts observed during the 2002 october outburst as the recurrence time ; for the second burst , we estimate the recurrence time inferred from the burst train modelling based on the simulated time of the preceding event . we note that the inferred recurrence time of 16.55 hr for the second burst observed with _ rxte _ is comfortably within the overall range for the source , with the minimum separation of 12.6 hr set by a pair of bursts observed by _ bepposax _ @xcite . the derived ( or adopted ) properties for the sources of interest are listed in table [ tab : sources ] . the properties of the bursts selected for our sample are summarised in table [ tab : bursts ] . we plot the burst lightcurves for each of the sequences in figure [ fig : examples ] . the accompanying material . for this paper includes lightcurves ( averaged in cases where there are more than one _ rxte _ burst in the sequence ) for comparison with numerical model results . the burst lightcurves show clear differences related to the inferred composition and accretion rate . for 4u 1820@xmath23303 , the burst lightcurves are short , with durations ( defined as the interval over which the luminosity exceeds 10% of the maximum ) of 15 s or so . each of these bursts exhibits strong photospheric radius expansion , and the quality of the spectral fits during the radius expansion episodes are relatively poor ( reduced-@xmath77 values in the range 28 ) . as @xmath0 increases , the burst timescale increases slightly ( @xmath78 to @xmath79 s ) , which qualitatively supports the presence of a small amount of h in the accreted fuel . with smaller recurrence time , there is less time to burn the accreted h , and the mass fraction at ignition will be larger . the bursts from sax j1808.4@xmath233658 , although also thought to be powered primarily by he , have a notably different morphology , with an extended radius - expansion phase at roughly constant luminosity . the timescale variation for sax j1808.4@xmath233658 is in the opposite direction , with the burst timescale and duration becoming shorter at higher accretion rates . here the recurrence time is long enough that any accreted h has already been exhausted at the base , so that lower accretion rate allows a larger pile of he to accumulate . the resulting increased fluence thus requires a longer interval emitting at the eddington luminosity , with some of the additional energy likely exported as kinetic energy of outflowing material . the bursts from gs 1826@xmath2324 are different again , showing the much slower ( @xmath80 s ) rises and long rp - process powered tails characteristic of these mixed h / he bursts . the peak luminosity is significantly smaller than for the he - rich bursts , and durations are approximately a factor of 5 longer . the burst profiles show only very subtle variations from epoch to epoch , with the peak flux decreasing slightly , while the timescale increases , as the accretion rate increases . the superburst profile is principally distinguished from the shorter , h or h / he events by the orders - of - magnitude longer timescale , with a duration of more than 2 hr . lccccccccc & @xmath44 & burst & @xmath19 & @xmath81 & & @xmath0 & @xmath20 & @xmath82 & + epoch & ( @xmath45 ) & ids@xmath83 & ( hr ) & ( @xmath84 ) & @xmath18 & ( @xmath28 ) & ( @xmath85 ) & ( @xmath84 ) & @xmath10 + + 1998 jun & 6(1 ) & 7 & @xmath86 & @xmath87 & @xmath88 & 0.0513 & @xmath89 & @xmath90 & @xmath91 + 2000 sep & 11(7 ) & 1420 & @xmath92 & @xmath93 & @xmath94 & 0.0692 & @xmath95 & @xmath96 & @xmath97 + 2007 mar & 10(7 ) & 5865 & @xmath98 & @xmath99 & @xmath100 & 0.0796 & @xmath101 & @xmath102 & @xmath103 + + 2002 oct & 1(1 ) & 2 & @xmath104@xmath105 & 2.5412.298 & @xmath106 & 0.0472@xmath107 & @xmath108 & @xmath109 & @xmath110 + & 1(1 ) & 3 & 21.10 & 2.2981.946 & @xmath111 & 0.0432@xmath107 & @xmath112 & @xmath113 & @xmath114 + & 1(1 ) & 4 & 29.82 & 1.9461.826 & @xmath115 & 0.0384@xmath107 & @xmath116 & @xmath113 & @xmath117 + + 1997 may 4 & 3(1 ) & 1 & @xmath118 & @xmath119 & @xmath120 & 0.144 & @xmath121 & @xmath122 & @xmath123 + 2009 jun 12 & 2(2 ) & & @xmath124 & @xmath125 & 1.4981 & 0.226 & @xmath126 & @xmath127 & @xmath128 + + 2001 feb & 1(1 ) & & @xmath129 @xmath130 & @xmath131 & 1.5346 & 0.167 & @xmath132 @xmath133 & @xmath134 @xmath133 + + @xmath83 burst index number in the catalog of @xcite . + @xmath135 persistent flux in the energy range 325 kev + @xmath105 inferred from ignition model comparisons + @xmath107 calculated based on linear interpolation between observations falling within the burst interval + @xmath136 only two bursts detected , so may not have been part of a regular train + @xmath130 recurrence time upper limit of 1.75 yr based on the next earliest event , detected by the _ rxte_/asm @xcite + @xmath133 values taken from the `` optimal '' fits of @xcite 303 , sax j1808.4@xmath233658 , gs 1826@xmath2324 and 4u 1636@xmath23536 observed by the _ rossi x - ray timing explorer_. in each panel we plot the lightcurve at the indicated recurrence time , with flux converted to luminosity at the distances indicated in table [ tab : sources ] . the bursts contributing to each profile are listed in table [ tab : bursts ] ; where more than one profile is observed by _ rxte _ in the train , we show the average profile . note the change in @xmath137- and @xmath138-axis between the top panels ( 4u 1820@xmath23303 and sax j1808.4@xmath233658 ) and the lower two panels . ] we have assembled a sample of burst observations intended for comparisons with , and between , numerical models . comparisons of different model codes for the same test cases will allow estimates of the intrinsic uncertainty that affects burst simulations . these uncertainties may also depend on the particular type of burst being simulated ; that is , the variation in model predictions for some types of bursts may be larger than for others . for example , @xcite suggest that the simple analytic model adopted for the simulations of bursts from sax j1808.4@xmath233658offer comparable fidelity to more sophisticated models , including kepler . this is less true for simulations of h - rich bursts from gs 1826@xmath2324 observed at higher accretion rates , as the simple models do not include the effects of compositional or thermal inertia . it is also anticipated that detailed comparisons of the burst measurements here may allow more stringent constraints to be placed on the neutron star parameters , as has been demonstrated as a proof - of - principle in sections [ lcfit ] and [ outburst ] . the anticipated procedure to replicate each of the burst measurements in table [ tab : bursts ] is as follows 1 . set the surface gravity @xmath26 and neutron star radius @xmath12 to the values appropriate for the source of interest in table [ tab : sources ] 2 . set the composition in the accreted material , @xmath5 and @xmath24 3 . perform a model run with a constant accretion rate chosen from one of the entries for that source from table [ tab : bursts ] , and extract the burst luminosity as a function of time . re - scale the predicted luminosity appropriately for the redshift parameter @xmath25 for this source , listed in table [ sources ] , and also the burst anisotropy parameter @xmath60 , where known . for the second and third bursts from sax j1808.4@xmath233658 , the predicted recurrence time and burst fluence may be adjusted by the factor described in [ outburst ] to account for the effects of the declining accretion rate during each burst 5 . calculate from the simulated burst train the recurrence time @xmath139 , and from the burst lightcurve , the burst fluence , peak flux , @xmath10-value , and timescale 6 . compare these predictions with the observed lightcurve and the measured values in table [ tab : bursts ] . in the event that the model recurrence time is significantly longer ( shorter ) than the observation , the cause may be that the true accretion rate is different from the estimate due to the unknown degree of anisotropy @xmath73 of the persistent emission . in that case , the adopted accretion rate might be increased ( decreased ) , by roughly the ratio of the predicted and observed recurrence times , and the procedure repeated from step 3 onwards . one additional parameter that is commonly included in burst simulations is the heat flux from below the simulation region , commonly parameterised as a `` base flux '' @xmath71 . this quantity is not well constrained by observations , and furthermore depends on the accretion rate ( e.g. fig . 18 of * ? ? ? * ) ; most studies to date have adopted values in the range 0.10.15 mev nucleon@xmath140 . for the purposes of model - to - model comparisons we recommend adopting a value of 0.1 , but suggest that for the purposes of matching the observations , this parameter can be considered free within the range 0.10.8 mev nucleon@xmath140 , depending upon the choice of core neutrino emissivity model . the approach to replicate the superburst observation is necessarily different . modeling a superburst by accreting hydrogen / helium comes at great computational expense , because hundreds of hydrogen / helium flashes must be simulated to create the carbon fuel . furthermore , it remains challenging to reproduce the observationally inferred carbon mass fraction ( e.g. , * ? ? ? * ; * ? ? ? an alternative approach is to directly accrete a mixture of carbon and heavier elements ( e.g. , fe or ru ; * ? ? ? * ; * ? ? ? * ) . for the 2001 superburst from 4u 1636@xmath23536 , we recommend a fuel composition of @xmath141 @xmath142c and @xmath143 @xmath144fe @xcite . such simulations are able to describe the observed superburst light curves in detail , but generally predict larger ignition depths than observed @xcite . the ultimate goal of this study is to constrain the rates of particular nuclear reactions , or masses of particular isotopes , based on the burst observations . it has been shown that several reactions are likely to affect the shape of the burst lightcurve ( e.g. * ? ? ? similarly , nuclear masses have also been shown to have an effect @xcite . it is hoped that the carefully - assembled and calibrated data presented here offer the first step on the road towards precision nuclear physics from thermonuclear bursts . this work benefited from discussions at the international symposium on neutron stars in the multi - messenger era in may 2016 , and more broadly by support from the national science foundation under grant no . phy-1430152 ( jina center for the evolution of the elements ) . the authors are grateful for support received as part of the international team on nuclear reactions in superdense matter by the international space science institute in bern , switzerland . the data analysed here are derived from preliminary analysis results of the multi - instrument burst archive ( minbar ) , which is supported under the australian academy of science s scientific visits to europe program , and the australian research council s discovery projects and future fellowship funding schemes . lk is supported by nasa under award number nng06eo90a . 65 natexlab#1#1 , t. , van der hooft , f. , de jong , j. a. , van kerkwijk , m. h. , & van paradijs , j. 1998 , a&a , 332 , 561 , j. , cornelisse , r. , steeghs , d. , charles , p. a. , hynes , r. i. , obrien , k. , & strohmayer , t. e. 2006 , mnras , 373 , 1235 , d. & morgan , e. h. 1998 , nature , 394 , 346 , d. , morgan , e. h. , muno , m. p. , galloway , d. k. , wijnands , r. , van der klis , m. , & markwardt , c. b. 2003 , nature , 424 , 42 , j. , galloway , d. k. , in t zand , j. j. m. , tomsick , j. a. , barret , d. , chakrabarty , d. , frst , f. , boggs , s. e. , christensen , f. e. , craig , w. w. , hailey , c. j. , harrison , f. a. , romano , p. , stern , d. , & zhang , w. w. 2016 , apj , 818 , 135 , g. w. , li , f. k. , canizares , c. , hayakawa , s. , jernigan , g. , & lewin , w. h. g. 1977 , mnras , 179 , 651 , r. , in t zand , j. j. m. , verbunt , f. , kuulkers , e. , heise , j. , den hartog , p. r. , cocchi , m. , natalucci , l. , bazzano , a. , & ubertini , p. 2003 , a&a , 405 , 1033 , a. 2003 , apj , 595 , 1077 , a. & bildsten , l. 2000 , apj , 544 , 453 . 2001 , apjl , 559 , l127 , a. , macbeth , j. , zand , j. j. m. i. , & page , d. 2006 , apj , 646 , 429 , r. h. , amthor , a. m. , heger , a. , johnson , e. , keek , l. , meisel , z. , schatz , h. , & smith , k. 2016 , apj , 830 , 55 , j. l. , grres , j. , wiescher , m. , & davids , b. 2006 , apj , 650 , 332 foreman - mackey , d. , hogg , d. w. , lang , d. , & goodman , j. 2013 , publications of the astronomical society of the pacific , 125 , pp . 306 , m. y. 1988 , apj , 324 , 995 , m. y. , hanawa , t. , & miyaji , s. 1981 , apj , 247 , 267 , d. k. & cumming , a. 2006 , apj , 652 , 559 , d. k. , cumming , a. , kuulkers , e. , bildsten , l. , chakrabarty , d. , & rothschild , r. e. 2004 , apj , 601 , 466 , d. k. , muno , m. p. , hartman , j. m. , psaltis , d. , & chakrabarty , d. 2008 , apjs , 179 , 360 , j. , gursky , h. , schnopper , h. , parsignault , d. r. , heise , j. , brinkman , a. c. , & schrijver , j. 1976 , apjl , 205 , l127 , c .- c . & keek , l. 2016 , apj , 819 , 47 , a. , cumming , a. , galloway , d. k. , & woosley , s. e. 2007 , apjl , 671 , l141 , j. j. m. , heise , j. , muller , j. m. , bazzano , a. , cocchi , m. , natalucci , l. , & ubertini , p. 1998 , a&a , 331 , l25 , j. j. m. , verbunt , f. , heise , j. , bazzano , a. , cocchi , m. , cornelisse , r. , kuulkers , e. , natalucci , l. , & ubertini , p. 2004 , in proceedings of the 2nd bepposax conference : `` the restless high - energy universe '' , amsterdam , 59 may 2003 , ed . j. van den heuvel , r. a. m. j. wijers , & j. j. m. in t zand , vol . 132 , 486495 , j. j. m. , homan , j. , keek , l. , & palmer , d. m. 2012 , a&a , 547 , a47 , r. , mels , w. a. , brinkman , a. c. , galama , m. y. , goulooze , h. , heise , j. , lowes , p. , muller , j. m. , naber , a. , rook , a. , schuurhof , r. , schuurmans , j. j. , & wiersma , g. 1997 , a&as , 125 , 557 , k. , swank , j. h. , giles , a. b. , stark , m. j. , strohmayer , t. , zhang , w. , & morgan , e. h. 1996 , proc . spie , 2808 , 59 , l. , ballantyne , d. r. , kuulkers , e. , & strohmayer , t. e. 2014 , apj , 789 , 121 , l. , cumming , a. , wolf , z. , ballantyne , d. r. , suleimanov , v. f. , kuulkers , e. , & strohmayer , t. e. 2015 , mnras , 454 , 3559 , l. , galloway , d. k. , int zand , j. j. m. , & heger , a. 2010 , apj , 718 , 292 , l. & heger , a. 2011 , apj , 743 , 189 . 2016 , mnras , 456 , l11 , a. r. & watson , m. g. 1986 , nature , 323 , 105 , e. 2009 , the astronomer s telegram , 2140 , e. , den hartog , p. r. , in t zand , j. j. m. , verbunt , f. w. m. , harris , w. e. , & cocchi , m. 2003 , a&a , 399 , 663 , e. , in t zand , j. , homan , j. , van straaten , s. , altamirano , d. , & van der klis , m. 2004 , in x - ray timing 2003 : rossi and beyond , ed . p. kaaret , f. k. lamb , & j. h. swank , vol . 714 ( melville , ny : aip ) , 257260 , n. , heger , a. , & galloway , d. k. 2016 , apj , 819 , 46 , w. h. g. , van paradijs , j. , & taam , r. e. 1993 , ssr , 62 , 223 , n. , budtz - jrgensen , c. , westergaard , n. j. , brandt , s. , rasmussen , i. l. , hornstrup , a. , oxborrow , c. a. , chenevez , j. , jensen , p. a. , laursen , s. , andersen , k. h. , mogensen , p. b. , rasmussen , i. , om , k. , pedersen , s. m. , polny , j. , andersson , h. , andersson , t. , kmrinen , v. , vilhu , o. , huovelin , j. , maisala , s. , morawski , m. , juchnikowski , g. , costa , e. , feroci , m. , rubini , a. , rapisarda , m. , morelli , e. , carassiti , v. , frontera , f. , pelliciari , c. , loffredo , g. , martnez nez , s. , reglero , v. , velasco , t. , larsson , s. , svensson , r. , zdziarski , a. a. , castro - tirado , a. , attina , p. , goria , m. , giulianelli , g. , cordero , f. , rezazad , m. , schmidt , m. , carli , r. , gomez , c. , jensen , p. l. , sarri , g. , tiemon , a. , orr , a. , much , r. , kretschmar , p. , & schnopper , h. w. 2003 , a&a , 411 , l231 , r. & heyl , j. s. 2003 , apj , 599 , 419 , f. , psaltis , d. , gver , t. , baym , g. , heinke , c. , & guillot , s. 2016 , apj , 820 , 28 , d. , kaaret , p. , & corbel , s. 2008 , apj , 688 , 1288 , b. , marchant , p. , schwab , j. , bauer , e. b. , bildsten , l. , cantiello , m. , dessart , l. , farmer , r. , hu , h. , langer , n. , townsend , r. h. d. , townsley , d. m. , & timmes , f. x. 2015 , apjs , 220 , 15 , w. & terrell , j. 1984 , apjl , 284 , l17 , h. , aprahamian , a. , barnard , v. , bildsten , l. , cumming , a. , ouellette , m. , rauscher , t. , thielemann , f .- k . , & wiescher , m. 2001 , physical review letters , 86 , 3471 , h. , bildsten , l. , & cumming , a. 2003 , apjl , 583 , l87 , h. & ong , w .- j . 2016 , arxiv e - prints 1610.07596 , a. w. , lattimer , j. m. , & brown , e. f. 2013 , apjl , 765 , l5 , l. , white , n. e. , & priedhorsky , w. 1987 , apjl , 312 , l17 , j. , brown , e. f. , cumming , a. , cyburt , r. , & schatz , h. 2014 , apj , 791 , 106 , t. & bildsten , l. 2006 , in compact stellar x - ray sources , ed . w. h. g. lewin & m. van der klis ( cambridge university press ) , ( astro ph/0301544 ) , t. e. & brown , e. f. 2002 , apj , 566 , 1045 , t. e. & markwardt , c. b. 2002 , apj , 577 , 337 , y. 1989 , in proc . 23rd eslab symposium on two topics in x - ray astronomy , bologna , italy , 13 - 20 september , ed . j. hunt & b. battrick no . sp-296 ( estec , noordwijk , the netherlands : esa ) , 313 , t. w. j. , galloway , d. k. , rothschild , r. e. , & homer , l. 2008 , apj , 681 , 506 , l. 1994 , apj , 434 , 570 , p. , bazzano , a. , cocchi , m. , natalucci , l. , heise , j. , muller , j. m. , & in t zand , j. j. m. 1999 , apjl , 514 , l27 , j. , van der klis , m. , van amerongen , s. , pedersen , h. , smale , a. p. , mukai , k. , schoembs , r. , haefner , r. , pfeiffer , m. , & lewin , w. h. g. 1990 , a&a , 234 , 181 , r. 2001 , apjl , 554 , l59 , r. & van der klis , m. 1998 , nature , 394 , 344 , s. e. , heger , a. , cumming , a. , hoffman , r. d. , pruet , j. , rauscher , t. , fisker , j. l. , schatz , h. , brown , b. a. , & wiescher , m. 2004 , apjs , 151 , 75 , h. , galloway , d. k. , & price , d. j. 2013 , apj , 772 , 94 . 2015 , apj , 801 , 60 , m. , cumming , a. , & galloway , d. k. 2012 , apj , 749 , 69 , w. , lapidus , i. , swank , j. h. , white , n. e. , & titarchuk , l. 1997 , iauc , 6541 an important source of systematic uncertainty in the determination of persistent and burst luminosities is the source distance @xmath27 , and in combination , the degree of anisotropy of the burst and persistent emission , parameterised as @xmath60 and @xmath73 , respectively . anisotropy of the burst ( and persistent ) flux is expected given the strongly non - spherically symmetric geometry introduced by the accretion disk and binary companion . while the distances to burst sources have been estimated with moderate precision from observations of radius - expansion bursts ( e.g. * ? ? ? * ) , these measurements implicitly also depend on the burst anisotropy . strictly speaking , distances estimated in such a manner correspond to @xmath145 , and in general it is not possible to separate these quantities to derive the physical distance @xmath27 , independently . the definition of the anisotropy factor @xmath60 follows that of @xcite , such that for a total burst luminosity @xmath146 , the measured burst flux would be given by @xmath147 from the above relation , we understand that @xmath148 corresponds to isotropic emission in all directions , while @xmath149 indicates that the burst flux toward the observer is _ in excess _ of the isotropic value , i.e. preferential beaming in our direction . as pointed out by @xcite , the values of @xmath60 and @xmath73 depend upon the system inclination and the structure of the accretion disk , neither of which are well constrained . for 4u 1636@xmath23536 , we estimated the distance based on measurements of the burst emission from photospheric radius - expansion ( pre ) bursts in minbar , assuming that they reach the eddington limit . this follows the approach of @xcite , but the burst flux measurements have been updated with the latest pca response matrices , and the effects of deadtime in the burst spectra have been included . the overall effect on the distances ( compared to the previous analyses ) is a reduction of 510% . we also incorporated the assumed gravitational redshift at the surface , as listed in table [ tab : sources ] . the anisotropy of burst emission for 4u 1636@xmath23536 can be estimated from the inclination of the binary system with respect to the observer s line of sight ( e.g. , * ? ? ? optical observations constrain the inclination angle to the range @xmath150 @xcite , which implies @xmath151 for a thin disk ( larger values for smaller angles ; * ? ? ? the observation of a broad fluorescent iron line favors a large inclination angle @xcite , but this depends strongly on the interpretation of the reflection spectrum . similarly , different constraints can be derived from the reflection signal during the superburst from this source @xcite . therefore , an uncertainty of several tens of percents must be taken into account for the burst luminosity of 4u 1636@xmath23536 . @xcite derived an upper limit on the distance ( combined with the burst anisotropy ) for gs 1826@xmath2324 of @xmath152 kpc . while our lightcurve comparisons for this source do not take into account the blackbody normalisation , as was analysed by @xcite , our derived system constraints are otherwise broadly consistent with those authors . however , our best - fit value based on the comparison with model a028 of @xcite is 6.1 kpc , slightly higher than the inferred limit . one additional consistency check that we can make is to compare the expected eddington flux for the adopted system parameters , with that measured from the first photospheric radius - expansion burst from the source , observed in 2014 june with _ nustar _ @xcite . assuming that the degree of burst anisotropy @xmath60 does not change during radius expansion , and that the peak flux is achieved at touchdown ( when the expanding photosphere has returned to the neutron star surface ) , the expected flux would be ( e.g. * ? ? ? * ) @xmath153 . this is well within the confidence interval of @xmath154 derived for the radius - expansion burst observed by @xcite . in the absence of a more detailed comparison with kepler simulations , which would allow us to incorporate the results from all three burst trains presented in table [ tab : bursts ] , we adopt the current value , but note that the true value may be lower

let be a graph chosen uniformly at random from the family of all labeled planar graphs with vertices and edges . in the paper we study the component structure of . combining counting arguments with analytic techniques , we show that there are two critical periods in the evolution of . the first one , of width , is analogous to the phase transition observed in the standard random graph models and takes place for , when the largest complex component is formed . then , for , when the complex components cover nearly all vertices , the second critical period of width occurs . starting from that moment increasing of mostly affects the density of the complex components , not its size .

1006.0444

since the seminal work of tutte @xcite maps and graphs on 2-dimensional surfaces have become widely studied combinatorial objects in discrete mathematics . the enumerative and structural problems around maps , i.e. _ embedded _ graphs on a surface , are relatively well settled . starting from the number of planar maps computed by tutte @xcite , the number of rooted maps on surfaces was found by bender , canfield , and richmond @xcite and other classes of maps were extensively enumerated since then . such enumeration results have been used to study typical properties of random maps on surfaces , e.g. the size of the largest components by banderier _ et al . _ @xcite . another important aspect of maps is that they allow nice bijections to the so - called _ well - labeled _ trees . the bijection between planar maps and the well - labeled trees was first studied amongst others by schaeffer @xcite , which was extended by bouttier , di francesco , and guitter @xcite to maps on an orientable surface with positive genus . these bijections are the corner stone of profound results on the topological structure of scaling limits of random maps by chassaing and schaeffer @xcite , le gall @xcite , and schramm @xcite . on the other hand , analogous problems related graphs that are _ embeddable _ on a surface are still wide open . the enumerative properties of random planar graphs have attracted much attention since the work of mcdiarmid , steger , and welsh @xcite who studied random labeled planar graphs with a given number of vertices . let @xmath7 be the number of labeled planar graphs on @xmath1 vertices . mcdiarmid , steger , and welsh @xcite showed amongst other results that the quantity @xmath8 converges to a limit @xmath9 as @xmath10 , which is called the growth constant . an upper bound @xmath11 , based on the triangulations and probabilistic methods , was obtained by osthus , prmel , and taraz @xcite . a lower bound @xmath12 was given by bender , gao , and wormald @xcite who studied the number of labeled 2-connected planar graphs through the singularity analysis of generating functions arising from the decomposition of graphs along connectivity . using a similar method gimnez and noy @xcite proved that @xmath13 , where @xmath14 are explicitly computable constants ( e.g. @xmath15 ) . as for the number @xmath16 of graphs embeddable on a surface with positive genus , mcdiarmid @xcite showed that its growth constant is the same as that of planar graphs while chapuy _ et al . _ @xcite found @xmath16 for all orientable surfaces of positive genus @xmath17 proving that @xmath18 , where @xmath19 and @xmath9 is the growth constant of the labeled planar graphs . mcdiarmid and reed @xcite studied various typical properties of random graphs on surfaces , e.g. subgraph containment and maximum degree . frieze @xcite asked about the asymptotic behavior of the number @xmath20 of labeled planar graphs on @xmath1 vertices with @xmath2 edges . note first that if @xmath21 for some @xmath22 , then a typical graph with @xmath1 vertices and @xmath2 edges is planar , i.e. @xmath23 ( cf . @xcite or @xcite ) gerke _ et al . _ @xcite proved the existence of its growth constant , in the sense that for @xmath24 with @xmath25 , the quantity @xmath26 converges to a limit @xmath27 as @xmath10 . the asymptotic formula for @xmath24 was found by gimnez and noy @xcite who showed that for @xmath28 , there are analytic constants @xmath29 such that @xmath30 . in this paper we deal with the case when @xmath31 . note that for such @xmath32 @xmath33 i.e. , @xmath34 , and a ` naive ' generating function approach does not lead to the asymptotic formula for @xmath20 . we use @xmath20 to study the asymptotic behavior of the uniform random planar graph @xmath0 by which we mean a graph chosen uniformly at random among all labeled planar graphs with @xmath1 vertices and @xmath2 edges ; thus , each of these graphs occurs as @xmath0 with probability @xmath35 . from the results in @xcite it follows that if @xmath36 is bounded away from both @xmath37 and @xmath38 , then @xmath0 has a well ordered structure , for instance , it has a large component of size @xmath39 , and all planar graphs of finite size appear as its subgraphs . thus , it corresponds to late stages of the evolution of the standard uniform random graph @xmath40 , the graph chosen uniformly at random among all graphs with @xmath1 vertices and @xmath2 edges . our goal is to study the typical size and structure of components in @xmath0 in a more interesting range , when @xmath41 . it turns out that , somewhat surprisingly , @xmath0 exhibits two critical ranges , which occur at @xmath4 and @xmath5 . the first critical period corresponds to a phase transition phenomenon observed in the plethora of different random graph models . let us recall some results on one of the most widely used random graph model @xmath40 . it follows from the papers of erds and rnyi @xcite , bollobs @xcite , uczak @xcite , uczak , pittel , and wierman @xcite , janson _ et al . _ @xcite , and janson @xcite ( see also janson , uczak , and ruciski @xcite ) that the giant component ( i.e. the unique largest component ) suddenly emerges at @xmath4 , and nowadays this spectacular phenomenon is well studied and understood . if @xmath42 and @xmath43 , then , a.a.s . ( i.e. with probability tending to 1 as @xmath1 approaches @xmath44 ) @xmath40 consists of isolated trees and unicyclic components , and the largest component is a tree of size @xmath45 . on the other hand , if @xmath46 , then a.a.s . @xmath40 contains exactly one component with more edges than vertices of size @xmath47 , while all other components are of size @xmath48 . furthermore , if @xmath49 , then a.a.s . @xmath40 contains a topological copy of @xmath50 and thus it is not planar , while , as we have mentioned , for @xmath51 , a.a.s . @xmath40 consists of isolated trees and unicyclic components , so it is clearly planar . another random structure relevant to the behavior of @xmath0 is the uniform random forest @xmath52 ( i.e. a forest chosen uniformly at random among all labeled forest with @xmath1 vertices and @xmath2 edges ) . uczak and pittel @xcite found that although the giant component in @xmath52 emerges at @xmath4 , as for @xmath40 , the critical behavior of these two models are somewhat different . let @xmath42 . if @xmath51 , then the structure of both @xmath52 and @xmath40 are similar ; in particular , the size of the largest tree in @xmath52 is a.a.s . however in the supercritical phase , when @xmath49 , the giant tree of @xmath52 is a.a.s . of size @xmath53 , which is roughly half of the size of the largest component of @xmath40 , while the second largest tree of @xmath52 is of size @xmath3 which does not depend much on @xmath54 provided @xmath55 , i.e. it is by far larger than the second largest component of @xmath40 for @xmath56 , which is of size @xmath57 . in the paper we show that as far as @xmath42 , where @xmath49 and @xmath58 is bounded away from @xmath59 , the behavior of @xmath0 is similar to that of @xmath52 . namely , a.a.s . the size of the largest complex component is of the order @xmath53 , while the second largest component has @xmath3 vertices . however , unlike in the case of @xmath52 for which @xmath60 , for @xmath0 we may have @xmath61 , so the rate of growth of the size of complex components must change at some point . we prove that it occurs when @xmath5 , more precisely , if @xmath62 and @xmath63 , then the complex components of @xmath0 have a.a.s . @xmath64 vertices altogether , while for @xmath65 they contain @xmath66 vertices for some computable constant @xmath67 . let us mention that the condition @xmath68 is a result of the proof method we have used and most likely can be replaced by @xmath69 . furthermore , our method can say quite a lot about structure of the largest component very much in the spirit of uczak @xcite . the rest of the paper is organized as follows . in the next section we describe the main idea of our argument . then , in section [ sec : cubicplanar ] we present the first analytic ingredient of the proof : counting specially weighted cubic planar multigraphs using generating functions . here we also describe how to use this result to bound the number of planar multigraphs with minimum degree three . then , in the next section , we estimate the number of planar graph with @xmath70 vertices and @xmath71 edges in which each component has more edges than vertices . finally , in the main chapter of this paper , we use a direct counting to study the number @xmath20 and the asymptotic properties of @xmath0 for different values of @xmath2 . as we have already mentioned most of results concerning the asymptotic behavior of @xmath20 is based on the generating function method . thus , for @xmath72 , one can study the function @xmath73 and deduce the asymptotic behavior of @xmath24 from the behavior of @xmath74 near its singularities . note however that since @xmath75 for @xmath76 , the coefficients in the expansion of @xmath74 tend to zero too fast to be handled by standard methods of generating function analysis . on the other hand , the condition that a graph is planar is very hard to grasp by purely combinatorial means . thus , in the paper we use a combination of analytic and combinatorial tools . from a planar graph we extract its kernel , which is its only part responsible for the planarity and is dense enough to be treated by generating functions method . then we use a technically involved but rather natural counting argument to find asymptotic properties of @xmath0 . in order to make the above description precise , we introduce some definition . the _ excess _ @xmath77 of a graph @xmath78 is the difference between the number of its edges and the number of its vertices . we call components of a graph with positive excess , i.e.those which have at least two cycles , _ complex _ components , and we say that the graph is complex if all its components are complex . the core of a graph @xmath78 , denoted @xmath79 , is the maximal subgraph of @xmath78 with the minimum degree two . the kernel of @xmath78 , denoted by @xmath80 , is obtained from the core by removing all isolated cycles and replacing each path whose internal vertices are all of degree two by an edge . note that for a graph @xmath78 the kernel is a multigraph which can have multiple edges and loops . however , the excess of both @xmath79 and @xmath80 are the same . note also that @xmath80 has clearly the minimum degree three . if @xmath80 is cubic , we say that @xmath78 is _ clean_. we define the _ deficiency _ @xmath81 of @xmath78 as the sum of degrees of vertices of @xmath80 minus three times the number of vertices of @xmath80 . therefore , a graph is clean if and only if its deficiency is zero . our argument is based on a simple observation that @xmath78 is planar if and only if @xmath80 is planar . we use analytic methods to count the number of possible candidates for the kernel , and then follow purely combinatorial argument . to this end , in section [ sec : cubicplanar ] we apply the singularity analysis to extract the asymptotic number of possible clean kernels of the planar graphs ( theorem [ thm : cubic - asymptotics ] ) . the proof is similar to that used in the case of counting cubic planar graphs presented in bodirsky _ et al . _ @xcite , but in order to find the number of all planar graphs , we need to count cubic planar multigraphs with some special weight function dependent on the number of loops and multiple edges of a multigraph . then , we use a simple combinatorial idea to generalize our estimates to the number of kernels with non - zero deficiency . in the remaining part of our argument we follow an idea from uczak @xcite ( see also janson , uczak , and ruciski @xcite ) and construct complex planar graphs @xmath78 from their kernels . thus , we first choose the kernel , then put on its edges vertices of degree two obtaining the core of the graph , and add to it a forest rooted on vertices of the core . this procedure lead to the estimate given in theorem [ thm : complex ] . we also remark that such a typical complex planar graph on @xmath70 vertices consists of large component and , perhaps , some small components of combined excess @xmath82 . finally , in the main part of the paper we count the number of planar graphs by splitting it into a complex planar graph , the number of which we have just found , and the part which consists of isolated trees and unicyclic components , whose number is well known ( see britikov @xcite ) . we compute @xmath20 for different values of @xmath83 . at the same time , we get information on the typical structure of @xmath0 such as the size of the largest component , its excess , and core . in this section we study the family of cubic planar weighted multigraphs , which plays a crucial role in studying the kernel of a random planar graph . we then consider the family of ` supercubic ' planar graphs that are planar weighted multigraphs with minimum degree at least three and a positive deficiency , which is indeed the set of possible kernels of complex planar graphs . in this section we count the number of all labeled cubic planar weighted multigraphs , where each multigraph with @xmath84 loops , @xmath85 double edges , and @xmath86 triple edges gets weight @xmath87 . for @xmath88 , let @xmath89 be the number of all labeled @xmath70-vertex connected cubic planar weighted multigraphs on @xmath1 vertices and @xmath90 be the corresponding exponential generating function defined by @xmath91 note that @xmath92 for odd @xmath1 and also for @xmath93 except that we set @xmath94 by convention . it is well - known ( e.g. @xcite ) that @xmath95 the function @xmath96 is defined by the following system of equations : @xmath97 this system of equations is obtained by following the lines of sections 36 in bodirsky _ et al . _ @xcite , where cubic planar _ simple _ graphs were studied , so below we just outline the argument . the starting idea is that given a connected cubic planar weighted multigraph @xmath78 , we select an arbitrary edge @xmath98 in @xmath78 and orient the edge @xmath98 , to obtain a _ rooted _ counterpart @xmath99 . more precisely , the rooted cubic graph @xmath100 obtained from a connected cubic multigraph @xmath101 consists of @xmath101 and an ordered pair of adjacent vertices @xmath54 and @xmath102 . the oriented edge @xmath103 is called the root of @xmath99 . denote by @xmath104 a graph obtained from @xmath99 by deleting the root of @xmath99 . we have the following lemma analogous to lemma 1 in @xcite . [ lem : structure ] a rooted cubic graph @xmath105 has exactly one of the following types . * @xmath106 : the root is a self - loop . * @xmath107 : @xmath108 is disconnected . * @xmath54 : @xmath108 is connected , but there is a cut edge in @xmath108 that separates @xmath54 and @xmath102 . * @xmath109 : @xmath108 is connected , there is no cut edge in @xmath108 separating @xmath54 and @xmath102 , and either @xmath103 is an edge of @xmath108 , or @xmath110 is disconnected . * @xmath111 : @xmath108 is connected , there is no cut - edge in @xmath108 separating @xmath54 and @xmath102 , @xmath78 is simple , and @xmath110 is connected . ( 0,0 ) [ fig : all_types],title="fig : " ] # 1#2#3#4#5 @font ( 6450,1820)(-14,-1115 ) ( 1393,-249)(0,0)[lb ] ( 1398,105)(0,0)[lb ] ( 1081,-1051)(0,0)[lb ] ( 301,-215)(0,0)[lb ] ( 2678,554)(0,0)[lb ] ( 2674,-698)(0,0)[lb ] ( 2251,-1051)(0,0)[lb ] ( 4050,-703)(0,0)[lb ] ( 4044,582)(0,0)[lb ] ( 5934,-703)(0,0)[lb ] ( 5826,563)(0,0)[lb ] ( 1,-1051)(0,0)[lb ] ( 3781,-1051)(0,0)[lb ] ( 5626,-1051)(0,0)[lb ] the generating function for the family @xmath112 of @xmath106-graphs is denoted by @xmath113 . the other generating functions in ( [ eq : genfunctions ] ) are analogously defined according to their corresponding types . furthermore , the above system of equations follows from decomposition of graphs . note that the number of labeled connected cubic planar weighted multigraphs with one distinguished oriented edge is counted by @xmath114 . the difference between the system of equations above and that in bodirsky _ et al . _ @xcite arises as follows . in @xcite the term @xmath115 does not appear in @xmath114 . the reason is that each graph in @xmath112 enumerated by @xmath113 works merely as a building block in @xmath99 when considering simple graphs , while it may also appear as a connected component ( whose root edge is a loop ) when considering multigraphs . note however that when considering weights , a graph on 2 vertices in @xmath112 ( that is , a rooted `` dumbbell '' consisting of two vertices , an edge , and two loops one of which is marked as a root ) gets a weight 1/2 when it is used as a building block ( one loop disappears in the building operation ) , but 1/4 when used as an isolated component ( due to two loops ) . in a similar way , a graph on 2 vertices in the family @xmath116 enumerated by @xmath117 ( which is indeed a triple edge , one of which is oriented ) gets a weight 1/2 when it is used as a building block , but 1/6 when used as an isolated component . analogous to section 6 in @xcite , one can use the singularity analysis to obtain the following asymptotic estimates . [ thm : cubic - asymptotics ] for @xmath1 even , @xmath118 where all constants are analytically given , @xmath119 is the dominant singularity of @xmath96 , and @xmath120 . furthermore , @xmath121 for odd @xmath1 . we note that the first digits of @xmath122 are @xmath123 , while the growth constant for the labeled _ simple _ cubic graphs is close to @xmath124 @xcite . the difference in growth is due to the fact that , unlike in the non - planar case , in average the cubic planar graphs contain large number of multiple edges and loops . let @xmath125 denote a random graph chosen uniformly at random from the family @xmath126of all labeled cubic planar weighted multigraphs on @xmath1 vertices , where each multigraph with @xmath84 loops , @xmath85 double edges , and @xmath86 triple edges gets weight @xmath87 . using the asymptotic estimation of @xmath127 , @xmath128 , we obtain the following results on the size @xmath129 of the largest component of @xmath130 . [ lem : cubic - giant ] 1 . uniformly over @xmath131 , and @xmath132 , @xmath133 2 . there are constants @xmath134 and @xmath135 such that for all integers @xmath136 and @xmath137 @xmath138 \(a ) if @xmath139 , where @xmath140 , then the graph obtained from @xmath130 by deleting the largest component ( which is a connected cubic planar weighted multigraph on @xmath141 vertices ) is an arbitrary cubic planar weighted multigraphs on @xmath142 vertices . by theorem [ thm : cubic - asymptotics ] we have @xmath143 \(b ) we let @xmath144 denote @xmath145 , the number of vertices outside of the largest component . the part ( a ) implies that there is a constant @xmath146 such that for each @xmath1 and each @xmath147 , @xmath148 and hence @xmath149 now it suffices to show that @xmath150 . let @xmath151 with @xmath152 . then , the vertex set of @xmath78 can be partitioned into two sets @xmath153 , @xmath154 , such that @xmath155 , @xmath156 , and there is no vertex between @xmath153 and @xmath154 . hence , since for all sufficiently large @xmath1 we have @xmath157 , for large @xmath1 we get @xmath158 this completes the proof of part ( b ) . let @xmath159 denote the family of labeled planar multigraphs @xmath78 on vertex set @xmath160=\{1,2,\dots , n\}$ ] with @xmath161 edges which have minimum degree at least three . therefore the deficiency @xmath81 of @xmath162 equals @xmath107 . moreover , let @xmath163 where @xmath164 counts loops in @xmath78 , and @xmath165 stands for the number of edges with parity @xmath166 for each @xmath167 . since each @xmath168 is cubic , the asymptotic behavior of @xmath169 is determined by ( [ eq : generalcubic ] ) . the following lemma gives bounds for @xmath170 for @xmath171 . [ lem : cubic - deficiency ] let @xmath172 and @xmath173 be integers such that @xmath174 is even . then we have @xmath175 where @xmath169 is given by ( [ eq : generalcubic ] ) . the idea of the proof is to generate multigraphs in @xmath159 from the graphs from @xmath176 by contracting edges incident to vertices @xmath177 . more precisely , for each of the vertices @xmath177 we choose one of the incident edges @xmath178 , @xmath179 , and contract it , i.e. we replace the vertices @xmath180 by one vertex @xmath181 which is adjacent to all neighbors of @xmath166 and @xmath182 . finally , we relabeled @xmath181 by label @xmath183 . sometimes this procedure fails to give a multigraph in @xmath176 ( e.g. when an edge is nominated by both its ends , or some of the edges @xmath184 form a cycle ) , nonetheless each multigraph from @xmath159 can clearly be obtained from some graph from @xmath159 in the above process . now let us show the upper bound for @xmath170 . choose @xmath185 and select edges @xmath184 , @xmath179 , in one of at most @xmath186 ways . suppose that by contracting all edges @xmath184 , @xmath179 , we get a multigraph @xmath187 . note that the weight of the multigraph @xmath188 could only increase by at most @xmath189 in the case when all vertices @xmath179 belong to different components of size two ( then we replace a triple edge which contributes to the weight @xmath190 by two loops of total weight @xmath191 ) . finally , we claim that there are at least @xmath192 graphs @xmath193 which differ from @xmath78 only by labelings of vertices @xmath194 , i.e. each @xmath187 is counted in this procedure at least @xmath192 times . indeed , let us remove all labels @xmath194 from vertices of @xmath78 getting a graph @xmath195 in which @xmath107 ` dummy ' vertices are not labeled . we try to relabel those vertices with @xmath194 . take any vertex @xmath196 of @xmath188 of degree larger than three . then , @xmath196 is adjacent in @xmath195 to @xmath166 dummy vertices , where @xmath197 . thus , we can label neighbors of @xmath196 with @xmath166 labels from @xmath194 in at least @xmath198 ways . now take another vertex which has been already labeled and choose labels for its dummy neighbors , and so on , until all dummy vertices gets their labels . clearly , the number of way of doing that is bounded from below by @xmath199 . hence , using ( [ eq : generalcubic ] ) , we get @xmath200 in order to get a lower bound for @xmath170 we count only multigraphs @xmath187 with maximum degree four . note that the number of vertices of degree four in @xmath188 is @xmath107 . each vertex @xmath201 of degree four in @xmath188 we split into two vertices : one of them we label with @xmath201 the other we leave as a ` dummy ' vertex which has not get labeled so far . we add an edge between the two vertices , which we mark as used . we can make such a split into at most six possible ways . now , we can choose labels for dummy vertices into one of possible @xmath202 ways ( note that each dummy vertex is uniquely identified by the other end of the used edge ) . consequently , from each @xmath188 we get at most @xmath203 different graphs from @xmath159 with @xmath107 disjoint edges marked . note however that splitting a vertex we may increase the weight of the graph by at most six ( if we split the quadruple edge into two double edges ) , thus the total weight of different multigraphs obtained from @xmath188 is bounded from above by @xmath204 . consequently , @xmath205 recall that a graph is called complex if all its components have positive excess . in this section , we derive the asymptotic number of all labeled _ complex _ planar graphs with given size and excess ( theorem [ thm : complex ] ) . in order to do that we first estimate the number of such graphs with given deficiency ( lemma [ lem : complex - supercubickernel ] ) . in this section we estimate the number @xmath206 of all labeled complex planar graphs @xmath78 on @xmath70 vertices with @xmath207 and @xmath208 . to this end , observe first that the core @xmath79 of a complex graph @xmath78 can be obtained from @xmath78 by pruning each vertex of degree one in @xmath78 recursively , i.e. to get @xmath79 we have to delete the tree - like part of @xmath78 rooted at vertices of @xmath79 . then , in order to find the kernel @xmath80 of @xmath78 , one needs to replace each path in @xmath79 whose internal vertices are all of degree two by a single edge . note that both @xmath79 and @xmath80 have the same excess as @xmath78 , and @xmath80 is a planar multigraph of minimum degree at least three . therefore , if @xmath209 and @xmath210 , the number of vertices in @xmath80 , denoted by @xmath211 , equals @xmath212 and the number of edges , denoted by @xmath213 , equals @xmath214 . in order to find @xmath20 we reverse the above procedure and first count all possible kernels of graphs , then study the number of cores which lead to these kernels , and finally add to them the rooted forest to obtain all possible graphs @xmath78 . more precisely , one can construct all graphs @xmath78 on @xmath70 vertices with @xmath71 edges , whose kernel is a planar multigraph graph with minimum degree at least three and deficiency @xmath107 , in the following way . * choose the vertex set @xmath215 of the core ( for some @xmath216 ) , and then select the vertex set @xmath217 of the kernel from the vertex set of the core . it can be done in @xmath218 ways . * select a kernel of order @xmath211 among all the possible candidates for kernels ( i.e. cubic planar weighted multigraphs on vertex set @xmath219 $ ] with @xmath213 edges ) , and then map @xmath219 $ ] to @xmath217 in their relative order , i.e. we map @xmath142 to @xmath220 . there are @xmath221 ways of doing that . * order the edges of the kernel lexicographically , each edge with a direction from the one end point with the smaller label to the other end point with the larger label . for multiple edges with the same ends take any order , and choose one of two possible directions for each loop ( the weights we assign when we counted candidates for the kernel was chosen precisely to assure that in this way we avoid double counting ) . now make a directed path of length @xmath213 consisting of the kernel edges , according to this order , and insert the core vertices that are not in the kernel on the edges of the kernel in such a way that each loop gets at least two core vertices , and at least @xmath222 edges from @xmath142 edges incident to the common end points get at least one core vertex . let @xmath223 be such that @xmath224 , where @xmath84 denotes the number of loops and @xmath225 multiple edges with parity @xmath142 for @xmath226 in @xmath80 ( hence @xmath227 ) . this can be done in @xmath228 ways . * plant a rooted forest on the core vertices . according to cayley s formula , one can do it in @xmath229 ways . as a consequence , we have @xmath230 applying lemma [ lem : cubic - deficiency ] we obtain the following estimate . [ lem : complex - supercubickernel ] let @xmath231 and @xmath232 be integers . let @xmath9 , @xmath17 be the constants so that the assertion of theorem [ thm : cubic - asymptotics ] holds . then @xmath233 for some @xmath234 , @xmath235 , and @xmath236 , @xmath227 . moreover , the typical size of the core in a randomly chosen complex planar graph @xmath78 on @xmath70 vertices with @xmath209 and @xmath210 is a.a.s . @xmath237 note that from lemma [ lem : cubic - deficiency ] we have @xmath238 where @xmath239 therefore from ( [ eq : cd ] ) we get @xmath240 thus , it is enough to estimate the quantity @xmath241 where @xmath242 and @xmath243 satisfy @xmath235 and @xmath227 . below we use several times stirling s formula @xmath244 and the following consequence of maclaurin expansion of @xmath245 @xmath246 to derive an asymptotic formula for ( [ eq : cdeficiency - upper ] ) , note that @xmath247 and @xmath248 next , we rewrite the sum over @xmath166 in ( [ eq : cdeficiency - upper ] ) as @xmath249 where the function @xmath250 is defined as @xmath251 we observe that the main contribution to ( [ eq : iterms ] ) comes from the terms @xmath252 , where @xmath253 for such @xmath166 s , we have @xmath254 and @xmath255 thus , we get @xmath256 finally , ( [ eq : cdeficiency - upper ] ) and ( [ eq : iterms - final ] ) yield @xmath257 the last part of the assertion follows from the fact that the main contribution to the sum ( [ eq : cdeficiency - upper ] ) comes from the terms @xmath258 , where @xmath259 is given by ( [ eq : add1 ] ) . in this section we estimate the number @xmath260 of labeled complex planar graphs with @xmath70 vertices and @xmath71 edges . [ thm : complex ] let @xmath9 , @xmath17 , @xmath261 be the constants for which the assertion of theorem [ thm : cubic - asymptotics ] holds and let @xmath70 , @xmath262 be integers . * there exists a function @xmath263 with @xmath264 , for which @xmath265 * the number @xmath266 of labeled _ connected _ complex planar graphs with @xmath70 vertices and @xmath71 edges is given by a similar formula , with @xmath17 replaced by @xmath261 . * a graph chosen uniformly at random among all complex planar graphs with @xmath70 vertices and @xmath71 edges has a.a.s . deficiency @xmath267 and the core of size @xmath268 . in particular , if @xmath269 , then a.a.s . such a random graph is clean . * if @xmath270 , then a graph chosen uniformly at random among all complex planar graphs with @xmath70 vertices and @xmath71 edges has a.a.s . @xmath82 components , among which there is a giant component of size @xmath271 . furthermore , a.a.s . each small component has @xmath272 vertices and the probability that such a graph contains exactly @xmath111 such components is bounded away from both 0 and 1 for every @xmath273 . using the asymptotic estimate of @xmath206 from lemma [ lem : complex - supercubickernel ] , we get the following . @xmath274 where @xmath234 and @xmath236 satisfy @xmath235 and @xmath227 . define a function @xmath275 as @xmath276 we observe that the main contribution to @xmath277 comes from @xmath278 and therefore @xmath279 finally , taking @xmath280 ( and thus @xmath264 ) completes the proof of ( i ) . in order to show ( ii ) one should repeat computations from the proof of lemma [ lem : complex - supercubickernel ] and the one given above , for graphs with _ connected _ kernels . therefore @xmath281 should be replaced by the number of _ connected _ cubic planar weighted multigraphs vertices , which , by ( [ eq : concubic ] ) , is equal to @xmath282 . to see ( iii ) observe the main contribution to ( [ eq : maxdeficiency ] ) comes from @xmath278 . finally , lemma [ lem : cubic - giant ] states that a randomly chosen cubic planar graph a.a.s . contains a giant component of size @xmath39 , and using exact counts it is easy to show that the number of small components has a non - degenerate distribution . since for @xmath270 we can count graphs up to a constant factor , a similar statement is true also for supercubic weighted multigraphs . now ( iv ) follows from the fact that the trees rooted in one edge of the kernel have in average @xmath272 vertices altogether . unfortunately , since we can only estimate the number of supercubic graphs up to a factor of @xmath283 , we can not prove the assertion of theorem [ thm : complex](iv ) in the case when @xmath284 . nonetheless , we think that it is true also in much wider range and that the following conjecture holds . * giant conjecture . * _ the assertion of theorem [ thm : complex](iv ) holds for every @xmath285 . _ in this section we derive the asymptotic number @xmath20 of labeled planar graphs with @xmath1 vertices and @xmath2 edges and investigate how the size of the largest component in @xmath0 , its excess and the size of its core change with @xmath2 . throughout the section , we let @xmath9 , @xmath17 be the constants for which the assertion of theorem [ thm : cubic - asymptotics ] holds . by @xmath286 we denote the number of vertices in the @xmath142-th largest component of @xmath0 . let @xmath287 ( resp . @xmath288 ) stand for the excess ( resp . the number of vertices in the core ) of the subgraph of @xmath0 which consists of its complex components , and let @xmath289 denote its size . finally , let @xmath290 and @xmath291 denote the excess and the size of the core of the largest component of @xmath0 , respectively . before studying @xmath0 we recall in the next section some properties of the uniform random graph @xmath40 which are relevant for our argument . let @xmath292 denote the number of vertices in the @xmath142-th largest component of @xmath40 , let @xmath293 stand for the excess of the largest component of @xmath40 , and let @xmath294 be the number of vertices in the core of the largest component of @xmath40 . the following results on the largest components were proved by uczak @xcite and uczak , pittel , and wierman @xcite ( see also janson , uczak , and ruciski @xcite ) . [ thm : subgnm ] let @xmath42 , where @xmath295 . if @xmath296 , then for @xmath142 fixed , a.a.s . furthermore , a.a.s . the @xmath142-th largest component of @xmath40 is a tree . [ thm : criticalgnm ] let @xmath42 , where @xmath295 . if @xmath298 , then for @xmath142 fixed , a.a.s . @xmath299 . the total excess of the complex components of @xmath40 is a.a.s . @xmath82 , and the probability that the @xmath142-th largest component of @xmath40 has excess @xmath111 is bounded away from zero for every fixed @xmath137 , and @xmath300 . furthermore , if @xmath40 contains some complex components , then a.a.s . they have @xmath3 vertices in total and @xmath301 . [ thm : supercriticalgnm ] let @xmath42 , where @xmath295 . if @xmath302 , then a.a.s . @xmath303 , while for @xmath226 fixed , @xmath304 . moreover , a.a.s . the @xmath142-th largest component of @xmath40 is a tree , provided @xmath226 . the structure of the giant component of @xmath40 was studied by uczak @xcite . [ thm : gnmexcesscore ] if @xmath42 , where @xmath302 but @xmath305 , then a.a.s . @xmath306 and @xmath307 . the following threshold for the property that @xmath40 is planar was proved by uczak , pittel , and wierman @xcite . [ thm : planariy ] let @xmath308 for a constant @xmath309 . then the probability that @xmath40 is planar tends to a limit @xmath310 as @xmath311 , where @xmath312 , @xmath313 , and @xmath314 . a different proof of the above result can be found in janson _ et al . _ @xcite , who also showed that @xmath315 . we shall use theorems [ thm : subgnm ] , [ thm : criticalgnm ] and [ thm : planariy ] in the proofs of theorems [ thm : subcritical ] and [ thm : critical ] . we do not use theorems [ thm : supercriticalgnm ] and [ thm : gnmexcesscore ] in our proofs below , but we decide to invoke them here to show how our results differ from those for @xmath40 . for a constant @xmath316 , let us define @xmath317 note that @xmath318 decreases monotonically with @xmath319 as @xmath320 and @xmath321 for large @xmath309 . in our argument we use also the following result of britikov @xcite . here and below by @xmath322 we denote the probability that @xmath40 contains no complex component . in other words , @xmath323 , where @xmath324 denotes the number of labeled graphs with @xmath1 vertices and @xmath2 edges , which contain no complex components . [ thm : britikov ] let @xmath42 , where @xmath295 . then the following holds . 1 . if @xmath296 , then @xmath325 . 2 . if @xmath298 , where @xmath309 is a ( not necessarily positive ) constant , then @xmath326 . 3 . if @xmath302 , then @xmath327 . the main ingredient of our argument is a simple observation that each graph can be uniquely decomposed into the complex part and the remaining part which consists of isolated tees and unicyclic components . moreover , it is the complex part which determines whether the graph is planar . consequently , the number @xmath20 of labeled planar graphs on @xmath1 vertices with @xmath2 edges is given by @xmath328 thus , the estimate of @xmath260 ( theorem [ thm : complex ] ) and that of @xmath329 ( theorem [ thm : britikov ] ) yield the asymptotic estimate of @xmath20 . moreover , the leading terms of ( [ eq : main ] ) give us information on the size of the complex part of the graph and thus , by theorem [ thm : complex ] , on the size of the largest component of @xmath0 and its internal structure . on the other hand , the size of the largest non - complex component can be deduced from theorems [ thm : subgnm ] and [ thm : criticalgnm ] . the behavior of @xmath0 in the subcritical case follows directly from theorem [ thm : subgnm ] . [ thm : subcritical ] let @xmath42 , where @xmath305 . if @xmath296 , then @xmath330 furthermore , for fixed @xmath142 a.a.s . @xmath331 and the @xmath142-th largest component of @xmath0 is a tree . theorem [ thm : planariy ] states that a.a.s . @xmath40 is planar . therefore , we have @xmath332 and the structure of @xmath0 follows from theorem [ thm : subgnm ] . the critical period is only slightly harder to deal with than the previous one as far as we estimate @xmath20 only up to a constant factor . [ thm : critical ] let @xmath42 . if @xmath298 for a constant @xmath316 , then @xmath333 furthermore , for @xmath142 fixed , a.a.s . @xmath334 , and , if @xmath0 contains complex components , then @xmath335 , @xmath336 , and @xmath337 . theorem [ thm : planariy ] states that the probability that @xmath40 is planar tends to a limit which is strictly between 0 and 1 . hence @xmath338 the assertion on the structure of @xmath0 is a direct consequence of theorem [ thm : criticalgnm ] . the evolution of @xmath0 in the ` early supercritical ' period starts to be more interesting . note that the result below estimates @xmath20 up to a factor of @xmath339 . [ thm : super ] let @xmath42 , where @xmath305 and @xmath340 . then @xmath341 furthermore , a.a.s . @xmath342 , while for any fixed @xmath226 we have @xmath334 . in addition , a.a.s . @xmath343 and @xmath344 . from theorem [ thm : complex ] and ( [ eq : stirling ] ) , we get @xmath345 in addition , the estimate @xmath346 and lemma [ thm : britikov ] give @xmath347 therefore , we get @xmath348 now let @xmath349 . then the sum in ( [ eq : firstrange1 ] ) depending on @xmath350 becomes @xmath351 where the function @xmath352 is defined as @xmath353 the main contribution to ( [ eq : planar - lterm ] ) comes from the terms @xmath354 , where @xmath355 furthermore , we have @xmath356 and @xmath357 this implies @xmath358 and hence ( [ eq : firstrange1 ] ) becomes @xmath359 we shall sum over @xmath70 in ( [ eq : firstrange2 ] ) , or , more specifically , over @xmath360 for @xmath361 , where @xmath362 will shortly be determined . letting @xmath361 , we estimate the summands in ( [ eq : firstrange2 ] ) as @xmath363 therefore ( [ eq : firstrange2 ] ) becomes @xmath364 where the function @xmath365 is defined as @xmath366 observe that the main contribution to @xmath367 , and therefore to ( [ eq : firstrange - sumr ] ) , comes from the terms @xmath368 . since @xmath369 , from lemma [ thm : britikov ] ( ii ) and definition [ def : nu ] we have @xmath370 , when @xmath371 . thus , the sum over @xmath368 in ( [ eq : firstrange - sumr ] ) can be replaced by an integral over @xmath372 , and we get @xmath373 as as consequence , the first part of the theorem follows . for the second part of the assertion note that in this case the main contribution to the sum ( [ eq : firstrange1 ] ) follows from @xmath350 s close to @xmath374 given in ( [ eq : l ] ) and @xmath375 , and so a.a.s . therefore , by theorem [ thm : complex ] , a.a.s . @xmath0 is clean , its kernel has @xmath376 vertices , and @xmath344 . as far as @xmath36 is bounded away from both @xmath377 and @xmath37 we can prove results similar to theorem [ thm : super ] but , since now @xmath378 , we can estimate @xmath20 only up to a constant factor . [ thm : middle ] if @xmath379 for a constant @xmath380 , then @xmath381 furthermore , a.a.s . @xmath382 , @xmath383 , and @xmath384 . following the lines of the proof of theorem [ thm : super ] , but with @xmath54 replaced by @xmath385 , yields the assertion . using theorems [ thm : subcritical]-[thm : middle ] one can find that the threshold for the property that @xmath0 has the chromatic number four is @xmath386 . [ cor : mid ] let @xmath387 . 1 . if @xmath388 , then a.a.s . @xmath389 . 2 . if @xmath390 , then a.a.s . @xmath0 contains a copy of @xmath391 , and as a consequence , @xmath392 . here we only sketch the argument . let @xmath379 , @xmath380 . then a.a.s . the kernel of @xmath0 has deficiency @xmath393 and @xmath394 vertices ( see theorem [ thm : complex ] and [ thm : middle ] ) . furthermore , on the edges of the kernel we need to place @xmath3 vertices of the core . thus , the probability that on some edge we place fewer than five vertices is @xmath395 , and so there are a.a.s . at most @xmath396 edges of the kernel of @xmath0 which contain fewer than three vertices of the core . moreover , none of the vertices of the kernel is incident with more than one such edge . it is easy to see that such a graph can be colored using three colors . a similar argument shows that @xmath397 for @xmath398 . on the other hand , from the formula for @xmath20 for @xmath390 by gimnez and noy @xcite and chebyshev s inequality it follows that for such an @xmath2 the graph @xmath0 contains a copy of @xmath391 ( in fact it a.a.s . contains @xmath399 copies of @xmath391 , in which three vertices have degree three in @xmath0 ) . let us remark that dowden @xcite studied the probability that @xmath0 contains a given subgraph and determined its asymptotic behavior depending on the ratio @xmath36 . in particular , theorem [ cor : mid ] ( ii ) is relevant to his theorem 17 . in the previous section we showed that as far as @xmath379 , and @xmath400 , the size of the largest component grows with @xmath2 , but its density does not depend much on the value of @xmath32 and in the whole range is of the order @xmath401 . clearly , this situation must change when the size of the largest component is @xmath402 . indeed , starting from some point , the increase in the number of edges of @xmath0 must contribute to the density of the largest component , since when @xmath62 , for @xmath102 large enough , we should expect @xmath403 . our next result states that this change occurs when @xmath5 . [ thm : secondrange ] set @xmath62 , where @xmath404 . 1 . let @xmath405 if @xmath63 , but @xmath406 , then @xmath407 furthermore , a.a.s . @xmath408 , @xmath409 , and @xmath410 . 2 . let @xmath106 be the unique positive solution of the equation @xmath411 . if @xmath412 for @xmath413 , then @xmath414 furthermore , a.a.s . @xmath415 , @xmath416 , and @xmath417 . 3 . let @xmath418 . if @xmath419 , then @xmath420 furthermore , if in addition @xmath68 , a.a.s . @xmath421 , @xmath422 , and @xmath423 . in order to show ( i ) we follow the argument presented in section [ sec : supercritical ] . thus , from theorem [ thm : complex ] and lemma [ thm : britikov ] , we get @xmath424 where the function @xmath425 is defined as @xmath426 in order to estimate ( [ eq : secondrange1 ] ) observe that the maximum of the sum @xmath427 is taken at @xmath428 ; more precisely , the main contribution to the sum comes from the terms @xmath429 , where @xmath430 . thus , @xmath431 where we sum over @xmath430 . note also that @xmath432 , and so if @xmath433 , then @xmath434 . next note that the main contribution to the sum over @xmath350 in ( [ eq : secondrange1 ] ) comes from the terms @xmath354 , where @xmath435 , so ( [ eq : secondrange1 ] ) can be estimated by @xmath436 where the function @xmath437 is defined as @xmath438 the behavior of @xmath437 depends on the range of @xmath102 . thus , we consider three cases , corresponding to the periods described in the three parts of the assertion of theorem [ thm : secondrange ] . first , let @xmath63 , but @xmath406 . the main contribution to the sum @xmath437 comes from the terms @xmath354 with @xmath439 , where @xmath440 in this case we have @xmath441 and ( [ eq : sec1 ] ) follows . in order to get the information on the structure of @xmath0 for this range one should apply theorem [ thm : complex ] and observe that @xmath442 now let @xmath443 for some constant @xmath413 . in this case , the main contribution to the sum @xmath437 comes from the terms @xmath444 with @xmath445 , where @xmath106 is the unique positive solution of the equation @xmath411 . we have @xmath446 and @xmath447 therefore , ( ii ) follows from ( [ eq : main2 ] ) and theorem [ thm : complex ] . finally , let @xmath419 , but @xmath404 . then the main contribution to the sum @xmath437 comes from the terms @xmath354 with @xmath448 , where @xmath449 we have @xmath450 which , together with ( [ eq : main2 ] ) , gives ( [ eq : sec2 ] ) . the second part of ( iii ) follows from ( [ eq : main2 ] ) , theorem [ thm : complex ] , and the observation that @xmath451 we also remark that we expect the estimates for @xmath287 and @xmath288 to hold also for @xmath452 but the error term @xmath453 in ( [ eq : sec2 ] ) becomes too large to provide a precise information on the structure of @xmath0 . let us also point out that if giant conjecture is true , then @xmath454 and @xmath455 in the whole range of @xmath2 . let us first propose some heuristic which explains the behavior of @xmath0 as described in theorems [ thm : subcritical][thm : secondrange ] . note that the number @xmath20 of planar graphs with @xmath1 vertices and @xmath2 edges can be computed in two different ways . our estimates were based on the formula @xmath456 where we extracted from the graph its complex part . but one could also use the formula @xmath457 where we first identify in the graph the largest component of @xmath70 vertices and @xmath71 edges , which typically is complex and unique , and then we supplement it by a random planar graph of @xmath458 vertices and @xmath459 edges . however , theorem [ thm : complex](iv ) states that ( at least for small @xmath350 ) a graph chosen at random from all complex planar graphs with @xmath70 vertices and @xmath71 edges consists of the giant component of size @xmath271 and , possibly , some small components of finite complexity and size @xmath272 . if the giant conjecture is true , it is in fact the case for all values of @xmath285 . thus , the planar graph which is outside the largest component must contain just a few ( if any ) components which are complex . it happens only if its density is such as the density of the standard uniform graph model in the critical period . consequently , in ( [ eq : main3 ] ) , we must have @xmath460 and , since in the critical period the sizes of all complex components of a random graph on @xmath1 vertices are of the order @xmath461 , @xmath462 let us check what it gives when @xmath42 with @xmath49 to ensure that @xmath0 contains a complex and unique giant component . if @xmath463 , then from ( [ eq : heur1 ] ) we get @xmath464 and , by ( [ eq : heur2 ] ) , @xmath465 which fits perfectly the estimates from theorem [ thm : super ] and [ thm : middle ] . now let @xmath466 , @xmath404 . then , as before , from ( [ eq : heur1 ] ) and ( [ eq : heur2 ] ) we infer that @xmath467 and @xmath468 note that when @xmath469 is large , then , clearly , @xmath470 , and the error term @xmath471 in ( [ eq : heur3 ] ) is equal to the main term @xmath472 . in order to compute this ` threshold value ' of @xmath102 one should use ( [ eq : heur4 ] ) and substitute to it @xmath470 and @xmath473 . it gives @xmath474 as the threshold value for the property that @xmath475 . consequently , if @xmath419 , then we have @xmath476 , as stated in theorem [ thm : secondrange](i ) , while for @xmath63 , we have @xmath470 and , by ( [ eq : heur4 ] ) , @xmath477 , which agrees with theorem [ thm : secondrange](iii ) . since for @xmath478 , where @xmath76 , the random graph @xmath0 is a.a.s . quite sparse ( see theorem [ thm : middle ] ) , while for @xmath479 , @xmath480 , it is quite dense ( e.g. a.a.s . it contains a copy of any given planar graph ) , it seems that the most interesting period in the evolution of @xmath0 is for @xmath481 . this intuition is confirmed by theorem [ cor : mid ] , but clearly a lot remains to be done . for instance , it seems that the correct threshold function for the property that @xmath392 is @xmath482 , more precisely , we conjecture that the following holds . * conjecture * _ if @xmath483 , then a.a.s . @xmath397 , while if @xmath484 , then a.a.s . @xmath485 and so @xmath392 . _ let us briefly justify the above claim that @xmath391 emerges in @xmath0 when @xmath482 . first of all , in order to have a single copy of @xmath391 in @xmath0 we need a lot of copies of @xmath391 in the kernel of the largest component of @xmath0 . the number of copies of @xmath391 in the kernel should be of the same order as the number of vertices of degree four in the kernel , which , in turn , is expected to be of the order of deficiency of the graph . once we have a copy of @xmath391 in the kernel , the probability that after placing vertices of the core at the edges of the kernel none of these vertices will be put at one of six edges of @xmath391 is @xmath486 . hence , if our estimates for @xmath487 and @xmath488 from theorem [ thm : secondrange](iii ) remain valid for all values of @xmath2 such that @xmath489 , then the expected number of copies of @xmath391 s in @xmath0 is of the order @xmath490 thus , this number is bounded away from zero for @xmath491 . let us also add a few words on the models of random planar graphs different from @xmath0 . one of the most natural one is the graph obtained by the random planar process , when we add to an empty graph on @xmath1 vertices @xmath2 edges one by one each time choosing a new edge uniformly at random from all pairs which preserve planarity of the graph ( see gerke _ et al . _ @xcite ) . in this model the structure of components is similar to that of a standard graph @xmath492 for an appropriately chosen @xmath493 . another model of random planar graph is the binomial random graph @xmath494 , when we look at properties of @xmath495 conditioned on the fact that it is planar . equivalently , one can view @xmath494 as the graph chosen from the family @xmath496 of all planar graphs on @xmath1 vertices in such a way that each @xmath497 appears as @xmath494 with the probability @xmath498 where @xmath499 denotes the number of edges of @xmath78 , and @xmath500 since clearly for every property @xmath501 @xmath502 once we determine the typical number of edges in @xmath494 the problem of finding properties of @xmath494 reduces to studying these properties for @xmath0 . from the estimates of @xmath20 given in theorems [ thm : subcritical][thm : middle ] and [ thm : secondrange ] it follows that if @xmath503 then a.a.s . @xmath494 has @xmath504 edges ; if @xmath505 , then a.a.s . @xmath506 ; for @xmath507 we are in the second critical period , i.e. a.a.s . @xmath5 ; finally for @xmath508 we have a.a.s . @xmath509 . note that a large part of the evolution of @xmath494 , when @xmath510 , corresponds to the period of evolution of @xmath0 when @xmath386 which , as we have already remarked , is crucial for many properties of @xmath0 . another interesting model is a random cluster model @xmath511 on planar graphs when for every labeled planar graph with vertex set @xmath160 $ ] we put @xmath512 where @xmath513 is a parameter , @xmath514 stand for the number of components in @xmath78 , and @xmath515 is the normalizing factor . in a similar way one can define @xmath516 adding factors @xmath517 to the right hand sides of ( [ eq : c:1 ] ) and ( [ eq : c:2 ] ) . it is well known that the additional cluster factor @xmath517 in , say , the standard model @xmath495 leads to an interesting phenomena such as the discontinuous phase transition which occurs in @xmath518 for @xmath519 ( cf . luczak and uczak @xcite ) . unfortunately , no such event can be observed in the planar case . the evolution of @xmath511 is quite similar to that of @xmath520 . the reason is quite simple : the giant complex component of @xmath0 is very sparse until it reaches the size @xmath402 and so the number of components is always close to @xmath521 and , as calculations show , can not be influenced much by the presence of the additional factor @xmath517 . the asymptotic behavior of @xmath516 does not depend very much on the value of @xmath522 either except for the scaling : the number of edges of @xmath516 is roughly the same as for @xmath523 . a part of this work was carried out when both the authors visited the mittag - leffler institute and completed during the first author s stay at adam mickiewicz university within the dfg heisenberg programme . we would like to thank these institutions for their support . the second author is supported by the foundation for polish science . s. gerke , c. mcdiarmid , a. steger , and a. weil , random planar graphs with @xmath1 nodes and a fixed number of edges , in _ proceedings of the sixteenth annual acm - siam symposium on discrete algorithms ( soda05 ) _ , pages 999 1007 , 2005 . t. uczak , the phase transition in a random graph , in _ combinatorics , paul erdo"s is eighty _ , vol . 2 , bolyai soc . math . , 2 , j. bolyai math . budapest , 1996 , 399422 . component behavior near the critical point of the random graph ,

we develop a linear method for solving the nonlinear differential equations of a lumped - parameter thermal model of a spacecraft moving in a closed orbit . our method , based on perturbation theory , is compared with heuristic linearizations of the same equations . the essential feature of the linear approach is that it provides a decomposition in thermal modes , like the decomposition of mechanical vibrations in normal modes . the stationary periodic solution of the linear equations can be alternately expressed as an explicit integral or as a fourier series . we apply our method to a minimal thermal model of a satellite with ten isothermal parts ( nodes ) and we compare the method with direct numerical integration of the nonlinear equations . we briefly study the computational complexity of our method for general thermal models of orbiting spacecraft and conclude that it is certainly useful for reduced models and conceptual design but it can also be more efficient than the direct integration of the equations for large models . the results of the fourier series computations for the ten - node satellite model show that the periodic solution at the second perturbative order is sufficiently accurate . lcl & = & outward - facing area of _ _ i__th - node , m + & = & thermal capacitance diagonal matrix , j / k + & = & thermal capacitance of _ _ i__th - node , j / k + & = & eigenvector of , k + & = & _ _ j__th - component of eigenvector , k + & = & driving function vector , k / s + & = &th fourier coefficient of , k / s + & = & complex conjugate of , k / s + & = & driving function forth - mode , k / s + & = & time derivative of , k / s + & = & driving function forth - node , k / s + & = &th fourier coefficient of , k / s + & = & driving function for second order temperature vector , k / s + & = & jacobian matrix , s + & = & jacobian matrix _ ij_-element , s + & = & number of steps in numerical integration of odes + & = & conduction coupling matrix , w / k + & = & matrix of conductances from linearized radiation coupling terms , w / k + & = &th - node conductance from linearized environment - radiation terms , w / k + & = & conduction coupling matrix _ ij_-element , w / k + & = & _ ij_-element of conductance matrix from linearized radiation terms , w / k + & = & number of samples for discrete fourier transform + & = & number of nodes + & = & eigenvector matrix for _ j _ , k + & = & _ _ i__th - component of-eigenvector forth - mode , k + & = & auxiliary heat input vector , k / s + & = & heat input to _ _ i__th - node , w + & = & mean value of over period , w + & = & _ _ i__th - node coefficient of radiation to the environment , w / k + & = & radiation coupling matrix , w / k + & = & time , s ( min in figures ) + & = & temperature vector , k + & = & vector of `` errors '' w.r.t . esatan solution , k + & = & steady - state temperature vector , k + & = & temperature of _ _ i-__node , k + & = & cosmic microwave background radiation temperature , k + & = & _ _ n__th order term in _ _ j__th - node temperature expansion , k + & = & stationary solution temperature vector , k + & = & _ _ n__th order of stationary solution temperature vector , k + & = & time derivative of _ _ i__th - node temperature , k / s + & = & steady - state temperature of _ _ i__th - node , k + & = & orbital period , s + & = & matrix of independent solutions to homogeneous equations , k + & = & solar absorptivity + & = & kronecker s delta + & = & antisymmetric part of , s + & = & _ ij_-element of , s + & = & infrared emissivity of _ _ i__th - node outward - facing surface + & = & perturbation parameter + & = & eigenvalue of _ j _ forth mode , s + & = & perturbation of _ j_-eigenvalue forth mode , s + & = & stefan - boltzmann constant , wmk +

1202.4372

the thermal control of a spacecraft ensures that the temperatures of its various parts are kept within their appropriate ranges @xcite . the simulation and prediction of temperatures in a spacecraft during a mission are usually carried out by commercial software packages . these software packages employ `` lumped parameter '' models that describe the spacecraft as a discrete network of nodes , with one energy - balance equation per node . the equations for the thermal state evolution are coupled nonlinear first - order differential equations , which can be integrated numerically . given the thermal parameters of the model and its initial thermal state , the numerical integration of the differential equations yields the solution of the problem , namely , the evolution of the node temperatures . however , a detailed model with many nodes is difficult to handle , and its integration for a sufficiently long time of evolution can take considerable computer time and resources . therefore , it is very useful to study simplified models and approximate methods of integrating the differential equations . many spacecraft missions , in particular , satellite missions , consist of an initial transient part and then a stationary part , in which the spacecraft just goes around a closed orbit , in which the heat inputs are periodic . these periodic heat inputs are expected to induce periodic temperature variations , with a maximum and a minimum temperature in each orbit . this suggests a conservative approach that consists in computing only the temperatures for the _ hot _ and _ cold cases _ of the given orbit , defining them as the two steady cases with the maximum and minimum heat loads , respectively . naturally , the real temperature variations in the orbit are smaller , because there is not enough time for the hot and cold cases to establish themselves . in fact , the temperature variations can be considerably smaller , to such a degree that it is necessary to integrate the differential equations , at least approximately . the differential equations for energy balance are nonlinear due to the presence of radiation couplings , which follow the stefan - boltzmann quartic law . a common approach to these equations involves a linearization of the radiation terms that approximate them by heat conduction terms @xcite . this approach transforms the nonlinear equations into standard linear heat conduction equations . but this approach has not been sufficiently justified , is of a heuristic nature and does not constitute a systematic approximation . in fact , nonlinear equations are very different from linear equations and , in particular , a periodic driving may not induce periodic solutions but much more complex solutions , namely , _ chaotic _ solutions . therefore , we have carried out in preceding papers a full nonlinear analysis of spacecraft thermal models @xcite . the conclusion of the analysis is that the complexities of nonlinear dynamics , such as multiple equilibria and chaos , do not appear in these models . while the existence of only one equilibrium state can be proved in general , the absence of chaos under driving by variable external heat loads can only be proved for a limited range of magnitudes of the driving loads . this range presumably includes the magnitudes involved in typical spacecraft orbits . the proofs in refs . and are constructive and are based on a perturbation method that is expected to be sound when the _ linear _ equations corresponding to the first perturbative order constitute a good approximation of the nonlinear equations . this implies that the fully nonlinear solution describes a weakly nonlinear oscillator . since the perturbative approximation is mathematically rigorous and systematic , it is worthwhile to study in detail the scope of the perturbative linear equations and , furthermore , to compare them with previous linear approaches of a heuristic nature . the main purpose of this paper is to study the linear method of predicting the thermal behavior of spacecraft in stationary orbits ( sect . [ linear ] and [ fourier - anal ] ) and to test it on a minimally realistic thermal model of a satellite in a circular orbit . since the general one and two - node models analyzed in refs . and , respectively , are too simple , we define in this paper a ten - node thermal model of a small moon - orbiting satellite ( sect . [ ten - node ] ) . this model is simple enough to allow us to explicitly show all the quantities involved ( thermal couplings and capacities , heat inputs , etc . ) and it is sufficient for illustrating the main features of the linear approach . as realistic thermal models have many more nodes , we consider in sect . [ scale ] the important issue of scalability of the method and , hence , its practical applications . computational aspects of the steady - state problem have been studied by krishnaprakas @xcite and by milman and petrick @xcite , while computational aspects of the direct integration of the nonlinear equations for the unsteady problem have been studied by krishnaprakas @xcite . here we focus on the linear equations for the stationary but unsteady case and survey its computational aspects . a note on notation : in the equations that contain matrix or vector quantities , sometimes we use component notation ( with indices ) while other times we use compact matrix notation ( without indices ) , according to the nature of the equations . a lumped - parameter thermal model of a continuous system consists of a discrete network of isothermal regions ( _ nodes _ ) that represent a partition of the total thermal capacitance and that are linked by thermal conduction and radiation couplings @xcite . this discretization reduces the integro - differential heat - transfer equations to a set of energy - balance odes , one per node , which control the evolution of the nodes temperatures @xcite : @xmath66 - r_i\,({t_i}^4- t_{0}^4 ) , \quad i= 1,\ldots , n,\end{aligned}\ ] ] where @xmath30 is the number of nodes and @xmath34 contains the total heat input to the @xmath17th - node from external radiation and from internal energy dissipation ( if there is any ) . the conduction and radiation coupling matrices are denoted by @xmath24 and @xmath67 , respectively ; they are symmetric ( @xmath68 and @xmath69 ) and @xmath70 ; so there are @xmath71 independent coupling coefficients altogether , but many vanish , usually . the temperature @xmath72 k is the temperature of the environment , namely , the cosmic microwave background radiation . the @xmath17th - node coefficient of radiation to the environment is given by @xmath73 , where @xmath0 denotes the outward facing area , @xmath74 its ( infrared ) emissivity , and @xmath75 is the stefan - boltzmann constant . the constant term @xmath76 can be included in @xmath34 or ignored altogether , if each @xmath77 . equations ( [ ode ] ) coincide with the ones implemented in commercial software packages , for example , esatan@xmath44 @xcite . there is no systematic procedure for finding the analytical solution of a system of nonlinear differential equations , except in some particularly simple cases . of course , nonlinear systems can always be integrated numerically with finite difference schemes . methods of this kind are employed in commercial software packages . when a nonlinear system can be approximated by a linear system and , hence , an approximate analytic solution can be found , this solution constitutes a valuable tool . actually , one can always resort to some kind of perturbation method to linearize a nonlinear system . therefore , we now study the rigorous linearization of eqs . ( [ ode ] ) based on a suitable perturbation method , and we also describe , for the sake of a comparison , a heuristic linearization , which actually is best understood in light of the results of the perturbation method . if we assume that the heat inputs @xmath78 in the energy - balance eqs . ( [ ode ] ) are periodic , namely , that there is a time interval @xmath79 such that @xmath80 , then it seems sensible to study first the effect of the mean heat inputs in a period . this _ averaging _ method , introduced in refs . and , relies on the fact that the _ autonomous _ nonlinear system of odes for constant @xmath81 can be thoroughly analyzed with analytical and numerical methods . for example , it is possible to determine that there is a unique steady thermal state and that it is ( locally ) stable @xcite . the actual values of the steady temperatures can be found efficiently with various numerical methods @xcite . furthermore , the eigenvalues and eigenvectors of the jacobian matrix of the nonlinear system of odes provides us with useful information about the dynamics , in particular , about the approach to steady - state : the eigenvectors represent independent thermal modes and the eigenvalues represent their relaxation times @xcite . once the averaged equations are solved , the variation of the heat inputs can be considered as a driving of the averaged solutions . thus , we can define the driving function @xmath82 where @xmath83 denotes the mean value of @xmath78 over the period of oscillation . a weak driving function must not produce a notable deviation from the averaged dynamics . in particular , the long - term thermal state of an orbiting spacecraft must oscillate about the corresponding steady - state . to embody this idea , we introduce a formal perturbation parameter @xmath84 , to be set to the value of unity at the end , and write eqs . ( [ ode ] ) as @xmath85 - \frac{r_{i}}{c_i } \,{t_i}^4 , \quad i= 1,\ldots , n,\end{aligned}\ ] ] then , we assume an expansion of the form @xmath86 when we substitute this expansion into eqs . ( [ eode ] ) , we obtain for the zeroth order of @xmath84 @xmath87 - \frac{r_{i}}{c_i } \,t_{(0)i}^4\ , , \quad i= 1,\ldots , n,\ ] ] that is to say , the averaged equations . the initial conditions for these equations are the same as for the unaveraged equations . for the first order in @xmath84 , we obtain the following system of linear equations : @xmath88 here , @xmath89 is the jacobian matrix @xmath90 where @xmath91 is the solution of the zeroth order equation . equations ( [ linode - pert ] ) are to be solved with the initial condition @xmath92 . the elements of the jacobian matrix at a generic point in the temperature space are calculated to be : @xmath93.\end{aligned}\ ] ] this matrix has interesting properties . first of all , it has negative diagonal and nonnegative off - diagonal elements . in other words , @xmath94 is a @xmath95-matrix @xcite . furthermore , it fulfills a semipositivity condition that qualifies it as a nonsingular @xmath96-matrix @xcite . since the eigenvalues of an @xmath96-matrix have positive real parts , the opposite holds for @xmath20 , namely , its eigenvalues have negative real parts . one more interesting property of @xmath94 , related to semipositivity , is that it possesses a form of _ diagonal dominance _ : it is similar to a diagonally dominant matrix and the similarity is given by a positive diagonal matrix . naturally , this property is shared by @xmath20 . these properties are useful to prove some desirable properties of the solutions of eqs . ( [ linode - pert ] ) . the chief property of @xmath20 is that @xmath94 is a nonsingular @xmath96-matrix . in particular , it implies that @xmath97 is non - negative and , therefore , that the perron - frobenius theory is applicable to it @xcite . the relevant results to be applied are : ( i ) perron s theorem , which states that a _ strictly _ positive matrix has a unique real and positive eigenvalue with a positive eigenvector and that this eigenvalue has maximal modulus among all the eigenvalues ; ( ii ) a second theorem , stating that if a @xmath95-matrix that is a nonsingular @xmath96-matrix is also `` irreducible '' , then its inverse is strictly positive . the irreducibility of @xmath20 follows from the symmetry of the matrices @xmath27 and @xmath39 @xcite . as the positive ( perron ) eigenvector of @xmath97 is the eigenvector of @xmath20 that corresponds to its smallest magnitude eigenvalue , it defines the slowest relaxation mode ( for a given set of temperatures ) . therefore , in the evolution of temperatures given by eqs . ( [ bode - pert ] ) , steady - state is eventually approached from the zone corresponding to simultaneous temperature increments ( or decrements ) . the matrix @xmath98 in eqs . ( [ linode - pert ] ) is obtained by substituting @xmath99 for @xmath100 in eqs . ( [ jij ] ) and ( [ jii ] ) . then , the nonhomogeneous linear system with variable coefficients , eqs . ( [ linode - pert ] ) , can be solved by variation of parameters @xcite , yielding the expression : @xmath101 where @xmath102 is a matrix formed by columns that are linearly independent solutions of the corresponding homogeneous equation , with the condition that @xmath103 ( the identity matrix ) . the difficulty in applying this formula lies in computing @xmath102 , that is , in computing the solutions of the homogeneous equation . moreover , this computation demands the previous computation of the solution for @xmath91 . since we are only interested in the stationary solutions of the heat - balance equations rather than in transient thermal states , it is possible to find an expression of these solutions that is more manageable than eq . ( [ t_1 ] ) . the transient thermal state relaxes exponentially to the stationary solution , which is a _ limit cycle _ of the nonlinear equations , technically speaking @xcite . therefore , the stationary solution is given by the solution of eqs . ( [ linode - pert ] ) with the _ constant _ jacobian matrix calculated at the steady - state temperatures , which we name @xmath104 . ) in which @xmath105 . ] this solution is simply @xcite : @xmath106 \cdot f(t-\t ) \,d\t , \label{t_1_jcst}\ ] ] with @xmath20 calculated at the point @xmath45 . furthermore , the periodic stationary solution is obtained by extending the upper integration limit from @xmath41 to infinity : @xmath107 \cdot f(t-\t ) \,d\t . \label{t_1_lim}\ ] ] this function is indeed periodic , unlike the one defined by eq . ( [ t_1_jcst ] ) , so it is determined by its values for @xmath108 $ ] . note that @xmath109 . for numerical computations , it can be convenient to express the integral from 0 to @xmath110 as an integral from 0 to @xmath79 , taking advantage of the periodicity as follows : @xmath111 \cdot f(t-\t)\,d\t = \sum_{n=0}^\infty \int_{n\t}^{(n+1)\t } \exp\left[\t j \right ] \cdot f(t-\t)\,d\t = \\ & & \sum_{n=0}^\infty \exp(n\t j ) \int_0^{\t } \exp\left[\t j \right ] \cdot f(t-\t)\,d\t = [ i-\exp(\t j)]^{-1 } \int_0^{\t } \exp\left[\t j \right ] \cdot f(t-\t)\,d\t\end{aligned}\ ] ] ( the series converges because the eigenvalues of @xmath20 have negative real parts ) . in the last integral , the argument of @xmath112 can be transferred to the interval @xmath113 $ ] : @xmath114 \cdot f(t-\t)\,d\t = \int_0^{t } \exp\left[\t j \right ] \cdot f(t-\t)\,d\t + \int_t^{\t } \exp\left[\t j \right ] \cdot f(t-\t+\t)\,d\t,\ ] ] where @xmath108 $ ] . note that the one - period shift in the argument of the last @xmath112 is necessary for the argument to be in @xmath113 $ ] . some remarks are in order . first of all , we have assumed that there is one asymptotic periodic solution of the nonlinear eqs . ( [ eode ] ) and only one ( a unique limit cycle ) . equivalently , we have assumed that the perturbation series converges . this assumption holds in an interval of the amplitude of heat input - variations @xmath112 @xcite . besides , for the integrals in eq . ( [ t_1_lim ] ) and the following equations to make sense , it is required that @xmath115 \ra 0 $ ] as @xmath116 . this is guaranteed , because the eigenvalues of @xmath20 have negative real parts , as is necessary for the steady - state to be stable . in fact , the eigenvalues are expected to be negative real numbers and @xmath20 is expected to be diagonalizable but both properties are not rigorously proven @xcite ( however , see sect . [ heur ] ) . if @xmath20 is diagonalizable , that is to say , there is a real matrix @xmath31 such that @xmath117 is diagonal , then the calculation of the integrals is best carried out on the eigenvector basis , given by the matrix @xmath31 . using this basis , eq . ( [ t_1_lim ] ) is expressed as @xmath118_i(t ) = \sum_{a=1}^n p_{ia } \int_0^\infty \exp\left[\t \l_a \right ] \sum_{j=1}^n p^{-1}_{aj } f_j(t-\t)\,d\t , \quad i= 1,\ldots , n , \label{t_1_eigenvec}\ ] ] where the first sum runs over the eigenvectors and their corresponding eigenvalues @xmath119 . expression ( [ t_1_eigenvec ] ) allows us to compare the contribution of the different thermal modes . in particular , for the fast modes , such that @xmath120 is large , we can use watson s lemma @xcite to derive the asymptotic expansion : @xmath121 f_a(t-\t)\,d\t = \frac{f_a(t)}{-\l_a } - \frac{\dot{f}_a(t)}{\l_a^2 } + \mfunction{o}\left(\frac{1}{\l_a^3 } \right),\ ] ] where @xmath122 . when @xmath120 is large , the first term suffices ( unless @xmath14 is also large , for some reason ) ; and the first term is small , unless @xmath12 is large . in essence , if the fast modes are not driven strongly , they can be neglected in the sum over @xmath13 in eq . ( [ t_1_eigenvec ] ) . for second order in @xmath84 , a straightforward calculation @xcite yields the following linear equation : @xmath123 where @xmath98 is the same jacobian matrix that appears in the first - order eq . ( [ linode - pert ] ) and @xmath124 the initial condition for eq . ( [ linode - pert2 ] ) is @xmath125 , as for eqs . ( [ linode - pert ] ) . therefore , the first - order and second - order equations have identical solutions in terms of their respective driving terms , although @xmath126 , eqs . ( [ tf ] ) , is a known function of @xmath41 only when the lower order equations have been solved . the integral expression , eqs . ( [ t_1_lim ] ) , of the stationary solution @xmath127 is also valid for @xmath128 , after replacing @xmath112 with @xmath126 and using in eq . ( [ tf ] ) the stationary values @xmath129 and @xmath130 ( which make @xmath126 periodic ) . it is possible to carry on the perturbation method to higher orders , and it always amounts to solving the same linear equation with increasingly complicated driving terms that involve the solutions of the lower order equations . the example of sect . [ ten - node ] shows that , in a typical case , @xmath128 is a small correction to @xmath127 , and further corrections are not necessary . this confirms that the perturbation method is reliable for a realistic case . a linearization procedure frequently used in problems of radiation heat transfer @xcite consists of using the algebraic identity @xmath131 to define an effective conductance for the radiation coupling between nodes @xmath17 and @xmath132 . the equation @xmath133 defines the effective conductance @xmath134 for specified values of the node temperatures @xmath46 and @xmath100 . for an orbiting spacecraft , the natural base values of the node temperatures are the ones that correspond to the steady - state solution of the averaged equations , namely , @xmath135 in the special case of radiation to the environment , @xmath136 can be replaced with linear terms @xmath137 such that @xmath138 for @xmath139 the resulting linear equations are : @xmath140 these equations have only conduction couplings , so they are a discretization of the partial differential equations of heat conduction . as a linear system of odes , the standard form is @xmath141 where @xmath20 ( the jacobian matrix ) is now given by : @xmath142.\end{aligned}\ ] ] the linear system of eqs . ( [ stdlinode ] ) can be solved in the standard way , yielding : @xmath143 \left({t}(0 ) + \int_0^t \exp\left[-\t j \right ] \cdot q(\t)\,d\t \right),\ ] ] where we have introduced the vector @xmath144 , with components @xmath145 . we can also express the solution in terms of the driving function @xmath146 : @xmath147 { t}(0 ) + \int_0^t \exp\left[(t-\t ) j \right ] \cdot \left(f(\t ) + \langle q \rangle \right ) d\t = \\ \exp\left[t j \right ] { t}(0 ) + \int_0^t \exp\left[(t-\t ) j \right ] \cdot f(\t ) d\t + j^{-1}(\exp\left[t j \right ] -i ) \langle q \rangle\end{aligned}\ ] ] for large @xmath41 , this solution tends to the periodic stationary solution @xmath148 \cdot f(t-\t ) d\t - j^{-1 } \langle q \rangle , \label{t_lim}\ ] ] assuming that @xmath149 \ra 0 $ ] as @xmath150 . this is a consequence of the structure of @xmath20 , as in the preceding section . in the present case , the eigenvalues of @xmath20 , beyond having negative real parts , are actually negative real numbers , as we show below . the total conductance matrix @xmath151 is symmetric but this does not imply that @xmath20 is symmetric . nevertheless , if we define @xmath152 , the matrix @xmath153 is symmetric , because its off - diagonal matrix elements are : @xmath154 hence , the matrix @xmath155 , similar to @xmath20 , has real eigenvalues . furthermore , @xmath155 is diagonalized by an orthogonal transformation ; that is to say , there is an orthogonal matrix @xmath156 such that @xmath157 is diagonal . therefore , the thermal modes are actually _ normal _ ; that is to say , the modes , which are the eigenvectors of @xmath20 and hence the columns of the matrix @xmath158 , are related to the eigenvectors of @xmath159 , which are normal and are given by the columns of @xmath156 . alternatively , one can say that the eigenvectors of @xmath20 are normal in the `` metric '' defined by @xmath2 ; namely , @xmath160 which can be written in matrix form as @xmath161 . naturally , the orthogonality of modes greatly simplifies some computations . furthermore , the symmetry of the conductance matrix implies that the sum in eq . ( [ linode ] ) can be written as the action of a _ graph laplacian _ @xcite on the temperature vector . naturally , the graph is formed by the nodes and the linking conductances . a graph laplacian is a discretization of the ordinary laplacian and is conventionally defined with the sign that makes it positive semidefinite . the zero eigenvalue corresponds to a constant function , that is , a constant temperature , in the present case . a vector with equal components , say , equal to @xmath162 , is the positive ( perron ) eigenvector of the matrix . with more generality , the laplacian of a graph can be defined as a symmetric matrix with off - diagonal entries that are negative if the nodes are connected and null if they are not @xcite . this definition does not constrain the diagonal entries and , therefore , does not imply that a graph laplacian is positive semidefinite . it can be made positive definite ( or just semidefinite ) by adding to it a multiple of the identity matrix , which does not alter the eigenvectors . of course , the eigenvector corresponding to the smallest eigenvalue does not have to be constant , but the perron - frobenius theorem @xcite tells us that it is positive . by this general definition of a graph laplacian , the matrix @xmath163 is a different laplacian for the same graph , and eqs . ( [ stdlinode ] ) contain the action of this laplacian on the vector @xmath164 . notice that this general definition of a graph laplacian is connected with the definition of a @xmath95-matrix @xcite and , actually , a symmetric @xmath95-matrix is a graph laplacian . if such a matrix is positive definite , then it is equivalent to a _ stieltjes matrix _ , namely , a symmetric nonsingular @xmath96-matrix @xcite . the general jacobian obtained in sect . [ pert - eq ] is also such that @xmath94 and also @xmath165 are both nonsingular @xmath96-matrices , but they need not be symmetric . to investigate the accuracy of the approximation of the radiation terms by conduction terms , let us compare the periodic solution given by eq . ( [ t_lim ] ) with the first - order perturbative solution found in sect . [ pert - eq ] , namely , @xmath166 . of course , the jacobian matrices in the respective integrals differ , as do the temperature vectors added to the integrals , namely , @xmath45 or @xmath167 . while @xmath45 corresponds to the authentic steady - state of the nonlinear averaged equations , @xmath168 corresponds to the steady - state of eqs . ( [ linode ] ) after averaging , which is a state without significance , since we have already used the set of temperatures @xmath45 of the authentic steady - state to define the radiation conductances @xmath169 in eq . ( [ linode ] ) . therefore , the only sensible linear solution is the perturbative solution @xmath166 , even if we replace the jacobian matrix given by eq . ( [ jij ] ) and ( [ jii ] ) with the one given by eq . ( [ njij ] ) and ( [ njii ] ) . in our context , the notion of radiation conductance actually follows from the symmetry of the matrices @xmath170 or @xmath155 . therefore , the most natural definition of radiation conductance probably is @xmath171 , that is , the symmetrization of the term @xmath172 in eq . ( [ jij ] ) . this symmetrization has been tested by krishnaprakas @xcite , considering the steady - state problem for models with up to @xmath173 nodes and working with various resolution algorithms . he found that the effect of symmetrization is not appreciable . to estimate the effect of the antisymmetric part of the matrix @xmath174 , namely , @xmath175 , on the eigenvalue problem for the jacobian , we proceed as follows . we formulate this eigenvalue problem in terms of the matrix @xmath155 , so that it is an eigenvalue problem for a symmetric matrix perturbed by a small antisymmetric part . this problem is well conditioned , because the eigenvectors of the symmetric matrix ( the columns of the matrix @xmath156 ) are orthogonal . in particular , the perturbed eigenvalues are still real . furthermore , the first - order perturbation formula for the eigenvalue @xmath119 associated with an eigenvector @xmath4 @xcite yields : @xmath176 vanishing because the perturbation matrix @xmath177 is antisymmetric . so the nonvanishing perturbative corrections begin at the second order in the perturbation matrix , and , in this sense , they are especially small . given that @xmath178 is a periodic function , it can be expanded in a fourier series . to derive this series , let us first introduce the fourier series of @xmath7 , @xmath179 inserting this series in the integral of eq . ( [ t_1_lim ] ) and integrating term by term , we obtain the fourier series for @xmath178 . alternatively , we can substitute the fourier series for both @xmath178 and @xmath7 into eqs . ( [ linode - pert ] ) , where @xmath20 is taken to be constant ; then we can solve for the fourier coefficients of @xmath178 . the result is @xmath180 the fourier coefficients @xmath8 are obtained by integration : @xmath181 given that @xmath7 is a real function , @xmath182 furthermore , @xmath183 implies @xmath184 so @xmath7 is defined by the sequence of fourier coefficients for positive @xmath9 . this sequence must fulfill the requirement that @xmath185 so a limited number of the initial coefficients may suffice . actually , for numerical work , eq . ( [ fourier - coef ] ) can be conveniently replaced by the discrete fourier transform @xmath186 which only requires sampling of the values for @xmath7 , but also only defines a finite number of independent fourier coefficients , because @xmath187 . notice that we usually have available just a sampling of the heat inputs at regular time intervals , rather than the analytical form of @xmath34 . to calculate the exact number of independent fourier coefficients provided by eq . ( [ d_fourier - coef ] ) , we must take into account eqs . ( [ f * ] ) and ( [ f0 ] ) . if @xmath29 is an odd number , the independent fourier coefficients @xmath8 are the ones with @xmath188 ; that is to say , there are @xmath189 independent real numbers . if @xmath29 is even , the independent fourier coefficients are the ones with @xmath190 , and @xmath191 is real , so there are @xmath189 independent real numbers as well . for definiteness , let @xmath29 be odd . then , we can express @xmath7 as @xmath192.\ ] ] of course , the values of @xmath7 at @xmath193 are the sampled values employed in eq . ( [ d_fourier - coef ] ) , but the expression is valid for any @xmath108 $ ] and constitutes an interpolation of the sampled values . naturally , the higher the sampling frequency @xmath29 , the more independent fourier coefficients we have and the more accurate the representation of @xmath7 is . as is well known , the fourier series of a function @xmath7 that is piecewise smooth converges to the function , except at its points of discontinuity , where it converges to the arithmetic mean of the two one - sided limits @xcite . however , the convergence is not _ uniform _ , so that partial sums oscillate about the true value of the function near each point of discontinuity and `` overshoot '' the two one - sided limits in opposite directions . this overshooting is known as the gibbs phenomenon , and , in our case , produces typical errors near the discontinuities of the driving function @xmath112 . these discontinuities are due to the sudden obstructions of the radiation on parts of the aircraft that occur at certain orbital positions , for example , when the sun is eclipsed . is always continuous but it undergoes sharp variations at some times . these sharp variations can be considered as discontinuities , especially , if the function is sampled . ] section [ dns ] shows that the gibbs phenomenon at eclipse points can be responsible for the largest part of the error of the linear method when the discrete fourier transform is used . the approximation of @xmath178 provided by the @xmath29 samples of @xmath7 is , of course , @xmath194 , \label{d_fourier - sol}\ ] ] and is valid for any @xmath108 $ ] . however , if we are only interested in @xmath195 at @xmath196 , @xmath197 we can compute these values with the inverse discrete fourier transform @xmath198 i/\t - j\right)^{-1 } \cdot \hat{f}(m)\ , , \label{i_d_fourier - tr}\ ] ] where , for @xmath199 @xmath200 , and where @xmath201 gives the remainder of the integer division by @xmath29 . this inverse discrete fourier transform can be more convenient for a fast numerical computation . regarding computational convenience , the discrete fourier transform , be it direct or inverse , is best performed with a fast fourier transform ( fft ) algorithm . the classic fft algorithm requires @xmath29 to be a power of two @xcite ; in particular , it has to be even . the function @xmath178 , computed by fourier analysis from @xmath29 samples of @xmath34 , is to be compared with the one computed by a numerical approximation of the integral formula , eq . ( [ t_1_lim ] ) , in terms of the same samples . naturally , we can use instead of the integral over @xmath202 $ ] the integral over @xmath203 $ ] below eq . ( [ t_1_lim ] ) . this integral can be computed from the @xmath29 samples of @xmath7 by an interpolation formula , say the trapezoidal rule . it is not easy to decide whether this procedure is more efficient than fourier transforms . considering that the substitution of the continuous fourier transform , eq . ( [ fourier - coef ] ) , by the discrete transform , eq . ( [ d_fourier - coef ] ) , is equivalent to computing the former with the trapezoidal rule , the integral formula may seem more direct . in particular , this formula allows us to select the values of @xmath41 for which we compute @xmath178 independent of the sampling frequency , so we can choose just a few distinguished orbital positions and avoid the computation of all the @xmath189 integrals ( one is removed by the condition @xmath204 ) . note that the computation of all of the independent @xmath8 with eq . ( [ d_fourier - coef ] ) is equivalent to the computation of precisely @xmath189 integrals . however , the efficiency of the fft reduces the natural operation count of this computation , of order @xmath205 , to order @xmath206 ; so its use can be advantageous , nevertheless . it goes without saying that the second - order perturbative contribution @xmath207 to the stationary solution is given by the right - hand side of eq . ( [ d_fourier - sol ] ) with the fourier coefficients @xmath8 replaced by the fourier coefficients of the function @xmath19 defined in sect . [ second - order ] . to test the previously explained methods , we construct a small thermal model of a simple spacecraft , namely , a ten - node model of a moon - orbiting satellite . our satellite ten - node model supports a basic thermal structure and is simple enough for allowing one to explicitly display the main mathematical entities , e.g. , the matrices @xmath24 , @xmath67 and @xmath20 . the satellite consists of a rectangular parallelepiped ( a cuboid ) of square base plus a small cylinder on one of its sides that simulates an observation instrument , as represented in fig . in addition , at a height of two thirds of the total height , there is an inner tray with the electronic equipment . the dimensions of the cuboid are @xmath208 , and the cylinder has a length of 0.1 m and a radius of 0.04 m. the satellite s frame is made of aluminum alloy , using plates 1 mm thick , except the bottom plate , which is 2 mm thick . this plate plays the role of a radiator and its outer surface is painted white to have high solar reflectance . the cylinder is made of the same aluminum alloy , as well as the tray ; they are 0.5 mm and 2 mm thick , respectively . the sides of the satellite , except the one with the instrument , are covered with solar cells , which increase the sides thickness to 2.25 mm . the thermal model of the satellite assigns one node to each face of the cuboid , one more to the cylinder and another to the tray , that is , eight nodes altogether . furthermore , to conveniently split the total heat capacitance of the electronic equipment , it is convenient to add two extra nodes with ( large ) heat capacitance but with no surface that could exchange heat by radiation . nodes of this type are called `` non - geometrical nodes '' . in the present case , they represent two boxes with equipment placed above and below the tray , respectively . we order the ten nodes as shown in fig . the lower box ( node 10 ) is connected to the radiator by a thermal strap . given the satellite s structure and assuming appropriate values of the specific heat capacities , it is possible to compute the capacitances @xmath209 with the result given in table [ tab1 ] . using the value of the aluminum - alloy heat conductivity and assuming perfect contact between plates , we compute the conduction coupling constants @xmath27 between nodes @xmath210 . the remaining conduction coupling constants are given reasonable values , shown in eq . ( [ kij ] ) . the computation of the radiation coupling constants @xmath211 and @xmath212 and indeed the computation of the external radiation heat inputs requires a detailed _ radiative model _ of the satellite , consisting of the geometrical _ view factors _ and the detailed thermo - optical properties of all surfaces . this radiative model allows us to compute the respective _ absorption factors _ @xcite . .node capacities and mean heat inputs with their associated steady - state temperatures . [ cols=">,>,>,>",options="header " , ] the thermo - optical properties of the surfaces are assumed to be as realistic as possible , given the simplicity of the thermal model . all radiation reflection is assumed to be diffuse , as is common for many types of surfaces . the inner surfaces are painted black and have high emissivity , @xmath213 , to favor the uniformization of the interior temperature . the outer surfaces are of three types . the three sides covered with solar cells also have high emissivity , @xmath214 , to favor the cooling of the solar cells . on the other hand , they have high solar absorptivity , @xmath215 . of this 0.75 , 0.18 is processed into electricity and the remaining 0.57 dissipates as heat in the solar cells . the top surface , the surface with the cylinder , and the cylinder itself ( its two sides ) have low emissivity , @xmath216 , and low solar absorptivity , @xmath217 , which are chosen to simulate the effect of a _ multilayer insulator_. in contrast , the bottom surface simulates a radiator , with @xmath218 and @xmath217 ( like an _ optical solar reflector _ ) . all of these thermo - optical properties are summarized in fig . [ opt ] . for the computation of the corresponding absorption factors , we employ the ray - tracing monte - carlo simulation method provided by esarad@xmath44 ( esatan@xmath44 s radiation module ) @xcite . taking into account the above information , one obtains the following conduction ( in w / k ) and radiation ( in w / k@xmath38 ) matrices : @xmath219 the satellite s thermal characteristics are defined by the data set @xmath220 but the radiation heat exchange depends on the nodal temperatures , which in turn depend on the heat input . as explained in sect . [ pert - eq ] , the appropriate set of nodal temperatures corresponds to the steady - state for averaged heat inputs , given by the algebraic equation that results from making @xmath221 in eq . ( [ bode - pert ] ) . since we need the external heat inputs and , therefore , the orbit , we proceed to define the orbit characteristics . we choose a circular equatorial orbit @xmath222 m above the moon s surface , such that @xmath223 s. the radiation heat input to the satellite consists , on the one hand , of the solar irradiation and the moon s albedo , and , on the other hand , of the moon s constant ir radiation . we take 0.12 for the mean moon s albedo and 270 k for the black - body equivalent temperature of the moon . there is also heat produced by the dissipation of electrical power in the equipment ( nodes 9 and 10 ) . for the sake of simplicity , the dissipation rate is assumed to be constant , equal to the mean electrical power generated in an orbit . in a part of the orbit , the moon eclipses the sun , so the satellite receives no direct sunlight or albedo , although there is always ir radiation from the moon . the satellite is stabilized such that the cylinder ( the `` observation instrument '' ) always points to the moon and the longer edges are perpendicular to the orbit . the radiation heat input can be computed by taking into account the given orbital characteristics and the satellite s thermo - optical characteristics , in particular , the absorption factors . it has been computed with esarad@xmath44 , taking 111 positions on the orbit , that is , at intervals of one minute . in fig . [ orbit ] , all 111 positions are plotted . the initial position of the satellite is at the subsolar point and it moves towards the east . the total external radiation heat input to the first eight nodes ( the ones that receive radiation ) is plotted in fig . [ heat - in ] ( only at every other position , for clarity ) . note the symmetry between nodes 1 and 3 , which denote the lateral faces , covered with solar cells . node 4 corresponds to the back side , also covered with solar cells . so the external radiation load on it has a similar time variation , but it is displaced . the solar radiation absorbed by all solar cells results in an orbital mean power rate of 10.5 w , dissipated in the equipment and split between nodes 9 and 10 , which receive 4.35 and 6.15 w , respectively . the external radiation absorbed by the side with the cylinder ( node 2 ) is considerably smaller than the radiation absorbed by the sides with solar cells , due to the low value of @xmath224 ( and of @xmath225 , as well ) for the corresponding surface . the bottom and top outer surfaces , which belong to nodes 5 and 6 , respectively , have view factors for the external radiation that are much less favorable than those of the side surfaces . nevertheless , the amount of lunar ir radiation absorbed by the bottom surface , due to its high @xmath225 , is such that the orbital mean of the external heat input to node 5 is , in fact , larger than the one for node 2 ( see table [ tab1 ] ) . naturally , node 7 , with no outer surfaces , does not absorb any external radiation . + to determine the hot and cold cases of the orbit , we compute the total heat load on the satellite for each position in the orbit , finding a maximum of 90.59 w at position 14 and a minimum of 18.65 w at any position in the eclipse , during which all the heat loads stay constant . the solution of the corresponding steady - state problems at position 14 and at a position in the eclipse yields the two sets of nodal temperatures ( for the given node order ) : hot : : @xmath226 . cold : : @xmath227 . the results in sect . [ j&p - sol ] show that the periodic thermal state does not reach these extreme temperatures , which could endanger the performance of the satellite . we compute the averages @xmath83 and substitute for them and the data set @xmath228 in eq . ( [ bode - pert ] ) to find the steady state temperatures . the results are given in table [ tab1 ] . then , according to eqs . ( [ jij ] ) and ( [ jii ] ) , the jacobian matrix is ( in s@xmath21 ) @xmath229 by inspection , one can check that it has nonnegative off - diagonal and negative diagonal elements , that is to say , @xmath94 is a @xmath95-matrix . it is also _ diagonally dominant _ , namely , @xmath230 their inverses ( in absolute value ) give us the typical relaxation times of the corresponding thermal modes . thus , we deduce that relaxation time of the fastest mode is about 55 s , whereas the relaxation time of the slowest one is @xmath231 s. the latter time is similar to @xmath232 s. the eigenvalues are real numbers and , furthermore , @xmath20 is diagonalizable , because the eigenvalues are different . both properties also follow from @xmath155 being almost symmetric : its antisymmetric part , @xmath233 , is relatively small , namely , @xmath234 , where the matrix norm is the frobenius norm ( other standard matrix norms yield similar values ) . therefore , the notion of `` radiation conductance '' ( sect . [ heur ] ) is appropriate in this case , as concerns its use in the linear equations . the thermal modes are almost normal , namely , the eigenvector matrix @xmath31 is such that @xmath161 with an error @xmath235 the most interesting eigenvector of @xmath20 is , of course , the positive ( perron ) eigenvector , which corresponds to the slowest mode . the normalized positive eigenvector is @xmath236 note that the temperature increments are of a similar magnitude , except the ones of node 7 and , especially , nodes 9 and 10 , which are associated , respectively , for the tray and the boxes of electronic equipment . the next mode , corresponding to the eigenvalue @xmath237 , has one negative component ( the ninth ) , and the remaining modes have more than one . to calculate @xmath178 , we choose the fourier series of eq . ( [ d_fourier - sol ] ) or , rather , the inverse discrete fourier transform of eq . ( [ i_d_fourier - tr ] ) , which can be computed with a fft algorithm . the fourier coefficients @xmath8 can also be computed with the fft , according to eq . ( [ d_fourier - coef ] ) . once the vector @xmath238 at the 111 positions is available , the set of nodal temperatures corresponding to the first - order perturbative solution is @xmath239 plotted in fig . [ tevol ] . a measure of the accuracy of this perturbative calculation is given by the second - order calculation in the next section . the truncation of the fourier series imposed by the sampling of @xmath112 also is a source of error , unrelated to perturbation theory . the piecewise smoothness of the function @xmath178 suggests that the error is small ( but see sect . [ dns ] ) . + it is also interesting to see if the first - order perturbative calculation is affected by neglecting the fastest modes : according to the analysis at the end of sect . [ pert - eq ] , these modes are expected to contribute in proportion to their relaxation times . the fastest mode relaxes in about 55 s , a short but non - negligible time . as a consequence , its contribution to @xmath238 , which we find to have a maximum magnitude of 0.8 k , is small but non - negligible . but we can deduce that still faster modes , which would appear in a thermal model of the satellite with more nodes , are hardly necessary . from the engineering standpoint , note that this satellite thermal model is successful , insofar as it predicts that all nodal temperatures stay within adequate ranges . in particular , nodes 9 and 10 , corresponding to the boxes with electronic equipment , stay within the range from @xmath240 to 23 @xmath241 . these nodes are inner nodes with large thermal capacity and , hence , are protected against the larger changes in the external heat inputs . in contrast , the outer nodes are very exposed and undergo considerable variation in temperature , with especially sharp changes at the beginning and end of the eclipse . + according to sect . [ second - order ] , the second - order perturbative correction @xmath242 to the periodic stationary solution is obtained by the same procedure as that for @xmath243 but using a different driving function @xmath126 that is computed from @xmath45 and from @xmath244 itself . the computations are straightforward and they yield the correction plotted in fig . [ t2evol ] . this correction is always negative , because the negative term in the expression for @xmath126 , eq . ( [ tf ] ) , dominates over the positive term . the equation ( [ t_1_lim ] ) for @xmath238 and the corresponding equation for @xmath242 are both linear , so @xmath238 and @xmath242 are proportional to the respective driving functions ; and we can compare their magnitudes by comparing those driving functions , say , comparing typical values of @xmath245 and @xmath246 . this latter quantity can be roughly estimated as @xmath247 w @xmath248 w , whereas @xmath249 w ( table [ tab1 ] ) . their ratio is about 25 , which roughly agrees with the ratio of @xmath238 to @xmath242 , as can be seen by comparing fig . [ tevol ] to fig . [ t2evol ] . the order of magnitude of the second - order correction suggests that higher orders are not necessary , as we show next . of course , the periodic solution @xmath250 can also be obtained by direct integration of the nonlinear eqs . ( [ ode ] ) with an adequate solver , based on runge - kutta or other methods @xcite . since the nonlinear analysis proves that the solution of eqs . ( [ ode ] ) will converge to the stationary periodic solution @xcite , the numerical solver must be run until this periodicity is established . periodicity can be enforced by comparing the nodal temperatures at the beginning and the end of every period and demanding that they be equal , within some tolerance . to do this , esatan@xmath44 provides the routine solcyc @xcite . we have employed this routine to obtain the nonlinear equations cyclic solution ( which is established in 10 periods if the tolerance is set to 0.001 ) . to compare the cyclic solution obtained by the linear method with the `` exact '' solution @xmath250 obtained by solcyc , we quantify the deviation by the vector of `` errors '' @xmath251.\ ] ] the largest components of @xmath252 are plotted in fig . [ terror ] ( the remaining components stay in the range @xmath253\,^{\circ}\mathrm{c}$ ] ) . there seem to be two branches for each node , but , in fact , it is an effect produced by the high - frequency oscillations of each @xmath254 . notice that the error is generally small when compared with @xmath207 ; namely , @xmath255 for each @xmath256 , except near the two positions corresponding to the beginning and end of the eclipse , where the heat inputs have discontinuities ( fig . [ heat - in ] ) . these discontinuities induce oscillations of certain amplitude about the true values of @xmath250 , due to the gibbs phenomenon in the discrete fourier transform ( sect . [ fourier - anal ] ) . to suppress the oscillations , we would need a specific method for treating the gibbs phenomenon ; for example , we could use a smoother method for fourier series summation , such as cesro summation @xcite the ten - node model studied in sect . [ ten - node ] is too small to pose a computational problem , whether we employ the linear method or directly integrate nonlinear eqs . ( [ ode ] ) . to assess the practical applicability of the linear method , we must study how it scales to realistic sizes and then compare it with the direct integration of eqs . ( [ ode ] ) . naturally , the first step in the method is to compute the steady - state temperature @xmath257 but this computation , arguably , is not a substantial part of the whole process and we do not consider it ( it may take , say , between a few percent and one fith of the whole process , depending on the circumstances ) . the size of an orbiting spacecraft thermal model can be scaled with respect to the number of nodes , @xmath30 , or the number of different positions in the orbit , @xmath29 . however , while realistic models must have many nodes , an @xmath29 of about one hundred can generally be suitable . note , in particular , that the condition that @xmath258 in the ten - node model is likely to be reversed in realistic models . for these reasons , the scaling with respect to @xmath30 is more relevant . for the moment , let us neglect any special feature of eqs . ( [ ode ] ) , such as possible coefficient sparsity . then , the computational complexity of numerically integrating those equations is of order @xmath259 , where @xmath23 is the number of time steps taken . on the other hand , the complexity of the numerical integration of linear eqs . ( [ linode - pert ] ) is also of order @xmath260 [ excluding the computation of @xmath98 ] . we can employ the explicit integral , eq . ( [ t_1 ] ) , which can be calculated with the trapezoidal rule , for example , but this does not reduce the complexity of the computation . however , for the stationary solution , eq . ( [ t_1_lim ] ) , its expression as an integral over a period @xmath79 sets the number of time steps to @xmath29 , which is advantageous if @xmath262 , where @xmath23 now is the number of steps necessary for some initial temperatures to relax to the stationary solution . since we have to evaluate the integral in eq . ( [ t_1_lim ] ) for several times @xmath41 , it is preferable to use the fft , as discussed in sect . [ fourier - anal ] , so that the total operation count is of order @xmath263 . however , we have not taken into account matrix operations of considerable complexity . for example , the discrete fourier transform , eq . ( [ d_fourier - sol ] ) , involves the _ inverse _ of an @xmath264 matrix ( @xmath94 plus a multiple of the identity matrix ) , and the inversion of a matrix is generally a process of order @xmath265 @xcite . if we have to employ a process of order @xmath265 , we may as well diagonalize @xmath20 , because the diagonalization of a matrix also is generally a process of order @xmath265 @xcite and the diagonalization of @xmath20 has several uses . let us assume that we carry out this diagonalization and determine the independent thermal modes , which can then be employed to express eq . ( [ t_1_lim ] ) as eq . ( [ t_1_eigenvec ] ) or to simplify the matrix operations in eq . ( [ d_fourier - sol ] ) . if we roughly compare the computational complexity of order @xmath265 with the numerical integration complexity of order @xmath259 , we deduce that the diagonalization is worthwhile when @xmath266 . using the above estimate @xmath267 and taking as relaxation time @xmath268 ( the rough value for the ten - node model ) , we deduce that the determination of thermal modes can be useful just as a computational procedure for models with a few hundred nodes . let us now consider that @xmath20 is surely a sparse matrix , so iterative matrix methods can take advantage of this characteristic . in fact , there are two degrees of sparsity in @xmath20 , associated with conduction or radiation coupling terms . the conductance matrix @xmath24 has to be very sparse , because conduction is a local process , so each node can only be coupled to a few nodes . in contrast , radiation is a nonlocal process and couples any pair of nodes that has a nonvanishing view factor . in addition to the different sparsity of conduction and radiation coupling matrices , there are two other circumstances that make them different : ( i ) we are assuming that the conductive coupling matrix just depends on material properties , so it is independent of the reference temperatures @xmath52 ; ( ii ) the conduction coupling terms are significantly larger than the radiation coupling terms for the natural values of those temperatures . all of this suggests separating the conduction and radiation parts of @xmath20 , and then diagonalizing the conduction part . based on the study in sect . [ heur ] , the diagonalization of the conduction part of @xmath20 boils down to the diagonalization of the ( generalized ) laplacian matrix @xmath269 and this matrix is sparse . suitable iterative algorithms to perform this diagonalization are , for example , the lanczos @xcite or the davidson @xcite algorithms . these algorithms are particularly useful when only a few of the largest or smallest eigenvalues are needed . this is indeed our case , as only the slower modes are expected to contribute to @xmath238 and @xmath270 once the conduction part of @xmath20 has been diagonalized , in the sense that the lowest eigenvalues and eigenvectors of the corresponding laplacian matrix are known , the radiation part of @xmath20 can be treated as a perturbation , using matrix perturbation methods @xcite . iterative matrix methods , combined with matrix perturbation methods or other methods , if necessary , can reduce the order @xmath265 to @xmath271 or even to almost linear and so allow us to diagonalize the jacobians for the largest values of @xmath30 that appear in current thermal spacecraft models . of course , the sparsity of thermal coupling matrices also facilitates the direct integration of the nonlinear eqs . ( [ ode ] ) . however , their nonlinearity prevents one from taking advantage of the above mentioned approximations methods , for example , the reduction to the small subspace of slow modes , or the splitting into conduction and radiation in which the latter is treated as a matrix perturbation . moreover , the linearization is useful in various respects . for example , the overall relaxation time , given by the eigenvalue of smallest magnitude , can be effectively bounded by inequalities @xcite and some of these bounds can be found with little computational effort . we have studied the evolution of the thermal state of an orbiting spacecraft and developed a linear approach to this problem that is based on a rigorous perturbative treatment of the exact nonlinear equations . the first - order perturbation equations , eqs . ( [ linode - pert ] ) , constitute the basic linear system , which can be applied to higher orders after calculating the corresponding driving terms . as the jacobian matrix of the nonlinear equations has negative eigenvalues , the linear equations describe the relaxation to a stationary thermal state , namely , a periodic solution that is independent of the initial conditions and only depends on the external heat input . this relaxation is similar to the relaxation to steady - state under constant external heat load . we have shown that the perturbative treatment reveals the scope of a common linearization procedure of a heuristic nature , in which the nonlinear equations are rendered linear by the definition of radiation conductances ( sect . [ heur ] ) . if one previously calculates , with the correct nonlinear eqs . ( [ bode - pert ] ) , the reference steady - state condition that corresponds to the average external heat input , the deviation from that steady - state is well approximated by the linear equations with radiation conductances . the jacobian matrix corresponding to radiation conductances , obtained in eqs . ( [ njij ] ) and ( [ njii ] ) , is related to a symmetric matrix and , therefore , is easier to diagonalize . furthermore , this relation implies that the thermal modes are normal , like the vibrational modes of a mechanical system . although the notion of radiation conductance is just an approximation , it serves nonetheless to show that the jacobian matrix is diagonalizable and has real eigenvalues . the diagonalization of the jacobian matrix is useful for the computation of the stationary thermal state and also provides information on the relaxation to that state , because the relaxation times of the thermal modes are the inverses of the eigenvalues . these times span a considerable range , but the longest times are much more significant than the shortest times , because the latter depend on the details of the lumped - parameter thermal model employed whereas the former are essentially independent of it . in fact , a thermal model that has more nodes and therefore more details also has more thermal modes ; but the slowest modes , which correspond to temperature changes in large parts of the spacecraft , are hardly affected by the details , whereas the fast modes can be significantly altered . the slowest mode , in particular , corresponds to a simultaneous but non - uniform increase ( or decrease ) of the temperature throughout the spacecraft and is hardly altered by small - scale changes . the computation of the stationary thermal state with the linear method relies on an explicit integral , eq . ( [ t_1_lim ] ) , or a fourier expansion , eq . ( [ d_fourier - sol ] ) . given a sampling of the thermal driving function at equal time intervals , the periodic solution can be obtained through two discrete fourier transforms : a direct transform to get the fourier coefficients of the driving function and an inverse transform of the coefficient vector multiplied by a suitable matrix ( sect . [ fourier - anal ] ) . of course , the discrete fourier transforms are best performed with a fft algorithm . this computation is more efficient than the numerical computation of the integral , eq . ( [ t_1_lim ] ) , if we need the values of the temperatures at all the given sampling times . however , the fourier transform presents the gibbs phenomenon , associated with sudden variations of the heat loads , as occur at eclipse times , for example . the gibbs phenomenon introduces errors , but these errors could be suppressed with special methods . the computation of the thermal modes and the stationary thermal state for a satellite ten - node thermal model confirms the validity of the linear method for a minimal but realistic model . the relaxation times span a considerable range , between 55 seconds and nearly one hundred minutes . of course , the latter time must be almost independent of the particular thermal model used , whereas the former has no intrinsic significance , and , if the number of nodes grew , that time would shrink ( thus further expanding the range of relaxation times ) . the slowest mode corresponds to node temperature increments with the same sign ( positive by convention ) , whereas the increments corresponding to other modes have both signs . the periodic variation in the external heat input ( fig . [ heat - in ] ) excites the thermal modes and produces a definite pattern of stationary temperature oscillations , well approximated by the first - order solution ( fig . [ tevol ] ) . the second - order correction is small compared to the first - order solution , but it is worth computing , as it reaches 1.7 k. higher order corrections are essentially negligible , but the error due to the gibbs phenomenon at the eclipse positions reaches 0.6 k ( at the most ) . focusing on the computational aspects of the linear approach , we have studied how it scales with the number @xmath30 of nodes and the number @xmath29 of sampling positions on the orbit . if the jacobian matrix is dense , the complexity of the corresponding matrix operations is of order @xmath265 . it is convenient to employ just one matrix operation , namely , the diagonalization of the jacobian matrix , because then only a few of the slowest modes are needed for the remaining operations , so these have negligible complexity . the complexity of a direct numerical integration of the nonlinear equations is of order @xmath259 , @xmath23 being the number of time steps necessary for relaxation . for a low - altitude orbit , @xmath23 is expected to be on the order of one thousand , as for our moon - orbiting satellite . therefore , the linear method would be computationally effective as just an integration method only for models with a few hundred nodes . at any rate , the jacobian matrix can be assumed to be sparse , and its conduction part can be assumed to be especially sparse , in addition to being the larger part of the jacobian matrix and also being independent of the orbit . as the orbit may be subjected to changes in the planning of a mission , a convenient strategy probably is to diagonalize the conduction part at the outset and , when needed , add the radiation part within some approximation scheme . this strategy can be far more efficient than integrating the nonlinear equations each time . moreover , the strength of the linear approach lies with the insight that it provides about the thermal behavior of the spacecraft , as embodied by the decomposition of its thermal modes , of which only the slowest ones are significant . these significant modes can actually be obtained with a reduced thermal model using few nodes . therefore , the linear approach is especially useful in the context of reduced models . furthermore , it provides a method for model reduction based on the mode decomposition : this decomposition can be used to group nodes . indeed , there is a technique for graph partitioning based on the eigenvalues and eigenvectors of the laplacian matrix of the graph @xcite . according to sect . [ heur ] , this technique is applicable to the jacobian matrix , but the details of this application are beyond the scope of the present paper and are left for future work . finally , our linear approach can surely be applied to other cyclic heating processes that involve radiation heat transfer . we thank isabel prez - grande for bringing ref . to our attention . f. kreith , _ radiation heat transfer for spacecraft and solar power plant design_. intnal . textbook co. , scranton , penn .

we consider 4-dimensional supersymmetric models inspired by deconstruction of 5-dimensional supersymmetric orbifold models and high dimensional non - supersymmetric models with wilson line gauge symmetry breaking . we discuss the models with bi - fundamental link fields where the gauge symmetry can be broken down to the pati - salam , , flipped or the standard model like gauge symmetry . we also propose an model with bi - fundamental link fields where the gauge symmetry is broken down to the pati - salam gauge symmetry , and an model with bi - spinor link fields where the gauge symmetry is broken down to the flipped gauge symmetry . in these two models , the pati - salam and flipped gauge symmetry can be further broken down to the standard model gauge symmetry , the doublet - triplet splittings can be obtained by the missing partner mechanism , and the proton decay problem can be solved . we also study the gauge coupling unification . we briefly comment on the interesting variation models with gauge groups and in which the proton decay problem can be solved . anl - hep - pr-04 - 66 + hep - ph/0407180 +

hep-ph0407180

supersymmetry ( susy ) provides an elegant solution to the gauge hierarchy problem , and grand unified theories ( guts ) give us a simple understanding of the quantum numbers of the standard model ( sm ) fermions . in addition , the success of gauge coupling unification in the minimal supersymmetric standard model ( mssm ) strongly supports the possibility of the susy gut . other appealing features include that the electroweak symmetry breaking is induced by radiative corrections due to the large top quark yukawa coupling , and that tiny neutrino masses can be naturally explained by the see - saw mechanism . therefore , susy gut is one of the most promising candidates that describe the known fundamental interactions in nature except gravity . however , there are problems in the 4-dimensional susy gut : the doublet - triplet splitting problem , the proton decay problem , the fermion mass problem , and the gut gauge symmetry breaking mechanism . during the last few years , orbifold guts have been studied extensively @xcite . the main point is that the supersymmetric gut models exist in 5 or 6 dimensional space - time , and they are broken down to 4-dimensional @xmath0 supersymmetric sm like models for the zero modes through compactification on various orbifolds . this is because the orbifold parity projects out the zero modes of some components in vector multiplet and hypermultiplets . the gut gauge symmetry breaking problem , the doublet - triplet splitting problem and the fermion mass problem have been solved elegantly by orbifold projections . the proton decay problem can be solved because we can define a continuous @xmath9 symmetry . other interesting phenomenology , like flavour symmetry from the @xmath10 symmetry , gauge - fermion unification , gauge - higgs unification , and gauge - yukawa unification , have also been studied @xcite . on the other hand , deconstruction was proposed about three year ago to latticize the gauge theories in higher dimensions @xcite . the idea of deconstruction is interesting because it gives a uv completion to the higher dimensional theories . applying this idea to orbifold susy guts , we are able to construct interesting 4-dimensional susy guts where the problems in the usual 4-dimensional models can be solved , in other words , the merits of orbifold guts can be preserved @xcite . in addition , inspired by deconstruction of orbifold susy guts , we can construct new models which can not be obtained from orbifold guts while still retain the nice properties in orbifold models . deconstruction of the orbifold @xmath11 models were discussed in ref . @xcite . in these models the doublet - triplet splitting problem and the proton decay problem can be solved . deconstruction may also provide insight into fermion masses and mixings @xcite . since the number of fields in these models is finite , the corrections to gauge couplings can be reliably calculated , and there exist threshold corrections to the differential runnings of the gauge couplings @xcite . elements of deconstruction can be found in earlier papers @xcite . in this paper , we construct 4-dimensional @xmath0 supersymmetric @xmath1 models inspired by deconstruction of the 5-dimensional @xmath0 supersymmetric orbifold @xmath1 models and high dimensional non - supersymmetric @xmath1 models with wilson line gauge symmetry breaking . we study @xmath2 models with bi - fundamental link fields where the gauge symmetry can be broken down to the pati - salam ( ps ) , @xmath3 , flipped @xmath4 or the sm like gauge symmetry . however , we need to fine - tune the superpotential in the models , in order to have the doublet - triplet splitting . in addition , we propose an @xmath12 model where the gauge symmetry can be broken down to ps gauge symmetry , and an @xmath6 model with bi - spinor link fields ( @xmath13 ) and ( @xmath14 ) , in which the gauge symmetry is broken down to the flipped @xmath15 gauge symmetry . in these models , the gauge symmetry can be further broken down to the sm gauge symmetry , and the doublet - triplet splitting is naturally realized through the missing partner mechanism . the proton decay due to dimension-5 operators is thus negligible and the proton lifetime due to dimension-6 operators is well above the current experimental bound because the gut scale is at least a few times @xmath16 gev . therefore , there is no proton decay problem . we also discuss the gauge coupling unification in these two models . furthermore , we briefly comment on the interesting variation models with gauge groups @xmath7 and @xmath17 where the proton decay problem can be solved , and the @xmath6 model with bi - spinor link fields in which the gauge symmetry is broken down to the @xmath3 gauge symmetry . we first give a brief review of the orbifold susy guts and the non - susy guts with wilson line gauge symmetry breaking , and deconstruction of both types of models in section [ sec : review ] . we then discuss the models where the gauge symmetry can be broken down to the diagonal ps gauge symmetry in section [ sec : patsal ] . we comment on the models inspired by deconstructions of non - susy guts with wilson line gauge symmetry breaking in section [ sec : noncom ] . in section [ sec : bispinor ] , we consider gauge symmetry breaking with bi - spinor link fields . we summarize our results in section [ sec : conclu ] . in the 5-dimensional orbifold susy guts , the 5-dimensional manifold is factorized into the product of ordinary 4-dimensional minkowski space - time @xmath18 and the orbifold @xmath19 . the corresponding coordinates are @xmath20 ( @xmath21 ) and @xmath22 . the radius for the fifth dimension is @xmath10 . the orbifold @xmath19 is obtained by @xmath23 moduloing the equivalent class @xmath24 where @xmath25 . there are two fixed points , @xmath26 and @xmath27 . the @xmath0 supersymmetric theory in 5-dimension have 8 real supercharges , corresponding to @xmath28 supersymmetry in 4-dimension . in terms of the physical degrees of freedom , the vector multiplet contains a vector boson @xmath29 with @xmath30 , two weyl gauginos @xmath31 , and a real scalar @xmath32 . in the 4-dimensional @xmath0 supersymmetry language , it contains a vector multiplet @xmath33 and a chiral multiplet @xmath34 which transform in the adjoint representation of group @xmath35 . the 5-dimensional hypermultiplet consists of two complex scalars @xmath36 and @xmath37 , and a dirac fermion @xmath38 . it can be decomposed into two chiral mupltiplets @xmath39 and @xmath40 , which are in the conjugate representations of each other under the gauge group . the general action for the group @xmath35 gauge fields and their couplings to the bulk hypermultiplet @xmath41 is @xcite @xmath42 \nonumber\\ & & + \int{d^5x } \left [ \int{d^4\theta } \left ( { \phi}^c e^v { \bar \phi}^c + { \bar \phi } e^{-v } \phi \right ) \right.\nonumber\\&&\left . + \int{d^2\theta } \left ( { \phi}^c ( \partial_5 -{1\over { \sqrt 2 } } \sigma ) \phi + { \rm h. c. } \right)\right]~.~\,\end{aligned}\ ] ] under the parity operator @xmath43 , the vector multiplet transforms as @xmath44 @xmath45 for the hypermultiplet @xmath41 and @xmath46 , we have @xmath47 @xmath48 where @xmath49 is @xmath50 , @xmath51 and @xmath52 are respectively the numbers of the fundamental index and anti - fundamental index for the bulk multiplet @xmath41 under the bulk gauge group @xmath35 . for example , if @xmath35 is an @xmath53 group , for fundamental representation , @xmath54 , @xmath55 , and for adjoint representation , @xmath54 , @xmath56 . moreover , the transformation properties for the vector multiplet and hypermultiplets under @xmath57 are the same as those under @xmath43 . for @xmath58 , to break the @xmath11 gauge symmetry , we choose the following @xmath59 matrix representations for the parity operators @xmath43 and @xmath57 @xmath60 under the @xmath57 parity , the gauge generators @xmath61 ( @xmath62 , @xmath63 , ... , @xmath64 ) for @xmath11 are separated into two sets : @xmath65 are the generators for the sm gauge group , and @xmath66 are the generators for the broken gauge group @xmath67 @xmath68 the zero modes of the @xmath69 gauge bosons are projected out , thus , the 5-dimensional @xmath0 supersymmetric @xmath11 gauge symmetry is broken down to the 4-dimensional @xmath0 supersymmetric sm gauge symmetry for the zero modes . for the zero modes and kk modes , the 4-dimensional @xmath0 supersymmetry is preserved on the 3-branes at the fixed points , and only the sm gauge symmetry is preserved on the 3-brane at @xmath27 . for @xmath70 , the generators @xmath61 of @xmath1 are imaginary antisymmetric @xmath71 matrices . in terms of the @xmath72 identity matrix @xmath73 and the pauli matrices @xmath74 , they can be written as tensor products of @xmath72 and @xmath75 matrices , @xmath76 and @xmath77 as a complete set , where @xmath78 and @xmath79 are the @xmath80 real anti - symmetric and symmetric matrices . the generators of the @xmath81 are & & _ 0 a_3 , _ 0 a_2 , _ 0 a_x + & & _ 2 s_3 , _ 2 s_2 , _ 2 s_x , the generators for flipped @xmath15 are & & _ 0 a_3 , _ 0 a_2 , _ 1 a_x + & & _ 2 s_3 , _ 2 s_2 , _ 3 a_x , and the generators for @xmath82 are @xmath83 where @xmath84 and @xmath85 are respectively the diagonal blocks of @xmath78 and @xmath79 that have indices 1 , 2 , and 3 , while the diagonal blocks @xmath86 and @xmath87 have indices 4 and 5 . @xmath88 and @xmath89 are the off diagonal blocks of @xmath78 and @xmath79 . we choose the @xmath90 matrix for @xmath43 as @xmath91 to break the @xmath1 down to @xmath3 , we choose @xmath92 to break the @xmath1 down to flipped @xmath15 , we choose @xmath93 and to break the @xmath1 down to the ps gauge symmetry , we choose @xmath94 for the zero modes , the 5-dimensional @xmath0 supersymmetric @xmath1 gauge symmetry is broken down to the 4-dimensional @xmath0 supersymmetric @xmath3 , flipped @xmath15 and the ps gauge symmetries . including the kk modes , the 3-branes at the fixed points preserve the 4-dimensional @xmath0 supersymmetry , and the gauge symmetry on the 3-brane at @xmath27 is @xmath3 , flipped @xmath15 and the ps gauge symmetries , for different choices of @xmath57 . we can also break the @xmath1 down to the sm like ( @xmath95 ) gauge symmetry , by choosing the following matrix representations for @xmath43 and @xmath57 @xmath96 or @xmath97 or @xmath98 other than the orbifold guts , wilson line gauge symmetry breaking can be applied to break high dimensional gauge symmetries @xcite . because supersymmetry can not be broken with this mechanism , only non - supersymmetric models are considered . first , let us consider the 5-dimensional space - time @xmath99 with coordinates @xmath20 , and @xmath22 . the radius for the fifth dimension is @xmath10 . the gauge fields are denoted as @xmath100 where @xmath30 . because @xmath101 belongs to the fundamental group of @xmath23 , we can define a @xmath101 parity operator @xmath102 for a generic bulk multiplet @xmath103 @xmath104 where @xmath105 and @xmath106 . for a @xmath11 model , if we choose @xmath102 to be equal to @xmath57 in eq . ( [ eq : pppsu5 ] ) , the @xmath11 gauge symmetry is broken down to the sm gauge symmetry for the zero modes . for a @xmath1 model , if we choose @xmath102 the same as @xmath57 in eqs . ( [ eq : psu5 ] ) , ( [ eq : pfsu5 ] ) or ( [ eq : pps ] ) , the @xmath1 gauge symmetry is broken down to the @xmath3 , flipped @xmath4 , or the ps gauge symmetry respectively , for the zero modes . similarly , we can consider the 6-dimensional space - time @xmath107 , with coordinates @xmath20 , @xmath22 , and @xmath108 . the radii for the fifth and sixth dimensions are @xmath109 and @xmath110 respectively . we can define two @xmath101 parity operators @xmath102 and @xmath111 for a generic bulk multiplet @xmath112 @xmath113 @xmath114 where @xmath115 , @xmath116 , @xmath106 , and @xmath117 . by choosing the following @xmath102 and @xmath111 exactly the same as @xmath43 and @xmath57 in eqs . ( [ eq : psmx1 ] ) , ( [ eq : psmx2 ] ) or ( [ eq : psmx3 ] ) , we can break the @xmath1 down to the sm like gauge symmetry . inspired by the high dimensional non - supersymmetric guts with wilson line gauge symmetry breaking , we can construct the 4-dimensional @xmath0 supersymmetric @xmath118 models where the gauge symmetry @xmath118 can be broken down to a diagonal subgroup of @xmath35 or the sm like gauge symmetry . susy gut models with wilson line gauge symmetry breaking have the problem that the 5-dimensional @xmath0 supersymmetry or the 6-dimensional @xmath28 supersymmetry can not be broken down to the 4-dimensional @xmath0 supersymmetry . however , we can start from the 4-dimensional @xmath0 supersymmetry , which is exactly the advantage of the 4-dimensional supersymmetric gut models inspired by the deconstructions . for the deconstruction of the orbifold susy guts , we break the @xmath119 to a diagonal @xmath120 as the following . suppose there are @xmath121 nodes , and the gauge symmetry on the first @xmath122 nodes is @xmath35 while the gauge symmetry on the last node is @xmath123 . the nodes are connected with @xmath122 bi - fundamental fields @xmath124 , and there is no link field between the first and last nodes . when the @xmath124 s acquire uniform vevs , @xmath125 , for all @xmath126 s . the gauge bosons of the group @xmath123 have a @xmath127 mass matrix , which has a zero eigenvalue , while the @xmath128 gauge bosons have a @xmath129 mass matrix , which does not have a zero eigenvalue @xcite . as the @xmath128 gauge bosons become heavy , @xmath119 is effectively broken down to @xmath123 for the massless fields . additional fields will be needed on the first and @xmath121-th nodes to cancel anomalies . alternatively , @xmath121 bi - fundamental fields @xmath124 connect the @xmath121 nodes to form a loop . with the same uniform vevs , the @xmath123 gauge bosons have @xmath127 mass matrices , while the @xmath128 gauge bosons have @xmath129 mass matrices . only one set of @xmath123 gauge bosons remain massless . the anomalies are always canceled in this setup , thus no additional field is required . from now on we always discuss the simplified case of two copies of the initial @xmath35 group @xmath130 where @xmath131 , with @xmath132 and @xmath133 being the gauge fields and generators of @xmath134 , and @xmath135 and @xmath136 the gauge fields and generators of @xmath137 . we assume that the gauge couplings are equal at the breaking scale for simplicity . the generalization to more copies of the @xmath35 group is straightforward . with @xmath138 and @xmath139 being two bi - fundamental fields , the covariant derivatives are [ eq : covderi ] d_u_1 = _ u_1 - i a_^ t_u_1 + i b_^ t_u_1 , + d_u_2 = _ u_2 + i a_^ t_u_2 - i b_^ t_u_2 , and the effective action for the scalar fields is [ eq : action ] s&=&d^4x \{-1 tr f_1 ^ 2 + tr[(d _ u_1)^ d^ u_1 ] - 1 tr f_2 ^ 2 + . u_2)^ d^ u_2]+ ... } . for simplicity , we only write down the effective action for the scalar fields . inspired by the high dimensional non - supersymmetric guts with wilson line gauge symmetry breaking , we can use the vevs of the bi - fundamental fields that are not commutating with a subset of the generators to break the original group @xcite . let us suppose the initial gauge group is @xmath140 and @xmath123 is a subgroup of @xmath35 . the term @xmath141 + tr[(d_{\mu } u_2)^{\dagger } d^{\mu } u_2]$ ] in eq . ( [ eq : action ] ) produces a mass matrix of the form [ eq : gmmatrx ] ( cc a_^&b_^ ) ( cc tr[u_1^t_^ t_ u_1+t_^ u_2^ u_2 t_ ] & -tr[u_1^ t_^ u_1 t_ + t_^ u_2^ t_ u_2 ] + -tr[t_^ u_1^ t_ u_1 + u_2^ t_^ u_2 t_ ] & tr[t_^ u_1^ u_1 t_ + u_2^ t_^ t_ u_2 ] ) ( c a_^ + b_^ ) . after @xmath138 and @xmath139 acquire vevs , the @xmath124 s in the mass matrix will be replaced by @xmath142 . the vev @xmath142 has the property that @xmath143 is always proportional to the identity matrix . the normalized generators satisfy the relation @xmath144 for @xmath53 or @xmath145 for @xmath146 , thus the diagonal terms in matrix in eq . ( [ eq : gmmatrx ] ) are always proportional to @xmath147 , while the off - diagonal terms depend on the commutation relations between the vevs of the bi - fundamental fields and the generators . if the vev of @xmath138 commutes with all the generators of @xmath35 , while the vev of the @xmath139 commutes with the generators of @xmath123 group , @xmath148 , and anti - commutes with those of @xmath128 , @xmath149 ^ t_a < u_1 > = t_a ^ t_a < u_1 > = t_a , ^ t_a < u_2 > = t_a ^ t_a < u_2 > = - t_a , the mass matrices become proportional to ( cc a_^a & b_^b ) ( cc 1 & -1 + -1 & 1 ) ( c a_^a + b_^b ) ( cc a_^a & b_^b ) ( cc 1 & 0 + 0 & 1 ) ( c a_^a + b_^b ) , respectively . while the former matrix has a @xmath150 eigenvalue , the latter does not . thus the gauge symmetry @xmath140 is broken down to the diagonal subgroup @xmath123 . before presenting our models , we would like to emphasize that we consider the 4-dimensional @xmath0 supersymmetry in all the models discussed in this paper . one way to realize the deconstruction of orbifold supersymmetric @xmath1 models is to start with the initial gauge group @xmath151 . it is broken down to the diagonal ps gauge symmetry by bifundamental link fields . the doublet - triplet splitting can be achieved by the missing partner mechanism , and the proton decay problem is solved because the gauge coupling unification scale is a few times @xmath16 gev . a simple variation of this model is the @xmath7 model . in addition , we can consider the @xmath6 model , where the gauge symmetry can be broken down to the diagonal ps gauge symmetry , which is inspired by the deconstruction of 5-dimensional wilson line gauge symmetry breaking . however , the doublet - triplet splitting problem can not be solved without fine - tuning in superpotential . to break @xmath152 to the diagonal ps gauge symmetry , we rely on two isomorphisms , @xmath153 and @xmath154 . we introduce the bi - fundamental fields @xmath138 and @xmath139 with the following quantum numbers @xmath155 suppose , for example , we have the superpotential @xmath156 where @xmath157 and @xmath87 are singlets , and the link fields acquire vevs < u_1 > = v ( c i_66 + 0_46 + ) , < u_2 > = v ( c 0_64 + i_44 + ) , where @xmath158 is the @xmath159 identity matrix , and @xmath160 is a @xmath161 matrix where all the entries are zero . the @xmath162 mass matrix for the ps gauge bosons is @xmath163 with eigenvalues @xmath164 . the masses of the @xmath165 gauge bosons are simply @xmath166 . if generalized to a model with gauge group @xmath167 , the ps gauge bosons have a @xmath127 mass matrix which has a zero eigenvalue , whereas the non - ps gauge bosons have a @xmath129 mass matrix , for which the zero eigenvalue is absent . three families of the sm fermions form three @xmath168 spinor representations of @xmath1 , and we introduce a 10-dimensional higgs field @xmath169 . to break the ps gauge symmetry and at the same time require a left - right symmetry , , the coupling @xmath170 of the @xmath171 equals to @xmath172 of the @xmath173 above the ps gauge symmetry unification scale ( @xmath174 ) , we introduce the fields @xmath175 , @xmath176 , @xmath177 , @xmath178 , @xmath179 , @xmath180 , @xmath181 and @xmath182 . their quantum numbers under the gauge symmetry @xmath183 are given in the following @xmath184 note that @xmath175 , @xmath176 , @xmath177 , @xmath185 can form one pair of @xmath168 and @xmath186 under @xmath1 . the ps gauge symmetry is broken down to the sm gauge symmetry when the right - handed neutrino scalar component in @xmath175 and its charge conjugation in @xmath187 obtain vevs at the @xmath174 scale . the superpotential is @xmath188 where @xmath189 s are yukawa couplings , and @xmath190 s are the usual yukawa couplings for the sm fermions . moreover , similarly to the doublet - triplet splitting via the missing partner mechanism , the color triplets in @xmath169 and @xmath179 obtain the gut scale masses while the doublets are massless because of the superpotential term @xmath191 . subsequently , the proton decay due to the dimension-5 operator is negligible because the mixing of the color triplets through the @xmath192 term is very small . limits on the contributions from the dimension-6 proton decay operators require that the gauge coupling unification scale be larger than @xmath193 gev @xcite , which is obviously satisfied in our model where the gut scale is at least a few times @xmath16 gev . furthermore , the @xmath192 term for the higgs doublets can be generated from the superpotential @xmath194 , as is similar to the next to the minimal supersymmetric standard model @xcite . we also assume that the masses for @xmath180 , @xmath175 and @xmath187 are around @xmath174 , and the masses for @xmath179 and higgs triplets in @xmath169 are about @xmath195 . we do not discuss the fermion masses and mixings , as they are out of the scope of this paper . for simplicity , we assume the universal masses @xmath196 gev for the supersymmetric particles . from @xmath197 to @xmath198 , the gauge couplings evolve the same way as those in the sm , and the beta functions are @xmath199 , where @xmath200 , @xmath201 and @xmath202 are for the gauge symmetries @xmath203 , @xmath171 and @xmath204 , respectively . from @xmath198 to the ps gauge symmetry unification scale @xmath174 , the beta functions are just those of the mssm , _ i.e. _ , @xmath205 . at the ps gauge symmetry unification scale @xmath174 , the gauge couplings for @xmath203 and @xmath204 are _ 1 ^ -1(m_ps ) = _ 1 ^ -1(m_z ) - ( ) - ( ) , _ 3 ^ -1(m_ps ) = _ 3 ^ -1(m_z ) - ( ) - ( ) . the @xmath173 gauge coupling @xmath172 is related to @xmath206 and @xmath207 by _ r^-1(m_ps ) = _ 1 ^ -1(m_ps ) - _ 3 ^ -1(m_ps ) . above the ps scale , including the contributions from fermions , ps gauge multiplets , two higgs doublets @xmath208 , @xmath175 @xmath187 , @xmath209 , @xmath210 , and @xmath180 , the beta functions @xmath211 become b^ii & = & b(m ) + b^ps(v ) + b(h^d ) + b ( ) + b(c_2 ) + & = & ( 6,6,6 ) + ( -6,-6,-12 ) + ( 1,1,0 ) + ( 4 , 4 , 4 ) + ( 0,0,1 ) = ( 5,5,-1 ) , where @xmath212 , @xmath213 and @xmath214 are the beta functions for @xmath173 , @xmath171 and @xmath215 , respectively . in the region above @xmath216 , one set of @xmath217 gauge multiplets , @xmath179 and higgs triplets @xmath218 in @xmath169 appear , then the beta functions are b^iii & = & b^ii + b^so(10)/ps(v ) + b(c_1 ) + b(h^t ) + & = & ( 5,5,-1 ) + ( -18,-18,-12 ) + ( 0 , 0 , 1 ) + ( 0 , 0 , 1 ) = ( -13,-13,-11 ) . we show the runnings of the gauge couplings near the unification scale in the left panel of fig . [ fig : uni_ps1 ] . note that the runnings of @xmath170 and @xmath172 coincide above @xmath174 because we have required the left - right symmetry . the gauge coupling @xmath219 of @xmath220 unify with @xmath172 and @xmath170 at the scale @xmath221 gev and the ps scale occurs at @xmath222 gev . thus , there is no proton decay problem . right : the gauge coupling unification for the left - right non - symmetric ps model near @xmath223 gev . [ fig : uni_ps1],title="fig:",width=302 ] gev . right : the gauge coupling unification for the left - right non - symmetric ps model near @xmath223 gev . [ fig : uni_ps1],title="fig:",width=302 ] in the above construction , we maintain the left - right symmetry above the @xmath174 scale . if we no longer require the left - right symmetry , we can drop the additional fields @xmath209 and @xmath224 from the table in eq . ( [ so10epf ] ) . in this setup , the discussion of the doublet - triplet splitting is unchanged from that of the above symmetric case . the rge runnings of the gauge couplings are different due to the simpler field content . more specifically , the beta functions are , @xmath225 and @xmath226 . above @xmath227 gev , @xmath172 and @xmath170 have different rge runnings , as is evident in fig . [ fig : uni_ps1 ] right panel . with a gut scale of @xmath228 gev , the proton lifetime is safely above the current experimental bound . if we want to solve the proton decay problem , we can also consider the model with gauge group @xmath7 , which can not be obtained but is inspired by the orbifold gut . to break @xmath229 to ps gauge symmetry , we introduce bi - fundamental field @xmath138 as @xmath230 the link fields acquire the following vevs < u_1 > = v ( c i_66 + 0_46 + ) . the @xmath7 gauge symmetry is broken down to the ps gauge symmetry . the discussions on the proton decay , ps gauge symmetry breaking and the gauge coupling unification are similar to those in the above subsection , so , we will not repeat them here . however , we would like to emphasize that the initial gauge group is smaller than that in the above subsection . inspired by the deconstruction of 5-dimensional @xmath231 model with wilson line gauge symmetry breaking , we consider the model with the simplest gauge group @xmath6 . let us introduce bi - vector fields @xmath138 and @xmath139 with the following quantum numbers @xmath232 the effective action is the same as that in eq . ( [ eq : action ] ) . if the @xmath138 and @xmath139 fields acquire vevs < u_1 > = v _ 0 ( 1 , 1 , 1,1,1 ) , < u_2 > = v _ 0 ( -1,-1,-1,1,1 ) , the @xmath162 mass matrix for the ps gauge bosons is @xmath233 with @xmath234 defined in eq . ( [ eq : sm32 ] ) . the squared masses for the ps group gauge bosons are either @xmath150 or @xmath235 . the mass matrix of the non - ps ( @xmath217 ) gauge bosons is @xmath236 it has degenerate eigenstates with squared masses equal to @xmath237 . three families of the sm fermions form three @xmath168 spinor representations under the first @xmath1 while they are singlets under the second @xmath1 . we also introduce a higgs field @xmath169 . to give masses to the color triplets in the higgs field @xmath169 , we introduce a 10-dimensional field @xmath238 , and to break the ps gauge symmetry , we introduce the fields @xmath239 , @xmath240 , and two singlets @xmath181 and @xmath182 . the quantum numbers for these particles are @xmath241 the superpotential is @xmath242 where @xmath243 is the @xmath1 unification scale . the doublet - triplet splitting can be obtained through the fine - tunning superpotential @xmath244 , as in the usual @xmath11 model . because it is less interesting than the usual @xmath1 model from the phenomenological point of view , we do not study it in detail . for the @xmath247 and @xmath248 models from the deconstruction of 5-dimensional supersymmetric orbifold @xmath1 models , the discussions on how to break the gauge symmetry down to the diagonal flipped @xmath249 or @xmath250 gauge symmetry are identical to those of section [ sec2.1 ] . two link fields with simple diagonal vevs will leave the diagonal subgroup gauge bosons massless and give the other gauge bosons masses of order @xmath251 . we will consider the models inspired by deconstruction of the 5-dimensional @xmath1 models with wilson line gauge symmetry breaking , where the @xmath1 gauge symmetry can be broken down to the flipped @xmath246 and @xmath250 , and the 6-dimensional @xmath1 model where the @xmath1 gauge symmetry can be broken down to the sm like gauge symmetry . starting with the @xmath252 , we use two bi - vectors @xmath138 and @xmath139 with quantum numbers @xmath253 . in order to break the gauge symmetry down to the diagonal @xmath81 , the vevs of the link fields are assumed to be < u_1 > & = & v _ 0 ( 1,1,1,1,1 ) , + < u_2 > & = & v _ 2 ( 1,1,1,1,1 ) , and to obtain the flipped @xmath15 , @xmath139 vev is replaced with < u_2 > = v _ 2 ( 1,1,1,-1,-1 ) . note the similarity between the vevs and the parity operators in eqs . ( [ eq : psu5 ] ) and ( [ eq : pfsu5 ] ) . as @xmath254 commutes with @xmath148 and @xmath149 , while @xmath255 commutes with @xmath148 and anti - commutes with @xmath149 , the gauge bosons of the unbroken subgroup @xmath256 or flipped @xmath15 have a mass matrix identical to the right hand side of eq . ( [ eq : so102 ] ) while the @xmath128 gauge bosons have a mass matrix the same as the right hand side of eq . ( [ eq : so2nops ] ) . one set of the @xmath123 gauge bosons remain massless and the @xmath128 gauge bosons acquire masses of @xmath257 . moreover , the @xmath258 can be broken to the intersection of two maximal subgroups , @xmath259 . if we introduce four bi - vector link fields with vevs & = & v _ 0 ( 1,1,1,1,1 ) , + < u_2 > & = & v _ 2 ( 1,1,1,1,1 ) , + < u_3 > & = & v _ 0 ( 1,1,1,1,1 ) , + < u_4 > & = & v _ 0 ( 1,1,1,-1,-1 ) , where all @xmath124 s are @xmath260 under the original group @xmath2 . it is easy to see that @xmath138 and @xmath139 are the same as the link fields we used for breaking @xmath261 to @xmath81 , and @xmath262 and @xmath263 are the same as those for breaking @xmath2 to the ps group . to clarify the notation , we assign @xmath264 for the mass matrices of the gauge bosons in the intersection of the @xmath3 and ps gauge symmetry , which is @xmath265 , @xmath266 for those in @xmath81 but not in ps gauge group , @xmath267 for those in the ps gauge group but not in @xmath81 , and @xmath268 is reserved for the mass matrices of the rest gauge bosons belong to neither @xmath3 or ps gauge group . these mass matrices are given by m_smx^2 = 4 g^2 v^2 ( cc 2&-2 + -2&2 + ) , m_5 ^ 2 = m_ps^2 = 4 g^2 v^2 ( cc 2&-1 + -1&2 + ) , m_r^2 = 4 g^2 v^2 ( cc 2&0 + 0&2 + ) . it is obvious that only @xmath264 has 0 eigenvalue , and the others do not . with the combination of these two sets of link fields , the @xmath269 gauge symmetry is broken to the @xmath270 gauge symmetry . similarly , we can break @xmath269 down to @xmath95 by choosing & = & v _ 0 ( 1,1,1,1,1 ) , + < u_2 > & = & v _ 2 ( 1,1,1,-1,-1 ) , + < u_3 > & = & v _ 0 ( 1,1,1,1,1 ) , + < u_4 > & = & v _ 0 ( 1,1,1,-1,-1 ) , or < u_1 > & = & v _ 0 ( 1,1,1,1,1 ) , + < u_2 > & = & v _ 2 ( 1,1,1,1,1 ) , + < u_3 > & = & v _ 0 ( 1,1,1,1,1 ) , + < u_4 > & = & v _ 2 ( 1,1,1,-1,-1 ) . in these models , the doublet - triplet splitting is still a problem without fine - tuning in the superpotential , thus it is difficult to solve the proton decay problem induced by the dimension-5 proton decay operators . in this section , we demonstrate that , by using bi - spinor link fields ( @xmath13 ) and ( @xmath271 ) , we are able to break the gauge symmetry @xmath6 down to the diagonal @xmath3 and flipped @xmath15 gauge symmetry , and at the same time solve the doublet - triplet splitting problem via the missing partner mechanism in the model with the flipped @xmath15 breaking chain . in this approach , we do not rely on the commutation relations between the link field vevs and the generators . we introduce two bi - fundamental fields @xmath138 and @xmath139 with following quantum numbers under @xmath272 @xmath273 then , the covariant derivative can be rewritten as , using @xmath274 defined in appendix [ apdx ] , d_u_1 = _ u_1 - u_1 + u_1 , + d_u_2 = _ u_2 + u_2 - u_2 . the effective action for scalar components is the same as that in eq . ( [ eq : action ] ) , and the bi - spinors @xmath138 and @xmath139 acquire the following vevs = < u_2 > = v diag(0,0,0,1,0,0,0,1,1,1,1,0,0,0,0,1 ) . note that the assignment of the @xmath275 s are consistent with the diagonal ( @xmath276 ) and ( @xmath277 , @xmath277 ) in the ( @xmath278 ) , see apppendix [ apdx ] . the gauge fields in eq . ( [ eq : gauge16 ] ) have the mass matrices of the following forms m_^2 = m_v^2= m_w_l^2= m_y^2 = g^2 v^2 x_1 , m_a^2 = m_w_r^2 = m_x^2 = 2 g^2 v^2 x_2 . thus , the massless fields are the 5 states in the @xmath279 fields , 6 @xmath280 , 2 @xmath281 and 12 @xmath282 states , total of 25 independent states . they are the gauge bosons of the diagonal @xmath81 group . for the flipped @xmath15 , the @xmath283 and @xmath277 consist of different fermion fields . the vevs of @xmath138 and @xmath139 are chosen as = < u_2 > = v diag(0,0,0,1,0,0,0,1,0,0,0,1,1,1,1,0 ) . the mass matrices become m_^2 = m_v^2= m_w_l^2= m_a^2 = g^2 v^2 x_1 , m_y^2 = m_w_r^2 = m_x^2 = 2 g^2 v^2 x_2 , and the massless states are 12 @xmath284 , 2 @xmath281 , 6 @xmath280 and the 5 states in @xmath279 s . these 25 massless states are the gauge bosons of the diagonal flipped @xmath15 . three families of the sm fermions form three @xmath168 spinor representations under the first @xmath1 while they are singlets under the second @xmath1 . we also introduce a 10-dimensional higgs field @xmath169 , and one pair of higgs @xmath239 and @xmath240 in the spinor representation . moreover , we introduce one pair of the fields @xmath285 and @xmath286 in the spinor representation , two singlet fields @xmath181 and @xmath182 . the quantum numbers for these particles are @xmath287 the superpotential is @xmath288 where @xmath289 is the unification scale for the flipped @xmath15 , and @xmath290 is the planck scale . the ( @xmath291 ) multiplets in @xmath239 and @xmath285 , and the ( @xmath292 ) multiplets in @xmath293 and @xmath286 are assumed to have the @xmath1 unification scale masses from the first two terms in the superpotential , while the @xmath294 multiplet @xmath175 in @xmath239 , the @xmath295 multiplet @xmath296 in @xmath285 , the @xmath297 multiplet @xmath176 in @xmath293 , and the @xmath298 multiplet @xmath299 in @xmath300 are still massless at @xmath301 . for simplicity , we assume that the masses @xmath302 for @xmath296 and @xmath299 are about @xmath303 gev from the last term in superpotential . under flipped @xmath15 , @xmath169 is decomposed into one pair of 5-plets @xmath304 and @xmath305 . to be explicit , we denote @xmath175 , @xmath187 , @xmath304 and @xmath305 as @xmath306 @xmath307 from the third , fifth and sixth terms in eq . ( [ superpotential33 ] ) , we obtain the superpotential below @xmath301 @xmath308 there is only one f - flat and d - flat direction , which can always be rotated along the @xmath309 and @xmath310 directions . hence we obtain that @xmath311 . in addition , the superfields in @xmath175 and @xmath176 are eaten or acquire large masses via the supersymmetric higgs mechanism , except for @xmath312 and @xmath313 . the superpotential @xmath314 and @xmath315 couple the @xmath312 and @xmath316 with the @xmath317 and @xmath318 respectively to form the massive eigenstates with masses @xmath319 and @xmath320 . we naturally have the doublet - triplet splitting due to the missing partner mechanism @xcite . because the triplets in @xmath304 and @xmath321 only have small mixing through the @xmath192 term , the higgsino - exchange mediated proton decay are negligible , _ i.e. _ , we do not have the dimension-5 proton decay problem . proton decay via the dimension-6 operators is well above the current experimental bounds . the gauge coupling unification for the flipped @xmath246 is realized by first unifying @xmath322 and @xmath207 at scale @xmath323 , then the gauge couplings of @xmath11 and @xmath324 further unify at @xmath301 . from @xmath197 to @xmath198 , the beta functions are @xmath325 , and from @xmath198 to @xmath326 gev , the beta functions are @xmath205 . from @xmath302 to the @xmath322 and @xmath207 unification scale @xmath323 , the beta functions are @xmath327 , because @xmath328 and @xmath299 contribute to the gauge coupling rge runnings . unification of @xmath322 and @xmath207 at @xmath323 gives the condition @xmath329 which can be solved for @xmath323 . the coupling @xmath330 of @xmath324 is related to @xmath206 and @xmath331 at @xmath323 by _ 1^-1(m_23 ) = _ 1 ^ -1(m_23 ) - _ 5 ^ -1(m_23 ) . between @xmath323 and @xmath301 , besides the 3 families of the standard model fermions @xmath332 , there are the @xmath15 gauge bosons in @xmath333 , @xmath169 , @xmath334 , @xmath335 , @xmath328 and @xmath299 , thus the beta functions for @xmath324 and @xmath11 are @xmath336 . in fig . [ fig : flipuni ] , we show the runnings of the gauge couplings near the unification scale @xmath301 . the unification scale @xmath323 for @xmath171 and @xmath204 is about @xmath337 gev , and the @xmath6 unification scale @xmath301 is about @xmath338 gev . if the masses @xmath302 for @xmath328 and @xmath339 are larger than @xmath323 , we will have @xmath340 at @xmath323 . because above @xmath323 the beta function of @xmath324 is positive while that of @xmath11 is negative , we can not achieve the unification of these two gauge couplings . in short , the intermediate mass scale @xmath302 is important for the gauge coupling unification @xcite . gev for flipped @xmath15 model . [ fig : flipuni],width=302 ] we introduce the link fields @xmath138 , @xmath139 , @xmath262 and @xmath263 with following quantum numbers @xmath342 we choose the following vevs for the link fields @xmath343 and @xmath344 , where @xmath345 is a @xmath346 matrix with elements at @xmath347 , @xmath348 , @xmath349 , @xmath350 , and @xmath351 being @xmath352 and all other elements @xmath150 , and @xmath353 is a column vector with @xmath352 as the @xmath354 element and all other elements @xmath150 . similarly to above subsection , we can break the @xmath355 gauge symmetry down to a diagonal flipped @xmath15 gauge symmetry . three families of the sm fermions form three @xmath168 spinor representations under the first @xmath1 while they are singlets and neutral under the @xmath356 . we also introduce a 10-dimensional higgs field @xmath169 , and one pair of higgs @xmath239 and @xmath293 in the spinor representation . moreover , we introduce one pair of the 5-plet fields @xmath357 and @xmath358 , one pair of the 10-plet fields @xmath328 and @xmath299 , one pair of the @xmath324 charged singlets @xmath359 and @xmath360 , and two singlet fields @xmath181 and @xmath182 . the quantum numbers for these particles are @xmath361 the superpotential is @xmath362 the discussions for the proton decay and the gauge coupling unification are similar to these in the previous subsection . we want to emphasize that the doublet - triplet splitting problem can be solved by the missing partner mechanism , and the solution of the proton decay problem follows . we have considered the 4-dimensional @xmath0 supersymmetric @xmath1 models inspired by the deconstruction of 5-dimensional supersymmetric orbifold @xmath1 models and high dimensional non - supersymmetric @xmath1 models with wilson line gauge symmetry breaking . we studied the @xmath2 models with bi - fundamental link fields where the @xmath2 gauge symmetry can be broken down to the ps , @xmath3 , flipped @xmath4 and the sm like gauge symmetry . however , we need to fine - tune the superpotential to obtain the doublet - triplet splitting . we proposed two interesting models : the @xmath363 model where the gauge symmetry can be broken down to ps gauge symmetry , and the @xmath6 model with bi - spinor link fields in which the gauge symmetry can be broken down to the flipped @xmath15 gauge symmetry . these intermediate gauge symmetries can be further broken down to the sm gauge symmetry . in these models , the missing partner mechanism can be implemented to solve the doublet - triplet splitting problem , then the higgsino - exchange mediated proton decay is negligible , and consequently the proton decay is mainly induced by the dimension-6 operators . with the gut scale being at least a few times @xmath16 gev , the proton lifetime is well above the current experimental bound . in addition , we discussed the gauge coupling unification in these two models : in the @xmath12 models , the gauge couplings unify at @xmath364 gev for a left - right symmetric model and at @xmath365 gev for a left - right non - symmetric model ; and in the @xmath6 model , the gauge couplings unify at about @xmath366 gev . furthermore , we briefly commented on the interesting variation models with gauge groups @xmath7 and @xmath367 where the proton decay problem can be solved in the similar manner , and the @xmath6 model with bi - spinor link fields in which the gauge symmetry can be broken down to the @xmath3 gauge symmetry . j. jiang thanks csaba balazs and cosmas zachos for helpful discussions . the research of c .- s . huang was supported in part by the natural science foundation of china , the research of j. jiang was supported by the u.s . department of energy under grant no . w-31 - 109-eng-38 , and the research of t. li was supported by the national science foundation under grant no . phy-0070928 . the generators and the assignment of the fermions in the * 16 * can be found in ref . we copy the @xmath368 , and rename it @xmath369 . the @xmath370 matrix can be re - written into four @xmath371 matrices , [ eq : gauge16 ] / = ( cccccccc _ 1 1&v_12&v_13&x_1 ^ 0&w_l^- & & & + v_12^*&_2 2&v_23&x_2 ^ - & & w_l^- & & + v_13^*&v_23^*&_3 3&x_3 ^ - & & & w_l^- & + _ 1 ^ 0&x_2^+&x_3^+&_4 4 & & & & w_l^- + w_l^+ & & & & _ 5 5&v_12&v_13&x_1 ^ 0 + & w_l^+ & & & v_12^*&_6 6&v_23&x_2 ^ - + & & w_l^+ & & v_13^*&v_23^*&_7 7&x_3 ^ - + & & & w_l^+&_1 ^ 0&x_2^+&x_3^+&_8 8 cccccccc 0&a_6 ^ 0&-a_5 ^ 0&-y_1^+&0&-y_6 ^ -&y_5 ^ -&-_1 ^ 0 + -a_6 ^ 0&0&a_4 ^ -&-_2 ^ 0&y_6 ^ -&0&-y_4^&-a_2 ^ - + a_5 ^ 0&-a_4 ^ -&0&-_3 ^ 0&-y_5 ^ -&y_4^&0&-a_3 ^ - + y_1^+&_2 ^ 0&_3 ^ 0&0&_1 ^ 0 & a_2 ^ -&a_3 ^ -&0 + 0&-a_3^+&a_2^+&-y_4^++&0&y_3 ^ 0&-y_2 ^ 0&-a_4^+ + a_3^+&0&a_1 ^ 0&-y_5^+&y_3 ^ 0&0&-y_1 ^ -&-_5 ^ 0 + -a_2^+&a_1 ^ 0&0&-y_6^+&y_2 ^ 0&-y_1 ^ -&0&-_6 ^ 0 + y_4^++&y_5^+&y_6^+&0&a_4^+&_5 ^ 0&_6 ^ 0&0 cccccccc 0&-_6 ^ 0&_5 ^ 0&y_1 ^ -&0&a_3 ^ -&-a_2 ^ -&y_4^ + _ 6 ^ 0&0&-a_4^+&y_2 ^ 0&-a_3 ^ -&0&_1 ^ 0&y_5 ^ - + -_5 ^ 0&a_4^+&0&y_3 ^ 0&a_2 ^ -&-_1 ^ 0&0&y_6 ^ - + -y_1 ^ -&-y_2 ^ 0&-y_3 ^ 0&0&-y_4 ^ -&-y_5 ^ -&-y_6 ^ -&0 + 0&y_6^+&-y_5^+&a_1 ^ 0&0&-_3 ^ 0&_2 ^ 0&a_4 ^ - + -y_6^+&0&y_4^++&a_2^+&_3 ^ 0&0&-y_1^+&a_5 ^ 0 + y_5^+&-y_4^++&0&a_3^+&-_2 ^ 0&y_1^+&0&a_6 ^ 0 + -a_1 ^ 0&-a_2^+&-a_3^+&0&-a_4 ^ -&-a_5 ^ 0&-a_6 ^ 0&0 cccccccc _ 9 9&-v_12^*&-v_13^*&-_1 ^ 0&w_r^- & & & + -v_12&_10 10&-v_23^*&-x_2^+ & & w_r^- & & + -v_13&-v_23&_11 11&-x_3^+ & & & w_r^- & + -x_1 ^ 0&-x_2 ^ -&-x_3 ^ -&_12 12 & & & & w_r^- + w_r^+ & & & & _ 13 13&-v_12^*&-v_13^*&-_1 ^ 0 + & w_r^+ & & & -v_12&_14 14&-v_23^*&-x_2^+ + & & w_r^+ & & -v_13&-v_23&_15 15&-x_3^+ + & & & w_r^+&-x_1 ^ 0&-x_2 ^ -&-x_3 ^ 3&_16 16 the 45 gauge bosons consist of 12 @xmath284 , 6 @xmath372 , 6 @xmath280 , 12 @xmath282 , 2 charge @xmath281 , 2 charge @xmath373 and 16 @xmath279 which can be rewritten as 5 independent fields , @xmath374 , @xmath375 , @xmath376 , @xmath377 and @xmath378 . the first family of the sm fermions forms a @xmath168 , _ 1= ( u_r , u_g , u_b,_e , d_r , d_g , d_b , e^-,d_r^c , d_g^c , d_b^c , e^+,-u_r^c ,- u_b^c,-_e^c)^t , similarly for the second and third families . as the @xmath1 is broken down to @xmath250 or flipped @xmath246 , the spinor representation @xmath168 is decomposed as + * |5 * + * 1 * where = ( q , u^c , e^+ ) , = ( d^c , l ) , = ^c for breaking to @xmath250 , and = ( q , d^c,^c ) , = ( u^c , l ) , = e^+ for breaking to flipped @xmath246 . l. j. hall , h. murayama and y. nomura , nucl . b * 645 * , 85 ( 2002 ) [ arxiv : hep - th/0107245 ] . n. haba , t. kondo , y. shimizu , t. suzuki and k. ukai , prog . theor . phys . * 106 * , 1247 ( 2001 ) [ arxiv : hep - ph/0108003 ] . t. asaka , w. buchmuller and l. covi , phys . b * 523 * , 199 ( 2001 ) ; [ arxiv : hep - ph/0108021 ] . t. li , nucl . b * 619 * , 75 ( 2001 ) [ arxiv : hep - ph/0108120 ] . r. dermisek and a. mafi , phys . d * 65 * , 055002 ( 2002 ) [ arxiv : hep - ph/0108139 ] . c. s. huang , j. jiang , t. li and w. liao , phys . b * 530 * , 218 ( 2002 ) [ arxiv : hep - th/0112046 ] . t. li , nucl . b * 633 * , 83 ( 2002 ) [ arxiv : hep - th/0112255 ] . b. kyae and q. shafi , phys . d * 69 * , 046004 ( 2004 ) [ arxiv : hep - ph/0212331 ] . h. d. kim and s. raby , jhep * 0301 * , 056 ( 2003 ) [ arxiv : hep - ph/0212348 ] . i. gogoladze , y. mimura and s. nandi , phys . b * 562 * , 307 ( 2003 ) [ arxiv : hep - ph/0302176 ] ; phys . * 91 * , 141801 ( 2003 ) [ arxiv : hep - ph/0304118 ] . t. li , jhep * 0403 * , 040 ( 2004 ) [ arxiv : hep - ph/0309199 ] . n. arkani - hamed , a. g. cohen and h. georgi , phys . rev . lett . * 86 * , 4757 ( 2001 ) [ arxiv : hep - th/0104005 ] . c. t. hill , s. pokorski and j. wang , phys . d * 64 * , 105005 ( 2001 ) [ arxiv : hep - th/0104035 ] . c. csaki , g. d. kribs and j. terning , phys . d * 65 * , 015004 ( 2002 ) [ arxiv : hep - ph/0107266 ] . h. c. cheng , k. t. matchev and j. wang , phys . b * 521 * , 308 ( 2001 ) [ arxiv : hep - ph/0107268 ] . t. li and t. liu , eur . j. c * 28 * ( 2003 ) 545 [ arxiv : hep - th/0204128 ] . see , , m. b. halpern and w. siegel , phys . d * 11 * , 2967 ( 1975 ) ; v. d. barger , w. y. keung and e. ma , phys . b * 94 * , 377 ( 1980 ) . n. arkani - hamed , t. gregoire and j. wacker , jhep * 0203 * , 055 ( 2002 ) . y. hosotani , phys . lett . b*126 * ( 1983 ) 309 ; phys . lett . b*129 * ( 1983 ) 193 ; e. witten , nucl . b*258 * ( 1985 ) 75 ; j. d. breit , b. a. ovrut and g. c. segre , phys . lett . b*158 * ( 1985 ) 33 . g. f. giudice and a. romanino , arxiv : hep - ph/0406088 . j. ellis , j. f. gunion , h. e. haber , l. roszkowski , and f. zwirner , phys . d * 39 * , 844 ( 1989 ) ; and references therein . recent references may be found in u. ellwanger , j. f. gunion , and c. hugonie , hep - ph/0111179 . c. s. huang , t. li and w. liao , nucl . b * 673 * , 331 ( 2003 ) [ arxiv : hep - ph/0304130 ] . j. l. lopez , d. v. nanopoulos and a. zichichi , arxiv : hep - ph/9307211 , and references therein . s. rajpoot , phys . d * 22 * , 2244 ( 1980 ) .

one of the major advances needed to realize all - optical information processing of light is the ability to delay or coherently store and retrieve optical information in a rapidly tunable manner . in the classical domain , this optical buffering is expected to be a key ingredient to managing the flow of information over complex optical networks . such a system also has profound implications for quantum information processing , serving as a long - term memory that can store the full quantum information contained in an optical pulse . here we suggest a novel approach to light storage involving an optical waveguide coupled to an optomechanical crystal array , where light in the waveguide can be dynamically and coherently transferred into long - lived mechanical vibrations of the array . under realistic conditions , this system is capable of achieving large bandwidths and storage / delay times in a compact , on - chip platform .

1006.3829

light is a natural candidate to transmit information across large networks due to its high speed and low propagation losses . a major obstacle to building more advanced optical networks is the lack of an all - optically controlled device that can robustly delay or store optical wave - packets over a tunable amount of time . in the classical domain , such a device would enable all - optical buffering and switching , bypassing the need to convert an optical pulse to an electronic signal . in the quantum realm , such a device could serve as a memory to store the full quantum information contained in a light pulse until it can be passed to a processing node at some later time . a number of schemes to coherently delay and store optical information are being actively explored . these range from tunable coupled resonator optical waveguide ( crow ) structures @xcite , where the propagation of light is dynamically altered by modulating the refractive index of the system , to electromagnetically induced transparency ( eit ) in atomic media @xcite , where the optical pulse is reversibly mapped into internal atomic degrees of freedom . while these schemes have been demonstrated in a number of remarkable experiments @xcite , they remain difficult to implement in a practical setting . here , we present a novel approach to store or stop an optical pulse propagating through a waveguide , wherein coupling between the waveguide and a nearby nano - mechanical resonator array enables one to map the optical field into long - lived mechanical excitations . this process is completely quantum coherent and allows the delay and release of pulses to be rapidly and all - optically tuned . our scheme combines many of the best attributes of previously proposed approaches , in that it simultaneously allows for large bandwidths of operation , on - chip integration , relatively long delay / storage times , and ease of external control . beyond light storage , this work opens up the intriguing possibility of a platform for quantum or classical all - optical information processing using mechanical systems . an optomechanical crystal @xcite is a periodic structure that constitutes both a photonic @xcite and a phononic @xcite crystal . the ability to engineer optical and mechanical properties in the same structure should enable unprecedented control over light - matter interactions . planar two - dimensional ( 2d ) photonic crystals , formed from patterned thin dielectric films on the surface of a microchip , have been succesfully employed as nanoscale optical circuits capable of efficiently routing , diffracting , and trapping light . fabrication techniques for such 2d photonic crystals have matured significantly over the last decade , with experiments on a si chip @xcite demonstrating excellent optical transmission through long ( @xmath0 ) linear arrays of coupled photonic crystal cavities . in a similar si chip platform it has recently been shown that suitably designed photonic crystal cavities also contain localized acoustic resonances which are strongly coupled to the optical field via radiation pressure @xcite . these planar optomechanical crystals ( omcs ) are thus a natural candidate for implementation of our proposed slow - light scheme . and @xmath1 , whose resonance frequencies differ by the frequency of the mechanical mode @xmath2 . both optical modes leak energy into the waveguide at a rate @xmath3 and have an inherent decay rate @xmath4 . the mechanical resonator optomechanically couples the two optical resonances with a cross coupling rate of @xmath5 . ( b)a simplified system diagram where the classically driven cavity mode @xmath1 is effectively eliminated to yield an optomechanical driving amplitude @xmath6 between the mechanical mode and cavity mode @xmath7 . ( c ) frequency - dependent reflectance ( black curve ) and transmittance ( red ) of a single array element , in the case of no optomechanical driving amplitude @xmath8 ( dotted line ) and an amplitude of @xmath9 ( solid line ) . the inherent cavity decay is chosen to be @xmath10 . ( inset ) the optomechanical coupling creates a transparency window of width @xmath11 for a single element and enables perfect transmission on resonance , @xmath12 . ( d ) energy level structure of simplified system . the number of photons and phonons are denoted by @xmath13 and @xmath14 , respectively . the optomechanical driving amplitude @xmath6 couples states @xmath15 while the light in the waveguide couples states @xmath16 . the two couplings create a set of @xmath17-type transitions analogous to that in eit.[fig : system],width=642 ] in the following we consider an optomechanical crystal containing a periodic array of such defect cavities ( see figures [ fig : system](a),(b ) ) . each element of the array contains two optical cavity modes ( denoted @xmath18 ) and a co - localized mechanical resonance . the hamiltonian describing the dynamics of a single element is of the form _ = _ 1+_2+_m^+h(+^)(+).[eq : h ] here @xmath19 are the resonance frequencies of the two optical modes , @xmath20 is the mechanical resonance frequency , and @xmath21 are annihilation operators for these modes . the optomechanical interaction cross - couples the cavity modes @xmath22 and @xmath23 with a strength characterized by @xmath5 and that depends linearly on the mechanical displacement @xmath24 . while we formally treat @xmath21 as quantum mechanical operators , for the most part it also suffices to treat these terms as dimensionless classical quantities describing the positive - frequency components of the optical fields and mechanical position . in addition to the optomechanical interaction described by equation ( [ eq : h ] ) , the cavity modes @xmath22 are coupled to a common two - way waveguide ( described below ) . each element is decoupled from the others except through the waveguide . the design considerations necessary to achieve such a system are discussed in detail in the section `` optomechanical crystal design . '' for now , we take as typical parameters @xmath25 thz , @xmath26 ghz , @xmath27 mhz , and mechanical and ( unloaded ) optical quality factors of @xmath28 ( room temperature)-@xmath29 ( low temperature ) and @xmath30 , where @xmath31 is the mechanical decay rate and @xmath32 is the intrinsic optical cavity decay rate . similar parameters have been experimentally observed in other omc systems @xcite . in practice , one can also over - couple cavity mode @xmath22 to the waveguide , with a waveguide - induced optical decay rate @xmath3 that is much larger than @xmath4 . for the purpose of slowing light , the cavity modes @xmath23 will be resonantly driven by an external laser , so that to good approximation @xmath33 can be replaced by its mean - field value . we furthermore consider the case where the frequencies are tuned such that @xmath34 . keeping only the resonant terms in the optomechanical interaction , we arrive at a simplified hamiltonian for a single array element ( see figure [ fig : system](b ) ) , h_=_1+_m^+_m(t)(e^-i(_1-_m)t+h.c.).[eq : hlinear ] here we have defined an effective optomechanical driving amplitude @xmath35 and assume that @xmath36 is real . mode @xmath23 thus serves as a `` tuning '' cavity that mediates population transfer ( rabi oscillations ) between the `` active '' cavity mode @xmath22 and the mechanical resonator at a controllable rate @xmath37 , which is the key mechanism for our stopped - light protocol . in the following analysis , we will focus exclusively on the active cavity mode and drop the `` @xmath22 '' subscript . a hamiltonian of the form ( [ eq : hlinear ] ) also describes an optomechanical system with a single optical mode , when the cavity is driven off resonance at frequency @xmath38 and @xmath39 corresponds to the sidebands generated at frequencies @xmath40 around the classical driving field . for a single system , this hamiltonian leads to efficient optical cooling of the mechanical motion @xcite , a technique being used to cool nano - mechanical systems toward their quantum ground states @xcite . while the majority of such work focuses on how optical fields affect the mechanical dynamics , here we show that the optomechanical interaction strongly modifies optical field propagation to yield the slow / stopped light phenomenon . equation ( [ eq : hlinear ] ) is quite general and thus this phenomenon could in principle be observed in any array of optomechanical systems coupled to a waveguide . in practice , there are several considerations that make the 2d omc `` ideal . '' first , our system exhibits an extremely large optomechanical coupling @xmath5 and contains a second optical tuning cavity that can be driven resonantly , which enables large driving amplitudes @xmath6 using reasonable input power @xcite . using two different cavities also potentially allows for greater versatility and addressability of our system . for instance , in our proposed design the photons in cavity @xmath22 are spatially filtered from those in cavity 2 @xcite . second , the 2d omc is an easily scalable and compact platform . finally , as described below , the high mechanical frequency of our device compared to typical optomechanical systems allows for a good balance between long storage times and suppression of noise processes . we first analyze propagation in the waveguide when @xmath41 is static during the transit interval of the signal pulse . as shown in the appendix , the evolution equations in a rotating frame for a single element located at position @xmath42 along the waveguide are given by & = & -+i_m+i((z_j)+(z_j))+(z_j),[eq : dadt ] + & = & -+i_m+(t).[eq : dbdt ] equation ( [ eq : dadt ] ) is a standard input relation characterizing the coupling of right- ( @xmath43 ) and left - propagating ( @xmath44 ) optical input fields in the waveguide with the cavity mode . here @xmath45 is the total optical cavity decay rate , @xmath46 is quantum noise associated with the inherent optical cavity loss , and for simplicity we have assumed a linear dispersion relation @xmath47 in the waveguide . equation ( [ eq : dbdt ] ) describes the optically driven mechanical motion , which decays at a rate @xmath48 and is subject to thermal noise @xmath49 . the cavity mode couples to the right - propagating field through the equation ( + ) ( z , t)=i(z - z_j)+ik_0,[eq : waveeqn ] where @xmath50 . we solve the above equations to find the reflection and transmission coefficients @xmath51 of a single element for a right - propagating incoming field of frequency @xmath52 ( see appendix ) . in the limit where @xmath53 , and defining @xmath54 , r(_k)=- , while @xmath55 . example reflectance and transmittance curves are plotted in figure [ fig : system](c ) . for any non - zero @xmath6 , a single element is perfectly transmitting on resonance , whereas for @xmath8 resonant transmission past the cavity is blocked . when @xmath56 , excitation of the cavity mode is inhibited through destructive interference between the incoming field and the optomechanical coupling . in eit , a similar effect occurs via interference between two electronic transitions . this analogy is further elucidated by considering the level structure of our optomechanical system ( figure [ fig : system](d ) ) , where the interference pathways and the `` @xmath17''-type transition reminiscent of eit are clearly visible . the interference is accompanied by a steep phase variation in the transmitted field around resonance , which can result in a slow group velocity . these steep features and their similarity to eit in a single optomechanical system have been theoretically @xcite and experimentally studied @xcite , while interference effects between a single cavity mode and two mechanical modes have also been observed @xcite . . the optical cavity modes @xmath57 of each element leak energy into the waveguide at a rate @xmath3 and have an inherent decay rate @xmath4 . the mechanical resonator of each element has frequency @xmath58 and is optomechanically coupled to the cavity mode through a tuning cavity ( shown in figure [ fig : system ] ) with strength @xmath6 . ( b ) the band structure of the system , for a range of driving strengths between @xmath8 and @xmath59 . the blue shaded regions indicate band gaps , while the color of the bands elucidates the fractional occupation ( red for energy in the optical waveguide , green for the optical cavity , and blue for mechanical excitations ) . the dynamic compression of the bandwidth is clearly visible as @xmath60 . ( c ) band structure for the case @xmath61 is shown in greater detail . ( d ) the fractional occupation for each band in ( c ) is plotted separately . it can be seen that the polaritonic slow - light band is mostly mechanical in nature , with a small mixing with the waveguide modes and negligible mixing with the optical cavity mode . zoom - ins of figures ( c ) and ( d ) are shown in ( e ) and ( f ) . [ fig : bandstructure],width=642 ] from @xmath51 for a single element , the propagation characteristics through an infinite array ( figure [ fig : bandstructure](a ) ) can be readily obtained via band structure calculations @xcite . to maximize the propagation bandwidth of the system , we choose the spacing @xmath62 between elements such that @xmath63 where @xmath64 is a non - negative integer . with this choice of phasing the reflections from multiple elements destructively interfere under optomechanical driving . typical band structures are illustrated in figures [ fig : bandstructure](b)-(f ) . the color coding of the dispersion curves ( red for waveguide , green for optical cavity , blue for mechanical resonance ) indicates the distribution of energy or fractional occupation in the various degrees of freedom of the system in steady - state . far away from the cavity resonance , the dispersion relation is nearly linear and simply reflects the character of the input optical waveguide , while the propagation is strongly modified near resonance ( @xmath65 ) . in the absence of optomechanical coupling ( @xmath8 ) , a transmission band gap of width @xmath66 forms around the optical cavity resonance ( reflections from the bare optical cavity elements constructively interfere ) . in the presence of optomechanical driving , the band gap splits in two ( blue shaded regions ) and a new propagation band centered around the cavity resonance appears in the middle of the band gap . for weak driving ( @xmath67 ) the width of this band is @xmath68 , while for strong driving ( @xmath69 ) one recovers the `` normal mode splitting '' of width @xmath70 @xcite . this relatively flat polaritonic band yields the slow - light propagation of interest . indeed , for small @xmath6 the steady - state energy in this band is almost completely mechanical in character , indicating the strong mixing and conversion of energy in the waveguide to mechanical excitations along the array . it can be shown that the bloch wavevector near resonance is given by ( see appendix ) k_0+++.[eq : keff ] the group velocity on resonance , @xmath71 can be dramatically slowed by an amount that is tunable through the optomechanical coupling strength @xmath6 . the quadratic and cubic terms in @xmath72 characterize pulse absorption and group velocity dispersion , respectively . in the relevant regime where @xmath73 , these effects are negligible within a bandwidth @xmath74 . the second term is the bandwidth over which certain frequency components of the pulse acquire a @xmath75-phase shift relative to others , leading to pulse distortion . this yields a bandwidth - delay product of ( , ( 6n^2)^1/3)[eq : bwdelayproduct ] for static @xmath6 and negligible mechanical losses . when intrinsic optical cavity losses are negligible , and if one is not concerned with pulse distortion , light can propagate over the full bandwidth @xmath68 of the slow - light polariton band and the bandwidth - delay product increases to @xmath76 ( see appendix ) . on the other hand , we note that if we had operated in a regime where @xmath77 , constructive interference in reflection would limit the bandwidth - delay product to @xmath78 , independent of system size . in the static regime , the bandwidth - delay product obtained here is analogous to crow systems @xcite . in the case of eit , a static bandwidth - delay product of @xmath79 results , where od is the optical depth of the atomic medium . this product is limited by photon absorption and re - scattering into other directions , and is analogous to our result @xmath80 in the case of large intrinsic cavity linewidth . on the other hand , when @xmath4 is negligible , photons are never lost and reflections can be suppressed by interference . this yields an improved scaling @xmath81 or @xmath82 , depending on whether one is concerned with group velocity dispersion . in atomic media , the weak atom - photon coupling makes achieving @xmath83 very challenging @xcite . in contrast , in our system as few as @xmath84 elements would be equivalently dense . we now show that the group velocity @xmath85 can in fact be adiabatically changed once a pulse is completely localized inside the system , leading to distortion - less propagation at a dynamically tunable speed . in particular , by tuning @xmath86 , the pulse can be completely stopped and stored . this phenomenon can be understood in terms of the static band structure of the system ( figure [ fig : bandstructure ] ) and a `` dynamic compression '' of the pulse bandwidth . the same physics applies for crow structures @xcite , and the argument is re - summarized here . first , under constant @xmath6 , an optical pulse within the bandwidth of the polariton band completely enters the medium . once the pulse is inside , we consider the effect of a gradual reduction in @xmath87 . decomposing the pulse into bloch wavevector components , it is clear that each bloch wavevector is conserved under arbitrary changes of @xmath6 , as it is fixed by the system periodicity . furthermore , transitions to other bands are negligible provided that the energy levels are varied adiabatically compared to the size of the gap , which translates into an adiabatic condition @xmath88 . then , conservation of the bloch wavevector implies that the bandwidth of the pulse is dynamically compressed , and the reduction in slope of the polariton band ( figure [ fig : bandstructure ] ) causes the pulse to propagate at an instantaneous group velocity @xmath89 without any distortion . in the limit that @xmath90 , the polaritonic band becomes flat and completely mechanical in character , indicating that the pulse has been reversibly and coherently mapped onto stationary mechanical excitations within the array . we note that since @xmath6 is itself set by the tuning cavities , its rate of change can not exceed the optical linewidth and thus the adiabaticity condition is always satisfied in the weak - driving regime . the maximum storage time is set by the mechanical decay rate , @xmath91 . for realistic system parameters @xmath26 ghz and @xmath92 , this yields a storage time of @xmath93s . in crow structures , light is stored as circulating fields in optical nano - cavities , where state of the art quality factors of @xmath94 limit the storage time to @xmath95 ns . the key feature of our system is that we effectively `` down - convert '' the high - frequency optical fields to low - frequency mechanical excitations , which naturally decay over much longer time scales . while storage times of @xmath96 ms are possible using atomic media @xcite , their bandwidths so far have been limited to @xmath97 mhz @xcite . in our system , bandwidths of @xmath95 ghz are possible for realistic circulating powers in the tuning cavities . the major source of error in our device will be mechanical noise , which through the optomechanical coupling can be mapped into noise power in the optical waveguide output . in our system , mechanical noise emerges via thermal fluctuations and stokes scattering ( corresponding to the counter - rotating terms in the optomechanical interaction that we omitted from equation ( [ eq : hlinear ] ) ) . to analyze these effects , it suffices to consider the case of static @xmath6 , and given the linearity of the system , no waveguide input ( such that the output will be purely noise ) . for a single array element , the optomechanical driving @xmath6 results in optical cooling of the mechanical motion @xcite , with the mechanical energy @xmath98 evolving as ( see appendix ) = -_m(e_m-_m|n_)-e_m+(e_m+_m).[eq : demdt ] the first term on the right describes equilibration with the thermal surroundings , where @xmath99 is the bose occupation number at the mechanical frequency and @xmath100 is the bath temperature . the second ( third ) term corresponds to cooling ( heating ) through anti - stokes ( stokes ) scattering , with a rate proportional to @xmath101 . the stokes process is suppressed relative to the anti - stokes in the limit of good sideband resolution @xmath102 . for an array of @xmath103 elements , a simple upper bound for the output noise power at one end of the waveguide is given by @xmath104 , where @xmath105 is the steady - state solution of equation ( [ eq : demdt ] ) . the factor of @xmath106 accounts for the optical noise exiting equally from both output directions , @xmath107 is the optically - induced mechanical energy dissipation rate , and @xmath108 describes the waveguide coupling efficiency . the term @xmath109 represents the transduction of mechanical to optical energy and is essentially the price that one pays for down - converting optical excitations to mechanical to yield longer storage times in turn , any mechanical noise gets `` up - converted '' to optical energy ( whereas the probability of having a thermal optical photon is negligible ) . in the relevant regime where @xmath110 , ( _ m|n_+()^2 ) . this noise analysis is valid only in the weak - driving regime ( @xmath67 ) @xcite . the strong driving regime , where the mechanical motion acquires non - thermal character @xcite and can become entangled with the optical fields @xcite , will be treated in future work . at room temperature , @xmath111 is large and thermal noise will dominate , yielding a noise power of @xmath112 nw per element for previously given system parameters and @xmath113 . this is independent of @xmath114 provided that @xmath110 , which reflects the fact that all of the thermal heating is removed through the optical channel.for high temperatures , the thermal noise scales inversely with @xmath58 , and the use of high - frequency mechanical oscillators ensures that the noise remains easily tolerable even at room temperature . thermal noise in the high - frequency oscillator can essentially be eliminated in cryogenic environments , which then enables faithful storage of single photons . intuitively , a single - photon pulse can be stored for a period only as long as the mechanical decay time @xmath115 , and as long as a noise - induced mechanical excitation is unlikely to be generated over a region covering the pulse length and over the transit time @xmath116 . the latter condition is equivalent to the statement that the power @xmath117 in the single - photon pulse exceeds @xmath118 . while we have focused on the static regime thus far , when thermal heating is negligible , realizing @xmath119 in the static case in fact ensures that the inequality holds even when @xmath120 is time - varying . physically , the rate of stokes scattering scales linearly with @xmath114 while the group velocity scales inversely , and thus the probability of a noise excitation being added on top of the single - photon pulse is fixed over a given transit length . in a realistic setting , the optomechanical driving amplitude @xmath6 itself will be coupled to the bath temperature , as absorption of the pump photons in the tuning cavities leads to material heating . to understand the limitations as a quantum memory , we have numerically optimized the static bandwidth - delay product @xmath121 for a train of single - photon pulses , subject to the constraints @xmath122 , @xmath123 , and @xmath124 . as a realistic model for the bath temperature , we take @xmath125 , where @xmath126 is the base temperature and @xmath127k is a temperature coefficient that describes heating due to pump absorption ( see appendix ) . using @xmath128 mk and @xmath129 , we find @xmath130 , which is achieved for parameter values @xmath131 , @xmath1321.1 ghz , and @xmath133 mhz . highlighted in the center . the unit cell contains two coupled l2 defect cavities ( shaded in grey ) with two side - coupled linear defect optical waveguides ( shaded in red ) . the different envelope functions pertain to the odd and even optical cavity supermodes ( green solid and dashed lines , respectively ) , the odd mechanical supermode ( blue solid line ) , and the odd and even optical waveguide modes ( red solid and dashed lines ) . the displacement field amplitude @xmath134 of the mechanical mode and in - plane electric field amplitude @xmath135 of the optical mode are shown in ( b ) and ( c ) , respectively , for a single l2 defect cavity . ( d ) bandstructure of the linear - defect waveguide ( grey ) and the zone folded superlattice of the entire coupled - resonator system ( red ) . the cavity mode ( green ) crosses the superlattice band at mid - zone , and the waveguide - cavity interaction is shown in more detail in the insets ( e ) and ( a ) for the even ( @xmath136 ) and odd ( @xmath137 ) supermodes , respectively . [ fig : omc_model],width=642 ] a schematic showing a few periods of our proposed 2d omc slow - light structure is given in figure [ fig : omc_model ] . the structure is built around a `` snowflake '' crystal pattern of etched holes into a silicon slab @xcite . this pattern , when implemented with a physical lattice constant of @xmath138 , snowflake radius @xmath139 , and snowflake width @xmath140 ( see figure [ fig : omc_model](a ) ) , provides a simultaneous phononic bandgap from @xmath141 to @xmath142 ghz and a photonic pseudo - bandgap from @xmath143 to @xmath144 thz ( see appendix ) . owing to its unique bandgap properties , the snowflake patterning can be used to form waveguides and resonant cavities for both acoustic and optical waves simply by removing regions of the pattern . for instance , a single point defect , formed by removing two adjacent holes ( a so - called `` l2 '' defect ) , yields the co - localized phononic and photonic resonances shown in figures [ fig : omc_model](b ) and ( c ) , respectively . the radiation pressure , or optomechanical coupling between the two resonances can be quantified by a coupling rate , @xmath145 , which corresponds to the frequency shift in the optical resonance line introduced by a single phonon of the mechanical resonance . numerical finite - element - method ( fem ) simulations of the l2 defect indicate the mechanical resonance occurs at @xmath146 , with a coupling rate of @xmath147 to the optical mode at frequency @xmath148 ( free - space optical wavelength of @xmath149 ) . in order to form the double - cavity system described in the slow - light scheme above , a pair of l2 cavities are placed in the near - field of each other as shown in the dashed box region of figure [ fig : omc_model](a ) . modes of the two degenerate l2 cavities mix , forming supermodes of the double - cavity system which are split in frequency . the frequency splitting between modes can be tuned via the number of snowflake periods between the cavities . as described in more detail in the appendix , it is the optomechanical cross - coupling of the odd ( @xmath150 ) and even ( @xmath151 ) optical supermodes mediated by the motion of the odd parity mechanical supermode ( @xmath152 ) of the double - cavity that drives the slow - light behaviour of the system . since @xmath152 is a displacement field that is antisymmetric about the two cavities , there is no optomechanical self - coupling between the optical supermodes and this mechanical mode . on the other hand , the cross - coupling between the two different parity optical supermodes is large and given by @xmath153 . by letting @xmath39 , @xmath154 and @xmath155 be the annihilation operators for the modes @xmath150 , @xmath151 and @xmath152 , we obtain the system hamiltonian of equation ( [ eq : h ] ) . the different spatial symmetries of the optical cavity supermodes also allow them to be addressed independently . to achieve this we create a pair of linear defects in the snowflake lattice as shown in figure [ fig : omc_model](a ) , each acting as a single - mode optical waveguide at the desired frequency of roughly @xmath156 ( see figure [ fig : omc_model](d ) . sending light down both waveguides , with the individual waveguide modes either in or out of phase with each other , will then excite the even or odd supermode of the double cavity , respectively . the waveguide width and proximity to the l2 cavities can be used to tune the cavity loading ( see appendix ) , which for the structure in figure [ fig : omc_model](a ) results in the desired @xmath157 . it should be noted that these line - defect waveguides do not guide phonons at the frequency of @xmath152 , and thus no additional phonon leakage is induced in the localized mechanical resonance . the full slow - light waveguide system consists of a periodic array of the double - cavity , double - waveguide structure . the numerically computed band diagram , for spacing @xmath158 periods of the snowflake lattice between cavity elements ( the superlattice period ) , is shown in figure [ fig : omc_model](d ) . this choice of superlattice period results in the folded superlattice band intersecting the @xmath150 ( @xmath39 ) cavity frequency @xmath159 at roughly mid - zone , corresponding to the desired inter - cavity phase shift of @xmath160 . a zoom - in of the bandstructure near the optical cavity resonances is shown in figures [ fig : omc_model](e ) and ( f ) . in figure [ fig : omc_model](e ) the even parity supermode bandstructure is plotted ( i.e. , assuming the even supermode of the double - waveguide is excited ) , whereas in figure [ fig : omc_model](f ) it is the odd parity supermode bandstructure . with @xmath161 , and the attenuation per unit cell @xmath162 ( solid black line ) are shown as a function of detuning of the pump beam from the pump cavity frequency . the dotted line is the approximate expression derived for the attenuation , @xmath163 . the grey region indicates the band gap in which the pump cavities can not be excited from the waveguide . the trade - off between small pump input powers and low pump attenuation factors is readily apparent in this plot . [ fig : inputpower],width=642 ] a subtlety in the optical pumping of the periodically arrayed waveguide system is that for the @xmath151 ( @xmath154 ) optical cavity resonance at @xmath164 , there exists a transmission bandgap . in order to populate cavity @xmath154 , then , and to create the polaritonic band at @xmath159 , the pump beam must be slightly off - resonant from @xmath164 , but still at @xmath165 . we achieve this by choosing a double - cavity separation ( @xmath166 periods ) resulting in a cavity mode splitting ( @xmath167 ) slightly smaller than the mechanical frequency ( @xmath146 ) , as shown in figures [ fig : omc_model](e ) and ( f ) . by changing the detuning between @xmath168 and @xmath58 , a trade - off can be made between the attenuation of the pump beam per unit cell , @xmath169 , and total required input power , shown in figure [ fig : inputpower ] . in [ app : input ] we show that the total attenuation per unit cell , is given by @xmath163 . interestingly , by using higher input powers such that @xmath170 , it is in principle possible to eliminate completely effects due to absorption in the array , which may lead to inhomogeneous pump photon occupations . the possibility of using optomechanical systems to facilitate major tasks in classical optical networks has been suggested in several recent proposals @xcite . this present work not only extends these prospects , but proposes a fundamentally new direction where optomechanical systems can be used to control and manipulate light at a quantum mechanical level . such efforts would closely mirror the many proposals to perform similar tasks using eit and atomic ensembles @xcite . at the same time , the optomechanical array has a number of novel features compared to atoms , in that each element can be deterministically positioned , addressed , and manipulated , and a single element is already optically dense . furthermore , the ability to freely convert between phonons and photons enables new possibilities for manipulating light through the manipulation of sound . taken together , this raises the possibility that mechanical systems can provide a novel , highly configurable on - chip platform for realizing quantum optics and `` atomic '' physics . this work was supported by the darpa / mto orchid program through a grant from afosr . dc acknowledges support from the nsf and the gordon and betty moore foundation through caltech s center for the physics of information . asn acknowledges support from nserc . m.h . acknowledges support from the u.s . army research office muri award w911nf0910406 . here we derive the equations of motion for an array of optomechanical systems coupled to a two - way waveguide . because each element in the array couples independently to the waveguide , it suffices here to only consider a single element , from which the result for an arbitrary number of elements is easily generalized . we model the interaction between the active cavity mode @xmath22 and the waveguide with the following hamiltonian , h_&=&_-^dk(ck-_1)-_-^dk(ck+_1)- + & & g_-^dz(z - z_j)(((z)+(z))+h.c.).[eqsi : hcavwg ] here @xmath171 are annihilation operators for left- and right - going waveguide modes of wavevector @xmath172 , and @xmath159 is the frequency of the cavity mode . for convenience we have defined all optical energies relative to @xmath159 , and assumed that the waveguide has a linear dispersion relation @xmath173 . the last term on the right describes a point - like coupling between the cavity ( at position @xmath42 ) and the left- and right - going waveguide modes , with a strength @xmath145 . the operator @xmath174 physically describes the annihilation of a right - going photon at position @xmath175 and is related to the wavevector annihilation operators by @xmath176 ( with a similar definition for @xmath177 ) . ( [ eqsi : hcavwg ] ) resembles a standard hamiltonian used to formulate quantum cavity input - output relations @xcite , properly generalized to the case when the cavity accepts an input from either direction . note that we make the approximation that the left- and right - going waves can be treated as separate quantum fields , with modes in each direction running from @xmath178 . this allows both the left- and right - going fields to separately satisfy canonical field commutation relations , @xmath179=\left[{\mbox{$\hat{a}_{l}$}}(z),{\mbox{$\hat{a}_{l}$}}(z')\right]=\delta(z - z')$ ] , while commuting with each other . thus each field contains some unphysical modes ( _ e.g. _ , wavevector components @xmath180 for the right - going field ) , but the approximation remains valid as long as there is no process in the system evolution that allows for the population of such modes . from the hamiltonian above , one finds the following heisenberg equation of motion for the right - going field , ( + ) ( z)=(z - z_j)+ik_0,[eqsi : waveeqn ] where @xmath50 . a similar equation holds for @xmath181 . the coupling of the cavity mode to a continuum of waveguide modes leads to irreversible decay of the cavity at a rate @xmath3 . below , we will show that @xmath3 is related to the parameters in the hamiltonian by @xmath182 . with this identification , one recovers eq . ( 5 ) in the main text . the heisenberg equation of motion for the cavity mode is given by = ig((z_j)+(z_j ) ) . to cast this equation into a more useful form , we first integrate the field equation ( [ eqsi : waveeqn ] ) across the discontinuity at @xmath42 , ( z_j^+ ) & = & ( z_j^-)+ , + ( z_j^- ) & = & ( z_j^+)+ . we can define @xmath183 and @xmath184 as the input fields to the cavity . it then follows that = ig(+)- , and thus we indeed see that the waveguide induces a cavity decay rate @xmath185 . in the case where the cavity has an additional intrinsic decay rate @xmath4 , a similar derivation holds to connect the intrinsic decay with some corresponding noise input field @xmath186 . from these considerations , and including the opto - mechanical coupling , one arrives at eq . ( 3 ) in the main text , = -+i_m+i((z_j)+(z_j))+(z_j).[eqsi : dadt ] finally , we consider the equation of motion for the mechanical mode given by eq . ( 4 ) in the main text , = -+i_m_1+(t).[eqsi : dbdt ] the zero - mean noise operator @xmath187 must accompany the decay term in the mechanical evolution in order to preserve canonical commutation relations of @xmath188 at all times . in the case where the decay causes the mechanical motion to return to thermal equilibrium with some reservoir at temperature @xmath100 , the noise operator has a two - time correlation function given by @xmath189 @xcite , where @xmath190 is the bose occupation number at the mechanical frequency @xmath58 . first we derive the reflection and transmission coefficients for a single element in the case of constant opto - mechanical driving amplitude @xmath6 . given the linearity of the system , it suffices to treat eqs . ( [ eqsi : waveeqn ] ) , ( [ eqsi : dadt ] ) , and ( [ eqsi : dbdt ] ) as classical equations for this purpose , and furthermore to set the noise terms @xmath191 , @xmath192 . for concreteness , we will consider the case of an incident right - going cw field in the waveguide . upon interaction with the opto - mechanical system at @xmath42 , the total right - going field can be written in the form a_r(z)=e^ikz - i_kt((-z+z_j)+t(_k)(z - z_j ) ) , while the left - going field is given by @xmath193 . here @xmath194 is the unit step function , @xmath195 is the detuning of the input field from the cavity resonance , and @xmath51 are the reflection and transmission coefficients for the system . at the same time , we look for solutions of the cavity field and mechanical mode of the form @xmath196 and @xmath197 . the coefficients @xmath198 can be obtained by substituting this ansatz into eqs . ( [ eqsi : waveeqn ] ) , ( [ eqsi : dadt ] ) , and ( [ eqsi : dbdt ] ) . this yields the reflection coefficient given by eq . ( 6 ) in the main text , while the transmission coefficient is related by @xmath55 . to calculate propagation through an array of @xmath103 elements , it is convenient to introduce a transfer matrix formalism . specifically , the fields immediately to the right of the opto - mechanical element ( at @xmath199 ) can be related to those immediately to the left ( at @xmath200 ) in terms of a transfer matrix @xmath201 , ( c a_r(z_j^+ ) + a_l(z_j^+ ) ) = m_om ( c a_r(z_j^- ) + a_l(z_j^- ) ) , where m_om= ( cc t^2-r^2 & r + -r & 1 ) . [ eqsi : mom ] on the other hand , free propagation in the waveguide is characterized by the matrix @xmath202 , ( c a_r(z+d ) + a_l(z+d ) ) = m_f ( c a_r(z ) + a_l(z ) ) , where m_f= ( cc e^ikd & 0 + 0 & e^-ikd ) .[eqsi : mf ] the transfer matrix for an entire system can then be obtained by successively multiplying the transfer matrices for a single element and for free propagation together . in particular , the transfer matrix for a single `` block '' , defined as interaction with a single opto - mechanical element followed by free propagation over a distance @xmath62 to the next opto - mechanical element , is given by @xmath203 , and the propagation over @xmath103 blocks is simply characterized by @xmath204 . before studying the propagation through the entire array , we first focus on the propagation past two blocks , @xmath205 . because we want our device to be highly transmitting when the optomechanical coupling is turned on , we choose the spacing @xmath62 between consecutive blocks to be such that @xmath206 , where @xmath64 is an integer . physically , this spaces consecutive elements by an odd multiple of @xmath207 , where @xmath208 is the resonant wavelength , such that the reflections from consecutive elements tend to destructively interfere with each other . this can be confirmed by examining the resulting reflection coefficient for the two - block system , r_2=-+o(_k^3 ) , where @xmath209 denotes matrix elements of @xmath210 . note that the reflection coefficient is now suppressed as a quadratic in the detuning , whereas for a single element @xmath211 is linear . in the above equation , we have made the simplifying approximation that @xmath212 , since in realistic systems the dispersion from free propagation will be negligible compared to that arising from interaction with an opto - mechanical element . now we can consider transmission past @xmath213 pairs of two blocks ( _ i.e. _ , @xmath103 elements in total ) . because the reflection @xmath214 is quadratic in the detuning , its effect on the total transmission is only of order @xmath215 ( because the lowest order contribution is an event where the field is reflected twice before passing through the system ) . thus , up to @xmath216 , the total transmission coefficient @xmath217 is just given by @xmath218 , where @xmath219 is the transmission coefficient for a two - block system . it is convenient to write @xmath220 in terms of an effective wavevector @xmath72 , which leads to eq . ( 7 ) in the main text . performing a similar analysis for the case where @xmath221 , where reflections from consecutive elements interfere constructively , one finds that the bandwidth - delay product for the system @xmath78 does not improve with the system size . in this section we derive the optical cooling equation given by eq . ( 9 ) in the main text . we begin by considering the system hamiltonian for a single element ( eq . ( 1 ) of main text ) , in the case where the tuning mode @xmath23 is driven on resonance and can be approximated by a classical field , @xmath222 , _ om=_1+_m^+h(+^)(_2e^-i_2t+_2^e^i_2t).[eqsi : h ] defining a detuning @xmath223 indicating the frequency difference between the two cavity modes , we can re - write eq . ( [ eqsi : h ] ) in a rotating frame , _ om=-_l+_m^+_m(+^)(+),[eqsi : hrot ] where @xmath224 ( we have re - defined the phases such that @xmath6 is real ) . in the weak driving limit ( @xmath67 ) , the cavity dynamics can be formally eliminated to arrive at effective optically - induced cooling equations for the mechanical motion @xcite . in particular , the opto - mechanical coupling terms @xmath225 and @xmath226 induce anti - stokes and stokes scattering , respectively . these processes yield respective optically - induced cooling ( @xmath227 ) and heating ( @xmath228 ) rates _ = . in the case where @xmath229 , the cooling process is resonantly enhanced by the cavity , yielding a cooling rate @xmath230 as given in the main text . also in this case , the optical heating rate is given by @xmath231 . this leads to the net cooling dynamics given by eq . ( 9 ) in the main text , = -_m(e_m-_m|n_th)-e_m+(e_m+_m).[eqsi : demdt ] because the optical cooling process removes phonons from the mechanical system via optical photons that leak out of the cavity , one can identify @xmath232 as the amount of optical power that is being leaked by the cavity in the anti - stokes sideband during the cooling process . similarly , the cavity leaks an amount of power @xmath233 in the stokes sideband . we have ignored this contribution in eq . ( 10 ) in the main text , because its large frequency separation ( @xmath234 ) from the signal allows it to be filtered out , but otherwise it approximately contributes an extra factor of @xmath23 to the last term in eq . finally , we remark that the expression for @xmath118 given by eq . ( 10 ) represents an upper bound in that it does not account for the possibility that the output spectrum from a single element may exceed the transparency bandwidth , which could cause some light to be absorbed within the system after multiple reflections and not make it to the end of the waveguide . for simplicity , here we work only with the classical equations so that the intrinsic noise terms in the heisenberg - langevin equations can be ignored . we begin by transforming eqs . ( [ eqsi : dadt ] ) and ( [ eqsi : dbdt ] ) to the fourier domain , 0 & = & ( i_k-)a + i_mb+i(a_r , in(z_j)+a_l , in(z_j)),[eqsi : dadt0 ] + 0 & = & ( i_k-)b+i_ma . to simplify the notation , we define operators at the boundaries of the unit cells ( immediately to the left of an optomechanical element ) given by @xmath235 , @xmath236 . it is also convenient to re - write the transfer matrix @xmath201 in the form m_om = ( cc 1- & - + & 1 + ) , [ eqsi : newmom ] with the parameter @xmath237 given by ( _ k ) = [ eqsi : beta ] . the transfer matrix @xmath238 describing propagation to the next unit cell can subsequently be diagonalized , @xmath239 , with the diagonal matrix @xmath240 given by d= ( cc e^ikd & 0 + 0 & e^-ikd ) .[eqsi : d ] physically , this diagonalization corresponds to finding the bloch wavevectors @xmath241 of the periodic system . the dispersion relation for the system can be readily obtained through the equation ( k(_k ) d ) = ( k d ) - i ( _ k ) ( k d)[eqsi : dispersion ] . writing @xmath172 in terms of @xmath242 , we arrive at @xmath243 . as described previously , the desirable operation regime of the system is such that the phase imparted in free propagation should be @xmath244 . for concreteness , we set here @xmath245 , satisfying this condition . for the frequencies @xmath242 of interest , which easily satisfy the condition @xmath246 and ignoring the intrinsic loss @xmath4 , the simple approximate dispersion formula ( k(_k ) d ) = - can be found . this dispersion relation yields two bandgaps , which extend from @xmath247 and @xmath248 , in the weakly coupled eit regime @xmath249 . we therefore have three branches in the band structure , with the narrow central branch having a width of @xmath250 . this branch has an optically tunable width and yields the slow - light propagation . the dispersive and lossy properties of the array can also be found by analyzing eq . ( [ eqsi : dispersion ] ) perturbatively . expanding eq . ( [ eqsi : dispersion ] ) as a power series in @xmath242 , we find k(_k ) = k_0+++ + o(_k^4),[eqsi : k ] which agrees with eq . ( 7 ) in the main text . in our system , the bloch functions are hybrid waves arising from the mixing of optical waveguide , optical cavity and mechanical cavity excitations . it is therefore of interest to calculate the hybrid or polaritonic properties of these waves , by studying the energy distribution of each bloch mode . the number of photons @xmath251 in the waveguide can be found by taking the sum of the left- and right - moving photons in a section of the device . over one unit cell , one obtains n _ =( |c_j|^2 + hybrid bloch wave may be found by considering the symmetry transformation used to diagonalize the unit - cell transmission matrix . defining @xmath252 to be the amplitude of the bloch mode of interest , one finds @xmath253 and @xmath254 , while from the properties of the symmetry matrix @xmath255 , @xmath256 . from here we can deduce the number of excitations in the waveguide @xmath251 , the optical cavity @xmath257 and the mechanical cavity @xmath14 for a given bloch wave amplitude : n _ & = & |c_j|^2 , + n_o & = & |a|^2 + & = & |c_j + d_j|^2 + & = & | s_11 + s_21|^2 |c_j|^2,[eqsi : nofinal ] + n_m & = & |b|^2 + & = & |a|^2 . we then define the fractional occupation in the mechanical mode by @xmath258 ( with analogous definitions for the other components ) . these relations were used to plot the fractional occupation and colored band diagrams shown in the main text . the technicalities associated with pumping the optomechanical crystal array system are subtly distinct from those in its atomic system analogue . due to the periodic nature of the structure and its strongly coupled property ( @xmath259 ) , a bandgap in the waveguide will arise at the frequency of the `` tuning '' or pump cavities . this prevents these cavities from being resonantly pumped via light propagating in the waveguide . this problem may be circumvented by making the pump frequency off - resonant from @xmath164 ( but still at @xmath38 ) , and by for example changing the splitting @xmath260 to be less than the mechanical frequency . interestingly , the periodic nature of the system also allows one to suppress attenuation of the pump beam through the waveguide ( which would cause inhomogeneous driving of different elements along the array ) . here we calculate the waveguide input powers required to drive the pump cavities . we then find the effect of the cavity dissipation rate @xmath4 on the beam intensity , to provide estimates for the power drop - off as a function of distance propagated in the system . finally we note that there is a trade - off between required pump intensity and pump power drop - off . in other words , by tuning the pump beam to a frequency closer to the pump cavity resonance , the required input power is reduced , but the attenuation per unit cell @xmath261 is increased . to calculate the waveguide input power , we start by considering the _ net _ photon flux in the right moving direction , _ r = |c_j|^2 - |d_j|^2 . using the properties of the bloch transformation matrix @xmath255 and equation ( [ eqsi : nofinal ] ) , this expression may written in terms of the number of photons in the optical cavity of interest ( @xmath262 ) , _ r = n_o , j , where _ 1(_k ) = [ eqsi : beta1 ] . the required input power @xmath263 for the system parameters studied in the main text is shown in figure [ fig : inputpower ] . to find the attenuation per unit cell @xmath264 , with @xmath265 , of this pump beam , we use a perturbative approach similar to that used to find the polaritonic band dispersion . by expanding the bloch - vector as @xmath266 and using equation ( [ eqsi : dispersion ] ) , we find @xmath267 , and @xmath268 implying that @xmath163 . this approximate expression is shown along with the exact calculated values for the attenuation in figure [ fig : inputpower ] . for the theoretical demonstration of our slow - light scheme , we confine ourselves to a simplified model of an optomechanical crystal where only the two - dimensional maxwell equations for te waves and the equation of elasticity for in - plane deformations of a thick slab are taken into account . these equations approximate fairly accurately the qualitative characteristics of in - plane optical and mechanical waves in thin slabs , and become exact as the slab thickness is increased . in this way , many of the intricacies of the design of high-@xmath269 photonic crystal cavities ( which are treated elsewhere @xcite ) may be ignored , and the basic design principles can be demonstrated in a slightly simplified system . the two - dimensional optomechanical crystal ( 2domc ) system used here utilizes the `` snowflake '' design @xcite , which provides large simultaneous photonic and phononic bandgaps in frequency . here we choose to use optical wavelengths in the telecom band , _ i.e. _ , corresponding to a free - space wavelength of @xmath270 . for this wavelength , we found that the crystal characterized by a lattice constant @xmath138 , snowflake radius @xmath139 , and width @xmath140 , shown in figure [ fig : bandstructure ] , should work well . we begin our design by focusing on the creation of a single optomechanical cavity on the 2domc , with one relevant optical and mechanical mode . this cavity is formed by creating a point defect , consisting of two adjacent removed holes ( a so - called `` l2 '' cavity ) . we calculate the optical and mechanical spectra of this cavity using comsol , a commercial fem package , and find a discrete set of confined modes . of these , one optical and one mechanical mode were chosen , exhibiting the most promising value of the opto - mechanical coupling strength @xmath145 ( see below for calculation ) . these modes are shown in figure [ fig : bandstructure](b ) and ( c ) , and were found to have frequencies @xmath271 and @xmath272 , respectively . from here we move to designing the nearly - degenerate double optical cavity system with large cross - coupling rates . as two separate l2 cavities are brought close to one another , their interaction causes the formation of even and odd optical and mechanical super - modes with splittings in the optical and mechanical frequencies . this splitting may be tuned by changing the spacing between the cavities . we take the even and odd optical modes of this two - cavity system as our optical resonances at @xmath164 and @xmath159 . the optomechanical coupling arises from a shift in the optical frequency caused by a mechanical deformation . our hamiltonian for the single cavity system can then be written as = ( ) + _ m , where @xmath273 is the quantized displacement of the mechanical mode , and @xmath274 is the characteristic per - phonon displacement amplitude . the deformation - dependent frequency @xmath275 may be calculated to first order in @xmath276 using a variant of the feynman - hellman perturbation theory , the johnson perturbation theory @xcite , which has been used successfully in the past to model optomechanical crystal cavities @xcite . the hamiltonian is then given to first order by = _ o + _ m + g ( + ) , where @xmath277 is the optical mode frequency in absence of deformation and g = . here @xmath278 , @xmath279 and @xmath280 are the optical mode electric field , optical mode displacement field and mechanical mode displacement field , respectively , @xmath281 is the thickness of the slab , and @xmath282 is the dielectric constant . these concepts can be extended to optically multi - mode systems , represented by the hamiltonian = _ i _ o , i + _ m + _ i , j g_i , j ( + ) , where now the cross - coupling rates can be calculated by the following expression : g_i , j = . we denote this expression for convenience as @xmath283 . for the modes of the l2 cavity shown in figures [ fig : bandstructure](b ) and ( c ) , the optomechanical coupling was calculated to be @xmath284 for silicon . when two cavities are brought in the vicinity of each other , super - modes form as shown in figure [ fig : bandstructure](a ) . we denote the symmetric ( @xmath285 ) and antisymmetric ( @xmath286 ) combinations by @xmath287 and @xmath288 . these modes can be written in terms of the modes localized at cavity 1 and 2 , e_= q_= . by symmetry , the only non - vanishing coupling term involving the @xmath152 ( antisymmetric mechanical ) mode is @xmath289 . assuming that the two cavities are sufficiently separated , we can approximate @xmath290 . for the super - modes of interest , @xmath291 . ) . ( c ) the optical band structure exhibits a single band that passes through the resonance frequency @xmath292 of the optical mode . the waveguide acts as a single - mode optical waveguide with field pattern @xmath293 shown in ( d ) . the @xmath293 component of the guided optical mode of opposite symmetry is shown in ( e ) for completeness . [ fig : waveguide],width=642 ] a line defect on an optomechanical crystal acts as a waveguide for light @xcite . here , the line defects used consist of a removed row of holes , with the rows above and below shifted towards one another by a distance @xmath294 , such that the distance between the centers of the snowflakes across the line defect is @xmath295 ( see figure [ fig : waveguide](a ) ) . the waveguide was designed such that mechanically , it would have no bands resonant with the cavity frequency ( see figure [ fig : waveguide](b ) ) and would therefore have no effect on the mechanical @xmath269 factors . optically , it was designed have a single band crossing the cavity frequency ( see figure [ fig : waveguide](c ) ) and would therefore serve as the single - mode optical waveguide required by the proposal . the band structure of the mechanical waveguide was calculated using comsol @xcite , while for the optical simulations , mpb @xcite was used . ( in arbitrary units ) , shows the leakage of photons out of the double - cavity system . this induces a loss rate on the optical modes , which was used to calculate the extrinsic coupling rate @xmath3 , plotted in ( b ) for various values of @xmath296 . ( c ) plot of the mechanical quality factor @xmath297 due to thermoelastic damping , as a function of the ambient temperature @xmath126 . ( d ) the time - harmonic component of the temperature field @xmath298 is plotted at various times during a mechanical oscillation period . [ fig : coupling_calcs],width=642 ] by bringing the optical waveguide near our cavity , the guided modes of the line - defect are evanescently coupled to the cavity mode , and a coupling between the two may be induced , as shown in figure [ fig : coupling_calcs](a ) . control over this coupling rate is achieved at a coarse level by changing the distance between the cavity and waveguide , _ i.e. _ , the number of unit cells between them . we found a distance of 6 rows to be sufficient in placing our coupling rate @xmath3 in a desirable range . at this point , a fine tuning of the coupling rate may be accomplished by adjustment of the waveguide width parameter @xmath294 , described previously . the achievable values of @xmath3 are plotted against @xmath294 in figure [ fig : coupling_calcs](b ) . for the final design , @xmath299 is used . to simulate this coupling rate , we performed finite - element simulations using comsol where we placed the waveguide near our cavity , and placed absorbing boundaries at the ends of the waveguide away from the cavity . the resulting time - averaged poynting vector @xmath300 is plotted in figure [ fig : coupling_calcs](a ) , showing how the power flows out of the system . the achievable storage times of our system are determined by the lifetimes of the mechanical resonances . since we use a phononic crystal , all clamping losses have been eliminated . however , other fundamental sources of mechanical dissipation remain , and here we provide estimates for one of these , the component due to thermoelastic damping ( ted ) @xcite . using the comsol finite - element solver @xcite , we solved the coupled thermal and mechanical equations for this system @xcite . in these simulations the change in the thermal conductivity and heat capacity of silicon with temperature were taken into account . the ted - limited quality factors , @xmath301 are plotted in figure [ fig : coupling_calcs](d ) . in these simulations , we see that for the mode simulated , @xmath301 surpasses @xmath302 at bath temperatures of @xmath303 . to illustrate some representative results of these simulations , we have plotted the change in temperature field @xmath304 from the ambient temperature versus the phase of the mechanical oscillation in figure [ fig : coupling_calcs](d ) . at @xmath305 , there are variations in temperature despite the displacement field @xmath280 being uniformly @xmath306 at this time . this shows that at these frequencies , the temperature does not follow adiabatically the displacement . as mentioned in the main text , in a realistic setting the optomechanical driving amplitude @xmath6 itself will be coupled to the bath temperature through absorption of optical pump photons in the tuning cavities . this optical pump heating of the structure is important in estimating the practical limits of the optomechanical system for quantum applications where thermal noise impacts system performance . as a realistic model for the bath temperature in our proposed silicon optomechanical crystal array , we take @xmath307 , where @xmath126 is the base temperature and @xmath308 is a temperature coefficient that describes the temperature rise in each cavity per stored cavity photon due to optical absorption . our estimate of @xmath308 for a thin - film silicon photonic crystal structure is as follows . the absorbed power for @xmath309 photons stored in a cavity is given simply by @xmath310 the optical ( intrinsic ) linewidth of the cavity . if we assume _ all _ of this power is being converted to heat , the change of temperature is @xmath311 , where @xmath312 is the effective thermal resistance of the silicon structure . there are a number of sources in the literature for @xmath312 in relevant photonic crystal geometries @xcite . we choose here to use the value for a two - dimensional crystal system in silicon , @xmath313 k / w , which yields a per photon temperature rise of @xmath127k assuming an intrinsic loss rate of @xmath314 rad / s ( @xmath315 ) . 10 url # 1`#1`urlprefix[2]#2 [ 2][]#2

we determine the 22 m luminosity evolution and luminosity function for quasars from a data set of over 20,000 objects obtained by combining flux - limited sloan digital sky survey optical and wide field infrared survey explorer mid - infrared data . we apply methods developed in previous works to access the intrinsic population distributions non - parametrically , taking into account the truncations and correlations inherent in the data . we find that the population of quasars exhibits positive luminosity evolution with redshift in the mid - infrared , but with considerably less mid - infrared evolution than in the optical or radio bands . with the luminosity evolutions accounted for , we determine the density evolution and local mid - infrared luminosity function . the latter displays a sharp flattening at local luminosities belowerg shz , which has been reported previously at 15 m for agn classified as both type-1 and type-2 . we calculate the integrated total emission from quasars at 22 m and find it to be a small fraction of both the cosmic infrared background light and the integrated emission from all sources at this wavelength .

1608.00465

as different processes in active galaxies ( agn ) result in emission in different energy ranges , a crucial class of information for understanding black hole , accretion disk , and jet systems is the intrinsic population characteristics of agn in widely separated wavebands ( e.g. * ? ? ? * ; * ? ? ? these tell us how agn have evolved in different wavebands over the history of the universe , as well as their luminosity and spectral distributions , and the correlations among the different waveband emissions . knowledge of these are necessary to constrain models of emission mechanisms as well as establish luminosity functions and integrated outputs at given wavelengths and their relation to those at other wavelengths . this work focuses on the mid - infrared population properties of quasars , an important class of agn , using a dataset of over 20,000 quasars seen with the wide field infrared explorer satellite ( _ wise _ * ? ? ? as we wish to focus on the mid - infrared wavelengths that are most distinct in emission from the near - infrared , we are interested in particular in the properties of quasars in the longest _ wise _ wavelength 22 @xmath0 m band . multiple strategies are possible for compiling a dataset of agn from an infrared survey to determine infrared population properties . on the one hand , one can select agn candidates using color or spectral - based selection in the infrared . this technique has been utilized with combined _ spitzer _ infrared space telescope and _ wise _ data by @xcite who determine luminosity functions at 5 @xmath0 m . alternately , one can select potential agn candidates from infrared colors and perform optical followups and/or matches to optical catalogs to classify objects and derive redshifts . this technique has been used with combined infrared space observatory and infrared astronomical satellite data by @xcite who , using spectral fits , determine 15 @xmath0 m luminosity functions . alternately , one could use optically identified quasars from the overlap of a deep optical survey area and a deep infrared survey . this has been utilized with _ data by @xcite and @xcite who present luminosity functions at 8 @xmath0 m and 3.6 , 4.5 , 5.8 , 8 , and 24 @xmath0 m , respectively . of these techniques the first two have the advantage of detecting agn which are obscured in the optical ( see e.g. * ? ? ? * ; * ? ? ? * ) but the potential disadvantages of missing agn which do not display the requisite infrared colors , and the inability to classify agn as specifically quasars or another type . the third technique has the advantage of being complete in the optical for a class of agn but could miss other optically obscured objects . some of the aforementioned works have the advantage of not being substantially flux - limited in the optical , either because of the use of infrared selection criteria for identification or the use of deep fields or both . however , in contrast to their advantages , all three of the previously listed compilation techniques have the disadvantage of having a small number of agn objects , numbering in the hundreds , or using photometric redshifts with complicated incompletness and selection considerations in the case of @xcite . they also may be subject to complex or incompletely understood selection effects in one or more wavebands such as in @xcite , and require spectral modeling in many cases . in this work we are interested in performing a complementary determination of agn population characteristics in the mid - infrared with a dataset covering a large portion of the sky with objects numbering in the tens of thousands , with complete spectroscopic redshifts , in which the selection effects are known and in which the crucial population characteristics are determined directly from the data non - parametrically with minimal modeling and assumptions . this can be done by restricting to quasars and using _ wise _ data , along with techniques which we have developed . given that the largest catalogs of quasars with redshifts are identified and catalogued by their optical spectral characteristics , a large dataset such as this used to evaluate population characteristics of quasars in another band depends on an optical survey as well and thus the limits of that survey . in order to evaluate the luminosity evolution in both mid - infrared and optical , and to separate and compare these effects , we require a dataset that has both infrared and optical fluxes to reasonable and known limits across a broad range of redshifts . the overlap of the _ wise _ satellite allwise catalog with the sloan digital sky survey ( sdss ) quasar catalog @xcite , can form such a dataset . when dealing with data from a large survey in the @xmath3 waveband , the luminosity function is usually obtained from a flux limited sample @xmath4 with @xmath5 denoting the flux limit of the @xmath6th object and the luminosity being @xmath7 , where @xmath8 is the luminosity distance and @xmath9 stands for the k - correction . for a pure power law emission spectrum of index @xmath10 defined as @xmath11 , one has @xmath12 . this simple form may be augmented by the presence of emission lines , as in the optical data in this work . in general , the determination of the full luminosity function and its evolution requires analysis of the bi - variate distribution @xmath13 . a correlation between @xmath14 and @xmath15 is known as luminosity evolution and would need to be taken into account when determining the distributions of the individual variables @xmath14 and @xmath15 . in the case of quasars here with the optical and some other band luminosity , because an optical measurement is necessary for quasar identification and spectroscopic redshift , we have at least a tri - variate function . we must take into account not only the correlations between the redshift and individual luminosities ( i.e. the two luminosity evolutions ) but also the possible intrinsic correlation between the two luminosities , before individual distributions can be determined ( e.g. * ? ? ? * ) . treating the infrared survey data as a stand - alone sample independent of the optical survey truncations and the relations between infrared and optical luminosities is not appropriate . @xcite pioneered new methods for determining the correlation of variables from a flux limited and more generally truncated dataset , which were expanded to multiwavelength and complicatedly truncated cases in works by singal et al . ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? our aim in this paper is to take all the selection and correlation effects into account in determination of the true evolution of optical and mid - infrared luminosities of quasars and to find their distributions , using an sdss @xmath16 _ wise _ dataset . in [ datasec ] we describe the infrared and optical data used . [ simlumf ] contains a general discussion of luminosity evolution and the sequence of the analysis . in [ evsec ] we apply the methods to achieve the luminosity - redshift evolutions and the correlation between the luminosities . we determine the density evolution in [ dev ] , and the local luminosity functions in [ local ] . in [ testass ] we investigate some of the assumptions used and their effect on uncertainty , and [ disc ] contains a discussion of the results . this work uses the standard @xmath17cdm cosmology with @xmath18kms@xmath2mpc@xmath2 , @xmath19 and @xmath20 . -band magnitude as described in [ datasec ] and includes applying the k - corrections provided by @xcite which include the continuum and emission line effects . objects plotted in red are those who do not have target flags indicating that they were flagged for spectroscopic followup based on optical colors or a radio match , or are flagged as extended sources , and are not considered for this analysis ( see [ datasec ] ) . of the remaining 62,276 objects , those plotted in blue are those which do not meet the k - corrected @xmath21 criterion ( see [ datasec ] ) , while black points are the objects that do meet the k - corrected @xmath21 criterion ( 21,600 objects ) and are used as the parent optical population in this analysis . it is seen that the k - corrected @xmath21 and flagged subset forms a catalog that has a somewhat smoother redshift distribution with a reduced a bias toward objects with @xmath22 ( although still with residual biases in the redshift distribution as discussed further in [ dev ] ) , and with a calculable limiting flux for every redshift . the solid line is the upper limiting flux corresponding to @xmath2315.0 . , width=336 ] criterion ( 21,600 objects ) which is used as the parent optical sample for this analysis . the effect of the non - uniform selection function with redshift is discussed in [ dev].,width=336 ] m rest frame absolute luminosity density for the quasars in the sdss @xmath16 _ wise _ sample used in this analysis ( 20,063 objects ) . to obtain the 22 @xmath0 m ghz luminosity density we use the luminosity distance obtained from the redshift and the standard cosmology and the k - correction discussed in [ datasec ] . for a few objects ( but only a few flor clarity ) the bottom of the red line indicates the lower limit luminosity for inclusion of the particular object in the survey.,width=336 ] but showing only the quasars in the main sdss @xmath16 _ wise _ sample used in this analysis ( 20,063 objects ) . , width=336 ] we use the well - established sdss data release 7 ( dr7 ) quasar catalog , which contains over 105,000 objects . we seek a subset with a well - defined flux - limit for inclusion for every object and a reduced bias in the redshift distribution . in the sdss dr7 , objects were identified as quasar candidates for spectroscopic follow - up if they displayed the requisite optical colors , or if identified via rosat x - ray data , or if selected by a so - called `` serendipity '' algorithm that identifies unusual colors in concert with a radio match , or if they had a radio match within @xmath24 @xcite . this means that the full dr7 quasar catalog can not be considered to be truly optically flux - limited in at any given magnitude limit . to obtain a set that is flux - limited in the optical band , we restrict the full set in two ways . first we allow only those objects that display the requisite optical colors for follow - up or are identified via a radio match within @xmath24 ( the latter criterion applies to less than 1 % of objects ) , and only those objects which are unextended . this corresponds to requiring that the target flag be set to `` 1 '' in one of three columns in the dr7 quasar catalog the low - z target selection flag ( # 55 ) , the high - z target selection flag ( # 56 ) , and the first selection flag ( # 57 ) and that a flag be set to `` 0 '' for extended morphology ( column # 33 ) , and reduces the full catalog from 105,760 to 62,276 objects . next we impose a k - corrected @xmath25-band magnitude limit of 19.1 at redshift 2 . as discussed in @xcite and @xcite this results in a catalog with a smoother redshift distribution with a reduced bias toward objects with @xmath262 ( although still with residual biases in the redshift distribution as discussed further in [ dev ] ) . it results in a catalog that is flux - limited at every redshift , although that flux limit is somewhat redshift dependent . the magnitude criterion reduces the set further to 21,600 objects . we consider this as the parent set of identified quasars for which we seek an infrared match . the luminosities of the dr7 quasars are shown in figure [ opts ] , and the raw redshift distributions of subsamples discussed here are shown in figure [ optsreds ] . we also note the presence of an upper limit @xmath25 band magnitude of 15.0 for inclusion in the catalog . however this criterion is not completely rigorous , as mentioned in @xcite , and corresponds to a luminosity higher than almost all sources . the allwise catalog is an extension to the _ wise _ general all - sky data release that combines data from the cryogenic and post - cryogenic ( neowise * ? ? ? * ) survey phases . it contains over 700 million objects observed by _ wise _ in the 3.4 , 4.6 , 12 and 22 @xmath0 m mid - infrared bandpasses ( known as w1 , w2 , w3 , and w4 respectively ) . we match the allwise catalog with the restricted set of sdss dr7 quasars with a two arcsecond matching radius criterion , resulting in 20,063 matches with a detection in the 22@xmath0 m band , defined as having a signal - to - noise ratio ( snr ) value greater than two . there are zero multiple matches with this matching radius . reducing the matching radius to one arcsecond results in only a slight reduction of 3% in the number of matches . therefore two arcseconds is an appropriate choice to include real matches while eliminating spurious ones . over 95% of the parent flux - limited optical set have an infrared match with a detection in the 22@xmath0 m band . for the k - corrections to calculate luminosities in the optical band , we adopt the full @xmath27 power law continuum _ plus _ emission line k - corrections presented in @xcite and discussed in 5 and table 4 of that work . the methods of this work can then account for any bias resulting from emission line effects , as long as they are included in the conversions from luminosity to flux , i.e. in the k - corrections . for the 22 @xmath0 m infrared k - correction , we note that studies have generally shown that quasar spectral energy distributions transition from being roughly flat ( @xmath28 ) in l@xmath29 space between 100 and @xmath30 10 @xmath0 m to having @xmath31 at shorter wavelengths ( e.g. * ? ? ? * ; * ? ? ? * ) . in light of this we adopt a 24@xmath0 m k - correction scheme in which the k - correction has @xmath32 [ @xmath33 = 1+@xmath15 ] for @xmath35 1.4 and @xmath36 [ @xmath33 = 2.4 ] for @xmath37 1.4 with a smooth leveling off . we adopt this population - averaged k - correction scheme in part to avoid the additional truncation and bias complications that would result from requiring an observation of every object in the sample in every other _ wise _ band in order to fit an infrared spectrum . figures [ influms ] and [ optlums ] show the infrared and optical luminosities versus redshifts of the quasars in the constructed sdss x _ wise _ sample . the limiting optical flux for inclusion in the data set is a function of redshift , and corresponds to the @xmath25-band magnitude 19.1 at @xmath15=2 k - corrected to the redshift in question . it thus has a slight dependence on redshift , as can be seen in figure [ optlums ] . on the other hand , the limiting 22 @xmath0 m infrared flux for inclusion in the survey , and therefore the limiting 22 @xmath0 m luminosity for inclusion of a particular object , depends on that object s snr in that waveband . in particular , since an snr of 2 is the lowest snr for an object to have a flux reported at 22 @xmath0 m , the lower limiting flux in that band ( w4 ) for an object @xmath6 to be included in the sample is @xmath38 the minimum limiting 22 @xmath0 m luminosity that an object could have to be included in the survey is a function of its lower limiting flux and redshift , with the standard luminosity - redshift relation @xmath39 and @xmath40 . some minimum limiting 22 @xmath0 m luminosities are shown in figure [ influms ] . we have also explored using subsequent data releases of sdss quasars from the baryon oscillation spectroscopic survey ( boss ) phase of the project @xcite . as the boss project is optimized to detect galaxies in a certain redshift range and not to produce a complete flux - limited quasar catalog while minimizing detection biases , we find that these catalogs are inferior to the dr7 catalog for our purposes . the boss catalog actually results in fewer matches to _ wise _ sources than the dr7 . we have also performed the analysis in this work with a significantly less restricted parent optical set , consisting of the dr7 quasars that simply have an @xmath25-band magnitude less than 19.1 and neglecting the target flags and the k - corrections to the magnitude for inclusion . this results in a uniform flux optical limit for inclusion , with a parent optical set of 63,492 objects and a matched 22 @xmath0 and optical set of 49,415 objects . the major conclusions for the luminosity evolutions , the correlations between the luminosities , and the local luminosity functions obtained using that alternate data set are quite similar to those obtained with the more restrictive data set , indicating a robustness of the results in this work . the luminosity function gives the number of objects per unit comoving volume @xmath41 per unit source luminosity , so that the number density is @xmath42 and the total number is @xmath43 . to examine luminosity evolution , without loss of generality , we can write a luminosity function in some waveband @xmath3 as @xmath44 where @xmath45 and @xmath46 describe the luminosity evolution and comoving density evolution with redshift respectively and @xmath47 stands for parameters that describe the shape ( e.g. power law indices and break values ) of the @xmath3 band luminosity funtion . in what follows we assume a non - evolving shape for the luminosity function ( i.e. @xmath48 , independent of @xmath49 and @xmath15 ) , which is a good approximation for determining the global evolutions . the later point is discussed in [ testass ] . once the luminosity evolution @xmath45 is determined using the methods described below we can obtain the mono - variate distributions of the independent variables @xmath50 and @xmath15 , namely the density evolution @xmath46 and `` local '' luminosity function @xmath51 . the total number of observed objects is then @xmath52 we consider this form of the luminosity function for luminosities in different bands , allowing for separate ( optical and infrared ) luminosity evolutions . we use a parameterization for the luminosity evolution with redshift @xmath53 which has been shown to be a goot fit for a dataset based on sdss with many @xmath54 objects @xcite . as discussed in that work , good value of @xmath55 is 3.7 , but the precise value does not matter for the analysis . with @xmath56 for positive values of @xmath57 the luminosities @xmath58 refer to the de - evolved values at @xmath59 , hence the name `` local '' . we discuss the determination of the evolution factors @xmath60 , which in this parameterization becomes a determination of @xmath57 , in [ evsec ] . the density evolution function @xmath46 is determined by the method shown [ dev ] . once these are determined we construct the local ( de - evolved ) luminosity function @xmath61 as in [ local ] . in general , determination of the evolution of the luminosity function of extragalactic sources with spectroscopic redshifts for any wavelength band except optical involves a tri - variate distribution because spectroscopic and most photometric redshift determination requires optical observations which introduces additional observational selection bias and data truncation . thus , in a case such as this , unless redshifts are known for all sources in an infrared survey from infrared data alone , we need to determine the combined luminosity function @xmath62 from a tri - variate distribution of @xmath15 and the fluxes in the optical and mid - infrared bands . if the optical and mid - infrared luminosities were statistically independent variables , then this luminosity function would be separable in the form of @xmath63 and we would be dealing with two bi - variate distributions . however , there may also be a correlation between the two luminosities . as described below , the methods employed here allow us to determine whether any pair of variables are independent or correlated . if it is determined that the luminosities are correlated ( see [ roevsec ] ) , the question must be asked how much of this luminosity correlation is intrinsic to the population and how much is induced in the data by flux limits and/or similar luminosity evolutions with redshift . determination of this is quite intricate as discussed in e.g. @xcite and appendix b of @xcite , and has not been explored sufficiently in the literature . while this will be the subject of future investigations , here we will consider both possibilities . at one extreme , if the luminosity correlation is intrinsic and not induced , one should seek a coordinate transformation to define a new pair of variables which are independent . this requires a functional form for the transformation . we define a new luminosity which we call a `` correlation reduced infrared luminosity '' @xmath64 , where the function @xmath65 describes the correlation between @xmath66 and @xmath67 and @xmath68 is a fiducial luminosity taken here as it is lower than the lowest 2500 luminosity considered in our sample , but results do not depend on the particular choice of numerical value . ] to be @xmath69erg s@xmath2hz@xmath2 . for the correlation function we will assume a simple power law @xmath70 where @xmath71 is a bulk power law correlation index to be determined from the data . this is essentially a coordinate rotation in the log - log luminosity space . as shown in [ evsec ] below , we can determine a best fit value for the index @xmath71 which orthogonalizes the new luminosities . given the correlation function we can then transform the data ( and its truncation ) into the new independent pair of luminosities @xmath72 and @xmath73 . the local luminosity functions of uncorrelated luminosities @xmath74 and @xmath75 can then be used to recover the local infrared luminosity function by a straight forward integration over @xmath75 and the true local optical luminosity function as @xmath76 the mid - infrared luminosities also undergo luminosity evolution with @xmath77^{\alpha } \label{gradform}\ ] ] ( cf equation [ rcrdef ] ) . at the other extreme , if the correlation between the luminosities is entirely induced by truncation effects and similar redshift evolutions , then the luminosity functions are separable into @xmath78 as described above and the analysis can proceed from there . as noted above , we will consider both possibilities here as extreme cases . it turns out that the major results obtained in both cases are very similar . here we first give a brief summary of the algorithmic strategy involved in these determinations , which was first proposed by efron and petrosian and has been expanded upon in recent works . this method uses a modified rank test to determine the best - fit values of parameters describing the correlation functions using the test statistic @xmath79 to test the independence of two variables in a dataset , say ( @xmath80 ) for @xmath81 . here @xmath82 is the dependent variable ( @xmath83 ) rank of the data point @xmath6 in a set associated with it , @xmath84 is the expectation value and @xmath85 is the variance , where @xmath86 is the number of objects in object @xmath6 s associated set . for untruncated data ( i.e. data truncated parallel to the axes ) the set associated with point @xmath6 includes all of the points with a lower ( or higher , but not both ) independent variable value ( @xmath87 ) . if the data is truncated one must form the _ associated set _ consisting only of those points of lower ( or higher , but not both ) independent variable ( @xmath88 ) value _ that would have been observed if they were at the @xmath88 value of point @xmath6 given the truncation _ ( see e.g. @xcite for a full discussion of these points ) . if ( @xmath80 ) are independent then the ranks @xmath82 should be distributed randomly and @xmath89 should sum to near zero . independence is rejected at the @xmath90 level if @xmath91 . to find the best fit correlation bewteen @xmath83 and @xmath88 the @xmath83 data are adjusted by defining @xmath92 and the rank test is repeated , with different values of parameters of the function @xmath65 until @xmath93 and @xmath88 are determined to be uncorrelated . in this analysis we can ignore the upper optical flux limit of sdss quasars discussed in [ datasec ] . the reason for this is that data truncations are only consequential in this analysis if the truncation is actually depriving the sample of data points that exist . as can be seen in figure [ opts ] , there are very few objects approaching the optical upper truncation limit , indicating that this truncation does not appreciably alter the sample from the underlying population . all of the truncations , therefore , are one - sided i.e. at the lower end of fluxes and luminosities . as an example of the determination of a correlation , here we will determine the observed correlation between the observed infrared and optical luminosities . assuming the correlation function between the luminosities of the form of equation [ rcrdef ] we calculate the test statistic @xmath89 from equation [ tauen ] as a function of @xmath71 . where @xmath89 is closest to zero corresponds to the values of @xmath71 that remove the correlation . figure [ taus ] shows the absolute value of the test statistic @xmath89 vs @xmath71 , from which we get the best fit value of @xmath94 with one @xmath95 range @xmath96 . as discussed in [ ifcorr ] , this correlation may be inherent in quasars or may be a result of the data truncations and similar positive redshift evolutions . the general yet quite nuanced and often overlooked question of determining whether an observed correlation between different waveband luminosities is intrinsic or induced will be explored in a future work . to complete the present analysis in the most robust manner , we will consider both possibilities . we shall see that it does not make a significant difference for the major conclusions of this work . it is interesting to note that the observed power - law correlation for the infrared and optical luminosities seen here is less than that for radio and optical luminosities ( @xmath97 ) seen in @xcite and @xcite . statistic as given by equation [ tauen ] as a function of @xmath71 for the observed relation @xmath98 , where @xmath67 and @xmath66 are the optical and infrared luminosities , respectively , for the quasars in the combined sample . the 1@xmath95 range for the best fit value of @xmath71 is where @xmath99 . it is seen that the observed optical and infrared luminosities are positively correlated , but with a sub - linear relation , although this may not be the true intrinsic correlation in this case , as discussed in [ roevsec ] . , width=336 ] the basic method for determining simultaneously the best fit intrinsic @xmath100 and @xmath101 , given the evolution forms in equation [ altevolution ] and properly taking into account the data truncations , is more complicated because we now are dealing with a three dimensional distribution ( @xmath102 ) and two correlation functions ( @xmath103 and @xmath104 ) , plus we can find the true intrinsic correlation in this case because the truncation effects in the luminosity - redshift space are known and redshift is the independent variable in both cases . since we have two criteria for truncation , the associated set for each object @xmath105 includes only those objects that are sufficiently luminous in both bands to have been in the survey if they were located at the redshift of the object in question . as discussed in [ datasec ] , for the optical data , this would be all objects with a luminosity greater than the limiting optical luminosity at the redshift of object @xmath105 given the optical flux limit as a function of redshift , while for the 22 @xmath0 m infrared data this is all objects whose luminosity is greater than their minimum limiting luminosity calculated at the redshift of object @xmath105 ( i.e. equation [ lumlimeq ] with @xmath106 ) . the luminosity cutoff limits for a given redshift must also be adjusted by factors of @xmath104 and @xmath103 . consequently , we have a two dimensional minimization problem , because objects will drop in and out of associated sets as @xmath104 and @xmath103 change , leading to changes in the calculated ranks in equation [ tauen ] . we form a test statistic @xmath107 where @xmath108 and @xmath109 are those evaluated considering the objects optical and mid - infrared luminosities , respectively . the favored values of @xmath100 and @xmath101 are those that simultaneously give the lowest @xmath110 and , again , we take the @xmath111 limits as those in which @xmath112 . figure [ alphas ] shows the 1 and 2 @xmath95 contours for @xmath110 as a function of @xmath100 and @xmath101 . we see that positive evolution in both infrared and optical wavebands is favored . the minimum value of @xmath110 favors an optical evolution with @xmath100 = 3.0 @xmath113 0.1 and an infrared evolution with @xmath101 = 2.4 @xmath113 0.1 . it should be noted that @xmath100 as determined here from the combined infrared - optical dataset is quite similar to that determined from both the a much larger optical only dataset with only a @xmath25-band magnitude cut ( 3.3 @xmath113 0.1 ) and a combined optical - radio dataset ( 3.0 3.5 depending on the radio flux limit assumed ) in @xcite , indicating that the truncations have been properly handled and the robustness of the method . we have previously verified this method including with monte carlo simulations as discussed in e.g. @xcite and @xcite . and @xmath114 contours for the simultaneous best fit values of @xmath100 and @xmath101 of the combined infrared - optical sample , for the forms of the luminosity evolutions given by equation [ altevolution ] . , width=336 ] if we consider that the infrared - optical luminosity correlation is entirely inherent in the underlying data , then the orthogonal luminosities are @xmath67 and @xmath115 ( see [ ifcorr ] ) and we can determine the best fit evolutions @xmath116 and @xmath104 . these results favor @xmath117 and @xmath118 . in this case the best fit infrared evolution can be recovered by equation [ gradform ] and it would have @xmath119 at low redshifts and a more complicated form at higher redshifts , which corresponds well with the results obtained from considering @xmath100 and @xmath101 as orthogonal . the optcal evolutions found by the two methods are in tension at the 1@xmath95 level although within 2@xmath95 agreement . for visualization in figure [ gs ] we plot the functions @xmath120 and @xmath121 vs. @xmath15 for the middle of the best - fit @xmath100 and @xmath101 values determined by this analysis . these results indicate that quasars have undergone a significantly lesser evolution in mid - infrared luminosity relative to optical luminosity , and indeed relative to radio luminosity where @xmath122 = 5.5 @xcite . we return to this point in [ disc ] . for comparison in the literature , @xcite state that optically identified quasars have an mid - infrared luminosity evolution , when fit to the form @xmath123 , of @xmath124 , while @xcite find an exponent with that functional form of 2.9 for luminosities at 15 @xmath0 m , both of which are somewhat stronger evolution than the result here for redshifts up to @xmath125 , at which point the functional forms diverge from the one employed here at higher redshifts . ( solid ) and @xmath121 ( dashed ) vs. @xmath15 for the functional form of equation [ altevolution ] and with the middle best - fit @xmath100 and @xmath101 values determined in this analysis . , width=336 ] next we determine the density evolution @xmath46 . one can define the cumulative density function @xmath126 which , following @xcite based on the method of @xcite which is equivalent to a maximum likelihood estimate , can be calculated by @xmath127 where the set of @xmath6 includes all objects with a redshift lower than or equal to @xmath15 , and @xmath128 is the number of objects with a redshift lower than the redshift of the object at redshift @xmath15 _ which are in that object s associated set_. in this case , the associated set is again those objects with sufficient optical and radio luminosity that they would be seen if they were at redshift @xmath15 . the use of only the associated set for each object accounts for the biases introduced by the data truncation . then the density evolution @xmath46 is @xmath129 however , to determine the density evolution , the luminosity evolution determined in [ roevsec ] must be taken out . thus , the objects optical and infrared luminosities , as well as the optical and infrared luminosity limits for inclusion in the associated set for given redshifts are scaled by taking out factors of @xmath104 and @xmath103 which are determined as above . the preceding method is fully adequate if there is a uniform selection function across redshift for quasars at a given flux . a non - uniform selection function will bias the redshift distribution by artificially removing objects at certain redshifts . as shown in e.g. figure 6 of @xcite the selection function of sdss quasars is not uniform across redshift . as stated in that work , by restricting to sources that are not extended , have been selected for spectroscopic followup based on either optical colors or a radio match , and are brighter than @xmath25=19.1 magnitude after applying the emission line k - correction , the redshift distribution can be made smoother but not unbiased . indeed this is visible in figure [ optsreds ] of this work . to correct for the bias due to the non - uniform selection function , we adopt the selection function derived in @xcite and correct the derived redshift distribution at each redshift for the incompleteness . in particular , if at the redshift and magnitude of any given object @xmath6 the survey is only a fraction @xmath130 complete then the differential density function at that redshift @xmath131 should be increased by a factor of 1/@xmath130 , and the cumulative density function @xmath132 at that redshift should be increased by a factor of 1/@xmath130 @xmath16 ( @xmath132 - @xmath133 ) . we apply these corrections , including a `` floor '' on the selection function of 0.333 as implemented in @xcite , to obtain the intrinsic density evolution functions @xmath134 and @xmath46 . the extent of the selection function correction can be seen in figure [ rholog ] . figures [ sigma ] and [ rholog ] show the cumulative and differential density evolutions , respectively . the normalization of @xmath46 is determined by equation [ inteq ] , with the customary choice of @xmath135 . the number density of quasars seems to peak at just below a redshift of 2 . this is later than found in @xcite and @xcite but similar to the results in @xcite , and @xcite and slightly earlier than the results in @xcite . we note that we plot @xmath46 only to redshift 3.2 because the number of objects is rapidly falling at that redshift and the distribution is more prone to errors resulting from small fluctuations in numbers . we also note that @xmath46 as plotted here contains a factor of dv / dz as in equation [ rhoeqn ] . we note that the analysis of this work in principle accesses the intrinsic redshift distribution for the full range of local luminosities ( @xmath136 , @xmath137 ) and redshift present in the data set , accounting for the various truncations . outside of the range of local luminosities and redshifts present in the data set the population could differ in some systematic way from within the range present in the data set . the data set is effectively unlimited in maximum luminosity , and contains local luminosities down to @xmath138erg s@xmath2hz@xmath2 in optical and @xmath139erg s@xmath2hz@xmath2 in mid - infrared , so outside of this range the redshift distribution could differ from the one obtained here . the result obtained here for the redshift distribution is the intrinsic bulk average for the full range of luminosities present in the data set . an analysis of if and how the redshift distribution differs for various ranges of mid - infrared and/or optical luminosities would require considering a different form than equation [ lumeq ] with @xmath48 for the full luminosity function in a band and is beyond the scope of this work , a point which we return to in [ testass ] . vs. redshift for the quasars calculated from the sample in this work . the normalization of @xmath134 is determined as described in [ dev ] . a spline fit to @xmath140 is used to determine @xmath141 by equation [ rhoeqn ] . , width=336 ] vs. redshift ( large stars ) for the for the quasars , calculated from the sample in this work . @xmath46 is defined such that @xmath142 . the normalization of @xmath46 is determined as described in [ dev ] . for reference we also overplot the form of @xmath46 if the selection function correction discussed in [ dev ] is not applied ( small red crosses ) . the offset between the two determinations at redshifts below @xmath143 is primarily because of the change in normalization of @xmath46 determined by equation [ inteq ] . , in a parallel procedure to determine the redshift distribution we can use the local ( redshift evolution taken out , or de - evolved ) luminosity ( and de - evolved luminosity thresholds ) to determine the ` local ' luminosity distributions , i.e. the luminosity functions @xmath144 , where again the subscript @xmath3 denotes the waveband , and the prime indicates that the luminosity evolution has been taken out . we first obtain a cumulative luminosity function @xmath145 which , following @xcite using the method of @xcite , @xmath146 , can be calculated by @xmath147 where @xmath105 runs over all objects with a luminosity greater than or equal to @xmath14 , and @xmath148 is the number of objects with a luminosity higher than the luminosity of object @xmath105 which are in object @xmath105 s associated set , which in this case consists of those objects which would be in the survey if they were at object @xmath105 s luminosity considering the luminosity limits for inclusion in both optical and mid - infrared . the luminosity function @xmath149 is @xmath150 in [ evsec ] we determined the luminosity evolutions for the optical luminosity @xmath67 and the mid - infrared luminosity @xmath66 . we can form the local optical @xmath151 and mid - infrared @xmath152 luminosity functions straightforwardly , by taking the evolutions out . as before , the objects luminosities , as well as the luminosity limits for inclusion in the associated set for given redshifts , are scaled by taking out factors of @xmath103 and @xmath104 , with @xmath101 and @xmath100 determined in [ evsec ] . we use the notation @xmath153 . for the local luminosity functions , we use the customary normalization @xmath135 . this normalization may be biased by around 8% due to quasar variability as discussed in [ testass ] . figure [ psiopt ] shows the local differential @xmath151 optical luminosity of the quasars calculated from the sample , and for comparison that determined for a combined optical - radio sample in @xcite . we would expect these to be very similar since presumably the underlying population is the same although subjected to different selection effects in the different samples . the optical luminosity function shows possible evidence of a break at @xmath154erg s@xmath2hz@xmath2 . fitting a broken power law above and below @xmath154erg s@xmath2hz@xmath2 yields values for the power law slopes of @xmath155 and @xmath156 below and above the break , respectively . for comparison , these values were found to be @xmath157 and @xmath158 for the combined optical - radio dataset , and @xmath159 and @xmath160 for the parent optical - only sdss dataset , in @xcite . as mentioned in that work , the optical luminosity function has been studied extensively in various agn surveys . for example , @xcite , using the 2df optical dataset ( but with no radio overlap criteria ) use a customary broken power law form for the luminosity function , with values ranging from @xmath1611.39 to @xmath1613.95 for different realizations , showing reasonable agreement . we note that as discussed in [ testass ] the normalization of the local optical luminosity function may be biased by as much as 12 percent . figure [ psiinf ] shows the local 22 @xmath0 m mid - infrared luminosity function @xmath152 calculated both the extremal cases of a ) assuming that the optical and infrared luminosities are truly independent ( stars ) and b ) assuming that all of the observed correlation in the luminosities is intrinsic and constructing the local infrared luminosity function from @xmath151 and @xmath162 with equation [ localrad ] ( diamonds ) . the two determinations vary somewhat and we take this to be the overwhelmingly dominant source of uncertainty in the reconstructed intrinsic local infrared luminosity function . it is seen that this local mid - infrared luminosity function contains a strong break around @xmath163erg s@xmath2hz@xmath2 with a dramatic flattening at luminosities below the break . the flattening is also seen by @xcite at 15 @xmath0 m for both type-1 and type-2 agn at approximately the same value of luminosity as seen here ( assuming a relatively flat spectrum from 15 @xmath0 m to 22 @xmath0 m ) , and at 24 @xmath0 m for agn in the heavily model - dependent analysis of @xcite . we discussed some implications of this flattening in [ disc ] . we note that a differential luminosity function which is flat at the faint end corresponds to a cumulative luminosity function which has a power law slope of -1 at the faint end , i.e. that the number of objects still increases with decreasing luminosity but not dramatically . for luminosities above the break we determine a power law slope of @xmath164 , with this relatively large range resulting from the difference in considering the luminosity correlation intrinsic versus induced . for comparison @xcite report values for the bright end power law slope at 15 @xmath0 m ranging between -2.13 and -3.15 depending on model assumptions . @xcite plot a local 24 @xmath0 m luminosity function for luminosities above @xmath1erg s@xmath2hz@xmath2 that is similar to the overplotted @xcite luminosity function in that range . . the large stars show the results from the sample in this work while the small red crosses show the results from a combined optical - radio dataset from @xcite . the normalization of the local luminosity functions is described in [ local ] . the power law slopes of @xmath151 are discussed in [ locopt ] . , width=336 ] m mid - infrared luminosity function @xmath152 for quasars , calculated from the combined infrared - optical sample . the diamonds show the results from considering that the infrared - optical luminosity correlation is entirely induced , while the stars show the results from considering the correlation to be entirely intrinsic , calculated with equation [ localrad ] . we consider the range between these extremal cases to be the overwhelming source of undertainty in the determination of @xmath152 . the power law slopes of @xmath152 and the flattening at local mid - infrared luminosities below @xmath165erg s@xmath2hz@xmath2 are disussed in [ localinf ] . , width=336 ] a few considerations point to the robustness of the results obtained here and previously . as discussed in [ method ] , results for the optical and infrared evolution factors determined here are consistent whether the infrared and optical luminosities are considered to be truly orthogonal or not . also as mentioned there the best - fit optical evolution factor @xmath100 determined with the combined infrared - optical sample here is quite similar to that determined previously with both the parent optical only sample and a combined radio - optical sample . as discussed in [ locopt ] , the local mid - infrared luminosity function shows the same dramatic flattening at low luminosities whether the infrared and optical luminosities are considered to be truly orthogonal or not . likewise the local optical luminosity function @xmath151 as determined with the combined infrared - optical sample here is quite similar to that determined previously with both the parent optical only sample and a combined radio - optical sample . reults here for the mid - infrared luminosity function parallel those in other works , as do the power laws for the redshift evolution of the infrared luminosity to the extent that results are comparable . additionally , as mentioned in [ datasec ] , we have also performed the analysis in this work with a significantly less restricted parent optical set , consisting of the dr7 quasars that simply have an @xmath25-band magnitude less than 19.1 and neglecting the target flags and the k - corrections to the magnitude for inclusion , resulting in a much parent larger set with a uniform flux optical limit for inclusion . the major conclusions for the luminosity evolutions , the correlations between the luminosities , and the local luminosity functions using that alternate data set are quite similar to those obtained with the more restrictive data set , indicating a robustness of the results . we also emphasize that even if the truncations that we apply on the parent optical quasar catalog to create a flux - limited data set ( including if flux limits vary by redshift ) need not correspond exactly to the actual flux limit of the survey , as long as the limit is consistently applied in both creating the catalog and performing the analysis this point is addressed in @xcite . we consider here some possible sources of error . _ luminosity dependent density evolution _ : a possible concern is that luminosity dependent density evolution , which is not explicitly considered in the functional forms for the luminosity functions used here ( i.e. equation [ lumeq ] with @xmath48 ) , may be necessary to represent the evolution of the luminosity function . as a test of whether the functional forms used here are inadequate for the considered analysis , we divide the data into high and low sets of de - evolved luminosity @xmath166 ( cutting on optical luminosity at a middle value of @xmath167erg s@xmath2hz@xmath2 ) and check the similarity of the computed density evolutions for the two sets and for the whole combined set . the density evolutions computed for both cuts are similar to each other , with the both the high half and the low half peaking in @xmath46 at @xmath168 . given the similarity of these distributions to each other and to that computed from the dataset as a whole , we conclude that we are justified in neglecting an explicit luminosity dependent density evolution form for the purposes of this analysis , which depends on a full luminosity function of the form of equation [ lumeq ] . while luminosity dependent evolution is a well - fit model explored in many works , the form of equation [ lumeq ] with @xmath48 is also adequate for the ranges of ( local ) luminosities and redshifts considered in this analysis , in particular for recovering in the bulk average intrinsic redshift distribution for objects with luminosities in the full range present in the data set . as mentioned in [ dev ] an analysis of if and how the redshift distribution differs for different combinations of mid - infrared and optical luminosities would require considering a different form than equation [ lumeq ] with @xmath48 for the full luminosity function in a band and is beyond the scope of this work . _ survey measurement errors _ : it is well known that measurement errors in magnitudes have the potential to bias the results if the number density of sources increases with decreasing flux , since it is more likely that sources will be erroneously included than excluded @xcite . the extent of this effect depends on the faint end source counts power law slope , and in the limit of flat differential source counts there is no bias . in the limit of measurement errors which are constant in fractional flux , an error will be introduced on the _ normalization _ of the source counts , and therefore on that of the luminosity functions , and can be approximated by [ @xmath169 where @xmath170 is the fractional error in flux and @xmath171 is the faint end differential source counts power law slope @xcite . although we have obtained the intrinsic local luminosity functions and not the source counts for quasars explicitly , we observe that in the sample the lowest fluxes correspond to roughly the lowest decade of local luminosities for both bands , so will approximate the faint end of the source counts power law slope with that of the faint end of the local luminosity functions to consider this effect . for the mid - infrared luminosity function @xmath172 so we would expect no appreciable error on the normalization of the mid - infrared luminosity function from this effect . for the optical luminosity function @xmath173 , and the typical reported sdss measurement errors are on the order of a few hundredths of a magnitude @xcite but we will conservatively adopt errors of 0.2 magnitudes in @xmath25 band to account for both measurement errors and the intrinsic rms scatter due to source variability . for the faint end magnitudes of 19.1 and this error , @xmath170 is around 0.16 , so the bias on the optical luminosity function _ normalization _ will be 12 percent or less . on the other hand , there will be an effect on the reconstructed power law _ slope _ of the luminosity functions only if the fractional measurement errors change systematically with luminosity . we can pursue an upper limit on this effect by considering a related quantity which is readily available from the data sets how fractional measurement errors in flux depend on flux . the optical data show only a modest dependence on @xmath25 band reported error with magnitude at magnitudes below 19.1 , with magnitude errors at most a factor of 1.5 higher at the highest magnitude end of that range than for the lowest magnitudes , which corresponds to a small differential fractional flux error . the 22 @xmath0 m data show a stronger dependence of the reported error in the measured magnitude on the magnitude leading to reported fractional errors in flux which are three to four times higher for the faintest fluxes as for the brightest . however data at a given flux corresponds to a wide range of luminosities , especially at the lowest fluxes . for example as mentioned above in both bands the objects with the lowest fluxes densely span an order of magnitude in local luminosity , which would significantly wash out the systematic dependence of fractional errors on flux when considering if fractional errors in luminosity vary systematically with luminosity . thus we consider that the dependence of fractional errors in luminosity on luminosity are small enough in both bands , which along with the above consideration that the faint end differential source counts are relatively flat in the 22 @xmath0 m band , leads to the conclusion that any effect on the power law slopes of either of the luminosity functions is negligible . _ redshift bias in the sdss quasar catalog _ : one could ask if the density evolution determined here is affected by any biases toward certain redshifts in the sdss quasar sample . as stated in @xcite , the main sources of bias in the redshift distribution of the sdss quasar sample are 1 ) the differing magnitude limits for @xmath22 , 2 ) the effects of emission lines on @xmath25 band flux at different redshifts , and , at a somewhat less important level , 3 ) the inclusion of extended sources at the lowest redshifts ( @xmath174 ) . as discussed in [ datasec ] , we have dealt with issues 1 and to some extent issue 2 by restricting the sample to a universal k - corrected @xmath21 magnitude limit and by adopting for calculating luminosities the @xcite k - corrections which include the effect of emission lines as well as the continuum spectrum , and issue 3 by not including extended sources . we further address issue 2 by incorporating the sdss quasar selection function derived in @xcite and correct the derived redshift distribution at each redshift for the incompleteness as discussed in [ dev ] . we do not believe that the density evolution determined is significantly affected by biases between the redshifts of 0 and 3 where the objects are overwhelmingly most common . there may be biases above redshift 3 but they do not affect the major conclusions of this work . we have used a general and robust method to determine the mid - infrared and optical luminosity evolutions and luminosity functions simultaneously for quasars using a sdss @xmath16 _ wise _ dataset , which combines 22 @xmath0 m infrared and @xmath25-band optical data for over 20,000 quasars ranging in redshifts from 0.08 to 4.97 . as discussed in [ intro ] quite different strategies can be used to assemble an infrared agn data sample for determination of infrared population characteristics . these strategies have advantages and disadvantages . here we have chosen to assemble a large sample of tens of thousands of objects with definite spectroscopic redshifts and known and straightforward flux truncations for inclusion , from which the true intrinsic population characteristics of interest can be determined directly and non - parametrically with limited modeling and assumptions . here we find , as discussed in [ method ] , that quasars have undergone significant luminosity evolution with redshift in the mid - infrared , but less than in the optical band , and , comparing to previous results ( e.g. * ? ? ? * ; * ? ? ? * ) , both of these evolutions are less dramatic than in the radio band . this provides an important input to constrain models of jet , accretion disk , and torus emission and their evolution over the history of the universe . for example , in the basic models of agn where the spin energy of the black hole is tapped to create the jets ( e.g * ? ? ? * ; * ? ? ? * ) , faster radio evolution than optical would indicate that the spin parameters of black holes were higher in the past since radio emission overwhelmingly results from jets . since mid - infrared emission in agn is some combination of emission from the dusty tori , the jets , and the host galaxies , the significantly less rapid evolution of infrared emission in comparison with radio would confirm that jet emission is a sub - dominant source of infrared emission . we also show in [ localinf ] that the local 22 @xmath0 m mid - infrared luminosity function of quasars @xmath152 shows a dramatic flattening at luminosities below @xmath175erg s@xmath2hz@xmath2 . a flattening of this sort is also seen in the 15@xmath0 m luminosity function of type-1 and type-2 agn by @xcite , and at 24 @xmath0 m by @xcite . such dramatic flattening of the luminosity function is not seen in optical wavebands at the luminosities probed by this analysis . however a similar flattening _ is _ seen in the local radio luminosity function at 1.4 ghz luminosities below @xmath176erg s@xmath2hz@xmath2 by @xcite . do these mid - infrared and radio luminosity function flattenings result from the same process ? a simple scaling of the radio emission to the mid - infrared with a synchrotron - like power law spectral index of @xmath177 would put the mid - infrared luminosity equivalent to the 1.4 ghz break at @xmath178erg s@xmath2hz@xmath2 . as this is not where the infrared break is observed , the two flattenings seemingly have different physical causes . this again points to jet emission being a sub - dominant component of the infrared emission . rather , it is some phenomenon of the tori or host galaxies that causes the relative scarcity of quasars with mid - infrared luminosities below @xmath175erg s@xmath2hz@xmath2 . given that that a differential luminosity function which is flat at the faint end corresponds to a cumulative luminosity function which has a power law slope of -1 at the faint end , we can conclude that the contribution of quasars to the integrated mid - infrared light output in the universe peaks at 22 @xmath0 m luminosities of around @xmath165erg s@xmath2hz@xmath2 . could the inferred flattening of @xmath152 result from selection effects , in particular the flux limit for optically - identified quasars in the data set used in this analysis ? in principle the analysis of this work accesses the true intrinsic distributions of local luminosities ( among other quantities ) and corrects for the effects of survey truncations . it is always possible that the population differs significantly _ intrinsically _ at combinations of @xmath136 , @xmath137 , and @xmath15 outside of those present in the data set from how it is at combinations represented in the data set . however , the data set spans two decades in local optical luminosity and local infrared luminosity with significant spread in the @xmath136 , @xmath137 plane . we believe it is most likely that the flattening in @xmath137 is intrinsic in the population of quasars , although can not rule out from this analysis alone an additional population of low infrared luminosity objects which almost universally have a low optical luminosity . [ [ integrated - emission - from - quasars - and - contribution - to - the - infrared - background - light ] ] integrated emission from quasars and contribution to the infrared background light ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ given the distributions calculated here , we can calculate the total integrated output of quasars in the unverse at 22 @xmath0 m . one should integrate the overall mid - infrared luminosity function @xmath179 times the flux corresponding to each luminosity and redshift over all redshifts and luminosities : @xmath180 in terms of the density evolution function @xmath46 and the mid - infrared luminosity portion @xmath181 this is @xmath182 in terms of the local mid - infrared luminosity function @xmath152 we would need the value of this function for the local luminosity corresponding to each luminosity and redshift combination : @xmath183 carrying out this integration results in a calculated value of @xmath184 w m@xmath185 sr@xmath2 hz@xmath2 . in @xmath186 units this is @xmath187 w m@xmath185 sr@xmath2 , which can be compared to e.g. the results obtained by @xcite at 15 @xmath0 m who report a value of ( 4.2 12.1 ) @xmath188 w m@xmath185 sr@xmath2 for type-1 agn and ( 5.5 14.6 ) @xmath188 w m@xmath185 sr@xmath2 for type-2 agn . it would be enlightening to compare the number obtained here for the integrated output of quasars at 22 @xmath0 m to the total cosmic infrared background light ( cib ) level at this wavelength . unfortunately a gap exists in reported measurement of the cib between 3.5 @xmath0 m and 60 @xmath0 m , with reported values of @xmath189 w m@xmath185 sr@xmath2 at the former @xcite and @xmath190 w m@xmath185 sr@xmath2 at the latter @xcite . taking an intermediate value between these two as an estimate of the cib at 22 @xmath0 m would indicate that quasars contribute less than one percent of the total integrated mid - infrared light output in the universe . alternately , one could also compare the value obtained here for the output of quasars to the total output calculated from source counts for all sources at this wavelength . @xcite calculate this output from all sources to be @xmath191 w m@xmath185 sr@xmath2 at 24 @xmath0 m , which , ignoring any spectral shape between 24 @xmath0 m and 22 @xmath0 m and given the value obtained here , would make quasars responsible for between 1% and 5% of the total output from 22 @xmath0 m sources . as discussed in [ roevsec ] , the mid - infrared and optical luminosities are highly correlated , but the power - law correlation index between the mid - infrared and optical luminosities ( @xmath300.8 ) is less than that found previously for the radio and optical luminosities ( @xmath301.0 ) . as discussed in [ ifcorr ] , the subject of luminosity - luminosity correlations is complicated and it is not straightforward to determine how much of these correlations are intrinsic to the waveband emissions in the population and how much are induced by similar redshift evolutions and the truncations of the data set . we will explore this issue in a future work . for the present , we can speculate that _ if _ these correlations are intrinsic ( or if the induced portion is roughly the same in the infrared - optical case as the radio - optical case ) , the radio - optical correlation being more powerful than the infrared - optical correlation could support the idea that the mass and/or spin of the black hole affect the size and/or temperature of the accretion disk and power of the jets more than they affect the size and/or temperature of the torus . a full understanding of the true nature of luminosity - luminosity correlations in agn , and an extension of these considerations to the x - ray band , will be useful in exploring these and other questions . 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/. 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 , and _ neowise _ , which is a project of the jet propulsion laboratory / california institute of technology . _ wise _ and _ neowise _ are funded by the national aeronautics and space administration . we thank the referee for very insightful comments . abazajian , k. , adelman - mccarthy , k. , ageros , m. et al . 2009 , , 182 , 543 babbedge , t. , rowan - robinson , m. , vaccari , m. et al . 2006 , , 370 , 1159 blandford , r. & znajek , r. 1977 , , 179 , 433 broderick , j. w. , & fender , r. p. 2011 , , 417 , 184 boyle , b. , shanks , s. , croom , r. , smith , l. , loaring , n. & heymans , c. 2000 , , 317 , 1014 brown , m. , brand , k. , dey , a. et al . 2006 , , 638 , 88 dermer , c. 2007 , , 659 , 958 eddington , a. 1940 , , 100 , 35 efron , b. & petrosian , v. 1992 , , 399 , 345 efron , b. & petrosian , v. 1999 , jasa , 94 , 447 , mill valley , ca : university science books 1989 finkbeiner , d. , davis , m. , & schlegel , d. 2000 , , 544 , 81 gallagher , s. , richards , g. , lacy , m. , hines , d. , elitzur , m. , & storrie - lombardi , l. 2008 , , 661 , 30 gorjian , v. , wright , e. , & chary , r. 2000 , , 536 , 550 hopkins , p. , richards , g. & hernquist , l. 2007 , , 654 , 731 kimball , a. , kellerman , k. , condon , j. , ivezic , z. , & perley , r. 2011 , , 739 , l29 lacy , m. , ridgway , s. , sajina , a. , petric , a. , gates , e. , urrutia , t. , & storrie - lombardi , l. 2015 , , 802 , 102 lafranca , f. , melini , g. , & fiore , f. 2010 , , 718 , 368 lonsdale , c. , polletta , m. , surace , j. et al . 2004 , , 154 , 54 lynden - bell , b. 1971 , , 155 , 95 mainzer , a. , bauer , j. , grav , t. et al . 2011 , , 731 , 53 maloney , a. & petrosian , v. 1999 , , 518 , 32 mateos , s. , alonso - herrero , f. , carrera , f. , blain , a. , severgnini , p. , caccianiga , a. , & ruiz , a. 2012 , , 426 , 3271 mateos , s. , alonso - herrero , f. , carrera , f. , blain , a. , severgnini , p. , caccianiga , a. , & ruiz , a. 2012 , , 434 , 941 matute , i. , lafranca , f. , pozzi , f. , gruppioni , c. , lari , c. , & zamorani , g. 2006 , , 451 , 443 miller , l. , peacock , j. , & mead , a. 1990 , , 244 , 207 papovich , c. , dole , h. , egami , e. et al . 2004 , , 154 , 70 pris , i. , petitjean , p. , aubourg , . 2014 , , 563 , a54 petrosian , v. 1992 , in statistical challenges in modern astronomy , ed . feigelson & g.h . babu ( new york : springer ) , 173 petrosian , v. 1973 , , 183 , 359 petrosian , v. , & singal , j. 2015 , in proc . iau s313 , extragalactic jets from every angle , eds . f. massaro , c. cheung , e. lopez , & a siemiginowska ( cambridge , uk : cambridge university press ) richards , g. , strauss , m. , fan , x. et al . 2006a , , 131 , 2766 richards , g. , strauss , m. , fan , x. et al . 2006b , , 131 , 2766 schmidt , m. 1972 , , 176 , 273 schneider , d. , richards , g , hall , p. et al . 2010 , , 166 , 470 shaver , p. , wall , j. , kellermann , k. , jackson , c. , & hawkins , m. 1996 , , 384 , 439 shupe , d. , rowan - robinson , m. , lonsdale , c. et al . 2008 , , 135 , 1050 singal , j. , petrosian , v. , lawrence , a. , & stawarz , . , 2011 , , 743 , 104 singal , j. , petrosian , v. , & ajello , m. 2012 , , 753 , 45 singal , j. , petrosian , v. , stawarz , . , & lawrence , a. 2013 , , 764 , 43 singal , ko , a. , & petrosian , v. 2014 , apj , 786 , 109 teerikorpi , p. 2004 , , 424 , 73 xu , c. , carol , j. , lonsdale , j. , shupe , d. , olinger , j. , & masci , f. 2001 , , 562 , 179 wright , n. , eisenhardt , p. , mainzer , a. et al . 2010 , , 140 , 1868

spin observables for the reaction are examined using three recent dynamical models and are compared to the general features of such observables deduced earlier by fasano , tabakin and saghai . these general features , such as the energy dependence of spin observables and the location of nodes in their angle dependence , are realized . several instructive surprises , which occur in this comparison to the conjectures of fasano et al . , are then discussed . the sensitivity of spin observables to isobar and _ t-_channel dynamics is analyzed and suggestions for selecting experiments which provide important dynamical information are presented . = cmr17 scaled 5

nucl-th9509027

the measurement of almost complete sets of spin observables has become technically feasible because newly developed polarized electron or photon beams and polarized targets offer high enough luminosity to permit measurement of the relevant observables . the reaction @xmath1 is particularly advantageous , since the angular asymmetry in the parity nonconserving weak decay @xmath2 provides a direct measurement of the @xmath3 s polarization . if the @xmath3 s spin state is measured along with the polarization of the beam ( @xmath4 ) or target ( p ) , then spin transfer and spin rotation observables can be measured . such spin - rich experiments are underway at elsa @xcite and planned at cebaf @xcite and graal @xcite . in an earlier publication @xcite(fts ) , the general structure of the full set of 16 observables for k@xmath5 photoproduction was examined . in that discussion , helicity amplitudes proved to be particularly useful for deducing general rules concerning the angular structure of the 15 spin observables . the sixteen observables ( the cross - section plus 15 spin observables ) were found to fall into four `` legendre classes , '' with four members in each class . the observables in each class have similar `` nodal structure '' possibilities e.g. , their values at 0@xmath6 and 180@xmath6 and their possible intervening nodes are of related nature . ( that classification procedure , along with the advantages of transversity amplitudes , has led to a reformulation of the general problem of determining which experiments constitute a complete set of measurements . a generalization to many reactions , including electroproduction and photoproduction of vector mesons , will be published separately @xcite . ) in addition , fts used various truncations , both in a helicity basis and in a multipole representation , to deduce rules concerning the nodal structure and energy evolution of all 15 spin observables . in this paper , those fts rules are confronted with specific dynamical models . the considered models@xcite are all based on an isobaric approach using diagrammatic techniques . these models include extended born terms and a very limited number of resonances in _ s- _ and _ u-_channels supplemented by contributions from the _ t-_channel ; namely , by @xmath7 and @xmath8 as well as @xmath9 meson exchanges . we now comment briefly on each model and give the relevant baryonic resonances . first , the model of adelseck and saghai(as ) @xcite is used to calculate the spin observables . their model , which represents a good fit @xmath10 to the existing data up to @xmath11 gev , uses the @xmath12 and @xmath13 baryon resonances . . ] in addition , su(3 ) constraints , based on their success in strong interaction dynamics , are used to limit an otherwise oversupply of possible fits to the data . then , we also study the work of williams , ji and cotanch(wjc ) @xcite , who examined electromagnetic production processes with photons(real / virtual ) energies up to 2.1 gev . that group included the n@xmath14 , n@xmath15 , and @xmath16 baryon exchanges , plus a significant incorporation of crossing symmetry requirements . the third dynamical model considered here is a new model from the saclay - lyon group called saly @xcite , which also gives good agreement with photoproduction data up to 2.1 gev and satisfies broken su(3 ) symmetry requirements . the saly model includes all of the above as baryon resonances plus the @xmath17 @xmath18 and @xmath19 baryon resonances . we emphasize that the first two models(as & wjc ) include only spin-1/2 baryonic resonances , while the third one(saly ) also includes spin-3/2 nucleonic resonances . spin 5/2 resonances are not included in any of these models . among these three models , the wjc model has the weakest kaon exchange ( _ t-_channel ) contributions , while as has the strongest . the wjc model produces weak _ t-_channel exchange because their main coupling constants @xmath20 and @xmath21 were not subject to su(3)-symmetry constraints and in fitting data they generate quite small values of both the _ t-_channel and @xmath22 couplings compared to those produced in the as and saly fits . t-_channel coupling constants come out to be smaller in saly than in as . this feature arises since the saly model includes spin-3/2 baryonic resonances and hence the need for the _ t-_channel exchanges is reduced , in line with duality ideas @xcite . the role of duality ( the interplay between s- and _ t-_channel strengths ) and its effect on spin observables is discussed in section iii.a.2 . can the general rules for spin observables conjectured in fts and summarized in section ii , be seen in the dynamical results ? what is the behavior of spin observables at lower energies and what role is played by particular baryonic resonances and by kaon exchanges ? how do the specific isobar and _ t-_channel dynamics of the three models affect spin observables ? are some spin observables particularly sensitive to interesting dynamics and are therefore particularly important to measure ? those questions are addressed in sections iii & iv . the general rules for the sixteen observables are described in detail in fts @xcite , which also includes the derivation of such observables from a density matrix approach . here we simply highlight and confront those rules . for convenience , the definition of the sixteen observables are recalled in appendix a. the legendre classes of the sixteen observables , which are labeled by @xmath23 , @xmath24 , @xmath25 , and @xmath26 , are : @xmath27 @xmath28 @xmath29 @xmath30 these observables , which are defined as the product of the usual spin observable and the cross - section function @xmath31 are called profile functions @xcite . the profile functions are proportional to bilinear products of amplitudes . observables are determined by the real and imaginary parts of the bilinear combination of helicity amplitudes @xmath32 for class @xmath33 observables by the real and imaginary parts of @xmath34 and for class @xmath35 observables by the real and imaginary parts of @xmath36 finally , the class @xmath37 observables are determined by the four combinations of the magnitudes @xmath38 in the above list , the first entry in each class is the cross - section or a single polarization observable @xmath39 the others are all double polarization observables , which appear ordered as beam - target @xmath40 beam - recoil @xmath41 with the last entry in each class being the target - recoil observables @xmath42 the angular dependence of the above observables are determined by expressing the four helicity amplitudes @xmath43 in terms of wigner rotation functions , with @xmath44 denoting the produced kaon s center - of - mass angle . it is then simple to deduce that @xmath45 class observables are to be expanded in the associated legendre , @xmath46 functions . thus , @xmath47 involve sums @xmath48 which vary as @xmath49 . hence , @xmath50 vanish at @xmath51 and @xmath52 similarly , class @xmath26 spin observables involve sums @xmath53 which vary as @xmath54 and hence vanish more sharply at those endpoints . " the @xmath23 class observables are not necessarily zero at the endpoints . " however , all other spin observables must vanish at @xmath55 and @xmath56 following this helicity amplitude procedure , fts deduced several rules : * * [ h1 ] * the observables @xmath57 and @xmath58 as functions of the scattering angle @xmath44 , must have an odd number of sign changing ( sc ) nodes , if they are nonvanishing observables . * * [ h2 ] * the observable @xmath59 can be nodeless or have an even number of sign - changing zeros . * * [ h3 ] * if the final - state @xmath60 amplitudes vanish , then the following relations must hold at all angles : and @xmath61 in ref . @xcite is corrected here . also , the @xmath33 theorem in section iv of fts is incorrectly stated as having @xmath62 for no @xmath63 amplitudes ; the corrected version is that @xmath64 when there are no @xmath65 contributions . ] @xmath66 these conditions are expected to occur near threshold and to rapidly change as the @xmath67 p - waves turn on . note that these observables are nonzero even when the @xmath68 amplitudes vanish . these properties follow directly from eqs . 4.1 - 4.6 of fts @xcite , since only @xmath69 and @xmath70 vanish if there are no @xmath65 amplitudes . ; \ \ \hat{p}+ \hat{h } = -2 { \rm i m } [ h_1 h_3^ * ] . $ ] ] * * [ h4 ] * the legendre class @xmath33 profile observables can have no more than @xmath71 intervening nodes ( i.e. , not counting the end point nodes ) . here , @xmath72 is the maximum @xmath73value expected at a particular energy . near threshold , these observables are thus expected to be nodeless , with a node developing as the @xmath74 amplitudes turn on , but no more than 2 nodes being possible , until @xmath75 states appear . note that the @xmath33 class observables are all nonzero even if @xmath65 amplitudes vanish . * * [ h5 ] * the @xmath35 class observables were expected to be small near threshold , with nodes developing only at higher energies , since the @xmath65 amplitudes are expected to be small near threshold and these observables depend on interference between @xmath76 and @xmath77 amplitudes . * * [ h6 ] * the @xmath26 class observables should be small and nodeless at lower energies . these observables also depend on interference between @xmath76 and @xmath65 amplitudes . we shall refer to the above statements as fts rules * h1 * @xmath78 * h6*. these remarks are summarized in table ii of fts . one interesting observation is that `` if @xmath79 or @xmath80 assume an even number of intervening zeroes , then at least one of these zeroes would be of non - sign changing '' ( nsc ) type . this feature , which will be examined in our later comparison with dynamical models , is a consequence of the requirement that both @xmath79 and @xmath80 have an odd number of sc nodes - which follows from basic parity requirements . similar remarks can be made about @xmath81 which must have an even number of sc nodes ; hence , an odd number of zeroes means that there must be an odd number of nsc zeroes . examples of these cases will be discussed later , when bifurcations appear in the nodal trajectory plots . since there are only four amplitudes for pseudoscalar meson photoproduction , only seven measurements of the sixteen observables are needed to extract unique amplitudes ( one overall phase is arbitrary ) . some observables provide redundant information and thus one needs to consult the known rules @xcite for selecting seven independent experiments . that issue is not considered in this paper , see ref . @xcite . additional rules concerning spin observables were discussed by fts , based on the possible truncation of multipole amplitudes . the advantage of expanding the k@xmath5 photoproduction amplitudes into multipoles @xmath82 is that the orbital angular moment , @xmath83 of the final @xmath84 state can be used to reduce the number of amplitudes , based on the existence of a centrifugal barrier . of course , this truncation does not include the possibility of dynamical effects , which could magnify selected orbital states . for example , a resonance could emphasize a particular partial wave or competing effects could attenuate selected waves . however , it is just the deviation from ordinary centrifugal - dominated behavior of spin observables that we hope will serve as the best indicator of such dynamical effects . several additional features of the sixteen kaon photoproduction observables were conjectured in fst , based on the suppression of higher orbital angular momentum states . these include : * * [ m1 ] * a cross - section peak at 0@xmath6 implies that the combination of the p - waves defined by @xmath85 ( called type-1 splitting ) and the s - wave amplitude @xmath86 have a relative phase angle of less than 90@xmath87 provided that : ( 1 ) @xmath882 amplitudes are negligible , and ( 2 ) @xmath89 . that is , to get a cross - section peak one needs not only p - waves to interfere with the s - wave multipole , but also the p - waves must have a nonzero ( type-1 ) splitting . * * [ m2 ] * a second type of p - wave splitting is defined by the linear combination of @xmath90 multipoles : @xmath91 ( called type-2 splitting ) . if @xmath92 waves can be neglected and there are p - waves , but they have zero type-2 splitting e.g. , @xmath93 then ( 1 ) the spin observable @xmath94 at @xmath95 and ( 2 ) @xmath96 and @xmath97 are zero at all angles . * * [ m3 ] * the type-2 splitting also leads to the following possible @xmath96 behavior . in order for the final @xmath3 polarization , @xmath96 , to have nodes , there must be nonzero type-2 p - wave splitting , @xmath98 and @xmath99 must not be collinear with @xmath100 unless @xmath101 multipoles contribute significantly . furthermore , if the type-2 p - wave splitting is nonzero and collinear with the @xmath86 multipole , then , in the absence of @xmath92 amplitudes , the polarization @xmath96 has a sc node at @xmath102 * * [ m4 ] * based on a scattering length expansion , it was anticipated that @xmath103 and @xmath104 are small and nodeless near threshold , with nodes developing only if @xmath105 amplitudes contribute . * * [ m5 ] * the beam - recoil observable @xmath106 and the target - recoil observable @xmath107 are both zero at all angles , unless there are @xmath108 multipoles and the stretched electric and magnetic multipoles are unequal e.g. , @xmath109 in magnitude and phase . * * [ m6 ] * the observables @xmath110 and @xmath111 are both of legendre class @xmath26 and are complementary in their nodal behavior , see eqs . 5.21 and 5.23 of fts @xcite . by complementary we mean that if one tends to have a node , the other does not so tend . * * [ m7 ] * for @xmath112 and @xmath113 to have sc nodes , type-2 p - wave splitting must be nonzero , until @xmath92 waves contribute significantly . * * [ m8 ] * if the legendre class @xmath33 observables @xmath114 and @xmath115 have zeroes then , as the momentum increases , these zeroes tend to be placed symmetrically about @xmath116 . * * [ m9 ] * near threshold , the class @xmath23 spin observable @xmath117 has a sc node at @xmath102 * * [ m10 ] * near threshold , the class @xmath23 spin observable @xmath118 has a sc node at @xmath102 we shall refer to the above statements as fts rules : * m1 * @xmath119*m10*. we are now ready to confront these fst rules with the spin observables found using three different dynamical models . the kaon photoproduction observables are shown for the adelseck - saghai(as ) @xcite , williams , ji and cotanch(wjc ) @xcite , and saly @xcite models in fig . 1 for @xmath120gev , and in fig . 2 for @xmath121gev . the observables versus angle @xmath122 , _ not _ the profile functions , are presented in figs . 1 & 2 . as emphasized in the introduction , these models are based on selected baryon resonances and exchanged mesons , with the physical basis for these models presented in refs . @xcite . here , we simply compare these recent dynamical models with the fts rules * h1 * @xmath78 * h6 * and * m1 * @xmath78 * m10*. note that these plots are organized with observables of a given legendre classes@xmath123 in columns , with the first row giving the cross - section and the single spin observables @xmath124 and @xmath125 the second row of plots in figs . 1 and 2 gives the beam - target @xmath126 ; the third row of plots shows the beam - recoil @xmath127 the fourth row of plots gives the target - recoil @xmath128 double spin observables . to facilitate comparison with the fts rules , the c.m . nodal angles ( in degrees ) for the three models are plotted versus the incident photon laboratory energy in fig . we call these nodal trajectory " plots . these curves have been calculated in the appropriate energy domain for each model , i.e. @xmath129 gev for as and @xmath130 gev for wjc and saly . the nodal angles are the angles at which a spin variable has a sign changing ( sc ) zero . a single node which moves with increasing beam energy appears as a single curve . a bifurcating curve shows the energy evolution of a non - sign changing zero into two sc nodes ( see the observable @xmath131 for a clear example ) . these nodal plots are organized by legendre class and also as : single spin observable(top row ) , double - spin beam - target(second row ) , beam - recoil(third row ) and target - recoil(bottom row ) . examination of figs . 13 allows one to confront the three dynamical results with the fts rules . the observables of legendre class @xmath132 and @xmath26 are all seen to vanish as expected at the endpoints ( 0@xmath6 and 180@xmath6 ) . only the set @xmath133 are nonzero at both of these endpoints . also note that the @xmath134 class observables approach the end point angles with nonzero slopes ; whereas , the @xmath26 class observables approach the endpoint angles with zero slope . thus the @xmath135 and @xmath136 properties are clearly seen in figs . 1 and 2 . spin observables for the three dynamical models are now examined at two energies , @xmath137 0.920 gev and 1.4 gev for the purpose of testing the fts rules . the spin variables at @xmath138 gev , which are near the @xmath139 threshold of @xmath140 gev , are shown in fig . 1 . the case of @xmath141 gev is shown in fig . 2 . at 0.92 gev , both @xmath117 and @xmath118 display the anticipated sign changing nodes , see the fts rules * h1 * , * m9 * and * m10*. the function @xmath131 is nodeless at low momenta and falls slightly below 1 for all three models , which indicates that only small @xmath63 amplitudes are in effect , see rules * h2 * and * h3*. thus , the fts suspicion that @xmath142 is an indication of small @xmath143 aligned @xmath144-waves is realized near threshold in these models . note the enlarged scale for the @xmath131 plot in fig . 1 . the consequences of assuming small aligned @xmath144-wave amplitudes e.g. , @xmath145 ( rule * h3 * ) , are realized for all three models at 0.92 gev . for example , we also see that @xmath146 and @xmath147 in fig . 1 . thus the rule * h3 * is fully realized near threshold . indeed , the small deviation from these fts rules can be used as a measure of the special sensitivity of each spin observable near threshold to @xmath148 aligned p - wave amplitudes . assuming that at 0.920 gev the @xmath33 observables are dominated by @xmath149 fig . 1 shows that the as model has @xmath76 amplitudes of the largest magnitude , that the wjc amplitudes are smaller and the saly model gives in general the smallest @xmath76 contributions . the @xmath23 observable @xmath131 also shows this 1/2 strength pattern . this behavior is consistent with the coupling constants @xcite of the three models . note that the fact that @xmath144 and @xmath61 are small compared to @xmath114 and @xmath115 indicates constructive interference between the helicity amplitudes @xmath69 and @xmath150 e.g. , they tend to be parallel near threshold . the next fts rule * h4 * , which is based on general helicity amplitude considerations , concerns the maximum number of nodes for observables in legendre class @xmath151 for just @xmath76 amplitudes , this class of observables should be nodeless near threshold , which is indeed the case at 0.920 gev , see fig . 1 . as the @xmath74 amplitudes turn on , due to either aligned p - waves ( 1 + 1/2 = 3/2 ) or unaligned d - waves @xmath152 the @xmath33 observables could develop intervening nodes , but not more than two , until the @xmath153 states turn on . indeed , at 1.4 gev some @xmath33 spin observables do develop nodes , see fig . 2 . in particular , the observables @xmath144 and @xmath61 , each develop one sc node for the saly model . for the saly and wjc models @xmath115 develops two nodes , while for as , @xmath115 stays at one node at 1.4 gev . these results comply with the fts restriction * h4 * that @xmath33 observables have no more than two nodes until @xmath75 amplitudes are strong . hence , at 1.4 gev the @xmath154 amplitudes are not explicitly seen in the nodal structure of the @xmath33 observables . the double nodal structure of @xmath115 is a possible indication of interesting @xmath155 dynamics in the proton to @xmath3 spin rotation function ( @xmath156 see appendix a ) . for the @xmath157 observables @xmath158 at 0.92 gev ( fig . 1 ) , two are nodeless @xmath159 and two @xmath160 have nodes in the vicinity of @xmath102 thus the node part of rule * h5 * works for @xmath161 but not for @xmath162 the @xmath35 observables depend on interference between @xmath76 and @xmath63 amplitudes ; hence , we can conclude that these @xmath163 nodes near @xmath116 yield important @xmath164 interference information for all three models . the reason for these nodes is presented later . for the @xmath35 observables at 1.4 gev ( fig . 2 ) , @xmath165 acquires one node for wjc and saly , while the as model remains nodeless . the observable @xmath166 stays nodeless for all three models at 1.4 gev . double nodes appear in @xmath167 for the saly and wjc results , while as stays at one node . the above double nodes seem to be located symmetrically above and below @xmath102 for @xmath168 saly gets three , while as and wjc stay with one node each , near @xmath116 and @xmath169 respectively . the mechanisms for these changes with energy are more readily understood by examination of the nodal trajectory plots of the next section . the @xmath26 observables @xmath170 at 0.92 gev ( fig . 1 ) comply with the rule * h6 * ; they are all nodeless and all , except perhaps @xmath171 are small . these observables , as in the @xmath35 case , depend on @xmath172 interference , which is why they are expected to be small near threshold . at 1.4 gev ( fig . 2 ) , @xmath173 acquires an even number of nodes(2 ) for as and wjc , while @xmath173 remains nodeless for saly . also @xmath174 and @xmath110 each acquire one node , but only for the saly model . the observable @xmath111 remains nodeless at 1.4 gev for all three models , but with much angular structure . all of these features agree with rule * h6*. however , the acquisition of an even number of nodes in @xmath173 for some models at and below 1.4 gev is a surprise , as will be discussed later . now let us return to the @xmath23 observables . at 1.4 gev ( fig . 2 ) , the @xmath175 spin observables vary considerably with angle and some have nodes . the rules that : @xmath59 must have an even number of nodes ( * h2 * ) , and @xmath117 and @xmath176 an odd number of nodes ( * h1 * ) , are clearly satisfied . however , the three models manage to satify these theorems in different ways , especially in the number of nodes . this suggests that their nodal structure " and the evolution of nodes with energy , is a source of specific dynamical information . all of the models display nodal structures consistent with the * h1 * @xmath78 * h6 * rules . to understand how these nodes develop with increasing energy , and the underlying reasons for the nodal structure of spin observables , it is more convenient to examine nodal trajectory " plots . we shall consider consequences of having a limited number of multipole amplitudes later . to follow the detailed development of nodes with increasing energy , we now examine the nodal trajectory " plots of fig . the nodal angles ( e.g. , angles at which sc nodes occur ) are plotted versus the photon s laboratory energy . + * ( a ) single spin observables * + let us start with the single spin observables ( @xmath177 and @xmath173 ) , which are presented in the top row of fig . 3 . ( the cross - section , and hence @xmath31 is nodeless . ) the recoil @xmath3 polarization , p in fig . 3 , has : no nodes for the as model ; one node starting at @xmath178 at 1.47 gev for the wjc case ; and an early node starting at @xmath178 at 0.96 gev for the saly model . both as and wjc models do not have spin 3/2 resonances , in contrast to the spin 3/2 isobar that is part of the saly model . that isobar accounts for the dramatic difference in the observable p ; namely , the early @xmath178 node for the saly model and the subsequent strong energy dependence of its single node is a reflection of the 3/2 , @xmath90 resonance @xmath179 at about @xmath180 gev , corresponding to the total energy of @xmath181 gev in the center - of mass frame . the other models rely on _ t-_channel exchanges for their @xmath63 strength at low energies . similar reasoning applies to the observable @xmath165 in fig . 3 , which also has an early @xmath178 node for the saly model . for wjc , the target polarization @xmath165 has a node due to @xmath182 strength , but the node displays smooth non - resonant evolution since the wjc model has no @xmath63 resonances . thus , the nodal structure of @xmath144 & @xmath165 are sensitive to explicit _ s-_channel spin 3/2 resonances , but are not so revealing concerning _ t-_channel contributions . the curves @xmath144 and @xmath165 for saly are therefore good examples of resonance - driven nodal trajectories and show how such plots can be used to extract detailed resonance dynamics . the same saly 3/2-resonances drive the spin observable @xmath61 , which has a particularly dramatic nodal trajectory as seen in fig . this case will be discussed in the beam - target section . for the photon asymmetry observable , @xmath183 the bifurcating behavior seen in fig . 3 for the as and wjc models is unexpected . this observable is _ not required _ to have an even number of nodes . ( in contrast , @xmath131 is restricted to an even number of nodes , which is the reason for its bifurcating nodal trajectory . ) _ to get two nodes in @xmath183 especially at low energies there has to be some j=5/2 amplitude strength . _ that strength is apparently not due to any 5/2 resonances , since none of the models have explicit @xmath184 resonances . instead , the @xmath184 strength at low energies arises from a _ t-_channel mechanism ! the polarized beam asymmetry @xmath173 shows this unexpected double nodal structure for the as model at lower energies and at higher energies for the wjc model . for the saly model , a single node appears at @xmath178 and moves rapidly to @xmath185 two nodes appear at higher energy and evolve smoothly . in the region that saly has one node , this single node and its rapid motion can be accounted for by its @xmath63 resonances . the double node at higher energy for saly occurs when it picks up @xmath184 strength , probably from _ t-_channel effects . the general form of the @xmath97 observable is @xmath186 where the terms @xmath187 and @xmath188 arise from interference between @xmath189 1/2 states , while the term @xmath190 arises only if @xmath191 states contribute ; for example , @xmath190 arises first from a @xmath192 amplitude interference . therefore , to get the double nodal structure seen for the as model near threshold , one needs a sizable @xmath190 term or , equivalently , at least @xmath184 multipoles . it is also clear from this general form that for small 5/2 , but sizable 3/2 , amplitudes , the term @xmath193 arises first from interference between @xmath194 and hence can give only one node . sc nodes , an observable needs to be described as @xmath195 times a polynomial of order @xmath196 in @xmath197 here @xmath198 depending on the legendre class @xmath199 the @xmath184 amplitudes could arise from a mechanism that boosts the orbital angular momentum to higher values . the striking of a virtual p - wave meson by a polarized incident photon ( t - channel exchange ) provides such a mechanism . for example , if a virtual @xmath200 kaon peals off from the nucleon in a p - wave , it can receive an extra orbital angular momentum boost when it is struck by the incident photon . the kaon then boosts to an @xmath201 state , which , when added to the @xmath3 s 1/2 spin , generates @xmath202 and @xmath203 strength , even at low energies , without an explicit 5/2 baryon resonance . thus , we learn that _ t-_channel , or kaon exchange provides @xmath202 and @xmath203 strength and therefore contributes to the @xmath204 and @xmath205 multipoles . this mechanism is illustrated in fig . the as model has particularly strong _ t-_channel couplings , which are generated , in a duality sense , to make up for the absence of the 3/2 resonances contained in the saly model . the wjc model , which has weaker coupling in the _ t-_channel processes , does ultimately also reveal a similar bifurcation in the @xmath173 observable , but not until higher energies of about 1.4 gev . although the saly model has stronger _ t-_channel strength than does the wjc model , it does not display the bifurcation one might expect , because it is dominated at low energy by the @xmath63 resonance . for the saly model the above role for _ t-_channel @xmath184 amplitudes is not seen . instead , in fig . 3 , the @xmath173 for saly has a @xmath178 node which travels rapidly toward @xmath55 in the region of the @xmath63 resonances of this model . this property is related to the resonance - driven evolution of the single node in the observables @xmath177 and @xmath61 for the saly model . in the @xmath173 case , the 3/2 resonance increases the size of the @xmath206 term and the rapid variation of @xmath193 with energy and its sign change explain how the 3/2 resonance causes the rapid @xmath207 nodal evolution . at higher energies the saly model does acquire 5/2 strength and thus a nonzero quadratic term , @xmath208 permits the appearance of two nodes , as seen in fig . thus , all of the @xmath173 curves arise from resonance and/or _ t-_channel mechanisms . we conclude that the observable @xmath173 provides a particularly sensitive test of resonances and of @xmath184 amplitudes and thereby of _ t-_channel or meson exchange processes . the basic idea of duality , which equates a sum over all _ t-_channel processes with a sum over all _ s-_channel resonances , shows that there is a close relationship between a dynamical model s content with respect to assumed resonances and the corresponding strength of the _ t-_channel exchanges needed to fit the data . thus , another way to describe the significance of @xmath173 is that it tests the duality structure of the dynamics . this result and the enhanced role of @xmath184 amplitudes were not anticipated by fts . + * ( b ) beam - target spin observables * + we now discuss the nodal trajectories for the beam - target observables , which are given in the second row of fig . 3 . for the observable @xmath209 a bifurcation of nodes appears in fig . recall that @xmath131 is restricted to an even number of nodes by rule * h2*. therefore , if a ( nsc)zero develops in this observable and with increasing energy this vanishing extremum acquires a non zero value , then it must produce two sc nodes : that is what we mean by a bifurcation . insight into this rule is gained by noting @xcite that @xmath210 has the form @xmath211 the coefficient @xmath190 is determined by amplitude products of the form @xmath212 and interferences @xmath213 thus @xmath190 can exist either by having a @xmath214 resonance ( as in saly ) or by enhanced _ t-_channel 3/2 strength , without invoking 5/2 terms . hence , the quadratic form can be realized for those two reasons by all models . indeed , all models ( fig . 3 ) display a bifurcation at about the same photon energy ; for as and wjc the term c is generated by the kaon exchange process ; whereas , for the saly case it is generated by the underlying 3/2 resonances . there are no nodes in @xmath131 until about 1.2 gev , even though the @xmath215 would seem to allow for low energy nodes . however , the @xmath216 term is dominated by @xmath217 while @xmath193 depends only linearly on @xmath86 s interference with @xmath90 multipoles . thus , the @xmath216 term dominates and without a sizable contribution from @xmath218 single nodes are prevented and one goes directly to double nodes at higher energies . the beam - target observable @xmath61 is related to the single spin observable @xmath144 which was discussed in the previous section . the rapid variation of the nodal location of @xmath61 for the saly model is due to the 3/2 resonance , but in a more dramatic fashion than for @xmath219 see fig . thus the double spin observable @xmath61 has enhanced 3/2-resonance dependence . this is another good example of nodal structure generated by resonances . the relation between the nodal trajectories for @xmath144 and @xmath61 is particularly instructive . we know that @xmath220 im@xmath221 where the @xmath69 and @xmath222 helicity amplitudes are zero unless @xmath65 amplitudes exist . thus , in the absence of @xmath65 amplitudes , we recover the theorem @xmath223 of rule * h3*. in the case of the two @xmath63 saly - resonances , the difference between @xmath144 and @xmath61 is determined by im@xmath224 and therefore by the associated amplitudes interference . the two saly - resonances drive both @xmath144 and @xmath61 in a very revealing manner . note that for the wjc model there are no @xmath63 resonances and consequently nodes for that case develop at higher energy and move from @xmath178 to smaller angles smoothly for both @xmath144 and @xmath225 this is an example of a non - resonant driven nodal evolution , where the @xmath63 strength probably arises first from _ t-_channel effects . hence , observation of both @xmath144 and @xmath61 could reveal the presence or absence of resonance dynamics . one would need to measure the polarization of the final @xmath3 _ via _ its decay and also measure @xmath61 which requires a linearly polarized photon plus a proton target polarized in the @xmath226 direction . in the present case , we are dealing with two nearby @xmath63 resonances albeit of different parity ; it would be of interest to explore the nature of nodal trajectories for isolated and/or dominant resonances such as are thought to occur in @xmath227 meson production @xcite . for @xmath166 we see no nodes until higher energies ( above 1.4 gev ) for all models . the as develops two nodes , followed by wjc with two nodes and then saly comes on with three nodes in the 1.8 gev region . this is clearly not driven by any of the resonances , they occur at lower energies . the form of @xmath166 is @xmath228 where from eq . d6 of fts the @xmath216 term is enhanced by the large @xmath86 multipole , which makes it difficult for the @xmath193 and @xmath190 terms , needed to generate nodes , to play a role . but this amplification of the @xmath216 term is also true for the observables @xmath144 and @xmath165 and they do exhibit nodes ; whereas , for @xmath166 there is a delay in the appearance of nodes . the @xmath193 term for @xmath166 must therefore conspire to be smaller than for these earlier cases ; it might be characteristic of the observable @xmath166 not to have nodes only after a critical energy is reached . when double nodes do turn on in @xmath229 for the as model , it is because @xmath216 and @xmath190 are comparable ; we need to invoke @xmath230 terms to get double nodes . to generate @xmath231 the as model must use its considerable 5/2 amplitudes , so again this suggests a _ t-_channel mechanism . the same reason explains the later appearance of 2 nodes for the wjc model , which is weakest in _ t-_channel strength . the saly model is the first to develop three nodes in @xmath166 , which means the @xmath232 term is effective , but not until higher energies where @xmath233 strength can be generated in several ways . the observable @xmath174 has a single node for only the saly model , as seen in fig . this observable is of the form @xmath234 where @xmath216 depends on p - wave interferences of the type ( @xmath235 and also s and d - wave interferences . the @xmath193 term , which is needed to generate a single node , depends on p - and d - wave interference ; indeed , in @xmath193 the d - waves ( @xmath202 and @xmath203 multipoles ) appear multiplied by p - wave amplitudes ( @xmath236 and @xmath237 multipoles ) . even above the resonance region , p - wave amplitudes ( @xmath236 and @xmath237 ) interfere with d - waves sufficiently to generate a @xmath193 that is comparable to @xmath216 only for the saly model , which yields the single node in @xmath174 seen in fig.3 . for the as and wjc model , the @xmath193 term does not have sizable p - d wave interference ; also , the s and d - wave interference keeps @xmath216 large enough , due to large s - waves , to make it difficult in general to generate nodes in @xmath238 a single node in @xmath174 provides evidence for significant p - d wave interference and possible p - wave enhancement . in many cases , especially at low energies , nodal trajectories are seen to reveal either resonance or _ t-_channel effects , and/or enhancements due to dominance of the @xmath239 multipole . + * ( c ) beam - recoil spin observables * + now consider the nodal trajectory plots for : @xmath240 at first glance , @xmath79 in fig . 3 seems impossible to understand , but it does have some simple features . all models have a @xmath116 node near threshold . this property can be understood from the general form for this observable @xmath241 where near threshold @xmath193 is dominated by @xmath217 while @xmath216 depends linearly on @xmath86 s interference with @xmath90 multipoles . thus near threshold the @xmath242 term dominates and gives the @xmath116 node near threshold . the saly 3/2 resonances show up in @xmath117 by rapidly moving that @xmath116 node first to smaller and then to larger angles . in the general form of the observable @xmath79 the cubic term @xmath243 appears even when a truncation to @xmath244 is used . the @xmath232 term involves interference between the aligned @xmath63 amplitudes . therefore , in addition to explaining the rapid motion of the lower energy node by 3/2 resonances , at @xmath2451.31 gev there is enough off - resonant 3/2 strength in the saly amplitudes to invoke the cubic term and hence yield the three nodes seen in fig . 3 . similarly , at @xmath246 1.23 gev the as model has considerable 3/2 strength , not by a resonance , but by _ t-_channel enhancement . thus , it exhibits a smooth ( non - resonant ) evolution of its low energy node , followed by an early turn - on of three nodes . this early turn - on of 3/2 amplitude appeared earlier in the bifurcations seen in @xmath173 and @xmath247 for wjc , there are only small 3/2 amplitudes ( it has no 3/2 resonances and its kaon exchange is small ) ; therefore , it exhibits a smooth evolution of the @xmath116 node toward zero angle , for @xmath248(fig . 3 ) we note an absence of nodes until energies above 1.4 gev . the as and then the saly model exhibit one simple node , while wjc turns on with three above 1.61 gev . the prevention of early nodes is understood from eq . d9 of fts , wherein the leading @xmath249 term is dominated by the s - wave multipole @xmath250 dominates the @xmath216 and @xmath251 enters linearly in the @xmath252 terms in the polynomial . that structure prevents nodes until higher energies . ] the spin observable @xmath253 has complicated nodal structure that can be understood from its general form @xmath254 where near threshold @xmath193 is dominant since it depends linearly on @xmath86 s interference with @xmath236 multipoles ; whereas , @xmath216 involves only @xmath108 waves . thus a @xmath116 node near threshold occurs for all three models , which although not required by fts rules , arises from dynamical dominance of s - waves near threshold . that node moves most rapidly for the saly model , due to its 3/2 resonances ; the saly model is also able to turn on the @xmath190 term to generate two nodes above 1.26 gev due to off - resonance @xmath236 strength . the as model displays a smooth evolution of the low energy @xmath116 node because it lacks a 3/2 resonance . the wjc exhibits structure in the evolution of its @xmath116 node and also acquires two nodes above 1.2 gev . this wjc evolution arises perhaps from the energy dependence of the s - wave amplitude and the onset of @xmath236 strength and interference with @xmath237 amplitudes for @xmath255 only saly and wjc exhibit single and smoothly evolving nodes . its general form @xmath256 has @xmath216 dominant because it involves the s - wave @xmath86 multipole , which accounts for the delay in the onset of nodes , which do occur when the @xmath236 p - waves allow @xmath193 to compete with @xmath257 + ( d ) * target - recoil spin observables * + now consider the nodal trajectory plots in fig . 3 for : @xmath258 for @xmath259 recall that fts * h3 * indicates that @xmath260 if @xmath77 amplitudes vanish . similar nodal structure for these two observables is then expected for models with small 3/2 amplitudes ; this holds true for the as case and , to a lesser extent , for the wjc model . for @xmath261 the s - wave dominance of the @xmath262 term prevents low energy nodes . by the time p - waves enter they are strong enough to invoke quadratic terms and generate double nodes ; hence the bifurcation for all nodes is seen . this is similar to the @xmath131 case , except there is no even node @xmath263 theorem . the form of @xmath264 is @xmath265 we have another case where the s - wave dominance of the @xmath193 term produces @xmath116 nodes for all models near threshold , followed by a rapid variation for the saly case due to its 3/2 resonance . once the 3/2 amplitudes turn on at higher energies , in part from _ t_-channel effects , then the @xmath190 term enters and double nodes appear . this is similar to the @xmath117 case . finally , with the form @xmath266 the s - wave makes @xmath216 dominant and nodes are therefore postponed until high energies ( 1.5 gev ) for wjc and as ( just barely seen , since this is the end of its region ) and at 1.55 gev for the saly model . many of the characteristics described above are determined by the dominance of s - waves and the subsequent appearance of p - waves . the most interesting cases are those that arise from isobar resonance and/or _ t_-channel effects . each multipole includes reference to the final k@xmath267 orbital angular momentum . hence , explicit truncations are suggested and additional rules * m1 * @xmath268 * m10 * can be deduced for observables near threshold . the assumption involved in deducing these additional rules is that centrifugal barrier suppression of amplitudes dominates the dynamics , and that resonance or other special dynamic effects can be neglected . however , dynamical models do include baryon resonances and particular _ t-_channel exchanges and therefore it is expected that the fts * m1 * @xmath269 * m10 * rules will be broken . in contrast , the helicity rules based on general symmetry principles are not and should not be broken . to facilitate comparison of the dynamical results of as , wjc , and saly to the rules offered by fts , it is best to examine the argand plots of multipoles(appendix b ) for the as , wjc and saly models . the @xmath270 and @xmath271wave multipoles are displayed in fig . 5a , while the @xmath272waves are in fig . note the scale is fixed in fig . 5a , but varies with model in fig . 5b with the as having the smallest scale . for convenience , the multipoles are phase rotated to give a real @xmath273 multipole at theshold . only that @xmath274wave multipole is nonzero at threshold ; all others start at zero and evolve as @xmath275 where @xmath276 is the @xmath277 linear momentum . the wjc model displays a rapid variation in @xmath86 due to its @xmath278 isobar ; the two other models have @xmath86 multipoles that decrease without resonance ( counter - clockwise looping ) structure . structure due to the @xmath279 of wjc is seen in the energy evolution of its @xmath280 multipole , with all of its other multipoles evolving nonresonantly . the as displays @xmath280 structure from its @xmath281 isobar . the @xmath282 multipole has nonresonant structure in the as model . for the saly case , that multipole displays resonance looping due to the added @xmath283 isobar , which also introduces some @xmath284 structure in the saly model . the saly model s @xmath280 also acquires resonant structure apparently due to its @xmath285 . the @xmath272wave multipoles ( fig . 5b ) are quite different for the three models . only the saly model has explicit @xmath272wave isobars and consequently that model displays resonant @xmath286 and @xmath287 @xmath203 looping . the @xmath202 multipoles @xmath288 evolve smoothly from zero . the small as @xmath272wave multipoles have some interesting structure , which does not arise from explicit @xmath272wave isobars . for the wjc model , there are no @xmath272wave resonances and the multipoles evolve without structure . note that the @xmath289channel mechanism illustrated in fig . 4 feeds mainly into the @xmath203 and @xmath202 multipoles . thus , aside from the saly case , these @xmath272wave multipoles are connected to that mechanism . the main properties of the argand plots of fig . 5 are the relative strengths of the @xmath290 multipoles and their relation to their input resonances and the strength of their _ t-_channel mechanisms . the @xmath239 multipole @xmath86 is naturally dominant near threshold . it decreases smoothly for the saly and as models , but due to the isobars in the wjc model , their @xmath86 multipole has a rapid looping at low energies . several characteristics are worth noting . in the as model : * the @xmath291 multipole is driven by the nucleon _ s-_channel term and a @xmath292 resonance ; * the @xmath293 multipoles are not driven by a resonance nor by _ t-_channel exchange ; * the @xmath294 multipole is driven by one @xmath295 resonance ; * the @xmath205 multipoles are driven by _ t-_channel exchange ; * the @xmath202 multipoles @xmath296 receive _ t-_channel contributions . in the wjc model : * the @xmath291 multipole is driven by the nucleon _ s-_channel term and the @xmath297 plus @xmath298 resonances ; * the @xmath293 multipoles are not driven by a resonance nor by _ t-_channel exchange ; * the @xmath294 multipole is driven by one @xmath299 resonance ; * the @xmath205 multipoles are driven by _ t-_channel exchange , * the @xmath202 multipoles @xmath296 receive _ t-_channel contributions . in the saly model : * the @xmath291 multipole is driven by the nucleon _ s-_channel term and by the @xmath292 resonance ; * the @xmath293 multipoles are driven by a @xmath300 resonance ; * the @xmath294 multipole is driven by the two @xmath301 resonances ; * the @xmath302 multipoles are driven by a @xmath303 resonance and by _ t-_channel exchange , * the @xmath202 multipoles @xmath296 receive _ t-_channel contributions . using these argand diagrams , we can now discuss the fts multipole rules . the cross - section peaking rule of fst , under the neglect of @xmath304 requires that @xmath305 and @xmath306 be within @xmath307 when viewed as vectors in the complex plane . when one plots @xmath308 and compares it to @xmath309 it is seen that the rule * m1 * holds true . although many fts multipole speculations do not occur because of the _ t-_channel mechanism , the rules * m9 * and * m10 * are true at low energies , since @xmath86 dominates and @xmath117 and @xmath115 do have sc nodes at @xmath102 other rules that hold true are * m6 * , which does get realized in the complementary nature of the nodes ; namely , @xmath110 has , while @xmath111 does not have nodes at least below 1.5 gev . the other fts multipole rules are not realized , because the fts truncation assumptions did not take into account that _ t-_channel exchange can introduce higher @xmath310 amplitudes . using fig . 5 , we can classify the relative roles of the multipoles for each model . for the three models , the relative importance of the multipoles , based on the maximum size of the multipole over the full energy range , is given in table i. the resonant multipoles and those that receive strength from _ t-_channel exchange are also indicated . the enhanced role for @xmath272waves is generated by one or another of these mechanisms depending on the dynamics of the model . these multipole strengths are consistent , of course , with the driving resonances and the _ t-_channel strengths of each model . the big surprise for fts is the importance of @xmath311 multipoles in the @xmath203 and @xmath202 states , due to kaonic exchanges . based on general symmetry requirements , fts @xcite deduced general rules for the 15 spin observables in the photoproduction of pseudoscalar mesons . these rules , supplemented by assumptions of smooth energy evolution and centrifugal barrier dominance can be used to define the `` normal '' behavior of spin observables . deviation from some of these rules indicate a serious violation of a symmetry , such as parity violation . deviations due to non - smooth energy evolution , or dominance of selected states , are of dynamical origin , as in the case of underlying hadronic resonances . as a prelude to analysis of future experimental results , we have confronted the fts analysis with three current models @xcite of the photoproduction of @xmath9 mesons . functions are wrong . as a consequence , their spin observables display a large and incorrect number of nodes near threshold @xcite . we appreciate receiving a copy of their code from c. bennhold and look forward to seeing their corrected parameters . this was a case where the general fts threshold rules @xcite served to detect an error and shows one way these rules can be useful . ] all of the predictions based on parity and angular momentum conservation are realized in these models , including statements about the even or odd number of sign - changing nodes . the only deviations noted are those possibly attributed to special dynamics , such as underlying resonances . indeed , an important conclusion is that observation of the nodes of spin observables , as they unfold with energy , offers a powerful way to extract specific dynamical resonance information . this conclusion is realized by addressing two crucial questions : @xmath312 what can the nodal structure of the _ forthcoming polarization data _ reveal about the highest spin of the intermediate state baryonic resonances required by the reaction mechanism ? @xmath313 in _ dynamical models _ , how can we disentangle the contributions due to genuine baryonic resonances from those mimicked by the kaonic exchanges in line with the duality hypothesis ? to summarize our findings , and in view of the envisioned polarization measurements @xcite we single out our most significant results on the reaction mechanism deduced by confronting the fts rules with specific models . the @xmath3-polarization asymmetry is technically the easiest to measured . here , the nodal structure of @xmath144 ( and also @xmath165 ) is mainly of resonance - driven nature and hence is sensitive to _ explicit s-_channel spin 3/2 resonances . the beam - asymmetry @xmath173 proves to be an appropriate observable in testing the validity of the _ duality hypothesis _ in the strangeness sector . this duality hypothesis is verified by investigating the underlying dynamics of three models ; wherein , 5/2 amplitude strength influences spin observables @xmath314 @xmath192 and/or @xmath315 interference , even though explicit 5/2 resonances are not included in the models . the 5/2 strength arises from _ t- _ rather than from _ s-_channel dynamics . the beam - asymmetry observable is also very suitable in investigating the role of _ explicit _ spin 5/2 resonances in the reaction mechanism . the beam - recoil asymmetries are found very attractive mainly with respect to their sensitivity to @xmath63 resonances . the richest information is embedded in the @xmath117 observable , obtained using a circularly polarized beam . the odd number of its nodes are produced through @xmath316 and @xmath317 mechanisms according to the dynamical ingredients of the models , e.g. , the explicit presence of spin 3/2 resonances _ versus _ the manifestation of the duality through the _ t-_channel exchanges . this property is also true for the beam - target observable @xmath131 . in addition , the cubic term in @xmath318 arises through the 3/2 amplitudes interferences . the other beam - recoil observable ( @xmath114 ) with circularly polarized beam shows high sensitivity to the @xmath63 aligned @xmath144-waves producing an _ explicit _ signal for the @xmath63 resonances . the asymmetries @xmath167 and @xmath110 corresponding to a linearly polarized photon beam contain important information on @xmath319 interferences . even at low energies , the @xmath167 observable shows deviations from fts rules for the three models because of _ dynamical _ effects . the next double polarization family , beam - target , contains some common features with the above asymmetries . namely , the observables connected to the circularly polarized beam , @xmath131 and @xmath166 , have behaviour comparable to @xmath117 and @xmath173 , respectively . for the linearly polarized beam observables the situation is slightly different . the @xmath61 asymmetry manifests characteristics similar to those of @xmath144 , with _ s-_channel spin 3/2 resonance effects amplified because of an enhancement due to a large @xmath86 multipole in the @xmath318 term , while the non - resonant driven nodal structure due to duality induces a very different evolution . the @xmath174 asymmetry is driven by an spin rich interference between 1/2 , 3/2 , and 5/2 terms . hence , the appearence of a single node in this observable provides evidence for significant @xmath320 wave interference and possible @xmath271waves enhancement . the last set of observables , target - recoil asymmetries , are characterized by the dominance of s - waves and the subsequent appearance of p - waves arising from isobar resonance and/or _ t_-channel effects , similar to the cases seen already for observables within other families . this redunduncy in information content is of course expected @xcite from analysis of the number of independent experiments . nodal angle versus @xmath321 trajectories , based on _ direct experimental information _ , rather than specific dynamical models , should provide a powerful tool in pinning down the reaction mechanism of the strangeness electromagnetic production processes and , hopefully , in the search for missing resonances @xcite . we wish to thank drs . c. bennhold , j. c. david , s. a. dytman , r. a. eisenstein , c. fayard , g. h. lamot , and f. piron , for their help and encouragement . we both wish to express our appreciation for the warm hospitality during visits to the university of pittsburgh(b.s . ) and to saclay(f.t . ) . the definition of the various spin observables are provided in the literature @xcite . for convenience , we present a brief discussion of the sixteen observables . the differential cross - section is defined by : @xmath322 with @xmath276 the final and @xmath323 the initial c.m . momenta . here we extract the angle dependent function @xmath324 which is used in fts to define `` profile functions . '' these profile functions are denoted by @xmath325 for any spin observable @xmath326 the profile functions are determined by bilinear products of amplitudes and therefore are useful for extracting amplitude information . of the sixteen observables , one is : ( 1 ) the cross - section function @xmath327 ; three are _ single spin observables _ : ( 2 ) the polarization of the produced @xmath328 @xmath329 ( 3 ) the polarized target asymmetry , @xmath330 and ( 4 ) the photon polarization asymmetry @xmath331 the remaining twelve spin observables are _ double spin observables_. these are further classified as involving : ( bt ) polarized beam , polarized target ; ( br ) polarized beam , polarized recoil @xmath3 ; ( tr ) polarized target , polarized recoil @xmath3 . each of these three types of double spin observables have four members . the ( br ) type is also called a spin transfer observable . the ( tr ) type is also called a spin depolarization observable , often denoted by the symbol d ; for example , the depolarization spin transfer variable where the incident p and the final @xmath3 spin directions are normal(n ) to the scattering plane is usually called @xmath332 . in the notation of fts that spin variable is called @xmath333 their notation is based on the spin correlation description , where the superscript indicates the two particles involved ( e.g. , the bt , br , or tr classification ) and the subscript indicates direction . these directions are denoted by either the initial unit vectors , @xmath334 or the final unit vectors , @xmath335 another convention used is that of normal n , sideways s , and longitudinal l directions : @xmath336 the relations between these spin - correlation ( double spin ) observables and the conventional set used here is for bt : @xmath337 @xmath338 for br : @xmath339 @xmath340 and for tr : @xmath341 @xmath342 the cartesian components refer to the spin axis for the baryons ; for the photon they indicate either linear or circular polarization states ( see ref . @xcite for a discussion ) . argand plots of the the electric @xmath343 and magnetic @xmath344 multipoles are obtained from the cgln amplitudes @xmath345 and @xmath346 by the following projection integrals : @xmath347 here @xmath348 @xmath349 are the legendre polynomials and the cgln amplitudes are given in ref . @xcite , where they have been calculated using diagrammatic techniques in an isobaric approach . research supported in part by the u. s. national science foundation . m. bockhorst et al . , z. phys . c * 63 * , 37 ( 1994 ) . r. a. schumacher , nucl . phys . * a585 * , 63c ( 1995 ) . j. p. bocquet et al . , in _ proceedings of the particles and nuclei xiii international conference _ , perugia , 1993 . c. g. fasano , f. tabakin and b. saghai , phys . c * 46 * , 2430 ( 1992 ) . f. tabakin nucl . phys . * a570 * , 311c ( 1994 ) ; cetin savkli , m. pichowsky and f. tabakin , _ polarization observables in vector meson photoproduction _ in preparation . r. a. adelseck and b. saghai , phys . c * 42 * , 108 ( 1990 ) . r. williams , c. ji , and s. cotanch , phys . c * 46 * , 1617 ( 1992 ) . david , c. fayard , g.h . lamot , and b. saghai , in _ proceedings of the 14th international few - body conference _ , williamsburg , 1994 , edited by f. gross , aip proceedings no 334 ( 1995 ) . see , e.g. , p. collins , _ an introduction to regge theory and high energy physics _ , ( cambridge university press , 1977 ) . i. s. barker , a. donnachie , and j. k. storrow , nucl . b25 * , 179 ( 1970 ) ; _ ibid _ * b95 * , 347 ( 1975 ) . j. ajaka , p. hoffmann - rothe , b. saghai , and f. tabakin , _ in preparation _ h. tanabe , m. kohno , and c. bennhold , phys . c * 39 * , 741 ( 1989 ) . b. saghai , in _ proceedings of the sixth workshop on perspectives at intermediate energies _ , trieste , 1993 , edited by s. boffi , c. ciofi degli atti , and m. giannini , world scientific ( 1994 ) . r. koniuk and n. isgur , phys . d * 21 * , 1868 ( 1980 ) ; s. capstick and w. roberts , _ ibid _ d * 49 * , 4570 ( 1994 ) . .the multipoles for the three models listed in order of their maximum size from threshold to the maximum energy range of the corresponding model , based on fig . 5 . the boldfaced multipoles display counter - clockwise resonance looping . the multipoles marked by a dagger receive strength from _ t-_channel exchange . [ cols="^,^,^,^,^,^,^ " , ]

the dust formation process in the winds of asymptotic giant branch stars is discussed , based on full evolutionary models of stars with mass in the rangemmm , and metallicities . dust grains are assumed to form in an isotropically expanding wind , by growth of pre existing seed nuclei . convection , for what concerns the treatment of convective borders and the efficiency of the schematization adopted , turns out to be the physical ingredient used to calculate the evolutionary sequences with the highest impact on the results obtained . low mass stars with mm produce carbon type dust with also traces of silicon carbide . the mass of solid carbon formed , fairly independently of metallicity , ranges from a fewm , for stars of initial massm , tom for mm ; the size of dust particles is in the rangem m . on the contrary , the production of silicon carbide ( sic ) depends on metallicity . for the size of sic grains varies in the range m , while the mass of sic formed is . models of higher mass experience hot bottom burning , which prevents the formation of carbon stars , and favours the formation of silicates and corundum . in this case the results scale with metallicity , owing to the larger silicon and aluminium contained in higher z models . at z= we find that the most massive stars produce dust massesm , whereas models of smaller mass produce a dust mass ten times smaller . the main component of dust are silicates , although corundum is also formed , in not negligible quantities ( ) . [ firstpage ] stars : abundances stars : agb and post - agb . ism : abundances , dust

1401.1332

during the last years the evolution experienced by stars of intermediate mass after the consumption of central helium has received a growing interest by the astrophysical community . this phase , known as asymptotic giant branch ( agb , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , is much shorter than the previous stages of core hydrogen and helium burning , but is extremely important for the role played by these sources in the pollution of the interstellar medium . during the agb evolution the stars loose all their external mantle , returning into the interstellar medium material whose chemical composition was altered by internal nuclear processes . agbs have been suggested as the main contributors to the pollution of the interstellar medium in globular clusters , giving rise to the formation of multiple populations @xcite . the gas ejected by agbs , possibly diluted with pristine gas , may stimulate the formation of one or more additional stellar components , overlapped to the original population present in the cluster @xcite . this scenario can explain most of the observational data , such as the chemical anomalies involving light elements @xcite and the photometric features of the main sequence and of the horizontal branch of globular clusters @xcite . the interest towards agbs stems also from the physical conditions of their circumstellar envelopes , particularly suitable to gas condensation into dust grains . the surface layers of these stars are sufficiently cool ( @xmath18k ) to allow dust formation at typical distances of 310 stellar radii from the surface , where the densities are still sufficiently large ( @xmath19 , * ? ? ? * ) to allow formation of meaningful quantities of dust . dust formation is a complex phenomenon , which is strongly interfaced with the dynamical and thermal structure of the winds of this class of objects @xcite . stellar pulsation is known to trigger the formation of periodic shocks , that favours the increase in the local density , thus stimulating gas condensation in cool regions in the stellar surroundings . radiation pressure on dust grains is currently believed as the dominant mechanism triggering mass loss in these structures @xcite . a self consistent description of the dust formation process would require full coupling of results from stellar evolution modelling with the description of the wind . a natural outcome of such a treatment would be the outwards velocity of gas and dust particles leaving the star , thus the mass loss rate experienced . because we are still far from achieving this target , the few models available in the literature treat the wind independently , based on the results coming from the integration of the equations of stellar equilibrium , from which the values of luminosity , effective temperature and mass loss rate of the star are obtained @xcite . these studies may be considered as a preliminary step towards a more complete treatment , where central object and external wind are described simultaneously . the reliability of the results obtained , in terms of the amount of dust formed , is hampered by the uncertainties associated to the wind model , for what concerns , e.g. , the initial velocity of gas particles , the opacities relevant to determine the radiation pressure on grains , the sticking coefficients of each dust species @xcite . agb modelling is also relevant for the results obtained . the type of grains formed , either silicates or carbonaceous particles , depends on the description of the processes able to alter the surface chemistry of agbs , i.e. third dredge up ( tdu ) and hot bottom burning ( hbb ) . the efficiency of either mechanisms is affected by the modelling of convection and in particular by the treatment of the convective / radiative interface @xcite and of the temperature gradient in regions unstable to convective motions @xcite . the description of mass loss is also extremely relevant , because it has a direct impact on the agb evolutionary time scale @xcite ; furthermore , the mass loss rate determines the density profile of the wind , thus the amount of dust formed . it is for all these reasons that results in the literature presented by different research groups , though using the same description of the wind and of the dust grains formation and growth , differ considerably in the mass and composition of dust formed @xcite . the scope of this work is to discuss how the properties of dust formed in the winds of agbs depend on the implementation of some fundamental features of the model , namely : i ) the macrophysics adopted to describe the agb evolution ; ii ) the thermodynamical description of the wind : iii ) the formation and growth of grains . we present a new grid of dust formation models assuming stellar metallicities of z=@xmath20 , which fills the gap among the chemistries of our previous explorations , i.e. z=@xmath21 @xcite , z=@xmath15 @xcite , z=@xmath22 @xcite . this allows to explore in details the role played by the chemical composition in the dust formation process . also , this metallicity will enable to compare model predictions with observational data in the small magellanic cloud @xcite and in metal rich galactic globular clusters @xcite . in addition to our previous investigations on this topic , we present agb models of different metallicities where i ) some extra mixing is assumed from any convective border ( including the base of the external mantle ) ; ii ) the description of mass loss is based on hydrodynamical wind models including carbon dust formation @xcite . the paper is organized as follows : in section 2 we present an overview of the status of the art of the modelling of dust formation around agbs ; sections 3 and 4 describe the inputs used to model the agb evolution of stars and the dust grain growth in the expanding winds ; carbon dust formation is discussed in section 5 , whereas silicates production in more massive agbs experiencing hbb is presented in section 6 ; finally , section 7 is devoted to an overall discussion of the results obtained . the pioneering explorations by the heidelberg group @xcite set the framework to describe the dust formation process in the winds of agbs . the wind is assumed to expand isotropically from the surface of the star , moving with velocities of a few km / s , entering into the dust formation layer , where gas ( and dust ) particles are accelerated via radiation pressure on the newly formed grains . the dust grains are assumed to grow owing to the gas molecules impinging on the already formed solid particles . the rate at which the grains grow scales with the density of the gas ; this defines a very narrow region where dust formation occurs . an asymptotic behaviour is reached , with the wind expanding outwards with large velocities ( tipically @xmath23 km / s ) . this is a simplified view of a much more complex situation , where large amplitude pulsations trigger periodic shocks @xcite : gas particles , rather than expanding at constant velocity , are projected on ballistic trajectories , where they contribute to the increase in the local density in regions very cool , thus extremely suitable to dust formation @xcite . a complete , self consistent treatment would require the simultaneous description of the periodic variation of the stellar properties and of the structure of the wind , such that condensation takes place in the regions where the shock favours the increase in the density . a natural outcome of such a treatment would be the determination of the mass loss rate , via the density and the velocities of the gas leaving the star @xcite . such a self consistent approach is still missing and the models currently available treat the wind independently , assuming as inputs the physical parameters of the central star , i.e. effective temperature , luminosity , mass loss rate and the surface chemical composition . the most relevant shortcoming of this approach is the lack of any feedback between the formation of dust grains and the mass loss rate , which is assumed apriori ; we will come back to this point in the following . because the structure of the wind is based on the properties of the central star , the effective temperature ( t@xmath24 ) , luminosity ( l ) and mass loss rate ( @xmath25 ) are key quantities in determining the amount of dust formed . condensation of gas molecules is favoured at low t@xmath24 s , because smaller temperatures in the wind allows the dust forming region to be closer to the surface of the star , in higher density zones . larger values of @xmath25 favour dust formation , because mass conservation requires the density in the wind to increase linearly with @xmath25 ( see eq . [ eqro ] below ) . luminosity has no direct impact on the dust formation process , but it affects the acceleration due to radiation pressure , which scales with l ( eq . [ eqv ] and [ eqgamma ] below ) . the surface chemistry of the star is relevant in determining which kind of dust is formed . the co molecule has an extremely large dissociation energy ( 11ev ) , which means that the least abundant between carbon and oxygen is bound in co molecules . in c rich environments there is no oxygen available , thus the only dust that can be formed is solid carbon , silicon carbide ( sic ) and iron @xcite . when the c / o ratio is below unity all the carbon present is locked into co : silicates , iron and corundum can form @xcite . the physical evolution of agbs , and of their surface chemistry , depends on some fundamental physical properties such as convection and mass loss @xcite . the treatment of the convective instability is the largest uncertainty affecting stellar evolution models . owing to the lack of a self consistent theory of turbulence , the transport of energy within regions unstable to convective motions is treated parametrically . the mixing length @xmath26 , the typical scale of convective motions , is assumed to be proportional to the pressure scale height h@xmath27 : @xmath28h@xmath27 . the parameter @xmath29 is calibrated to reproduce the current radius of the sun ; with the latest updates in the micro physics adopted , recent investigations give @xmath30 @xcite . the agb phase is the most sensitive to the convection model adopted . models based on the full spectrum of turbulence ( fst , * ? ? ? * ) evolve at larger luminosities , on more expanded configurations , in comparison with their counterparts calculated with the traditional mixing length theory @xcite . fst models loose their envelopes more rapidly , and experience on the average a smaller number of thermal pulses ( tp ) . the efficiency of the convective model adopted has also an important feedback on the hot bottom burning phenomenon , in models with mass m@xmath31m@xmath3 : the base of the convective envelope becomes sufficiently hot ( t@xmath32 mk ) to favour an advanced proton capture nucleosynthesis , with the destruction of the surface carbon , and , when the temperatures exceed @xmath33 mk , the depletion of oxygen and the production of aluminium via activation of the mg al chain . hbb is found in all models with core mass above @xmath34m@xmath3 when convection is modelled according to the fst formulation . also mlt models with @xmath35 achieve much more easily hbb conditions @xcite . the ignition of hbb is accompanied by the increase in the luminosity @xcite , which , in turn , triggers the increase in the mass loss rate . in models experiencing hbb the production of silicates is greatly favoured : the c star stage is inhibited by the destruction of the surface carbon , and the star evolves at large luminosities , loose mass at a high rate , which increase the dust formation rate . low mass agbs ( m below @xmath36m@xmath3 ) do not reach hbb conditions , independently of convection modelling . in these stars the only mechanism able to alter the surface chemistry is the third dredge up , which gradually enriches the surface in carbon , eventually making c / o to exceed unity . initially , when the mass loss rate is small and the c star stage is not yet reached , small quantities of silicates are produced ; in the final evolutionary stages carbon dust is produced in great quantities , owing to the large surface carbon abundances and the high mass loss rates experienced . while there is a general consensus on the sequence of events given above , the amount of carbon dust formed is still matter of debate , because the evolutionary stage at which the star becomes a c star and the extent of the surface carbon enrichment depend on the inwards penetration of the convective envelope during the tdu ( see discussion in * ? ? ? this is rendered uncertain by the poor knowledge of the behaviour of convective eddies near the convective / radiative interface , particularly of the extent of the extra mixing , i.e. the distance they travel within radiatively stable regions . the extra mixing from convective cores of stars in central burning phases can be calibrated via comparison with the observed main sequences of open clusters @xcite , by analysis of binary star data @xcite , and asterosismological investigations @xcite . the situation for the agb phase is more complex , as there is no observable that allows a straight calibration of the overshoot from the base of the envelope , and from the boundaries of the shell which forms at the ignition of each tp . the most widely argument used so far is the reproduction of the carbon star luminosity function in the magellanic clouds , which allows to determine the core mass at which tdu begins , and the extent of the inwards penetration of the convective envelope @xcite . these information can be used to estimate the extension of the mixed regions within the agbs interiors . the larger is the extra mixing , the more carbon dust is produced , because more and more carbon is dredged up to the surface regions . the description of the mass loss mechanism is also extremely relevant in determining the evolution of agbs @xcite . the rate at which agbs loose mass determines the duration of the whole evolutionary phase @xcite , but has also important consequences on the modification of the surface chemistry . in massive agbs , experiencing hbb , the mass loss rate is stricly correlated with the degree of p capture nuclesynthesis at the bottom of the surface envelope : lower rates allow a stronger nucleosynthesis , thus larger changes in the abundances of the elements involved in the cno , ne na and mg al cycles . in the low mass regime , large rates favour faster consumption of the envelope : the star undergoes a smaller number of tps , thus the carbon surface enrichment is reduced . this has a strong impact on the dust formation in the circumstellar envelopes . a direct effect of the mass loss description is associated with the higher density of the winds when the mass loss is large , which enhance dust production . also , as stated above , mass loss changes the surface chemical composition , and the mass fractions of those elements relevant to allow dust formation ( e.g carbon for carbon type dust ) : this clearly changes the amount of dust which can be formed . the evolutionary sequences presented here were computed by means of the aton code for stellar evolution . the interested reader may find in @xcite a detailed description of the numerical structure of the code . the most recent updates can be found , e.g. , in @xcite . we recall here the main physical inputs , most relevant for the topic of the present investigation . the arguments presented in the previous section outline how relevant the description of convection is for the agb phase , in terms of the evolution of the main physical quantities , number of thermal pulses experienced , change in the surface chemical composition due to hbb and tdu . in agreement with other investigations by our group , the temperature gradient within convective regions is found by the fst model , developed by @xcite . mixing of chemicals and nuclear burning were coupled with a diffusive approach , according to the scheme suggested by @xcite . the overshoot of convective eddies into radiatively stable regions is described by an exponential decay of the velocity @xmath37 beyond the formal border found via the schwarzschild criterion . the relation used is @xmath38 where the argument of the exponential is taken positive for overshoot from the upper border of the convective zone , whereas the minus sign is used for overshoot below the convective region . @xmath39 and @xmath40 are the values of velocity and pressure at the formal boundary . the extent of the overshoot is given by the free parameter @xmath41 . for the convective cores that form during the hydrogen and helium burning phases , and generally for the phases previous the agb evolution , we followed the calibration given in @xcite , and assumed @xmath42 . for what concerns the agb phase , we calibrated the extent of the extra mixing from the base of the convective envelope and from the borders of the shell that forms at the ignition of each thermal pulse based on the luminosity function ( lf ) of carbon stars in the large magellanic cloud @xcite . we find that adopting @xmath43 leads to a satisfactory agreement between the observed and the predicted lfs . with this assumption we recover the relation between initial mass and luminosity of the transition from o rich to c rich surface chemical composition given in @xcite . in agreement with our previous explorations focused on the agb evolution , we use for the phases preceeding the c star stage the formulation by @xcite . the blcker formula ( @xmath44 , where m , r and l are , respectively , the stellar mass , radius and luminosity ) is suitable to describe the steep increase of @xmath45 with luminosity that characterizes massive agbs ; however , it underestimates the mass loss suffered by c stars , because it is based on a description of the circumstellar envelope of mira variables that neglects the carbon dust formation process , and the radiation pressure on dust grains , which induces a great acceleration of the wind . to describe the c star phase we rely on the formulation by @xcite , giving the mass loss rates experienced by low mass agbs as a function of mass , luminosity and effective temperature . the expression used is @xmath46 , where typical values for the exponents ( partly depending on the metallicity ) are @xmath47 , @xmath48 , @xmath49 . the radiative opacities @xmath50 for temperatures above @xmath51 k were calculated using the opal online tool @xcite ; for smaller temperatures we used the aesopus tool described in @xcite . this choice allows to account for the increase in the opacity associated with the change in the surface chemistry determined by tdu . the pioneering study by @xcite shows how the modelling of agbs experiencing tdu is sensitive to the set of opacities adopted ; detailed discussions on this argument can be found in @xcite . the new models presented in this work have metallicity z=0.004 , initial helium y@xmath52 ; the mixture is alpha enhanced , with @xmath53fe@xmath54=+0.2 $ ] . a discussion of their main physical properties , and of the change in the surface chemistry determined by tdu and hbb , can be found in @xcite . here we focus on the dust formed in their surroundings . in addition to this new grid , we also present updated models for stellar metallicities z=0.001 , z=0.008 ( discussed in previous papers ) obtained with the new implementation of the extra mixing from convective regions during the agb phase and the mass loss description of the c star phase . the structure of the wind and the dust formation process is described following the schematization by @xcite . as discussed in the introduction , the thermodynamic structure of the wind and the dust formation process are entirely determined by the physical and chemical conditions at the surface of the central star . the latter is assumed to evolve independently of the properties of the wind : this is possible because the mass loss rate is determined via a parametric recipe , rather than by full coupling of star s and wind s descriptions . the wind is assumed to expand isotropically from the surface of the star , described by a mass m , luminosity l , radius r , and effective temperature @xmath55 . the thermodynamic structure of the wind is set by the relations giving the variation of density ( @xmath56 ) and temperature ( t ) with the distance @xmath57 from the center of the star . indicating with @xmath45 the rate of mass loss , we have @xmath58 where v is the velocity of the wind . the temperature of the gas in the wind decreases with @xmath57 according to the relation given by @xcite @xmath59 . \label{eqt}\ ] ] the optical depth entering eq.[eqt ] is found by the differential equation @xmath60 the boundary condition for eq.[eqtau ] is that @xmath61 vanishes at infinity ( for numerical reasons , we impose that this condition is reached at a distance of @xmath51r from the centre , where the asymptotic behaviour of all the quantities is definitively reached ) . the model is completed by the description of the acceleration of dust particles under the effects of gravitational attraction from the star and radiation pressure on dust grains : @xmath62 with @xmath63 for the wind velocity , we assume that it enters the dust forming region with a constant velocity @xmath64 km / sec , and is accelerated by radiation pressure . when dust formation is efficient , the results show only a modest dependence on the assumed value for @xmath65 . the system of equations given above can be closed if we know the opacity coefficient @xmath50 , needed to find @xmath66 ( eq.[eqgamma ] ) , and the optical depth ( eq.[eqtau ] ) . @xmath50 can be found by summing the gas and dust contributions . the former is calculated by means of the aesopus tool , whereas for the latter we calculate the individual contributions from the various dust species formed . in a carbon rich context the species of dust formed are solid carbon , iron and silicon carbide , whereas for m stars we consider the formation of silicates ( olivine , pyroxene and quartz ) , solid iron and corundum . this latter species was not present in our previous investigations . for each species it is defined a key element , the least abundant among the elements present in the molecules concurring to the condensation process ( see table 1 for the details of the formation reactions of the various dust species considered ) . the abundance of the key element in the wind is extremely relevant for the amount of dust of a given species that can be formed , because the growth rate ( @xmath67 ) of the grains scales with the number density of the key species , whereas the destruction rate ( @xmath68 ) is proportional to the vapour pressure of the molecule containing the key element in equilibrium with the solid particles . [ cols="<,<,^,^",options="header " , ] [ tabrates ] the growth of the grain size of species @xmath69 is calculated via the equation @xmath70 where @xmath71 is the volume of the nominal molecule in the solid . the vapour pressures needed to estimate the decay rates @xmath68 were found by calculating the free enthalpy of formation of the molecules involved . most of the computations are based on the release by @xcite , with the only exceptions of carbon and sic , that were kindly provided by prof . gail ( private communication ) . following @xcite , we do not consider any destruction reaction for solid carbon , but we assume that it can form only at temperatures below 1100k . the growth rates @xmath67 are proportional to the sticking coefficients @xmath72 of the molecule including the key elements on the solid grains . the choice of various coefficients is the same as in our previuos investigations . for corundum , owing to lack of laboratory data , we explore the range 0.11 . the composition of olivine and pyroxene depends on the relative fraction of magnesium and iron , indicated , respectively , with @xmath73 and @xmath74 ( see table 1 ) . the magnesium percentages in the olivine ( @xmath75 ) and pyroxene ( @xmath76 ) dust particles are found via the equations @xmath77 \label{eqxol}\ ] ] @xmath78 \label{eqxpy}\ ] ] where @xmath79 is defined as the relative gas abundance of magnesium with respect to the mg+fe sum . the quantity @xmath80 gives the difference between the exchange rate of iron by magnesium per unit surface area during collisions of mg with the grain surface and the rate of the reverse reaction , and is defined as in @xcite . the amount of dust formed is indicated by the fraction of gas condensed into solid grains . this is expressed by the fraction @xmath81 of the key element that is in the solid state . we have @xmath82 where @xmath83 is the initial size of the seed nuclei in the wind , @xmath84 is the number density of the key elements in the wind , normalized to the hydrogen density , and @xmath85 is the normalized density of the seed nuclei . we assumed for each species @xmath86 m ; the final results are independent of this assumption , provided that a miminum amount of dust , sufficient to accelerate the wind , is formed . the situation for @xmath85 is more tricky . the typical number densities of grains in the outflow of m stars of solar metallicity has been determined observationally , and is of the order of @xmath87 @xcite . the situation for c stars and for m stars of different metallicity could be different ( see the discussion in nanni et al . , 2013 ) , but presently any choice for @xmath85 would have a large degree of arbitrariness . the opacity coefficient is found via the equation @xmath88 where the @xmath89 s are the opacity coefficients for the various dust species formed . the optical constants used to determine the opacities are the same as in @xcite . the coupled system of differential equations for the wind and the dust formation of the various species is solved with an adams bashforth integrator @xcite . the equations are solved in the radial outward direction . all the r.h.s . [ eqagrain ] are set to zero until the first dust species become stable . [ refmodel ] + 10pt models with m@xmath90m@xmath3 form silicates at the beginning of the agb evolution , whereas in more advanced phases repeated tdu episodes lead to c / o@xmath91 . the surface carbon enrichment depends on the treatment of convective borders , particularly on the extension of the extra mixed region : this zone is stable , based on the schwarzschild criterium , yet it is reached by convective eddies , that cross the convective / radiative interface , pushed by inertia . -230pt locating the borders of convection zones during the agb phase is one of the long standing problems in the context of stellar evolution @xcite . certain trends have been identified by extant models , such as the increase in the extent of tdu in lower metallicity models @xcite , and the fact that tdu ceases when the mass of the envelope drops below a threshold value @xcite . yet the extension of mixed zones is still unknown from first principles : comparison with the observations of carbon stars in the galaxy @xcite and in the magellanic clouds @xcite indicate that standard models , where mixing is allowed only within zones satisfying the schwarzschild criterium , fail to reproduce the observed patterns , and that some extra mixing is required . the presence of extra mixing favours the increase in the surface carbon , for two reasons : a ) overshoot from the bottom of the convective envelope during the inwards penetration that follows the thermal pulse makes more carbon rich matter to be mixed with the surface layers ; b ) extra mixing from the borders of the convective shell that forms during each tp increases the strength of the pulse , in turn increasing the extent of the inwards penetration of the convective envelope @xcite . previous investigations fixed the conditions for tdu to occur either by imposing the temperature of the innermost layer reached by the penetration of the convective mantle @xcite , or by assuming a relation between the core mass at which tdu begins as a function of mass and metallicity @xcite . in our case the most straightforward way of introducing some overshoot is by assuming a non vanishing value of @xmath41 ( see eq . the impact of assuming some extra mixing on dust production during the agb phase was discussed on qualitative grounds in @xcite , by analysing the role played by @xmath41 ( see eq . 1 ) ; no calibration of the extra mixing was given . in this paper we use @xmath43 , according to the calibration of @xmath41 aimed at reproducing the luminosity function of carbon stars in the magellanic clouds , given in section 3.1 used to describe overshoot from convective cores during the h burning phase , given the different physical conditions in the two cases , and the much shorter duration of each individual tdu , compared to the time scales of core h burning . ] . -230pt the effects of extra mixing are shown in fig . [ 2msun ] , where we compare the results obtained with no extra mixing ( blue , dashed tracks ) , with those found with @xmath43 ( red , dotted ) . we will refer to these models as _ noov _ and _ over _ , respectively . the figure refers to the evolution of a 2m@xmath3 model of metallicity z=0.001 , published in @xcite . in the upper , left panel , we see that the carbon increase is much more pronounced in the _ over _ model : the surface carbon reaches a final mass fraction of x(c)@xmath92 , a factor @xmath93 larger than in the _ noov _ case . this has a straight consequence on the production of solid carbon , as we see in the upper , middle panel , showing the variation of the solid carbon size in the two cases : the size of carbon grains formed in the _ over _ model approaches @xmath94 m , whereas in the _ noov _ case they barely reach @xmath95 m . carbon dust formation occurs at an earlier phase in the _ over _ model , because tdu is more efficient in favouring the transition from m to c star ; in the _ noov _ case , the star looses 0.2m@xmath3 before the condition c / o@xmath96 is reached . -5pt the fraction of carbon condensed into dust ( right , upper panel ) is larger in the _ noov _ case , owing to the smaller amount of carbon available at the surface . the quantity of silicon carbide formed is less sensitive to the treatment of convective borders , as shown in the middle and right , lower panels of fig . [ 2msun ] . this is because the key element for sic is silicon , whose surface content is independent of the efficiency of tdu . clearly production of sic demands that the c star stage has already been achieved , thus sic grains start to form in earlier phases in the _ over _ model . to discuss the role played by the description of mass loss , we compare the results obtained with the treatment based on the blcker s formula with the more recent compilation by @xcite . this latter is more suitable to describe the mass loss mechanism in these evolutionary phases , as it is based on simulations of stellar winds driven by radiation pressure on carbon grains . the rates by @xcite show a great sensitivity to the effective temperature of the star , and show a rapid increase in the rate of mass loss as soon as carbon dust formation begins . the results obtained with the @xcite formula are indicated in fig . [ 2msun ] with black , solid , tracks . we used the same prescription of the extra mixing from convective borders as for the _ over _ case , thus we can analyse the role of mass loss treatment by comparing with the red , dotted lines in the six panels of fig . [ 2msun ] . the mass loss rate given by the @xcite formula are larger once the c star stage is reached , as can be seen in the left , bottom panel : the star experiences a smaller number of tps , thus the surface carbon enrichment ( shown in the left , upper panel ) is lower . despite the lower carbon available , more dust is formed : this is due to the larger density of the wind , favoured by the higher mass loss rate . we see in the middle , bottom panel that the higher @xmath45 also favours a greater production of sic . the overall comparison among results obtained with the different assumptions concerning the convective borders and the mass loss mechanism are shown in fig . [ fcarb ] ( solid carbon produced ) and in fig . [ fsic ] ( silicon carbide ) . the three panels refer to the metallicities z=@xmath21 ( left ) , z=@xmath20 ( middle ) , z=@xmath15 ( right ) . part of the results concerning z=@xmath21 and z=@xmath15 were published in @xcite . here we extend the computations to models calculated with the mass loss by @xcite , and to the metallicity z=@xmath20 ; for this latter z the case @xmath97 with no extra mixing was not explored . in the z=@xmath21 case we see that when extra mixing is neglected only models of initial mass m=@xmath98m@xmath3 become carbon stars and produce solid carbon . use of the mass loss rate by @xcite favours a larger dust production for m@xmath99m@xmath3 . for smaller masses the effect of the higher mass loss rates is counterbalanced by the lower carbon available , such that the results turn out to be approximately independent of the mass loss description . the z=@xmath20 models show a similar behaviour , the impact of mass loss increasing for larger masses . the amount of solid carbon produced ( and also the differences determined by the mass loss treatment ) reaches a maximum around for stars of initial mass around @xmath98m@xmath3 , and becomes progressively smaller as the mass increases , and approaches the threshold limit for hbb ignition . -230pt at z=@xmath15 , similarly to z=@xmath21 , we find that when no overshoot is considered only models with m@xmath100m@xmath3 produce solid carbon . use of the mass loss prescription by @xcite favours an increase in the total solid carbon produced of almost one order of magnitude for masses above @xmath101m@xmath3 . for smaller masses the differences are negligible , because use of the @xcite formula favours a rapid loss of the envelope as soon as the c star is achieved , which prevents further carbon to be dredged up to the surface . [ fsic ] shows that the production of silicon carbide is also dependent on the treatment of convective borders and the description of mass loss . the results are easier to be interpreted in this case , owing to the approximately constant value of the key element , i.e. silicon , which , unlike carbon , is not changed by tdu . models using the mass loss treatment by @xcite produce more silicon carbide , owing to the higher density in the wind . clearly models with no overshoot produce a smaller amount of sic , because they reach the c star stage in later phases , after part of the envelope was lost . we compare the results by @xcite , and the more recent investigation by @xcite with our models . this comparison is aimed to stress the differences in the agb modelling , because the three studies use the same description of the thermodynamics of the wind and of the dust formation process . in fig . [ fconfrontic ] we indicate the results by @xcite and @xcite , respectively , with magenta , full triangles and green , solid diamonds . the comparison for z=@xmath20 is limited to @xcite , as this metallicity was not discussed by @xcite . in all cases the amount of carbon dust formed increases with mass , because more massive stars experience more thermal pulses , dredge up more carbon at the surface , and also loose the external envelope at a higher rate , which increases the number density of gas particles in the wind . for z=@xmath21 the carbon dust produced by our models for m@xmath100m@xmath3 is a factor of @xmath93 smaller compared to @xcite . the interpretation of such a difference is not straightforward , because the models by @xcite are based on a synthetic modelling , and keep the same efficiency of tdu until the latest agb phases , when the mass of the envelope is greatly reduced . also , in our models we account for the modification of the surface chemistry in the computation of the low temperature opacities via the aesopus tool @xcite : this leads to a general cooling of the outer layers , which , in turn , favours the expansion of the external regions , a higher mass loss rate , and a faster consumption of the stellar envelope , before a great enrichment of surface carbon can be achieved @xcite . the smaller densities of carbon nuclei is probably the main reason for the lower amount of solid carbon produced by our models @xcite . the comparison with the results by @xcite also shows that the carbon dust produced by our models is lower , although the differences are smaller than with @xcite . these can be ascribed to the choice of the initial density of seed nuclei : we assumed @xmath102 , whereas @xcite scale @xmath85 with the carbon excess with respect to oxygen ( see session 4.1 in * ? ? ? turning to the metallicity of the new models presented here , z@xmath103 , we see in the top , middle panels of fig . [ fconfrontic ] that the difference with @xcite is smaller than for z=@xmath21 . the models by @xcite produce more dust for m@xmath1042.5m@xmath3 ; this is due to the different convective model adopted , because the use of the fst description partially limits the efficiency of tdu for the masses close to the limit for hbb . for @xmath105 the differences among the three compilations depends on the stellar mass . for m@xmath106m@xmath3 our yields are a factor @xmath93 smaller compared to @xcite and @xcite . these models experience a strong mass loss as soon as they become carbon stars , owing to the large sensitivity of the formulation by @xcite to the effective temperature of the star , and the small temperatures typical of agbs of this metallicity . consequently , they loose their envelope very rapidly , preventing the possibility that large quantities of carbon are dredged up to the surface . for masses above @xmath101m@xmath3 the carbon dust produced by our models is similar to @xcite , and smaller than @xcite ; this likely reflects the difference in the computation of the low t opacities in the c rich mixture , as discussed previously . the amount of silicon carbide formed is independent of the surface carbon enrichment , because in this case silicon is the key element . we see in the bottom panels of fig . [ fconfrontic ] that , independently of metallicity , more sic is formed in our case , compared to @xcite and @xcite . the reason for this is in the details of the growth of sic and solid carbon grains in the winds of c stars . sic is formed in more internal regions , and is rather transparent to the radiation , thus determining only a modest acceleration of the wind . carbon grains are formed in more external layers , and their large opacity favours a fast acceleration of the wind . the formation of carbon grains halts the growth of sic grains , because the acceleration of the wind determines a drop in the density of silicon particles in the wind . in the models by @xcite and @xcite the amount of carbon formed is larger , which , in turn , determines a lower content of sic . in the previous sections we stressed how dust production by c stars is sensitive to many uncertainties , the most relevant being the treatment of convective borders and the description of mass loss . we consider as our reference model the results obtained with the extra mixing during the agb phase calibrated as discussed in section 3.1 , and the mass loss treatment modelled as in @xcite . these results are indicated as black , full circles in fig . [ fcarb ] , [ fsic ] , [ fconfrontic ] . [ fgrains ] shows the variation during the agb life of the grain size of dust particles formed around stars of various masses and metallicities . full points indicate the dimension of solid carbon grains , whereas open points refer to sic . the choice of the current mass of the star as abscissa helps to understand how many grains of a given size are formed around the star during the whole evolution . each point refers to the grain size in the middle of each interpulse phase . for the metallicities @xmath107 and @xmath105 we show results for models of initial masses 1.5 , 2 , 2.5 , 3m@xmath3 ; in the @xmath108 case , since the 3m@xmath3 model shows that hbb is active and that the c star stage is not reached , we show the 1m@xmath3 case instead . concerning solid carbon , we see that for each mass the size of dust particles , a@xmath109 , increases during the evolution , owing to the increase in the surface carbon abundance and in the mass loss rate . values of @xmath110 span the range 0.070.2 @xmath111 m , although for masses m@xmath112 we find @xmath113 m . we see in the three panels of fig . [ fcarb ] that in the low mass regime , with m@xmath114m@xmath3 , the amount of carbon dust formed increases with mass , independently of @xmath115 : the higher is m , the larger is the surface carbon enrichment . in this interval of mass the @xmath108 models produce more carbon , because low@xmath115 models experience deeper tdus @xcite , and reach more easily the c star stage , owing to the smaller content of oxygen initially present at the surface . for masses m@xmath100m@xmath3 dust production is favoured in the higher metallicity models , because these evolve at smaller effective temperatures : dust formation occurs in layers closer to the stellar surface , where the densities are higher . the mass of solid carbon produced , m@xmath116 , ranges from @xmath6m@xmath3 to @xmath117m@xmath3 at @xmath108 , whereas at higher metallicities m@xmath118m@xmath3 for the two masses m=@xmath98m@xmath3 . in the three bottom panels of fig . [ fgrains ] we see that sic grains grow during the evolution until a threshold size is reached , after which no further growth occurs . the maximum dimension achieved by sic grains ranges from @xmath119 m for @xmath108 , to @xmath120 m for @xmath107 , up to @xmath121 m for @xmath105 . this behaviour can be explained by the saturation due to the lack of silicon particles in the wind . owing to the stability of sis molecule , in carbon rich environments we have n@xmath122$]n@xmath123 @xcite , where @xmath124 and @xmath125 are the number density of silicon and sulphur particles in the wind ( normalized to the hydrogen density ) , whereas f@xmath126 is the fraction of silicon condensed into sic . this expression poses an upper limit on the amount of sic that can form , corresponding to the value of f@xmath126 at which n@xmath127 vanishes : @xmath128 because in solar or @xmath129enhanced mixtures @xmath130 @xcite , f@xmath126 can not exceed @xmath131 . the mass of sic ( m@xmath126 ) produced increases with the stellar mass , as can be seen in fig . [ fsic ] , because of the larger mass returned into the interstellar medium by massive agbs during the evolution . metallicity models produce more sic , owing to the larger amount of silicon present in the stars . @xmath108 models produce an amount of sic @xmath132m@xmath133m@xmath134m@xmath3 , whereas at @xmath107 we find @xmath135m@xmath133m@xmath136m@xmath3 ; the largest production of sic occurs at @xmath105 , for which we have @xmath6m@xmath133m@xmath137m@xmath3 . the scaling of the mass of sic produced with the metallicity is approximately linear , as expected . -00pt stellar models with mass above @xmath93m@xmath3 experience hbb during the interpulse phases of the agb evolution . parameter to describe the hbb strength is the temperature at the bottom of the convective envelope , which determines the degree of nucleosynthesis experienced in those regions , and the modification of the surface chemistry . stars that experience hbb reach very high luminosities @xcite , much larger than predicted by the classic relation by @xcite . [ flumhbb ] shows the evolution of the luminosity of agb models of various initial masses and metallicities . the luminosity first increases in the early agb phases , then declines until the envelope is consumed . the initial trend is due to the increase in the core mass of the star , whereas the consumption of the envelope is the reason for the later decrease in the overall energy flux . higher mass models evolve on bigger cores , experience stronger hbb , and evolve at larger luminosities . models with smaller metallicity are substained by a more powerful cno burning shell : hbb conditions are reached more easily in lower @xmath115 models , that evolve at larger luminosities . we see in fig . [ flumhbb ] that for a given mass the @xmath107 models ( hereinafter z4m3 , see ventura et al . ( 2013b ) for a more exhaustive discussion of the physical properties of agb models with @xmath107 ) presented here evolve at a luminosity intermediate between models of the same mass of metallicity @xmath108 ( hereinafter z1m3 ) and @xmath105 ( z8m3 ) . -230pt the evolution of the surface chemistry of massive agbs can also be explained on the basis of the hbb experienced . [ fcohbb ] shows the variation of the surface abundances of oxygen ( solid tracks ) and carbon ( dotted lines ) for the same models shown in fig . [ flumhbb ] . the strong depletion of the surface carbon in the early agb phases is the signature of hbb , starting at @xmath138mk . the depletion of oxygen requires higher temperatures at the bottom of the convective zone ( t@xmath139mk ) , thus it takes place in more advanced phases . the z8m3 models experience a soft hbb , thus the reduction of the surface oxygen is modest ( at most a factor @xmath101 ) . in the z1m3 models the depletion of the surface oxygen is not monotonic with mass : stars with mass around @xmath140m@xmath3 produce the most o poor ejecta , the initial oxygen being destroyed by almost two orders of magnitude . more massive stars , though experiencing a stronger hbb , loose their surface envelope very rapidly , before a very advanced nucleosynthesis can be activated : their surface oxygen is slightly higher @xcite . in all the cases shown in fig . [ fcohbb ] we see that the surface c / o keeps below unity , thus preventing the possibility that the c star stage is reached , and , consequently , that carbon type dust is formed . -230pt the size of the dust grains of the various species that form during the agb evolution is shown in the three panels of fig . [ fsizehbb ] . for clarity reasons we show only the @xmath141m@xmath3 and @xmath142m@xmath3 cases , as representative of the lowest and highest masses experiencing hbb . in agreement with previous investigations , we find that olivine is the dominant species @xcite . for the z4m3 and z8m3 cases the dimension of the olivine grains formed increases with the stellar mass . the grain size @xmath143 is in the range @xmath144 m for @xmath105 , and @xmath145 m for @xmath107 . the z4m3 produce smaller amounts of olivine than their z8m3 counterparts , because although they evolve at larger luminosities and loose mass at a higher rate , their surface silicon ( which scales with the metallicity , and is scarcely touched by hbb ) is smaller . the z1m3 models produce even less olivine , @xmath143 never exceeding @xmath146 m . the formation of olivine grains in the more massive models of this metallicity stops at a certain stage during the agb evolution , because the strong depletion of the surface oxygen ( see the left panel of fig . [ fcohbb ] ) prevents the formation of water molecules , that are essential to form olivine grains . this behaviour was found and discussed in @xcite . + 10pt in the winds of m stars pyroxene is the second most abundant species after olivine . depending on the metallicity , the size of pyroxene grains ranges from @xmath147 m to @xmath146 m ( see fig . [ fsizehbb ] ) . quartz is the least abundant among the silicates , the grain size keeping below @xmath147 m . note that in the early agb phases the size of quartz grains is larger : this is because a smaller amount of olivine is formed , thus the acceleration experienced by the wind is smaller , which contributes to form more quartz . besides silicates , an additional species that forms in the winds of m stars is corundum , an al rich compound very stable @xcite , and extremely transparent to radiation @xcite . in our previous investigations the formation of corundum was neglected , although , owing to its stability , it is likely that considerable amounts of this dust species from in regions close to the surface of the star @xcite . the details of the corundum condensation process and a discussion of the related uncertainties , will be published in a separate paper ( dellagli et al . , submitted ) . the variation of the corundum grain size during the agb evolution is different compared to silicates : corundum particles form in larger and larger dimensions as the evolution proceeds , whereas the size of silicates follows the path traced by luminosity ( see figg . [ flumhbb ] and [ fsizehbb ] ) . this is an effect of hbb , which , via activation of the mg al nucleosynthesis , favours a gradual aluminium enrichment of the surface layers @xcite . the total mass of silicates and corundum is shown in fig . [ fsilhbb ] . for the metallicities @xmath107 and @xmath105 the mass of silicates produced is in the range @xmath21m@xmath133m@xmath148m@xmath3 and increases with the stellar mass , in agreement with the previous discussion on the grain sizes of olivine and pyroxene . the z1m3 models produce a smaller quantity of dust , owing to the scarcity of silicon available , but also because of the strong depletion of the surface oxygen : this is the reason for the dip around @xmath149m@xmath3 , that can be seen in the z1m3 line in the left panel of fig . [ fsilhbb ] . the comparison with the results by @xcite shows that the mass of silicates produced by our models is much larger , owing to the stronger hbb experienced , which , in turn , favours a large increase in the mass rate , hence in the density of the wind . this is particularly evident at the small metallicities , where the production of dust by the @xcite is modest , as a consequence of the scarcity of silicon in the surface layers of the star ( the models by @xcite of metallicity @xmath108 are not shown in the figure , as they would fall off the chosen scale for the vertical axis ) . the comparison with the results by @xcite is less straightforward . at @xmath105 the overall mass of silicates produced by our models is larger , though the two most massive models in the @xcite compilation produce a quantity of silicates comparable to ours . in the @xmath108 case no silicates is produced by @xcite models , because the models , though experiencing a soft hbb , eventually become carbon stars , thus produce carbon type dust . a word of caution is needed here . in the models presented here we follow the same schematization described in @xcite , in that we assume that the disintegration of olivine , and more generally of silicates , takes place by chemisputtering : @xcite discuss the stability of silicates , arguing against the possibility that pure thermal composition is responsible for the disintegration of silicates . this assumption is compatible with the silicates forming reactions given in tab . [ tabrates ] . conversely , @xcite assume that the destruction of silicates grains is triggered by a pure vaporization process . the difference between these two descriptions is that while in the chemisputtering case silicates are not stable at temperatures above @xmath150k , in the second hypothesis this threshold is lifted to @xmath151k . clearly in the latter case more silicates are formed , because the condensation process begins in regions closer to the surface of the star , at larger densities . discriminating among these two hypothesis is beyond the scope of the present investigation , however we are more favourable to the chemisputtering solution because : a ) condensation temperatures for silicates as high as 1500k are not observed in radiative transfer models for o rich circumstellar dust shells @xcite ; b ) observationally , the highest temperatures observed are @xmath152k for aluminum dust @xcite . the mass of corundum produced , @xmath153 , is also increasing with the stellar mass ( see right panel of fig . [ fsilhbb ] ) . because of the saturation conditions reached , there is practically no difference between the @xmath107 models and their @xmath105 counterparts : @xmath153 varies from a few @xmath154m@xmath3 for m@xmath155m@xmath3 , to @xmath156m@xmath3 for m@xmath157m@xmath3 . the @xmath108 models produce a smaller amount of corundum , owing to the low surface mass fraction of aluminium . -00pt in the previous sections we discussed separately the dust yields by stars with mass @xmath158 , that produce carbon dust , and by more massive agbs , that produce silicates . tdu favours the surface carbon enrichment for the lower masses , where hbb dominates for masses above @xmath159 . the total dust produced by stars of various masses and metallicities is shown in fig . [ fdusttot ] . for consistency with the new grid of agb models with @xmath107 presented here , the mass of carbon dust produced by @xmath108 and @xmath105 ( discussed in ventura et al . 2012a , b ) have been recalculated to account for : i ) a small extra mixing from the bottom of the convective envelope and ii ) the mass loss rate by @xcite , that is more suitable to describe mass loss for carbon stars . both these effects increase the quantity of carbon dust formed around m@xmath160m@xmath3 agbs . the extra mixing below the base of the envelope favours a more penetrating tdu , thus a higher surface carbon available . use of the @xcite mass loss leads to higher rates for carbon stars ; this , in turn , determines an increase in the density of the wind , and consequently a higher rate of dust formation . the quantity of dust formed depends on the initial mass of the star . the various lines in fig . [ fdusttot ] , indicating different metallicities , outline an initial increasing trend for masses m@xmath161m@xmath3 , owing to the higher number of tps experienced , which favour a larger carbon enrichment of the surface layers . in these stars @xmath162 of dust is under the form of solid carbon , with @xmath17 of sic and solid iron . the stars with mass around @xmath163m@xmath3 produce smaller quantities of dust , because hbb prevents the formation of carbon type dust , and their mass loss rates are not sufficiently large to trigger the formation of large amount of silicates . higher mass models produce more dust , because they evolve at larger luminosities , thus they loose mass at larger rates . most of the dust produced ( @xmath164 ) are silicates ( mostly olivine ) , with traces of cordundum and iron . the yields scale with the metallicity in the high mass domain , owing to the quantity of silicon available , which increases with z. in the low mass domain , where carbon dust formation occurs , the results are less sensitive to metallicity . the overall dust produced by the @xmath108 models is smaller , because the minimum mass at which hbb occurs is smaller , which limits the range of masses experiencing carbon enrichment . rs acknowledges that the research leading to these results has received funding from the european research council under the european union s seventh framework programme ( fp/2007 - 2013 ) / erc grant agreement n. 306476 . the authors are indebted to the anonymous referee , for the careful reading of the manuscript , that helped improving the quality of this work . 99 bertschinger e. , chevalier r. a. 1985 , apj , 167 , 190 bessel m. s. , wood p. r. , evans t. l. , 1983 , mnras , 202 , 59 blcker t. , 1995 , a&a , 297 , 727 blcker t. , schenberner d. , 1991 , a&a , 244 , l43 boothroyd a. i. , sachmann , i. j. 1988a , apj , 328 , 653 boothroyd a. i. , sachmann , i. j. 1988b , apj , 328 , 671 bowen g. h. , 1988 , apj , 329 , 299 bressan a. , marigo p. , girardi l. , salasnich b. , dal cero c. , rubele s. , nanni a. , 2012 , mnras , 427 , 127 briquet m. , morel t. , thoul a. , scuflaire r. , miglio a. , montalban j. , dupret , m. a. , aerts c. , 2007 , mnras , 381 , 1482 canuto v.m.c . , mazzitelli i. , 1991 , apj , 370 , 295 carretta e. , bragaglia a. , gratton r. , dorazi v. , lucatello s. , 2009 , a&a , 508 , 695 claret a. , 2007 , a&a , 475 , 1019 cloutman l.d . , eoll j.g . , 1976 , apj , 206 , 548 dercole a. , vesperini e. , dantona f. , mc millan s.l.w . , recchi s. , 2008 , mnras , 391 , 825 dercole a. , dantona f. , ventura p. , vesperini e. , mc millan s.l.w . , 2010 , mnras , 407 , 854 di criscienzo m. , dellagli f. , ventura p. , schneider r. , valiante r. , la franca f. , rossi c. , gallerani s. , maiolino , r. , 2013 , mnras , 433 , 313 dwek , e. , galliano , f. & jones , o. , 2007 , apj , 662 , 927d ferrarotti a.d . , gail h.p . , 2001 , a&a , 371 , 133 ferrarotti a.d . , gail h.p . , 2002 , a&a , 382 , 256 ferrarotti a.d . , gail h.p . , 2006 , a&a , 553 , 576 fleischer a. j. , gauger a. , sedlmayr e. , 1992 , a&a , 266 , 339 gail h. p. , sedlmayr e. , 1985 , a&a , 148 , 183 gail h.p . , sedlmayr e. , 1998 , faraday discussions , 109 , 303 gail h. p. , sedlmayr e. , 1999 , a&a , 347 , 594 golub g. h. , ortega j. m , 1992 , scientific computing and differential equations ( boston : academic press ) gratton r. , carretta e. , bragaglia a. 2012 , ar&av , 20 , 50 grevesse n. , sauval a. j , 1998 , ssrv , 85 , 161 groenewegen m. a. t. , 2004 , a&a , 425 , 595 groenewegen m. a. t. , whiteloak p. a. , smith c. h. , kerschbaum f. , 1998 , mnras , 293 , 18 groenewegen m. a. t. , sloan g. c. , soszynski i. , petersen e. a. , 2009 , a&a , 506 , 1277 herwig f. , 2000 , a&a , 360 , 952 herwig f. , 2005 , ar&a , 43 , 435 herwig f. , austin s.m . , 2004 , apj , 613 , l73 iben i. jr , renzini a. , 1983 , ara&a , 21 , 271 iglesias c. a. , rogers f. j. , 1996 , apj , 464 , 943 izzard r. g. , tout c. a. , karakas a. i. , pols o. r. , 2004 , mnras , 350 , 407 karakas a. i. , lattanzio j. c. , pols o. r. , 2002 , pasa , 19 , 515 karakas a. i. , 2011 , in asp conf . , why galaxies care about agb stars ii : shining examples and common inhabitants . proceedings of a conference held at university campus , vienna , austria , 16 - 20 august 2010 . edited by f. kerschbaum , t. lebzelter , and r. f. wing . san francisco : astronomical society of the pacific , 2011 . , 455 , p.3 karovicova i. , wittkowski m. , ohnaka k. , boboltz d. a. , fossat e. , scholz m. 2013 , a&a , 560 , a75 knapp g. r. 1985 , apj , 293 , 273 koike c. , kaito c. , yamamoto t. , shibai h. , kimura s. , suto h. , 1995 , icarus , 114 , 203 larsen s. s. , clausen j. v. , storm j. , 2000 , a&a , 364 , 455 lattanzio j. c. , boothroyd a. i. , 1997 . in _ astrophysical implications of the laboratory study of presolar materials _ , ed . t. bernatowitz , e. zinner , aip conf . 85 lucy l. b. , 1976 , apj , 205 , 482 marigo p. , 2002 , a&a , 387 , 507 marigo p. , girardi l. , 2007 , a&a , 469 , 239 marigo p. , aringer b. , 2009 , a&a , 508 , 1538 marigo p. , girardi l. , bressan a. , 1999 , a&a , 344 , 123 mattsson l. , wahlin r. , hfner s. , eriksson k. , 2008 , a&a , 484 , l5 montalban j. , miglio a. , noels a. , dupret , m. a. , scuflaire r. , ventura p. , 2013 , apj , 766 , 118 mowlavi n. , 1999 , a&a , 350 , 73 nanni a. , bressan a. , marigo p. , girardi l. , 2013 , mnras , 434 , 2390 ossenkopf v. , henning th . , mathis j. s. , 1992 , a&a , 261 , 567 paczynski b. , 1970 , acta astron . , 20 , 47 piotto g. , in iau symp . 258 , 2009 , the ages of stars , ed . e. e. mamajek , d. r. soderblom , r. f. g. wyse ( cambridge : cambridge university press ) , 233 renzini a. , voli m. , 1981 , a&a , 94 , 175 sharp c. m. , huebner w. f. , 1990 , apjs , 72 , 417 straniero o. , chieffi a. , limongi m. , busso m. , gallino r. , arlandini c. , 1997 , apj , 478 , 332 ventura p. , dantona f. , mazzitelli i. , gratton r. , 2001 , apj , 550 , l65 ventura p. , carini r. , dantona f. , 2011 , mnras , 415 , 3865 ventura p. , dantona f. , 2005a , a&a , 431 , 279 ventura p. , dantona f. , 2005b , a&a , 439 , 1075 ventura p. , dantona f. , 2009 , mnras , 499 , 835 ventura p. , dantona f. , 2011 , mnras , 410 , 2760 ventura p. , di criscienzo m. , schneider r. , carini r. , valiante r. , dantona f. , gallerani s. , maiolino r. , tornamb a. , 2012a , mnras , 420 , 1442 ventura p. , di criscienzo m. , schneider r. , carini r. , valiante r. , dantona f. , gallerani s. , maiolino r. , tornamb a. , 2012b , mnras , 424 , 2345 vandenberg , d. a. , bergbusch p. a. , dowler p. d. , 2006 , apjs , 162 , 375 ventura p. , marigo p. , 2009 , mnras , 399 , l54 ventura p. , marigo p. , 2010 , mnras , 408 , 2476 ventura p. , dantona f. , mazzitelli i. , 2000 , a&a , 363 , 605 ventura p. , zeppieri a. , mazzitelli i. , dantona f. , 1998 , a&a , 334 , 953 ventura p. , di criscienzo m. , carini r. , dantona f. , 2013a , mnras , 431 , 3642 ventura p. , di criscienzo m. , dantona f. , vesperini e. , dercole a. , tailo m. , dellagli f. , 2013b , mnras , in press wachter a. , schrder k. p. , winters j. m. , arndt t. u. , sedlmayr e. , 2002 , a&a , 384 , 452 wachter a. , winters j. m. , schrder k. p. , sedlmayr e. , 2008 , a&a , 486 , 497 wallerstein g. , knapp g. r. , 1998 , ara&a , 36 , 369 wood p. r. , 1979 , apj , 227 , 220